A health management system for catheter stent platforms based on digital twin technology

Abstract

This invention discloses a health management system for a jacket platform based on digital twin technology, comprising a data acquisition module, a marine data transmission module, a land data receiving module, a data preprocessing module, a distributed data computing module, and a digital twin module. It includes: a primary digital twin model for performing simulation calculations and early warning of jacket platform health indicators; a secondary digital twin model for providing further data support to a tertiary digital twin model; and a tertiary digital twin model for predicting the overall health status of the jacket platform. This invention is designed to address the marine environment and practical application needs of jacket platforms, focusing on data acquisition, transmission, preprocessing, and computation. It applies digital twin technology to the health management of jacket platforms, enabling comprehensive monitoring of the stress on the entire jacket and key locations, ensuring the safe and stable operation of the jacket platform.

Description

Technical Field

[0001] This invention relates to the field of health management technology for marine engineering equipment, and in particular to a health management system for a jacket platform based on digital twin technology. Background Technology

[0002] The environments in which jacket platforms are used are typically quite harsh, mainly in the following aspects:

[0003] Marine climate environment, including:

[0004] Jacket platforms are typically deployed in vast ocean areas and therefore must withstand the effects of various marine climates. These effects include, but are not limited to:

[0005] Storms: Storms in the ocean can bring strong winds and huge waves, causing enormous impacts and vibrations to the jacket platform.

[0006] Waves: Continuous wave action can cause fatigue damage to the platform, especially in deep-sea areas where wave energy is greater and poses a threat to the platform's stability.

[0007] Ocean currents: Strong ocean currents may scour and erode the jacket platform, affecting the stability of its foundation structure.

[0008] Sea ice: In cold waters, sea ice can cause compression and damage to platforms, especially jacket platforms operating in ice-covered areas, where the impact of ice on the platform needs to be carefully monitored.

[0009] Marine life and corrosion, including:

[0010] Marine organism attachment: Due to their long-term exposure in the ocean, jacket platforms are prone to becoming sites for marine organism attachment. These organisms may include shellfish, algae, etc. Their attachment not only increases the platform's weight and drag but may also cause corrosion and damage to the platform's structure.

[0011] Seawater corrosion: The salt and electrolytes in seawater are corrosive to the metal structure of the jacket platform. Parts that are submerged in seawater for extended periods are susceptible to corrosion, leading to a decline in material properties and even structural failure.

[0012] Geological conditions, including:

[0013] Seabed Soil: The foundation structure of a jacket platform is typically driven into the seabed soil, making the soil properties crucial to the platform's stability. Seabed soil properties vary greatly across different sea areas, ranging from soft, compressible soil layers to hard rock layers. These soil conditions place high demands on the design and construction of the jacket platform's pile foundations.

[0014] Earthquakes: Marine earthquakes are another natural hazard faced by jacket platforms. Earthquakes can trigger the movement and liquefaction of seabed soil, leading to instability of the platform's foundation structure.

[0015] Under the aforementioned harsh and complex sea conditions, comprehensive health management of the jacket platform and timely detection and handling of potential safety hazards are of great significance.

[0016] In recent years, digital twins, as a technology that fully utilizes models, data, intelligence, and integrates multiple disciplines, have been applied to the entire product lifecycle, serving as a bridge and link between the physical and information worlds. As one of the most dynamic scientific technologies under the Industry 4.0 framework, digital twins have promoted a further upgrade from intelligent manufacturing to intelligent health management. By combining with big data, the Internet of Things, artificial intelligence, and other technologies, it can achieve health management and maintenance throughout the entire lifecycle of equipment and products. This includes virtual-physical mapping of structural behavior, monitoring and diagnosis of operational status, prediction of future behavior, and optimization and control of workflows. Therefore, it has become a highly promising method for equipment health status monitoring, predictive assessment, and operation and maintenance management. How to utilize digital twin technology for comprehensive, accurate, and timely health management of jacket platforms has become an urgent technical problem to be solved. Summary of the Invention

[0017] This invention provides a catheter stent platform health management system based on digital twin technology to overcome the above-mentioned technical problems.

[0018] To achieve the above objectives, the technical solution of the present invention is as follows:

[0019] The data acquisition module is used to acquire high-frequency, real-time monitoring data of the jacket platform from different types of data acquisition devices through a combination of nested multi-threading and asynchronous multi-threading. The monitoring data of the jacket platform includes structural data of the jacket platform and marine environmental load monitoring data.

[0020] The data acquisition module is also used to acquire the jacket platform model data;

[0021] The marine data transmission module is used to transmit the data collected by the data acquisition module to the land data receiving and processing module through a set marine-land data transmission method.

[0022] The land data receiving module is used to receive data transmitted by the marine data sending module and to import and store the received data according to the set import and storage rules.

[0023] The data preprocessing module is used to perform preprocessing operations on the stored data, including: data noise reduction and data missing data filling.

[0024] The distributed data computing module includes several data computing sub-modules, and each data computing sub-module is used to perform one or more of the following processing on the data according to the data type of the preprocessed data: data decomposition, data query, and early warning calculation.

[0025] The digital twin module includes:

[0026] A Level 1 digital twin model, which is a SACS simulation model based on preprocessed data and corrected data of the duct stent platform model, is used for duct stent platform simulation calculations and early warning of duct stent platform health indicators;

[0027] A second-level digital twin model, which is an application database formed based on the data processed by the data distributed computing module and the first-level digital twin model, is used to further provide data support for the third-level digital twin model;

[0028] A three-level digital twin model, which is a deep learning model trained based on the two-level digital twin model, is used to predict the overall health status of the current jacket platform.

[0029] The data interaction management module provides a data display interface for the jacket platform, enabling users to obtain and manage jacket platform information in real time.

[0030] Beneficial effects: This invention includes a data acquisition module that uses a nested multi-threaded and asynchronous multi-threaded combination to collect high-frequency, real-time monitoring data and model data of the jacket platform from various data acquisition devices; a marine data transmission module that transmits data collected by the data acquisition module to a land-based data receiving and processing module using a predefined sea-land data transmission method; a land-based data receiving module that receives data transmitted from the marine data transmission module and imports and stores the received data according to predefined import and storage rules; a data preprocessing module that preprocesses the stored data, including data noise reduction and data missing information filling; several data calculation sub-modules that perform one or more processing operations, such as data computation, data query, and early warning calculation, based on the data type of the preprocessed data; and a digital twin module comprising a first-level digital twin model, a second-level digital twin model, and a third-level digital twin model to predict the overall health status of the current jacket platform.

[0031] This invention addresses the marine environment and practical application needs of the jacket platform by designing various aspects such as data acquisition, transmission, preprocessing, and calculation. It also applies digital twin technology to the health management of the jacket platform. By establishing a first-level twin model, a second-level twin model, and a third-level twin model, it achieves comprehensive monitoring of the stress on the entire jacket and key locations, ensuring the safe and stable operation of the jacket platform. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram of the structure of a catheter stent platform health management system based on digital twin technology in this invention;

[0034] Figure 2 This is a system block diagram of the sea-land long-distance data transmission system in an embodiment of the present invention;

[0035] Figure 3 This is a system data flow diagram of a land-sea remote data transmission system provided by the present invention in an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the ring boundary early warning method in an embodiment of the present invention;

[0037] Figure 5 This is a schematic diagram of the upper and lower limit early warning method in an embodiment of the present invention;

[0038] Figure 6 This is a schematic diagram of the settlement difference early warning method in an embodiment of the present invention;

[0039] Figure 7 This is a schematic diagram of the main leg weight center of gravity sensor and node arrangement of the multi-leg support structure of the jacket platform in an embodiment of the present invention;

[0040] Figure 8 This is a schematic diagram of the regular arrangement of the eight main legs in an embodiment of the present invention;

[0041] Figure 9 This is a schematic diagram of the irregular arrangement of the eight main legs in an embodiment of the present invention;

[0042] Figure 10 This is a structural diagram of the sensor outer protective cover model designed in an embodiment of the present invention;

[0043] Figure 11 This is a structural diagram of the sensor outer cover model after applying a load in an embodiment of the present invention;

[0044] Figure 12 This is a structural diagram of the strain sensor clamping block model designed in an embodiment of the present invention;

[0045] Figure 13 This is a structural diagram of the strain sensor outer cover and strain sensor clamping block designed in an embodiment of the present invention;

[0046] Figure 14 This is a schematic diagram of the main leg arrangement of the guide frame in an embodiment of the present invention;

[0047] Figure 15 This is a schematic diagram of the pile-soil py curve in an embodiment of the present invention;

[0048] Figure 16 This is a flowchart illustrating the SACS simulation model correction method in an embodiment of the present invention.

[0049] Figure 17 This is a rose diagram showing the combined wind speed and direction distribution in the Liuhua Sea area of ​​the South China Sea in an embodiment of the present invention.

[0050] Figure 18 This is a rose diagram showing the combined distribution of flow velocity and direction in an embodiment of the present invention.

[0051] Figure 19 This is a schematic diagram of the surface velocity distribution fitting (following the Gumbel distribution) in the summer of 2022 in an embodiment of the present invention;

[0052] Figure 20 This is a schematic diagram of the distribution fitting of measured wind speed data following a two-parameter Weibull distribution in an embodiment of the present invention.

[0053] Figure 21 This is a schematic diagram of wave distribution parameter fitting that follows a 3-parameter Weibull distribution in an embodiment of the present invention;

[0054] Figure 22 This is a schematic diagram of the fiber optic sensor arrangement in an embodiment of the present invention;

[0055] Figure 23 This is a schematic diagram illustrating the process of establishing and updating the secondary twin model in an embodiment of the present invention;

[0056] Figure 24 This is a schematic diagram of the force analysis of a two-dimensional model of a duct stent in an embodiment of the present invention;

[0057] Figure 25 The structural diagram shows the overall stress prediction model of the jacket structure and its members provided by this invention.

[0058] In the diagram: 1. Wireless node; 2. Main leg of the jacket; 3. Weight center of gravity strain sensor; 4. Jacket platform. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0060] This embodiment provides a catheter stent platform health management system based on digital twin technology, such as... Figure 1 As shown, it includes:

[0061] The data acquisition module is used to acquire high-frequency, real-time monitoring data of the jacket platform from different types of data acquisition devices through a combination of nested multi-threading and asynchronous multi-threading. The monitoring data of the jacket platform includes structural data of the jacket platform and marine environmental load monitoring data.

[0062] The data acquisition module is also used to acquire the jacket platform model data;

[0063] The marine data transmission module is used to transmit the data collected by the data acquisition module to the land data receiving and processing module through a set marine-land data transmission method.

[0064] The land data receiving module is used to receive data transmitted by the marine data sending module and to import and store the received data according to the set import and storage rules.

[0065] The data preprocessing module is used to perform preprocessing operations on the stored data, including: data noise reduction and data missing data filling.

[0066] The distributed data computing module includes several data computing sub-modules, and each data computing sub-module is used to perform one or more of the following processing on the data according to the data type of the preprocessed data: data decomposition, data query, and early warning calculation.

[0067] The digital twin module includes:

[0068] A Level 1 digital twin model, which is a SACS simulation model based on preprocessed data and corrected data of the duct stent platform model, is used for duct stent platform simulation calculations and early warning of duct stent platform health indicators;

[0069] A second-level digital twin model, formed through an application database, is used to further provide data support for the third-level digital twin model. The application database is formed based on the data processed by the data distributed computing module and the first-level digital twin model.

[0070] A three-level digital twin model, which is a deep learning model trained based on the two-level digital twin model, is used to predict the overall health status of the current jacket platform.

[0071] The data interaction management module provides a data display interface for the jacket platform, enabling users to obtain and manage jacket platform information in real time.

[0072] In a specific embodiment, high-frequency real-time acquisition of monitoring data for the jacket platform from different types of data acquisition devices is achieved through a combination of nested multithreading and asynchronous multithreading, including:

[0073] Multiple marine data acquisition main threads are created on an industrial control computer. Each of these marine data acquisition main threads is used to initiate the task of the industrial control computer to acquire monitoring data of the jacket platform from different types of data acquisition devices at high frequency in real time.

[0074] A media-level time controller sub-thread is nested within each of the main threads for ocean data acquisition. The industrial control computer executes the task of high-frequency real-time acquisition of monitoring data of the jacket platform from the data acquisition device based on the interval set by the media-level time controller sub-thread. Each of the data acquisition devices stores monitoring data transmitted from multiple sensors.

[0075] The media-level time controller sub-thread transmits the collected monitoring data to the marine data transmission module.

[0076] Specifically, the media-level time controller, also known as the Multimedia Timer, is an advanced timer service provided by the Windows API. Compared to system timers and control timers, it offers higher precision and lower latency, reaching the 1-millisecond level. This makes it suitable for scenarios requiring precisely timed task execution and meets the requirements of real-time data acquisition. Therefore, in this embodiment, a media-level time controller sub-thread is nested within each main thread of the marine data acquisition process to meet the millisecond-level real-time data acquisition needs of the jacket platform.

[0077] Specifically, in this embodiment, an asynchronous processing task is nested within the sub-thread of the media-level time controller. The industrial control computer extracts and analyzes monitoring data from the cache area based on the asynchronous processing task according to the set processing task time, and saves the analyzed monitoring data to the database for subsequent actual needs.

[0078] In a specific embodiment, the main thread for marine data acquisition includes: the main thread for the dynamic acquisition instrument, the main thread for the fiber optic demodulator, the main thread for the STS4 wireless test system acquisition system, the main thread for the radar equipment, the main thread for the anemometer, the main thread for the impressed current cathodic protection system, the main thread for the seabed integrated observation platform, and the main thread for memory monitoring.

[0079] The main thread of the dynamic acquisition instrument is used to start the task of the industrial control computer to acquire data from the accelerometer, tilt sensor and displacement sensor from the dynamic acquisition instrument;

[0080] The main thread of the fiber Bragg grating demodulator is used to start the task of the industrial control computer to collect data from the fiber Bragg grating sensor and the level sensor from the fiber Bragg grating demodulator.

[0081] The main thread of the STS4 wireless test system acquisition system is used to start the task of the industrial control computer to acquire weight center of gravity sensor data from the STS4 wireless test system acquisition system.

[0082] The main thread of the radar device is used to start the task of the industrial control computer to collect and monitor water level, wave height, wave peak wavelength, wave peak direction, flow velocity, flow direction, average wave height, average wavelength, average flow velocity, and average flow direction data from the radar device.

[0083] The main thread of the anemometer is used to start the task of the industrial control computer to collect wind speed and wind direction data from the anemometer.

[0084] The main thread of the impressed current cathodic protection system is used to start the task of the industrial control computer to collect data from the impressed current cathodic protection system, including the auxiliary anode release current / output voltage, reference electrode, reference electrode, sacrificial anode release current, and cathode current density sensor data.

[0085] The main thread of the seabed integrated observation platform is used to start the task of the industrial control computer to collect 75kHz flow velocity data, 75kHz flow direction data, 1200kHz flow velocity data, 1200kHz flow direction data, temperature, conductivity, pressure and salinity data from the seabed integrated observation platform.

[0086] Specifically, in this embodiment, the dynamic data acquisition device is equipped with nine accelerometers, comprising 27 data channels across the X, Y, and Z axes, with a sampling frequency of 100Hz, acquiring a total of 2700 data points per second; eight tilt sensors, comprising 16 data channels across the X and Y axes, with a sampling frequency of 10Hz, acquiring a total of 160 data points per second; and four displacement sensors, comprising eight data channels across the X and Y axes, with a sampling frequency of 10Hz, acquiring a total of 80 data points per second.

[0087] Specifically, the data interface in the dynamic data acquisition instrument is a dynamic library provided by the equipment manufacturer; the data format is a decimal number array; and the integration method is to acquire data through the dynamic library interface provided by the equipment manufacturer.

[0088] Specifically, in this embodiment, the multi-threaded operation adopts an asynchronous mode, and the main thread forms a nested asynchronous multi-threaded structure to prevent thread blocking. Simultaneously, a media-level time controller sub-thread is nested within each of the main ocean data acquisition threads. This allows control over the time interval of each data acquisition to be accurate to 1 millisecond, with an acquisition frequency of 100Hz (100 milliseconds). For example, the dynamic acquisition instrument pushes 10 sets of data each time, with each set containing 27 channels of acceleration data, 16 channels of tilt angle data, and 8 channels of displacement data, totaling 510 data points acquired each time. Acquiring data once every 100 milliseconds, 10 times per second corresponds to 100Hz, resulting in 5100 data points acquired per second. After acquiring and caching the data, an asynchronous processing task created within the media-level time controller parses and saves the acquired data. This ensures that the acquisition process of the main thread is not blocked while data parsing and processing are performed, including processing the raw data using specified calculation formulas and analyzing abnormal data. After data processing, a database connection is created, and the data is saved to the corresponding data table.

[0089] Specifically, current data acquisition requirements for jacket platform structure monitoring are generally divided into the following acquisition cycles, depending on the type of equipment: 1Hz (acquiring data once per second), 10Hz (acquiring data once every 10 seconds), 50Hz (acquiring data once every 50 seconds), and 100Hz (acquiring data once every 100 seconds). Multiplying these by the different data collected from each sensor, the number can exceed 100. Depending on the offshore platform, the data volume is approximately 3000 data points per second or more, and the daily data volume is around 10GB. Most commercially available software timer controls are based on second-level controls. Although they can be configured for millisecond-level control, there is always a time difference of 10-30 milliseconds, which cannot meet the high-speed data acquisition requirements of jacket platform structure monitoring, resulting in missed data acquisitions. Given the complexity of the marine environment, missed data acquisitions could lead to safety issues. Therefore, this embodiment uses an existing media-level time controller to control the data acquisition process.

[0090] Specifically, in this embodiment, the fiber Bragg grating demodulator is equipped with a total of 130 fiber Bragg grating sensors, including 130 temperature-compensated data points and 130 wavelength data points, totaling 260 data points; it is also equipped with four level sensors, including four temperature-compensated data points and four wavelength data points, totaling eight data points; the data protocol is TCP / IP (Transmission Control Protocol / Internet Protocol); the data format is set to hexadecimal data packets; the integration method is set to a handshake operation via TCP protocol, where one party initiates the handshake process and the other party listens. When the listening end detects a valid connection, it immediately starts the data transmission process, continuously sending data to the client at certain time intervals. By creating a media-level time controller to control the time interval of each data acquisition to an accuracy of 1 millisecond, with an acquisition frequency of 10Hz, one set of data is read each time, including 130 temperature-compensated data points, 130 wavelength data points, and 8 level data points, with 10 reads per second, for a total of 10 sets of 2680 data points. After the data is acquired and cached, the acquired data is parsed and saved through an asynchronous processing task created within the media-level time controller. This ensures that the acquisition process is not blocked while parsing the hexadecimal data into decimal data. Then, by using a specified lookup table and formula, the data is converted into stress-strain data and abnormal data is analyzed. After the data is processed, a database connection is created, and the data is saved to the corresponding data table.

[0091] Specifically, the STS4 structural wireless testing system's data acquisition system incorporates a weight center of gravity sensor. This device is imported and its data is read from a specified data file directory. Weight center of gravity sensor data is obtained through filtering and parsing. The data interface is set to .tdms file format, generated by the manufacturer's software. The integration method involves reading and parsing the .tdms file via a dynamic library interface, obtaining weight center of gravity sensor data through data parsing and filtering. Data files are generated every two minutes. A media-level time controller reads the latest .tdms data file from a designated folder directory every two minutes, parses the file format, and obtains the specified data content. An asynchronous processing task created within the media-level time controller parses and calculates the corresponding sensor data, analyzes abnormal data, and after processing, removes the read files to a designated directory for saving. Then, a database connection is created, and the data is saved to the corresponding data table.

[0092] Specifically, in this embodiment, the monitoring data of the integrated wave radar equipment is collected. The communication interface adopts the Modbus-RTU protocol, and the serial communication interface is RS485 / TCP. The data format is hexadecimal double-precision floating-point. The integration method is to send specified instructions to the slave device to obtain the corresponding data by using the ModbusRTU protocol. The data collection interval is once every two minutes, and one hexadecimal data is collected each time. Through the created media-level time controller, the collection frequency is controlled to be once every two minutes, and one set of data is pushed each time. After the data is acquired, the hexadecimal data is parsed by the asynchronous processing task created in the media-level time controller, the data is converted into decimal data, and abnormal data is parsed and analyzed. After the data is processed, a database connection is created, and the data is saved to the corresponding data table.

[0093] Specifically, in this embodiment, the system integrates and collects anemometer and wind direction monitoring data. The communication interface uses the Modbus-RTU protocol, and the data format is hexadecimal. The integration method involves establishing a connection with the hardware device using the Modbus RTU protocol to receive data pushed by the device in real time. The data acquisition interval is one hexadecimal data point per second. The acquisition frequency is controlled by a media-level time controller to be once per second, with each push containing one set of data, including wind speed and wind direction information. After acquiring the data, the hexadecimal data is parsed through an asynchronous processing task created within the media-level time controller. The data is converted into decimal data and subjected to logical processing, wind direction and wind speed conversion, and analysis of abnormal data. After processing the data, a database connection is created, and the data is saved to the corresponding data table.

[0094] Specifically, in this embodiment, monitoring data from the impressed current cathodic protection system is integrated and collected. The collected data includes data from 100 auxiliary anode release current / output voltage sensors, 24 reference electrodes (Ag), 24 reference electrodes (Zn), 13 sacrificial anode release current sensors, and 4 cathode current density sensors. The data interface is a WebAPI, and the data format is a JSON string. The integration method involves establishing a connection via the WebAPI to a specified IP address, acquiring specified configuration information and data files in batches every minute. The number of sensors varies depending on the project. A media-level time controller controls the acquisition frequency to acquire data once per minute, acquiring multiple sets of different types and sensor data each time. After acquiring the data, an asynchronous thread is created to control the data parsing and saving operations. The asynchronous processing task created within the media-level time controller parses the JSON data, performs logical processing, calculates various parameter values, analyzes abnormal data, and after processing the data, creates a database connection and saves the data to the corresponding data table.

[0095] Specifically, in this embodiment, monitoring data from a comprehensive seabed observation platform is integrated and collected. The collected data includes 75kHz current velocity data for layers 1-25, 75kHz current direction data for layers 1-25, 1200kHz current velocity data for layers 1-25, 1200kHz current direction data for layers 1-25, temperature, conductivity, pressure, and salinity data. The data interface is ModbusTCP; the data format is hexadecimal double-byte unsigned integer. The integration method involves establishing a connection with the hardware device using the ModbusTCP communication protocol, sending specified instructions according to a dictionary table provided by a third party, and reading the corresponding data. A media-level time controller controls the acquisition frequency to once per hour. According to the instruction set in the dictionary table, data acquisition instructions are sent sequentially to acquire the corresponding sensor data. After acquiring the data, an asynchronous processing task created within the media-level time controller parses the collected hexadecimal data, converts the hexadecimal data into decimal data for logical processing, analyzes abnormal data, and after processing the data, creates a database connection and saves the data to the corresponding data table.

[0096] In a specific embodiment, a power supply status monitoring main thread is created on the industrial control computer. The power supply status monitoring main thread is used to start the task of the industrial control computer to monitor the UPS power supply status in real time.

[0097] A first timer thread is nested within the main thread for power supply status monitoring. The industrial control computer acquires UPS power supply status data based on the interval set by the first timer thread, and determines whether to shut down the industrial control computer based on the acquired UPS power supply status data.

[0098] Specifically, a first timer thread is nested within the main thread for power supply status monitoring. The industrial control computer acquires UPS power supply status data based on the interval set by the first timer thread. By nesting asynchronous processing tasks within the first timer thread, the asynchronous processing tasks parse the hexadecimal UPS power supply status data and determine the UPS power supply status according to a preset UPS power supply status lookup table based on the parsing results. When the UPS power supply status is determined to be "discharging", the industrial control computer sends an instruction to read the UPS power and voltage. If the acquired UPS power and voltage reach the set threshold, the shutdown instruction in the script file is executed, thereby shutting down the Windows system.

[0099] Specifically, the working environment of the jacket platform may experience power outages due to typhoons, power supply issues, and uncontrollable on-site factors. During high-frequency writes of large amounts of data, a sudden power outage can lead to a series of problems, including hardware damage, software malfunctions, and unrecoverable database table corruption. Since most jacket platforms operate without network access, this significantly hinders subsequent hardware, software, and database maintenance and repair. Therefore, a UPS power supply system needs to be activated for temporary power after a power outage. However, due to the limited power capacity of UPS systems, which typically only last 2-3 days, if the power supply to the jacket platform is not restored in time, the UPS system may automatically shut down due to insufficient power. If the UPS system automatically shuts down, the industrial control computer may suffer software and hardware damage due to the sudden power loss. Therefore, a UPS power monitoring task is set up, and the monitoring results are compared with preset UPS parameters. The power supply status is compared with the UPS power supply status table. If the UPS power supply system is in charging state, it means that the power supply status of the jacket platform is normal and the UPS power supply system is not being used for power supply. If the UPS power supply system is in discharging state, it is necessary to obtain the specific power and voltage of the UPS power supply system and determine whether the power and voltage of the UPS power supply system have reached the set threshold value according to the UPS power supply status table. If the power and voltage of the UPS power supply system reach the threshold value, it means that the power supply capacity of the UPS power supply system can no longer meet the requirements. At this time, the industrial control computer will activate the low power system protection function, that is, gradually shut down the relevant software acquisition and monitoring programs and shut down the computer to avoid damage to hardware devices, software systems and database files caused by sudden power failure.

[0100] In a specific embodiment, a memory cleanup main thread is created on the industrial control computer, and the memory cleanup main thread is used to start the task of cleaning up the memory of the industrial control computer;

[0101] A second timer thread is nested within the main memory cleanup thread. The industrial control computer cleans up memory based on the second timer thread at set intervals, using a memory reclamation mechanism and the C# automatic garbage collection mechanism.

[0102] Specifically, in this embodiment, while each nested asynchronous multithread synchronously executes its own task, the software will gradually generate memory usage and be unable to release memory. Therefore, a second timer thread is created, and a time controller based on Windows time control is implemented through it to monitor the software's memory usage in real time. Asynchronous processing tasks are created within the time controller based on Windows time control. Since the C# automatic garbage collection mechanism may sometimes fail to achieve the expected results, this embodiment uses a memory reclamation mechanism and the C# automatic garbage collection mechanism to perform garbage collection every certain period of time, thereby ensuring that the acquisition system can run stably for a long time.

[0103] Specifically, garbage collection is an automatic memory management mechanism in C#. It is responsible for detecting and releasing the memory occupied by objects that are no longer in use. In C#, developers do not need to manually release memory; the garbage collector will automatically handle these tasks. This ensures that even under remote conditions, the pipeline platform can release memory in a timely manner, avoiding memory space filling and affecting data acquisition, thus ensuring the smooth progress of data acquisition.

[0104] Specifically, both the first timer thread nested within the main thread for power supply status monitoring and the second timer thread nested within the main thread for memory monitoring are designed using time controllers based on the Windows Time Control. In this embodiment, although the media-level time controller is more accurate, it also consumes more resources. The Windows Time Control-based time controller, on the other hand, is more lightweight, consumes fewer resources, and is suitable for use cases requiring data at the second level or higher. Therefore, a media-level time controller is used for data acquisition on the jacket platform, while a Windows Time Control-based time controller is sufficient for scenarios requiring data acquisition at the second level or higher, such as memory and power supply monitoring. By using a hybrid approach to time controllers, limited system resources can be utilized more efficiently.

[0105] In a specific embodiment, within the main thread of the dynamic data acquisition instrument, the asynchronous processing task processes the acquired data from the accelerometer, tilt sensor, and displacement sensor as follows:

[0106] ,

[0107] ,

[0108] ,

[0109] In the formula, 1 and These are the processed accelerometer sensor values ​​and the measured accelerometer sensor values, respectively. and These are the processed tilt sensor value and the measured tilt sensor value, respectively. 1 and These represent the processed displacement sensor value and the measured displacement sensor value, respectively; c represents zero drift; d represents gain; b represents bias.

[0110] In the main thread of the fiber Bragg grating demodulator, the asynchronous processing task processes the acquired data from the MOI-type fiber Bragg grating sensor as follows:

[0111] ,

[0112] In the formula, FG S is the strain factor; T Temperature sensitivity; λ 0S For calibrating the wavelength; λ S λ represents the wavelength of the measured strain. 1S λ is the initial strain wavelength; 0T For calibration wavelength; λ T The measured temperature-compensated wavelength; λ 1T Initial temperature-compensated wavelength; CTE S CTE is the coefficient of thermal expansion. T The coefficient of thermal expansion of the temperature probe;

[0113] The data collected from the ZO130 fiber Bragg grating sensor were processed as follows:

[0114] με =((λ-λ0)-( -λ T0 ) K T ) / K*,

[0115] In the formula, λ is the strain test wavelength; λ0 is the zero strain wavelength; For temperature-compensated wavelength; λ T0 K is the temperature-compensated zero-point wavelength; K* is the sensitivity coefficient; K T This is the temperature compensation coefficient.

[0116] In this embodiment, because the jacket platform operates in a long-term unattended environment, the stability of software execution, the accuracy of data, and the accuracy of data acquisition cycles are extremely important. Not only must the data accuracy be guaranteed, but the accuracy of the data acquisition cycle must also be ensured. Furthermore, the connection status of each hardware component and the running status of the software must be monitored in real time. Different types of monitoring devices need to be added as needed. The system must also be able to automatically run acquisition and monitoring programs without human intervention after the acquisition program is shut down or restarted due to an unexpected power outage. Therefore, to ensure that the acquisition software can start acquisition without re-login after a computer restart, this embodiment adopts a Windows service acquisition program and monitoring client architecture. The acquisition program operates as a Windows service, with the acquisition service set to automatic startup. When the computer system starts, the Windows acquisition service automatically starts. After the service starts, the data acquisition program can be started automatically without entering a password to log in to the computer system, ensuring automatic startup after a system restart and meeting the requirements of the unattended platform.

[0117] In this embodiment, the method is implemented based on the .NET development platform using the C# language and client-server architecture, using the .NET Framework 7.2, and running on the Windows 10 system. The C# language and development framework have significant advantages in long-term operation on the Windows system and in terms of memory usage and release, ensuring the long-term stable operation of the system.

[0118] In this embodiment, a third timer task is also created to monitor the service operation status of the current acquisition system periodically (once every 10 seconds). If the acquisition service is unexpectedly shut down or fails to start after the system is restarted, the watchdog timer task can start the acquisition service to ensure that the acquisition service can operate normally.

[0119] The working environment of the jacket platform may experience power outages due to typhoons, power supply issues, and other uncontrollable factors. During high-frequency writing of large amounts of data, a sudden power outage can lead to a series of problems, including hardware damage, software malfunctions, and unrecoverable database table corruption. Furthermore, most jacket platforms operate without network access, significantly hindering subsequent hardware, software, and database maintenance and repair. Therefore, this embodiment includes a main data export thread, nested within which is a fourth time controller sub-thread. The industrial control computer exports data based on the intervals set by this fourth time controller sub-thread. Specifically, due to the large volume of data in the industrial control computer system, the fourth time controller sub-thread sets different export times based on different device types, allowing the industrial control computer to export data as CSV files at different time intervals. For example, large tables are exported at 5-10 minute intervals, while smaller tables are exported at 1-hour intervals or daily intervals to prevent the exported data files from becoming too large to open.

[0120] Specifically, an exception handling mechanism is also set up in this embodiment. When the hardware device connection is abnormal, the industrial control machine will try to connect to the hardware repeatedly and generate abnormal data to save to the database. When the database connection is abnormal, the abnormal data will be saved to the specified directory in the form of text for timely troubleshooting.

[0121] In this embodiment, by setting up a combination of nested multithreading and asynchronous multithreading, the problems caused by the large number and variety of lower-level hardware, as well as the different communication protocols and data formats, are effectively solved. This addresses numerous issues such as communication anomalies, congestion, and failure to meet acquisition frequency targets due to hardware protocol conflicts. The goal is to achieve parallel and independent operation of each nested thread, with no interference between sub-threads within nested threads. Furthermore, the combination of nested and asynchronous multithreading better utilizes computer CPU resources, allowing each CPU core to be allocated to different threads for task processing. This significantly improves program throughput, enabling the host computer software to handle multiple tasks simultaneously when integrated with multiple lower-level hardware, thus improving program response speed and operating efficiency, and ultimately enhancing the overall processing power and execution efficiency of the software. In addition, nested multithreading programming offers advantages such as avoiding blocking, decoupling, improved error handling capabilities, and flexible task scheduling. By combining millisecond-level time controls, it can collect large amounts of data at high speed while enabling each device to connect, process, and save independently without interfering with each other, thus ensuring the integrity of each business process and data.

[0122] This embodiment addresses the complex offshore platform application environment involving multiple devices, multiple protocols, large amounts of data, and the need for high-frequency data acquisition. It proposes a method for acquiring detection data from the jacket platform, employing a combination of nested and asynchronous multithreading to acquire data in real time. This allows each acquisition task to run in parallel without interference, enabling the acquisition of large amounts of data from various monitoring devices, sensors, data transmission protocols, and data formats on the jacket platform. This avoids congestion and anomalies during the large-scale data acquisition process, ensuring the integrity of the jacket platform's data acquisition and facilitating subsequent research using the complete data.

[0123] In a specific embodiment, the defined sea-land data transmission method includes:

[0124] According to the data type of the monitoring data of the jacket platform, corresponding raw data files are formed. Based on the number of raw data files, the corresponding time span and the corresponding processing thread, the monitoring data of the jacket platform is split in chronological order according to the corresponding time span and formed into compressed files and corresponding verification information, timestamps and ID information.

[0125] Specifically, signals are sent to the land data receiving and processing module at fixed time intervals, and the current network status is judged based on whether the feedback signal sent by the land data receiving and processing module can be received. When the network status is normal, a compressed file and corresponding verification information are sent to the land data receiving and processing module, and a data transmission progress table representing the information of the sent compressed file and the real-time reception status of the compressed file is generated.

[0126] Specifically, by setting a time span, large data packets are broken down into several smaller data packets. Even when network conditions are poor and large data packets cannot be transmitted, some smaller data packets can still be sent, thus improving the transmission rate. This system segments files into independent small files. Taking a 100MB file as an example, by calculating the start and end times to determine a 100-minute span, and dividing the file into 100 small files according to a custom 1-minute span, the file will be sent. When the network is poor, even if only one small file is successfully transmitted, the receiving end can still process and use that file without having to wait for the file to be complete. This greatly improves efficiency compared to traditional full-data transmission and power-off resume transmission methods.

[0127] In a specific embodiment, the formula for calculating the configuration time span is as follows:

[0128] ,

[0129] in, Indicates the time span, This indicates network bandwidth.

[0130] Configure time spans to facilitate data splitting and transmission based on time spans.

[0131] When the network is not functioning properly, the aforementioned process of collecting maritime data and generating compressed files is executed until the network is functioning properly, at which point the generated compressed file and corresponding verification information are sent to the land data receiving and processing module.

[0132] Specifically, this embodiment also includes a sea-land communication module. During sea-land transmission, the sea-land communication module is used to monitor the data transmission status of the sea data sending module and the land data receiving module in real time. The sea-land communication module can confirm the transmission progress of the sea data sending end. First, the sea data sending module sends the compressed file and the corresponding verification information to the land data receiving module and records the names of the sent files in the data transmission progress table.

[0133] The land data receiving module receives and verifies the data. If the verification information is correct, it saves the data and sends feedback to the land-sea communication module that the file verification is correct and the data has been saved. After receiving the information, the land-sea communication module sends the information to the maritime data sending module. The maritime data sending module records the file as saved in the data transmission progress table. If there is an error, it sends feedback to the land-sea communication module, which then informs the maritime data sending module that the file verification failed and the data was not saved, and the file needs to be resent.

[0134] The land-sea communication module also functions as follows: after the land data receiving module saves the data to the database, it updates the data record table to confirm that the file has been saved. The land data receiving module then retrieves the data record table and sends the saving result to the maritime data sending module via the land-sea communication module. If the saving is successful, both the maritime data sending module and the land data receiving module delete the successfully saved data. If the saving fails, it first determines whether the failure is due to a data reception failure. If so, the maritime data sending module resends the data. If not, it retryes the saving process to further determine the cause of the failure. If the issue cannot be resolved by retrying (e.g., file corruption, human deletion, unknown garbled characters), a resend command is sent. If the issue can be resolved by retrying (e.g., system crash, unstable network conditions), it continuously retryes without sending failure information to the maritime data sending module, thus reducing network overhead and data transmission costs.

[0135] Specifically, in this embodiment, during the transmission of sea and land data, the overall communication of the sea and land communication module is implemented based on the Netty TCP communication framework. This framework can ensure the integrity of the data packets generated by the compression system under extremely high network fluctuations, and can achieve more than 100,000 TCP connections per machine, fully ensuring the high load availability of the system.

[0136] In a specific embodiment, the land data receiving module receives data transmitted by the maritime data sending module, and performs import and storage operations on the received data according to the set import and storage rules, including:

[0137] The system receives the transmitted compressed file and corresponding verification information, performs verification according to the verification information, determines whether the transmitted compressed file has been received, whether the compressed file is correct and complete, and saves the correct and complete compressed file to form a data record table; the data record table is used to characterize the information of the received compressed file and the real-time saving status of the corresponding compressed file.

[0138] The saved compressed file is decompressed according to the data record table, the decompressed data is stored in the database, and the saving status of the corresponding compressed file in the data record table is updated.

[0139] Specifically, in this embodiment, the data transmission status between the marine data transmission module and the land data receiving module is monitored in real time. After receiving the data reception result and verification result from the land data receiving module, feedback is sent to the marine data transmission module.

[0140] Specifically, after receiving the data storage status feedback from the land data receiving module, the system sends feedback to the maritime data sending module, which then performs corresponding operations based on the feedback, such as deleting data.

[0141] In a specific embodiment, creating multiple ocean data acquisition main threads includes:

[0142] Configure the corresponding data acquisition path, data storage path and time span according to the data type of the maritime data, calculate the number of corresponding processing threads according to the number of files of different data types of maritime data, and calculate the number of processing threads to be allocated to each file according to the time span of different data types.

[0143] Specifically, the number of processing threads is calculated based on the number of files of different data types in the maritime data, and the number of processing threads to be allocated to each file is calculated based on the time span of different data types, including:

[0144] The configuration field sets the number of files. If the number of files of a certain data type is greater than the configuration field, a basic thread is allocated to each file of that data type for processing. If the number of files of a certain data type is less than the configuration field, a basic thread is allocated to all files of that data type for processing.

[0145] A configuration field sets an upper limit for the time interval. If the time interval of data in a certain data file of a certain data type is greater than the set configuration field, the data in the file is split according to the time span, and the number of threads allocated after calculating the time span is obtained, and the split data is processed in parallel. If the time interval of data in a certain data file is less than the set configuration field, one thread is allocated to process the data in the file.

[0146] Specifically, in this embodiment, time spans are configured for different maritime data, as shown in Table 1.

[0147] Table 1

[0148]

[0149] Specifically, the marine data transmission module is used to collect marine data and form corresponding raw data files according to the data type. Each raw data file contains data from a sensor over a period of time. The module manages data transmission and constructs a data transmission schedule. The data transmission schedule records information about compressed files sent by the marine data transmission module and the real-time reception status of the compressed files. Based on the current schedule data and time span, the module calculates the time period for collecting new marine data.

[0150] Specifically, the specified time span includes the configured time type and span field. The types are 1: seconds, 2: minutes, 3: hours, and 4: days. For example, if the type is set to 2 and the span field is set to 10, it means that the file will be split into smaller files every 10 minutes at the minute level. Or, if the type is set to 1 and the span is set to 30, it means that the file will be split into smaller files every 30 seconds at the second level. By reading the start and end times of the file, the overall span of the data in the file is calculated, and the number of file segments that can be split into is calculated based on the custom time span.

[0151] The maritime data is saved into multiple raw data files according to the sensor type. Two different processing threads are allocated to the maritime data based on the number of raw data files, the corresponding time span, and the threads calculated based on the number of files and the time span, respectively. The allocated processing threads then split the maritime data files into several data packets in chronological order according to the time span. Each data packet contains maritime data for a specific time period, which can be directly accessed for subsequent processing after transmission. The data packets are first compressed to generate a ".csv" file, and then a second compression is performed using the Univocity method to compress the ".csv" file into a ".zip" file. The file size after these two compressions is only one-tenth of the original data file size. Verification information, timestamps, and ID information are generated from the compressed file.

[0152] In a specific embodiment, the scheme for calculating the number of processing threads corresponding to the number of files of different data types in maritime data, and calculating the number of processing threads to be allocated to each file based on the time span of different data types, is as follows:

[0153] The configuration field sets the number of files. If the number of files of a certain data type is greater than the configuration field, a basic thread is allocated to each file of that data type for processing. If the number of files of a certain data type is less than the configuration field, a basic thread is allocated to all files of that data type for processing.

[0154] A configuration field sets an upper limit for the time interval. If the time interval of data in a certain data file of a certain data type is greater than the set configuration field, the data in the file is split according to the time span, and the number of threads allocated after calculating the time span is obtained, and the split data is processed in parallel. If the time interval of data in a certain data file is less than the set configuration field, one thread is allocated to process the data in the file.

[0155] Specifically, a configuration field for the number of files is set. If the number of files of a certain data type is greater than the configuration field, then a basic thread is allocated to process each file of that data type. If the number of files of a certain data type is less than the configuration field, then a basic thread is allocated to process all files of that data type.

[0156] Taking a file count of 5 as an example, if there are 6 data files in the scanned directory, the number of files is greater than the configured number of files, i.e., 5. The system will automatically allocate a thread to process each of these 6 data files. If there are 4 data files in the scanned directory, the number of files is less than the configured number of files. In this case, a thread will be allocated to process each of the 4 data files.

[0157] The data is stored in a ".csv" format file, with one data entry per line, and each data entry contains time information;

[0158] A configuration field sets an upper limit for the time interval. If the time interval of data in a data file of a certain data type exceeds the set configuration field, the data in the file is split according to the time span, and the number of threads allocated after calculating the time span is obtained. The split data is then processed in parallel. If the time interval of data in a data file of a certain data type is less than the set configuration field, then one thread is allocated to process the data in that file.

[0159] Taking a configuration field with an upper limit of 60 minutes for the time interval and a time span of 5 minutes as an example, if a data file contains more than 60 minutes of data, then the time interval of the data in this data file exceeds the set configuration field. The data is split into 12 files according to the time span, that is, 12 threads are allocated to process these 12 data files at the same time, that is, the 12 data files are compressed at the same time.

[0160] In a specific embodiment, the scheme for splitting the maritime data according to the corresponding time span in chronological order and forming compressed files along with corresponding verification information, timestamps, and ID information is as follows:

[0161] The maritime data is split into data packets according to the corresponding time span and in chronological order. The data packets are then compressed twice to generate compressed files. In this embodiment, the data packets are first compressed into ".csv" format files, and then compressed into ".zip" format files using the Univocity method.

[0162] Verification information is generated based on the filename of the compressed file. The verification information includes the total file length, MD5 checksum, file name, file type, file date, and file path. A timestamp and ID information are added to the compressed file.

[0163] The verification information is encapsulated into header information in JSON format, and the compressed file is temporarily stored.

[0164] Specifically, the collected sensor data is converted into ".csv" format, compressed again into a ".zip" format archive, and named in a specific format, with the naming rule being "data type + start and end time", for example: 1_2023111812000020231118130000_.zip;

[0165] Based on the compressed filename, generate verification information, including the total file length, MD5 checksum, filename, file type, file date, and file path. Add a timestamp and ID information to the compressed file and store them in memory.

[0166] For each compressed file that needs to be transmitted, a hash calculation is performed, and the calculation result is encrypted with MD5. The encrypted result is used as the ID information of the compressed file, which is the digital signature of the compressed file. At the same time, a timestamp is added to each compressed file that needs to be transmitted. The timestamp is the time span of the data collection time in the compressed file.

[0167] The verification information is encapsulated into header information in JSON format, stored in memory, and awaits transmission.

[0168] Temporarily store compressed data files in ".zip" format;

[0169] Check if the network status is normal. If the network status is normal, send data by sending the encapsulated header information and the corresponding temporarily stored compressed file to the land data receiving module.

[0170] If the network status is abnormal, continue to perform data collection, export, compression, and encapsulation operations, and store the data in a temporary directory. Wait until communication is normal before sending it together. The temporary directory is a folder for storing temporary files.

[0171] Due to network instability in the maritime data transmission module, there is a possibility that the maritime data transmission module may be unaware that the land-based data receiving module has already received the data, resulting in duplicate transmissions. To address this issue, this system adds timestamps and ID information to files. Each time a file is sent, a header containing the ID and timestamp is sent. After sending the file, the sliced ​​file is then sent. At this point, the land-based data receiving module can determine whether the file has already been received based on the ID, timestamp, and data. If it has been received, the receiving of that file is abandoned, saving transmission resources and improving efficiency.

[0172] In a specific embodiment, such as Figure 2 and Figure 3 As shown, the land data receiving module receives the compressed file and corresponding verification information transmitted by the maritime data sending module, forms a data record table, verifies the verification information, determines whether the transmitted compressed file has been received, whether the compressed file is correct and complete, and saves the correct compressed file.

[0173] (1) Receive the compressed file and corresponding verification information transmitted by the marine data transmission module. Verify the verification information. If the verification result is correct, receive the compressed file transmitted by the marine data transmission module and save the data compressed package according to the header to form a data record table.

[0174] (2) If the verification result is incorrect, feedback is sent to the maritime data transmission module, which needs to resend the compressed file and the corresponding verification information;

[0175] (3) Compare the header data with the data record table to determine whether the verification information and the corresponding compressed file transmitted by the maritime data transmission module have been received. If they have been received, terminate the current reception. If they have not been received, repeat steps (1)-(2) to verify and save the verification information and update the data record table.

[0176] Specifically, the data receiving, data verification, and data decompression of the land data receiving module are separate and can be performed synchronously. This means that even if there is no network between the land and sea and data transmission is impossible, the land data receiving module can still verify the files previously transmitted by the sea data sending module, determine whether the file has been received, and perform operations such as decompression, data storage, and record table updates on the transmitted data, regardless of whether it has received data from the sea data sending module. Compared to the sequential receiving, verification, and decompression work, this greatly improves data processing efficiency.

[0177] Specifically, the land data receiving module can also verify the integrity of the data file. If the verification passes, the subsequent decompression and import will proceed. If the verification fails, the subsequent decompression operation will not be performed, and the maritime data sending module will be notified to resend this part of the data, thus improving data transmission efficiency while ensuring data integrity.

[0178] In a specific embodiment, such as Figure 2 and Figure 3 As shown, the land data receiving module is further configured to decompress the received and saved compressed files according to the data record table, and update the saving status of the corresponding compressed files in the data record table. Specific steps include:

[0179] According to the data record table, decompress the compressed file;

[0180] Based on the data transmission record table, the ".zip" format compressed package is decompressed into a ".csv" format file, and then the data is entered into the database;

[0181] According to the data record table, import the decompressed file data and update the save status of the corresponding compressed file in the data record table.

[0182] Specifically, the land data receiving module can update the data record table based on the imported data, and promptly delete the saved file data based on the updated data record table. Then, it feeds back to the maritime data sending module through the land-sea communication module, so that the maritime data sending module can promptly delete the data files that have been transmitted and stored, reduce disk usage, and avoid the risk of system crash.

[0183] In a specific embodiment, the step of performing noise reduction processing on the stored data based on the data preprocessing module includes:

[0184] Set a corresponding noise reduction method for the stored data to obtain the analysis data;

[0185] An adaptive noise reduction selection method is constructed. Based on this method, the analysis data is analyzed to determine whether to change the noise reduction method corresponding to the monitoring data. Then, the stored data is denoised using the determined noise reduction method, forming a noise-reduced dataset together with the analysis data. This dataset includes:

[0186] Set the amount of data to be analyzed and the noise criteria, including the standard deviation criterion and the noise quantity criterion:

[0187] The standard deviation criterion is set to n′ times the standard deviation. Data that falls outside the range of the mean plus or minus n′ times the standard deviation is considered noisy data.

[0188] Set a specific value for the noise quantity criterion. If the noise quantity in the analyzed data is lower than the noise data quantity, the nearest neighbor difference algorithm is used for data denoising; if the noise quantity in the analyzed data is higher than the noise data quantity, the bandpass filtering algorithm is used for data denoising.

[0189] Set an analysis cycle, process the monitoring data in the next analysis cycle according to the changed noise reduction method, obtain new noise reduction data, and combine it with the analysis data to form a noise reduction dataset.

[0190] Specifically, set corresponding noise reduction methods for the monitoring data:

[0191] The data acquisition frequency for strain data, weight center of gravity data, acceleration data, displacement data, and static horizontal data is relatively high (approximately 10 data points per second), and the acquisition process involves significant low-frequency interference and high-frequency noise. Therefore, a bandpass filtering algorithm is used for data noise reduction. By setting a reasonable bandpass range, low-frequency interference and high-frequency noise in the data are removed, thereby preserving the effective signal.

[0192] Since the wind direction and wind speed data are collected at a low frequency (about 1 data point per second) and are less affected by platform vibration, the nearest neighbor interpolation algorithm is used to remove data noise.

[0193] The noise in the reference electrode potential data and sacrificial anode data is caused by electromagnetic interference between sensors. This electromagnetic interference is irregular. In addition, since the values ​​of the reference electrode potential and the sacrificial anode current sensor change slowly during the lifespan of the jacket platform, the nearest neighbor interpolation algorithm is used to remove the data noise.

[0194] Based on previous engineering experience, this solution sets an initial noise reduction method for the monitoring data, providing analytical data for subsequent evaluation of the suitability of the noise reduction method.

[0195] In this embodiment, the number of data points analyzed is set to 10,000, the standard deviation criterion is set to 3 (a multiple of the standard deviation), and the noise data criterion is set to 5. The standard deviation of each of the 10,000 different data points is calculated. Data points falling outside the range of mean ± 3 times the standard deviation are identified as noise data. If the number of noise points is less than 5, the nearest neighbor interpolation algorithm is used for data denoising. If the number of noise points is greater than 5, the bandpass filtering algorithm is used for data denoising. The set period is 1 day. Every day, the data detected that day is judged to determine which denoising method to use for the data of the next day.

[0196] This solution constructs an adaptive noise reduction selection method. By analyzing the initial analysis data, it can determine whether the selected noise reduction method meets the requirements. If it does, no changes are made; if it does not, it is adjusted to a more suitable noise reduction method, which can make the obtained noise reduction data more accurate.

[0197] In a specific embodiment, the step of performing data missing data imputation processing on the preprocessed dataset based on the data preprocessing module includes:

[0198] Based on the preprocessed dataset, obtain the target sensor dataset to be supplemented;

[0199] A sensor data missing detection strategy is constructed, and the target sensor dataset is preprocessed. The preprocessed target sensor dataset is divided into training and test sets for short-term and long-term data missing types according to the sensor data missing detection strategy. The target sensor dataset includes the non-failed data of the target sensor itself and the non-failed data of other sensors.

[0200] A bidirectional recurrent neural network and a GRU unit are introduced to obtain the temporal relationship of the sensor data, and a fully connected layer is introduced to obtain the spatial relationship between the sensor data. A two-branch data processing model based on the bidirectional recurrent neural network and the GRU unit is constructed. The two-branch data processing model is used to determine different data missing types according to the sensor data missing type judgment strategy, and select the corresponding branch to process the input data.

[0201] The training sets for the two types of missing data are respectively input into the dual-branch data processing model for training to obtain the sensor data recovery model; the sensor data recovery model is used to recover missing data by using the corresponding sub-model according to different types of missing data.

[0202] The test set is input into the sensor data recovery model to obtain the predicted missing data. The sensor data recovery model is evaluated and optimized based on the error between the predicted value and the true value to obtain the optimized sensor data recovery model.

[0203] The optimized sensor data recovery model is input with the non-failure data of the target sensor itself and the non-failure data of other sensors. The model determines the type of missing data and outputs the missing data of the target sensor to supplement the missing data in the target sensor dataset, thus obtaining the preprocessed dataset.

[0204] In a specific embodiment, the dual-branch data processing model includes: an input layer, a missing data type discriminator, a network call module, a short-term missing data recovery model, a long-term missing data recovery model, and an output layer;

[0205] The missing data type discriminator is used to determine whether the missing data type is short-term or long-term based on the sensor data missing judgment strategy, and inputs the judgment result to the network call module to call the corresponding branch for training.

[0206] The network call module is used to call the short-term missing data recovery model or the long-term missing data recovery model to process the input data based on the judgment result of the missing data type discriminator.

[0207] The short-term missing data recovery model includes m fully connected layers, 1 GRU layer, and r Bi-GRU layer connected in sequence, used to recover short-term missing data;

[0208] The long-term missing data recovery model comprises m fully connected layers and n GRU layers connected in sequence, used to recover long-term missing data.

[0209] In a specific embodiment, training sets for two types of missing data are respectively input into the dual-branch data processing model for training to obtain a sensor data recovery model, including:

[0210] S31: Randomly determine a set of hyperparameters, namely m, n and r, and randomly initialize the weights and biases of each network layer in the dual-branch data processing model under these parameters;

[0211] S32: Input the training data set into the missing data type discriminator through the input layer for judgment. If it is a short-term missing data type, the judgment result is input into the network call module, and enter S33 to call the short-term missing data recovery model for training. If it is a long-term missing data type, the judgment result is input into the network call module, and enter S34 to call the long-term missing data recovery model for training.

[0212] S33: Input the short-term missing data training set into m fully connected layers, and perform dimensionality reduction and feature extraction on the input data through multi-layer linear and non-linear transformations to capture the spatial correlation between different sensors;

[0213] The features extracted through m fully connected layers are input into a single GRU layer to obtain the short-term temporal dependencies in the input features.

[0214] The features processed by one GRU layer are input into r Bi-GRU layers to obtain contextual information of a certain past time period and a certain future time period in the features;

[0215] The features extracted by the Bi-GRU layer are integrated and fused through the output layer to predict the missing data at the time point corresponding to the current data and use it as the missing data at the first time point.

[0216] Add the missing data from the first time point to the training set, repeat the training process, predict the missing data from the current time point, and use it as the missing data for the second time point.

[0217] The missing data corresponding to the predicted time point of the current data is continuously added to the training set for training, until the missing data of the x*-1th time point is added to the training set for training. The missing data corresponding to the current data is then predicted and used as the missing data of the x*th time point; x* represents the time point at the end of the short-term missing data.

[0218] S34: Input the long-term missing data training set into m fully connected layers, and perform dimensionality reduction and feature extraction on the input data through multi-layer linear and non-linear transformations to capture the spatial correlation between different sensors;

[0219] The features extracted through m fully connected layers are input into n GRU layers to obtain the long-term temporal dependencies in the input features.

[0220] The features extracted by n GRU layers are integrated and fused through the output layer to predict the missing data at the time point corresponding to the current data and use it as the missing data at the first time point.

[0221] Add the missing data from the first time point to the training set, repeat the training process, predict the missing data from the current time point, and use it as the missing data for the second time point.

[0222] The missing data corresponding to the predicted current data time point is continuously added to the training set for training, until the missing data of the y*-1th time point is added to the training set for training. The missing data corresponding to the current data time point is then predicted and used as the missing data of the y*th time point; y* represents the time point of the end position of the long-term missing data.

[0223] S35: Set the loss function and optimizer: Set MSE as the loss function and use the Adam optimizer for optimization;

[0224] S36: Calculate the loss function between the predicted data and the real data, calculate the gradient based on the loss function, and use the Adam optimizer to update the weights and biases of the model based on the calculated gradient;

[0225] S37: Set the numerical range of hyperparameters, and adjust the number of fully connected layers, GRU layers and Bi-GRU layers of the model according to the set numerical range based on the grid search method to obtain the sensor data recovery model.

[0226] In a specific embodiment, the sensor data missing determination strategy includes:

[0227] Determine the start and end times of the missing data periods in the training set, and calculate the duration of the missing data periods.

[0228] Determine the length of the available data period after the missing data. The available data period is the length of time between the end time of the missing data period and the start time of the next missing data period.

[0229] Set a time threshold for missing data and a time threshold for available data. When the duration of the missing data period is less than the time threshold for missing data and the duration of the available data period is greater than the time threshold for available data, the missing data of the target sensor is classified as short-term missing data.

[0230] When the duration of the missing data period is greater than the missing data time threshold, or the duration of the available data period is less than the available data time threshold, the missing data of the target sensor is classified as long-term missing data.

[0231] In a specific embodiment, the steps corresponding to the early warning calculation provided by the data distributed computing module include:

[0232] Based on the data characteristics of different noise reduction data in the noise reduction dataset, a rule for selecting an early warning method is constructed, and an early warning method corresponding to the noise reduction data is selected according to the early warning rule. A safety range and a warning value are set to obtain the early warning result.

[0233] Based on the data characteristics of different noise-reduced data in the noise-reduced dataset, a rule for selecting the early warning method is constructed, including:

[0234] S701: When monitoring data comes from the horizontal direction and it is necessary to combine the data values ​​of the horizontal direction for early warning, select the ring boundary early warning method;

[0235] Specifically, in this embodiment, the early warning of displacement data uses the ring boundary early warning method. Under the action of sea wind, waves and currents, the platform is prone to displacement. When the displacement is large, it may pose a threat to the safety of the platform. The displacement sensor mainly measures the displacement of the jacket platform along the horizontal X and Y directions. The jacket platform may not necessarily shift along the X or Y direction alone. Therefore, it is necessary to combine the X-direction displacement and the Y-direction displacement to determine the early warning.

[0236] Because the horizontal cross-section of the jacket is usually a rectangle with varying lengths and widths, the warning thresholds for the jacket's horizontal offset in the X and Y directions are often different. Therefore, the warning threshold formed by the combination of X and Y offsets is generally elliptical. A schematic diagram of the warning method is shown below. Figure 4As shown, each sensor is configured with (X / Y) coordinate values ​​for this warning type, 8 values ​​per level, for a total of 24 values. These 8 values ​​are connected to form three nested rings, creating a warning map. After data collection, the warning level is determined based on the data's position on the map. The data with the largest offset among all sensor data is selected, and a consecutive count index (e.g., 10 times) is configured. A warning is issued when a data point appears consecutively in a Level 1, Level 2, or Level 3 warning level for 10 consecutive times. A Level 3 warning range includes Level 1 and Level 2 warnings, and a Level 2 warning range includes Level 1 warnings. For example, if 10 consecutive data points fall within the "Level 3-Level 3-Level 3-Level 3-Level 3-Level 2-Level 3-Level 3-Level 3" warning area, a Level 3 warning is issued.

[0237] S702: When the monitored data is a specific value of a certain data type and exceeds the normal value range, an early warning will be issued, selecting the upper and lower limit warning method;

[0238] In this embodiment, the method for selecting upper and lower limit warnings for reference electrode potential data, sacrificial anode current data, acceleration data, and strain data includes:

[0239] The data collected by the reference electrode is the potential value of the measuring point. When the potential value of the measuring point is not within the normal range, such as the normal range of the potential value of the Ag / Agcl reference electrode, which is -800mV to -1100mV, it indicates that the cathodic protection status of the jacket platform is not good.

[0240] The data collected by the accelerometer is the acceleration at the measuring point. When the acceleration at the measuring point exceeds the normal range, it indicates that external factors (construction, earthquake, impact of unknown organisms, etc. on the jacket platform) may pose a safety hazard to the jacket, and corresponding measures should be taken in a timely manner.

[0241] The strain sensor collects data as strain values ​​at the measuring points. When the strain value at a measuring point exceeds the normal range, it indicates a safety hazard in the jacket platform structure. Upper and lower limit warning methods are as follows: Figure 5 and 6 As shown, after collecting data, the corresponding warning level is determined based on the position of the data in the graph. The data with the largest offset among all sensor data is selected, and a consecutive count index (such as 10 times) is configured. If the data point appears in the first, second, or third warning level for 10 consecutive times, a warning message is issued.

[0242] S703: When the monitoring data is a specific value of a certain data type at different locations in the vertical direction, and the difference between the specific values ​​at different locations exceeds the normal range, an early warning will be issued, and the settlement difference warning method will be selected.

[0243] In this embodiment, a settlement difference early warning method is selected for static level data. The static level is typically installed at the four corners of the jacket platform, and the collected data is the vertical movement distance of the measuring points. When the vertical distance difference between the measuring points is large, the jacket platform is at risk of tilting. The settlement difference early warning method is as follows: Figure 6 As shown in the figure, assuming that the static level of data point 1 is significantly lower than that of data point 3, the jacket platform is at risk of tilting in the direction of data point 1. After collecting the data, the corresponding warning level is determined based on the position of the data in the figure, and a consecutive number index (such as 10 times) is configured. If the data point appears in the first, second or third warning level for 10 consecutive times, a warning message is issued.

[0244] By selecting different early warning methods based on different data, more accurate early warnings can be provided according to different states, making it easier for managers to make adjustments for different problems.

[0245] Specifically, firstly, duct stent monitoring data is acquired, and corresponding noise reduction methods are set for the monitoring data. Analysis data is obtained, and an initial noise reduction method is set based on experience, providing analytical data for subsequent assessment of the suitability of the noise reduction method. Secondly, an adaptive noise reduction selection method is constructed. The analytical data is analyzed using this method to determine whether to change the noise reduction method corresponding to the monitoring data. The noise reduction data is then denoised using the determined method, forming a noise reduction dataset together with the analytical data. The adaptive noise reduction selection method analyzes the initial analytical data to determine if the selected noise reduction method meets the requirements. If it does, no change is made; otherwise, a more suitable noise reduction method is used, resulting in more accurate noise reduction data. Thirdly, early warning method selection rules are constructed based on the data characteristics of different noise reduction data within the noise reduction dataset. Early warning methods are selected according to these rules, setting safety ranges and alert values ​​to obtain early warning results. Selecting different early warning methods based on different data allows for more accurate early warnings based on different states, facilitating adjustments by management personnel for different issues.

[0246] In a specific embodiment, the data calculation includes calculating the weight center of gravity, and the specific steps include:

[0247] When the jacket platform is unloaded, the unloaded monitoring data of the main legs in the multi-leg support structure is obtained. The unloaded monitoring data is the strain value.

[0248] Specifically, in this embodiment, during the land construction phase of the jacket platform, the main legs of the multi-leg support structure are not affected by the loading force of the blocks. The strain value at this time is used as the zero-point reference and subsequently used to calculate the weight center of gravity of the blocks.

[0249] After the jacket platform is launched, the actual monitoring data under the action of the upper block is obtained, that is, the actual strain value of the main leg of the multi-leg support structure of the jacket platform under the action of the upper block.

[0250] Specifically, after the jacket platform is launched, upper modules such as drilling work areas, mechanical equipment areas, and living areas will be set up on the jacket platform.

[0251] Based on the no-load monitoring data and the actual monitoring data of the jacket platform under the action of the upper block, the axial force of the main leg of the multi-leg support structure of the jacket platform is obtained through the main leg axial force calculation formula.

[0252] Based on the axial force of the main leg of the multi-leg support structure of the jacket platform, the weight of the upper block is obtained, and the center of gravity coordinates of the upper block are obtained.

[0253] The weight of the upper module is used to determine whether the upper module of the jacket platform is overloaded, and the center of gravity coordinates of the upper module are used to determine whether the jacket platform is at risk of overturning, so as to achieve health monitoring of the jacket platform during service.

[0254] Specifically, this embodiment effectively solves the problem that during the underwater service phase of the jacket platform, due to the harsh and complex marine environment, the weight center of gravity sensor fails, resulting in inaccurate estimation of the weight center of gravity of the upper block of the jacket platform, thus making it impossible to effectively monitor the health status of the jacket platform, and protecting the safety of the lives and property of construction personnel on the jacket platform.

[0255] In a specific embodiment, the formula for calculating the axial force of the main leg is expressed as follows:

[0256] ,

[0257] ,

[0258] ,

[0259] In the formula: This indicates the axial force of the main leg of the multi-leg support structure of the jacket platform; Indicates the elastic modulus; This represents the cross-sectional area of ​​the main leg of the multi-leg support structure of the jacket platform; This indicates the serial number of the strain sensor deployed at the monitoring point, that is, the serial number of the strain value of the main leg of the multi-leg support structure of the jacket platform under no-load condition obtained at the location of the monitoring point. This indicates the total number of strain sensors deployed at the monitoring point, which is the total number of strain values ​​of the main legs of the multi-leg support structure of the jacket platform under no-load conditions obtained at the location of the monitoring point. Indicates the influence coefficient of the sensor housing; The first monitoring point was deployed at the location of the monitoring point. i Actual monitoring data from each strain sensor; The first monitoring point was deployed at the location of the monitoring point. i No-load monitoring data from a strain sensor; This indicates the diameter of the main leg of the multi-leg support structure of the jacket platform; This indicates the wall thickness of the main leg of the multi-leg support structure of the jacket platform; Represents a counting function; Indicates actual monitoring data The total number of strain sensors that are not zero.

[0260] Specifically, in one embodiment of the present invention, the eight main legs 2 of the duct frame platform 4 are arranged regularly in the multi-leg support structure, and the main legs 2 of the duct frame are equipped with wireless nodes 1, such as... Figure 7 and Figure 8 As shown, four weight center strain sensors 3 are installed at the monitoring points of the main legs of the multi-leg support structure. If all four weight center strain sensors can work normally, the axial force of the main legs of the multi-leg support structure is calculated as shown in Table 2.

[0261] Table 2

[0262]

[0263] Specifically, due to the harsh and complex marine environment, the weight center of gravity sensors are at risk of failure. In another embodiment of the present invention, if the fourth weight center of gravity sensor at the monitoring point fails and no strain data is detected, and the total number of weight center of gravity strain sensors with non-zero actual monitoring data is 3, then the axial force of the main leg of the multi-leg support structure is calculated as shown in Table 3:

[0264] Table 3

[0265]

[0266] When the first, second, and third weight center of gravity sensors in this embodiment fail, the axial forces of the main legs of the multi-leg support structure are calculated as shown in Tables 4, 5, and 6, respectively:

[0267] Table 4

[0268]

[0269] Table 5

[0270]

[0271] Table 6

[0272]

[0273] Specifically, elastic modulus The unit is Pa In this embodiment, the axial force of the main leg is calculated based on the average value of monitoring data from multiple valid weight center strain sensors. Invalid weight center strain sensors are considered to have no monitoring data and are therefore ignored in the axial force calculation. As shown in Tables 2 to 6, the calculated axial force values ​​for the same main leg at three and four monitoring points differ by less than ±2t, with an accuracy of ±0.02%, meeting the practical engineering requirement of ±0.1%. Therefore, using the axial force calculation formula of this embodiment, even if a weight center strain sensor fails, the overall monitoring results remain relatively consistent. Thus, the axial force calculation formula of this embodiment, by arranging multiple weight center strain sensors at monitoring points and averaging the multiple valid monitoring values, improves the tolerance to sensor failures and yields highly accurate main leg axial forces.

[0274] Specifically, when the main legs of a multi-legged support structure are arranged in other ways, such as when the eight main legs are arranged irregularly, for example... Figure 9 As shown, the axial force calculation formulas in this embodiment are all applicable.

[0275] In a specific embodiment, the weight of the upper block is obtained based on the axial force of the main leg of the multi-leg support structure of the jacket platform, in order to obtain the center of gravity coordinates of the upper block, including:

[0276] Preferably, the weight calculation formula for the upper module is as follows:

[0277] ,

[0278] In the formula: This indicates the weight of the upper component, in kg. The first leg support structure of the jacket platform Axial force of the main leg; This indicates the sequence number of the main leg of the multi-leg support structure of the jacket platform. This indicates the total number of main legs of the multi-leg support structure of the jacket platform; It is the acceleration due to gravity;

[0279] Preferably, the formula for calculating the centroid coordinates of the upper module is as follows:

[0280] ,

[0281] ,

[0282] In the formula: The X-axis coordinate representing the centroid of the upper block; The Y-axis coordinate representing the centroid of the upper block; The first leg support structure of the jacket platform k The X-axis coordinates of each main leg; The first leg support structure of the jacket platform k The Y-axis coordinate of each main leg;

[0283] Specifically, in this embodiment, a two-dimensional coordinate system is established on the plane where the multi-leg support structure connects to the guide frame platform where the upper block is set, that is, a two-dimensional coordinate system is established on the platform at the top of the multi-leg support structure, so as to calculate the center of gravity coordinates of the upper block.

[0284] Specifically, in the embodiments of the present invention, the weight and center of gravity of the upper module are calculated using axial force, as shown in Table 7:

[0285] Table 7

[0286]

[0287] In a specific embodiment, the data correction process includes correcting the strain data, and the specific steps include:

[0288] S11: Design a corresponding outer cover for the strain sensor according to its type, and establish a simulation model of the outer cover for the strain sensor.

[0289] S12: Construct a strain sensor clamping block model, and calculate the strain data of the strain sensor when different load conditions are applied to the two clamping bases of the strain sensor without the outer protective cover, based on the strain sensor clamping block model;

[0290] S13: Construct a strain sensor clamping block model with a strain sensor outer cover based on the simulation model of the strain sensor outer cover, and calculate the strain data of the strain sensor when different load conditions are applied to the two clamping bases of the strain sensor with the strain sensor outer cover based on the strain sensor clamping block model with a strain sensor outer cover.

[0291] S14: Calculate the correction coefficients of the strain sensor based on the two strain data obtained in S12 and S13. The correction coefficients include the axial force correction coefficient and the bending moment correction coefficient. Specifically, use the axial force correction coefficient and the bending moment correction coefficient to correct the strain sensor data to obtain the corrected axial force data and bending moment data.

[0292] Specifically, firstly, a corresponding outer protective cover for the strain sensor is designed according to its type, and a simulation model of the outer protective cover is established. A protective cover model is constructed, and the thickness of the protective cover is calculated, which prepares for subsequent assessment of the protective cover's impact on the sensor. Secondly, a strain sensor clamping block model is constructed. This model simulates the scenario where the strain sensor is welded to the sensor's rod and held by clamps. Based on the strain sensor clamping block model, the strain data of the strain sensor is calculated under different load conditions when the two clamping bases are not covered by the outer protective cover, providing reference data for subsequent calculation of correction coefficients. Furthermore, a model with strain gauges is constructed based on the simulation model of the strain sensor's outer protective cover. The strain sensor clamping block model with the outer protective cover is used to calculate the strain data of the two clamping bases of the strain sensor under different load conditions when the outer protective cover is present. The strain data is used to prepare for the subsequent calculation of the sensor correction coefficient. The correction coefficient of the strain sensor is calculated based on the two strain data obtained in S12 and S13. The correction coefficient includes axial force correction coefficient and bending moment correction coefficient, which can eliminate the influence of the sensor data acquisition by the outer protective cover. Finally, the data of the strain sensor is corrected using the correction coefficient to obtain the corrected axial force data and bending moment data, eliminating the influence of the outer protective cover and meeting the actual needs.

[0293] In a specific embodiment, the scheme for designing a corresponding sensor outer cover according to the type of strain sensor and establishing a simulation model of the strain sensor outer cover is as follows:

[0294] Based on the sensor's operating conditions and considering the influence of environmental factors, a corresponding sensor outer protective cover is designed, and a scale model of the outer protective cover is established:

[0295] In this embodiment, the strain sensor is a fiber Bragg grating sensor. Taking the fiber Bragg grating sensor installed on the steel structure of an offshore platform as an example, i.e., on the jacket support member, the outer cover of the sensor needs to be welded to the jacket platform to cover the sensor. The simulation modeling software used is ANSYS Workbench software, and the analysis process is as follows:

[0296] S801: Finite element analysis model of the component, i.e., constructing the sensor outer protective cover model, such as... Figure 10 As shown;

[0297] S802: Input boundary conditions, define the bottom edge reinforcement of the protective cover, and apply a load to the outer surface of the protective cover in conjunction with external environmental factors. The load direction is the normal direction of the outer surface of the protective cover. Figure 11 As shown;

[0298] S803: Calculate the maximum deformation and equivalent stress of the protective cover under the load. According to the design requirements, the maximum deformation shall not exceed the preset value and the equivalent stress shall not exceed the preset value (an additional safety factor shall be reserved). Perform strength verification calculation on the protective cover. Finally, it is found that when the thickness of the protective cover is 8mm, it meets the usage requirements. Therefore, the final dimensions of the sensor outer cover are set as follows: outer diameter 219mm semi-circular cover, cover length 600mm, and thickness 8mm.

[0299] In this embodiment, to protect the sensor from the influence of the external environment and avoid damage caused by complex external conditions, a protective cover needs to be added above the sensor after installation. Since the guide frame is a steel structure, the protective cover is installed by welding. A protective cover model is constructed, and the thickness of the protective cover is calculated to prepare for subsequent assessment of the protective cover's impact on the sensor.

[0300] In a specific embodiment, a strain sensor clamping block model is constructed. The scheme for calculating the strain data of the strain sensor under different load conditions when the two clamping bases of the strain sensor are not covered by the outer sensor cover, based on the strain sensor clamping block model, is as follows:

[0301] (1) The strain sensor clamp is welded to the rod where the sensor is located. Therefore, the ANSYS Workbentch software is used to construct the rod where the sensor is located and the strain sensor clamp model. The model is created with the positive x-axis in the software as the compressive stress direction. In this embodiment, the rod where the sensor is located is the guide frame rod.

[0302] The mesh elements are tetrahedral or hexahedral, with the clamping blocks and rods sharing nodes. The rod element size is no larger than 50mm, and the clamping block mesh element size is no larger than 1mm to ensure displacement measurement accuracy. A node is set at the center of each sensor clamping block, with an initial distance of 65mm between the center nodes of the two clamping blocks for each sensor group, arranged in a 360-degree circumferential direction. A total of 4 sensors and 4 sets of clamping blocks are arranged. The strain sensor clamping block model is shown below. Figure 12 As shown, around the guide frame member, each sensor is held by two clamping bases, i.e., clamping blocks, and different load conditions are applied to the sensor, including axial force load and bending moment load.

[0303] (2) Set the initial distance between the two clamping bases of the strain sensor. In this embodiment, the distance between the two clamping bases is set to 65mm according to actual requirements;

[0304] (3) Apply different axial force loads to one end of the guide frame member and calculate the displacement of the clamping base of the strain sensor when there is no sensor outer cover. The formula is:

[0305] ,

[0306] ,

[0307] in, and These represent the displacements of the two clamping bases under axial load. This indicates the different axial force loads applied. and These represent the distances from the two clamping bases to the end of the guide frame member where force is applied; Indicates the elastic modulus; This indicates the cross-sectional area of ​​the jacket support member;

[0308] Under different bending moment load conditions applied to one end of the jacket support member, the displacement of the strain sensor's clamping base is calculated when there is no sensor outer cover. The formula is as follows:

[0309] ,

[0310] ,

[0311] in, These represent the displacements of the two clamping bases under bending moment load. This indicates the different applied bending moment loads. The moment of inertia of the cross section of the jacket support member; Indicates the length of the jacket support member;

[0312] (4) According to The distance between the two clamping bases after the axial force load strain is calculated using the following formula:

[0313]

[0314] according to and The distance between the two clamping bases after the applied bending moment load strain is calculated using the following formula:

[0315] ,

[0316] (5) According to and When the sensorless outer protective cover is obtained, the strain data of the strain sensor under axial load is obtained by the following formula:

[0317] ,

[0318] in, This represents the displacement change of the two clamping bases under axial load. ,Right now ;

[0319] according to and When the sensorless outer protective cover is obtained, the strain data of the strain sensor under bending moment load is obtained by the following formula:

[0320] ,

[0321] in, This represents the displacement change of the two clamping bases under bending moment load. ,Right now .

[0322] In this embodiment, a model of the guide frame members and strain sensor clamping block without an outer protective cover is constructed. This facilitates the software to calculate the strain data of the strain sensor under different load conditions without an outer protective cover, providing reference data for subsequent calculation of correction coefficients.

[0323] In a specific embodiment, a strain sensor clamping block model with the strain sensor outer cover is constructed based on the simulation model of the strain sensor outer cover. The scheme for calculating the strain data of the two clamping bases of the strain sensor under different load conditions when the strain sensor outer cover is present, based on the strain sensor clamping block model with the strain sensor outer cover, is as follows:

[0324] (1) Model the outer cover at the clamping block according to the actual sensor outer cover arrangement position. The mesh division ensures that the cover and the rod share nodes. The rod unit size is no greater than 50mm, and the clamping block mesh unit size is no greater than 1mm to ensure displacement measurement accuracy. Set a node at the center position of each sensor clamping block. The initial distance between the center nodes of the two clamping blocks of each group of sensors is 65mm. A total of 4 sensors and 4 groups of clamping blocks are arranged in a 360° circumference. The distance between the measurement point and the force point is created according to the actual position. The strain sensor clamping block model with strain sensor outer cover is as follows: Figure 13 As shown, each sensor is held in place by two clamping bases, or clamping blocks.

[0325] (2) Set the initial distance between the clamping bases of the strain sensor. ;

[0326] (3) Apply different axial force loads to one end of the guide frame member and calculate the displacement of the clamping base of the strain sensor when the sensor outer cover is present. The formula is:

[0327] ,

[0328] ,

[0329] in, and These represent the displacements of the two clamping bases under bending moment load. This indicates the different applied bending moment loads. and These represent the distances from the two clamping bases to the end of the guide frame member where force is applied; Indicates the elastic modulus; This indicates the cross-sectional area of ​​the guide frame member when a sensor outer cover is installed.

[0330] (4) Apply different bending moment loads to one end of the guide frame member and calculate the displacement of the clamping base of the strain sensor when the sensor outer cover is present. The formula is:

[0331] ,

[0332] ,

[0333] in, These represent the displacements of the two clamping bases under bending moment load. Indicates the different applied bending moment loads; The moment of inertia of the cross section when the outer protective cover of the sensor is installed on the guide frame member;

[0334] (5) According to The distance between the two clamping bases after the axial force load strain is calculated using the following formula:

[0335] ,

[0336] according to and The distance between the two clamping bases after the applied bending moment load strain is calculated using the following formula:

[0337] ,

[0338] (6) According to and The strain data of the strain sensor under axial load, with the sensor outer protective cover in place, is obtained by the following formula:

[0339] ,

[0340] in, This represents the displacement change of the two clamping bases under axial load. ,Right now ;

[0341] according to and When the sensor has an outer protective cover, the strain data of the strain sensor under bending moment load is obtained by the following formula:

[0342] ,

[0343] in, This represents the displacement change of the two clamping bases under bending moment load. ,Right now .

[0344] In this embodiment, a strain sensor outer cover and a strain sensor clamping block model are designed to facilitate software calculation of the strain data of the strain sensor under different load conditions when the two clamping bases of the strain sensor are equipped with the strain sensor outer cover. This strain data is used to prepare for subsequent calculation of the sensor correction coefficient. In this embodiment, four sensors are set up in the strain sensor clamping block model with the strain sensor outer cover.

[0345] In a specific embodiment, the correction coefficients for the strain sensor are calculated, including axial force correction coefficients and bending moment correction coefficients, comprising:

[0346] S141: Create or import a model using Ansys Workbench software and perform simulation calculations using solid186 elements. Apply four axial pressure loads: 1800t, 1100t, 800t, and 500t. Calculate the axial force correction factor based on the obtained strain data. The formula is as follows:

[0347] ,

[0348] in, This indicates the axial force correction factor for the sensor housing. This represents strain data acquired without a protective shield. This indicates strain data acquired with a protective shield in place;

[0349] S142: Apply four bending moment loads: 600 t·m, 480 t·m, 300 t·m, and 180 t·m. Calculate the bending moment correction factor based on the obtained strain data. The formula is:

[0350] * ,

[0351] in, This represents the moment correction factor. This represents the bending moment correction coefficient at different points on the sensor. ; This represents the strain data obtained under bending moment load without a protective cover. This represents the strain data obtained under bending moment load with a protective cover.

[0352] In this embodiment, two correction coefficients for the sensor are calculated based on the sensor strain data calculated without the sensor outer cover and the sensor strain data calculated with the sensor outer cover. The influence of the sensor data acquisition by the sensor outer cover can be eliminated through the correction coefficients.

[0353] In a specific embodiment, the scheme for using the correction coefficient to correct the strain sensor data to obtain the corrected axial force and bending moment data is as follows:

[0354] The calculation formula is as follows:

[0355] ,

[0356] ,

[0357] in, This represents the axial force data detected by the strain sensor. This represents the bending moment data detected by the strain sensor; This indicates the corrected axial force data obtained from the axial force data detected by the strain sensor; This indicates the corrected bending moment data obtained from the bending moment data detected by the strain sensor.

[0358] Specifically, in this embodiment, the sensor data is divided by two correction coefficients to convert the sensor data into the required corrected axial force and bending moment data, thereby eliminating the influence of the outer protective cover and meeting the actual needs.

[0359] In a specific embodiment, the steps by which the land data receiving module imports the decompressed data into the database include:

[0360] Encode the data type of the decompressed data and obtain the data partition unit used for data storage;

[0361] Specifically, the decompressed data, i.e. the monitoring time series data, is partitioned according to the date range. Combined with the database tasks and stored procedure technology built into the database itself (such as MySQL database), a new storage data partition table is automatically created every day to ensure that the data for each natural day has a corresponding independent partition table.

[0362] Specifically, this embodiment also includes a data query buffer and a data query hot table area; the data query buffer is a storage area with a fixed storage length used to store buffered data;

[0363] The buffered data refers to the final data retained after the monitoring time-series data is split and written to the data partition table based on the time-level sequence data by the data writing module, and the data is written to the storage area simultaneously based on the buffer writing strategy; the length of the buffer is fixed at 30 records.

[0364] Buffered write strategy: Based on the first-in-first-out (FIFO) principle, data is written and deleted in chronological order according to a fixed storage length;

[0365] The data query hotspot area is used to store periodic data;

[0366] The periodic data refers to the periodically updated data that is written to the data query hot table area simultaneously when the monitoring time-series data is split and written to the data partition table based on the time-level sequence data through the data writing module. For example, in this embodiment, while writing data to the second-level data table, the data is simultaneously copied to the data query hot table area. In order to facilitate partition management, the data query hot table area stores the data volume of the most recent day. Through a scheduled task (executed once per period), only the most recent hot table data partition is retained, and the expired data partition table is deleted (in order to save hardware resources and ensure the effectiveness of execution, the execution period of the scheduled task is set to 1 hour) to periodically update the data in the data query hot table area.

[0367] Set up at least three time-level splitting sub-modules to split the monitoring time series data into time-level segments according to the time-level data splitting strategy and obtain minute-level, hour-level, and day-level sequence data respectively, thereby obtaining the split sequence data for each time level;

[0368] Specifically, in this embodiment, a first time-level splitting submodule, a second time-level splitting submodule, and a third time-level splitting submodule are set.

[0369] In a specific embodiment, the time-level data splitting strategy is as follows:

[0370] Get the current progress timestamp of each type of data in the corresponding decompressed data;

[0371] Using the current progress timestamp as the starting time point, and based on the splitting time level corresponding to the at least three time-level splitting sub-modules, determine the timestamp node as the ending time point, so as to realize the splitting of data at each time level and obtain the splitting sequence data of the corresponding time level.

[0372] For example, in this embodiment, the data splitting process based on the time-level data splitting strategy is as follows:

[0373] (1) Split into minute data tables

[0374] When splitting to a minute-level data table, the current progress timestamp of each type of data in the corresponding monitoring time series data is obtained; and the current progress timestamp is used as the starting time point, and the next minute 0 seconds is used as the ending time point to achieve the splitting of minute-level sequence data;

[0375] (2) Split into hourly data tables

[0376] When splitting to an hourly data table, the current progress timestamp of each type of data in the corresponding monitoring time series data is obtained; and the current progress timestamp is used as the starting time point, and the time from the current timestamp to the next hour 0 minutes 0 seconds is used as the ending time point, so as to achieve the splitting of hourly sequence data;

[0377] (3) Split into daily-level data tables

[0378] When splitting to a daily data table, the current progress timestamp of each type of data in the corresponding monitoring time series data is obtained; and the current progress timestamp is used as the starting time point, and the time from the current timestamp to 0:00:00 on the next day is used as the ending time point, so as to achieve the splitting of daily-level sequence data.

[0379] The split sequence data of each time level is written into the data partition unit according to the set writing rules, so as to realize the storage of monitoring time series data according to the data type, and then obtain the stored data of different time levels.

[0380] In a specific embodiment, the writing rules set include:

[0381] Use the start and end times as data query conditions, that is, use the split sequence data that is greater than the start time and less than the end time as the query conditions for time span data collection.

[0382] Write the maximum and minimum values ​​of the data corresponding to the split sequence data, along with the corresponding query conditions, into the data partitioning unit;

[0383] Update the end time to the current progress timestamp.

[0384] Using a timed task approach, the splitting progress of the compressed data by the at least three time-level splitting sub-modules is polled and obtained respectively. A progress query signal is obtained, and based on the progress query signal, after confirming that the splitting sequence data of the corresponding time level has been split, the time-level splitting sub-module is controlled to stop running.

[0385] Specifically, time-series data at various levels is polled using scheduled tasks. The standard time interval for executing second-level tasks is 1 second, for minute-level tasks it is 1 minute, for hour-level tasks it is 1 hour, and for day-level tasks it is 1 day. In this embodiment, to increase the frequency of data collection and shorten the time of insufficient business data, the task interval can be shortened and the number of task cycles can be increased. Preferably, in actual production practice, the standard time interval for executing second-level tasks is 3s to 5s. Specifically, a record table is established by the data splitting progress query module to record the data splitting progress, which enables querying of the splitting progress of the monitored time-series data. The record table consists of a key and a current_timestamp. The key is used to record the type of split data, and the current_timestamp is used to represent the timestamp of the currently split data.

[0386] The system polls and retrieves the splitting progress of the monitoring time-series data by the first, second, and third time-level splitting sub-modules, respectively, and obtains progress query signals. When the splitting progress of the corresponding splitting sub-module reaches the completion stage, that is, the corresponding time-level splitting sub-module has completed the splitting of all monitoring time-series data, a stop splitting signal is sent to the corresponding splitting sub-module; otherwise, the data splitting of the monitoring time-series data continues.

[0387] In this embodiment, the data query steps provided by the distributed data computing module include:

[0388] Receive query requests for data to be queried from preset clients;

[0389] Receive timestamp query priority for query requests for data to be queried from preset clients;

[0390] The response data corresponding to the query conditions of the query request is determined based on the query request for the data to be queried;

[0391] The timestamp query priority refers to the query matching rules established for different time span query conditions, specifically as follows:

[0392] If the query request for the data to be queried is within a one-minute query range, the response data corresponding to the second level will be called first.

[0393] If the query request for the data to be queried is within a one-hour query range, the response data corresponding to the minute level will be called first.

[0394] If the query request for the data to be queried is within a one-day query range, the response data corresponding to the hour level will be called first.

[0395] If the query request for the data to be queried is within a year, the response data corresponding to the day level will be called first.

[0396] In a specific embodiment, an API interface service unit is set up; the API interface service unit is provided with multiple API interface service sub-units for receiving query requests for query data to be queried sent by a preset client and querying the timestamp priority.

[0397] The API interface service subunit is used to determine the response data corresponding to the query conditions of the query request based on the query request for the data to be queried;

[0398] The timestamp query priority refers to the query matching rules established for different time span query conditions, specifically as follows:

[0399] If the query request for the data to be queried is within a one-minute query range, the API interface service sub-unit corresponding to the second level will be called first.

[0400] If the query request for the data to be queried is within a one-hour query range, the API interface service sub-unit corresponding to the minute level will be called first.

[0401] If the query request for the data to be queried is a data query condition within a day's query range, the API interface service sub-unit corresponding to the hour level will be called first.

[0402] If the query request for the data to be queried is within a year, the API interface service sub-unit corresponding to the day level will be called first.

[0403] This embodiment also includes an API interface service subunit for querying data in the data query hot table area. When dealing with query business, the most frequent and unavoidable query is setting default query conditions. In this embodiment, the current default query condition is preferably data from the last 30 minutes. Under this default query condition, the API interface service subunit for querying data in the data query hot table area is directly called to avoid the problem of low query efficiency caused by large data volume, thereby reducing the query workload and improving the efficiency of data query.

[0404] In a large amount of data that is not very relevant to decision-making, data is filtered by retaining time-series data, i.e., the maximum and minimum data per minute, hour, or day, to obtain the most valuable data for querying. This improves the efficiency of data querying while ensuring the accuracy of data querying. The data fast query module sets up different API interface service sub-units for different levels of query needs to query data at different time levels, which greatly improves the efficiency of querying.

[0405] Furthermore, using this system, data is split into business data at the hot table level, second level, minute level, hour level, and day level, and the data is written using a database partitioning method. Different levels of data are read according to the actual query business needs. When querying data, the average response time for each data query is about 40 milliseconds, which improves query efficiency by about 60 times.

[0406] In a specific embodiment, the construction steps of the first-level digital twin model include:

[0407] Obtain the design data of the jacket support members, and based on the design data, establish a SACS simulation model of the jacket support. The design data includes the pipe diameter and wall thickness of the jacket support members.

[0408] Obtain measured data of the jacket platform, and obtain the first three natural frequencies of the jacket platform based on the measured data. The measured data includes the average acceleration and displacement of the monitoring points of the jacket platform, as well as the axial force data of the main legs and horizontal supports of the jacket platform.

[0409] Specifically, this embodiment uses sensors to acquire in real time the average acceleration and displacement of the rods at different horizontal levels on the jacket platform, as well as the axial force data of the main legs and horizontal supports of all jackets, in order to obtain kinematic information such as acceleration and displacement at different heights of the jacket, as well as dynamic response information such as the axial force of the main legs and horizontal supports, providing basic data for subsequent data analysis.

[0410] Specifically, based on the measured data, the measured first three natural frequencies of the catheter stent are obtained, including the first natural frequency, the measured second natural frequency, and the measured third natural frequency.

[0411] Specifically, based on the collected measured data, the first, second, and third natural frequencies of the jacket are calculated to provide foundational data for subsequent model correction and to provide a basis for error assessment. The methods used to obtain the measured first, second, and third natural frequencies of the jacket are existing technologies in the field and will not be described in detail here.

[0412] The predicted first three natural frequencies of the jacket platform were obtained based on the SACS simulation model;

[0413] Specifically, this embodiment uses SACS simulation software to input design data such as pipe node positions, pipe diameter, wall thickness, and steel properties of the jacket structure to perform simulation calculations and obtain the predicted first three natural frequencies of the jacket structure. The method of establishing the SACS simulation model is a conventional technique for those skilled in the art and will not be described in detail here.

[0414] The SACS simulation model is corrected based on the measured first three natural frequencies and the predicted first three natural frequencies to obtain a first-order twin model.

[0415] In a specific embodiment, such as Figure 16 As shown, the correction of the SACS simulation model based on the measured first three natural frequencies and the predicted first three natural frequencies includes:

[0416] S111: Compare whether the error between the measured first three natural frequencies of the jacket platform and the predicted first three natural frequencies is greater than the set threshold. If it is greater, execute S112; otherwise, execute S114.

[0417] Specifically, the first error between the predicted first natural frequency and the measured first natural frequency, the second error between the predicted second natural frequency and the measured second natural frequency, and the third error between the predicted third natural frequency and the measured third natural frequency are obtained.

[0418] When max{first error, second error, third error} If the SACS simulation model is modified, it needs to be corrected; otherwise, it does not need to be corrected. The threshold value is set.

[0419] Specifically, the calculation formulas for the first error, the second error, and the third error in this embodiment are as follows:

[0420] ,

[0421] Specifically, in this embodiment, the first three natural frequencies measured in practice are compared with those calculated by the SACS simulation model to obtain the first error, the second error, and the third error. The maximum error among the three is then determined. When the maximum error is greater than or equal to 1%, the SACS simulation model needs to be corrected. This includes adjusting the center of gravity of the upper module of the jacket, the structural stiffness of the jacket, and the pile-soil py curve at the installation position of the jacket. The first three natural frequencies are then recalculated to ensure that the SACS simulation model is closer to the actual situation.

[0422] S112: Using the structural stress-strain calculation submodule and the jacket platform model calculation submodule, the weight center of gravity of the upper jacket block, the wall thickness of the jacket members, and the pile-soil py curve of the jacket installation position are calculated by the jacket platform model to correct the SACS simulation model and obtain the corrected SACS simulation model.

[0423] Specifically, in this embodiment, a coordinate system is established using the horizontal plane at the connection between the guide frame and the upper module. The arrangement of the main legs of the guide frame is as follows: Figure 14 As shown. When the SACS simulation model needs to be corrected, the weight center of gravity of the upper module is calculated by re-acquiring sensor monitoring data, thereby correcting the SACS simulation model.

[0424] In a specific embodiment, the formula used to correct the wall thickness of the guide frame members is as follows:

[0425] ,

[0426] in: Indicates the corrected wall thickness of the jacket support member; D 1 indicates the pipe diameter of the jacket support member after it leaves the factory; The actual quality of the jacket structure members after they leave the factory; This refers to the length of the jacket support member.

[0427] Specifically, the initial SACS simulation model in this embodiment is established based on the design data of the jacket. Since the data of the jacket, including the pipe diameter and wall thickness of the jacket members, will be different from the design data after the jacket is built and installed, when it is determined that the SACS simulation model needs to be corrected, the pipe diameter of the jacket at the time of manufacture is adopted, the wall thickness of the jacket members is recalculated, and the recalculated pipe diameter and wall thickness of the jacket members are assigned to the SACS simulation model to correct the SACS simulation model.

[0428] Specifically, after correction, all member information is re-entered into the SACS software to obtain the corrected SACS simulation model. It is important to note that this correction does not require real-time monitoring data input; it only uses the actual data from the jacket's factory shipment to correct the SACS simulation model once.

[0429] In a specific embodiment, the method used to correct the pile-soil py curve at the installation position of the jacket is as follows:

[0430] Specifically, such as Figure 15 As shown, the range for adjusting the pile-soil Py curve is determined based on marine exploration data. An iterative method is used to adjust the pile-soil Py curve, and the modified parameters are re-entered to calculate the new natural frequency.

[0431] Specifically, the cyclic load coefficient of the pile-soil py curve is iteratively adjusted. Standard soil resistance and initial modulus of internal friction angle This method allows for the acquisition of corrected pile-soil Py-curves, which bring the first three natural frequencies calculated by the SACS simulation model closer to the measured first three natural frequencies. The iterative calculation method is as follows:

[0432] S121: Construct an implicit function based on the cyclic load coefficient, standard soil resistance, and initial modulus of internal friction angle, and obtain the difference scheme of the implicit function as follows:

[0433] , , ,

[0434] In the formula, It is an implicit function of the first-order natural frequency; It is an implicit function of the second-order natural frequency; It is an implicit function of the third-order natural frequency; This represents the first-order natural frequency increment. This represents the second-order natural frequency increment; It is a third-order natural frequency component; express A The difference step size; express The difference step size; express The difference step size; A Indicates the cyclic load factor; Indicates standard soil resistance; Indicates the initial modulus of the internal friction angle;

[0435] Then, the discretized implicit function partial derivatives are:

[0436] , , ,

[0437] , , ,

[0438] , , ,

[0439] In the formula: , , , , , , , , These are all intermediate calculation parameters; Represents the discrete A The difference step size; Represents the discrete The difference step size; Represents the discrete The difference step size;

[0440] Therefore, the discretized difference scheme is as follows:

[0441] , , ,

[0442] in,

[0443] ,

[0444] ,

[0445] ,

[0446] ,

[0447] ,

[0448] ,

[0449] ,

[0450] ,

[0451] ,

[0452] In the formula: Implicit function representing the first-order natural frequency exist A Increment in direction; Implicit function representing the first-order natural frequency exist Increment in direction; Implicit function representing the first-order natural frequency exist Increment in direction; Implicit function representing the second-order natural frequency exist A Increment in direction; Implicit function representing the second-order natural frequency exist Increment in direction; Implicit function representing the second-order natural frequency exist Increment in direction; Implicit function representing the third natural frequency exist A Increment in direction; Implicit function representing the third natural frequency exist Increment in direction; Implicit function representing the third natural frequency exist Increment in direction;

[0453] The formulas used to obtain the cyclic load coefficient, standard soil resistance, and initial modulus of internal friction angle are as follows:

[0454] ,

[0455] ,

[0456] ,

[0457] ,

[0458] In the formula: This indicates the set iteration threshold;

[0459] Specifically, in this embodiment, the discretized py curve is... A , , Differential step size The values ​​of were designed and substituted into the SACS simulation model. , , , , , , , , Calculations are performed to obtain A The difference step size, The difference step size and The difference step size.

[0460] S122: Based on the differential step size of the cyclic load coefficient, the differential step size of the standard soil resistance, and the differential step size of the initial modulus of the internal friction angle, the corrected cyclic load coefficient, the corrected standard soil resistance, and the corrected initial modulus of the internal friction angle are obtained to obtain the corrected pile-soil Py curve at the installation position of the jacket. The formula used is as follows:

[0461] ,

[0462] In the formula: Indicates horizontal soil resistance; Indicates the initial modulus of the internal friction angle; Indicates the depth of effect.

[0463] S113: Reacquire the predicted first three natural frequencies of the jacket platform based on the corrected SACS simulation model, and use the reacquired predicted first three natural frequencies of the jacket platform as the current predicted first three natural frequencies, and execute S111.

[0464] S114: Complete the correction of the SACS simulation model.

[0465] Specifically, this embodiment incorporates the corrected model parameters into the new calculations, forming an iterative process until the inherent frequency error of the SACS simulation model meets the requirements. Ultimately, a high-fidelity first-level twin model is established, which can be applied to scenarios such as real-time monitoring, performance prediction, and decision support. The first-level twin model is continuously updated during subsequent operation and maintenance to reflect any new changes or external conditions, thereby laying the foundation for comprehensive health maintenance management of the jacket platform.

[0466] S2: Construct a second-level twin model based on the first-level twin model, such as... Figure 23 As shown, it includes:

[0467] The environmental load information of the jacket platform during the current monitoring period is obtained. The environmental load information includes: wind speed, wind direction, wave height, wave period, seawater flow direction, and seawater flow velocity.

[0468] Specifically, this embodiment acquires environmental load information such as wind speed, wind direction, wave height, wave period, seawater flow direction, and seawater flow velocity within the current monitoring time period (wave height, wave period, seawater flow direction, and seawater flow velocity for 2 minutes; wind speed and wind direction for 1 hour) through multiple data acquisition devices. The wind speed and wind direction monitoring data are obtained from an anemometer installed on the top of the jacket structure; the wave monitoring data (wave height and wave period) are obtained from a wave radar installed on the upper module of the jacket structure; and the seawater flow velocity and seawater flow direction are obtained from a current meter suspended below the horizontal plane. This environmental load information is used as input samples for the interpolation module to participate in the interpolation of the internal forces of the jacket structure members.

[0469] Based on the environmental load information and the first-level twin model during the current monitoring period, the maximum axial force and maximum bending moment of the jacket platform under the prior working conditions are calculated.

[0470] Based on the SACS simulation software, a finite element model containing all members of the jacket framework was established, and the maximum axial force of all members of the jacket framework under the prior working conditions was obtained. The maximum bending moments of the member about the principal axes of inertia in the y and z directions under the prior working conditions. ;

[0471] Specifically, this embodiment calculates and stores the internal forces (axial force and bending moment) of the members based on the jacket simulation model and prior operating conditions (i.e., preset environmental load information) in the SACS simulation software. In the SACS simulation software, the members of the jacket are arranged by member number. The preset environmental load information needs to include the extreme sea conditions of the target sea area. For example, for a jacket in the Liuhua sea area of ​​the South China Sea... Figure 17 and Figure 18 As shown, the preset sea state is as follows:

[0472] Wind direction: 0°, 60°, 150°, 240°, 270° and 330° (0° is due north, 90° is due east, and so on, the same below);

[0473] Flow direction: 0°, 45°, 225°, 270° and 315°;

[0474] Rationale: Considering the symmetry of the jacket platform structure (axis of symmetry is northwest-southeast), the wind direction interval is set at 30°, and the flow direction interval at 45°, reducing the number of flow directions to 5 and the number of wind directions to 6. Based on historical data, less frequent wind directions were eliminated, and adjustments were made based on the symmetry of the platform structure. The final selected wind direction conditions are 0°, 60°, 150°, 240°, 270°, and 330°, and the flow direction conditions are 0°, 45°, 225°, 270°, and 315°.

[0475] Flow velocities: 0.29 m / s, 0.44 m / s and 0.69 m / s;

[0476] Reason for selection: Three characteristic values ​​were chosen for surface flow velocity, representing the 30%, 60%, and 90% quantiles of the distribution function followed by the historical flow velocity data. The calculated characteristic values ​​for surface flow velocity were 0.29, 0.44, and 0.69 m / s, respectively. Figure 19 The figure shows a schematic diagram of the surface velocity distribution fitting in the summer of 2022.

[0477] Wind speeds (1 h, 10 m above ground): 5.56, 6.71, 7.85 and 9.23 m / s:

[0478] Rationale for selection: Wind speed, wave height, and period are generally considered to have a strong correlation; therefore, a joint distribution of these environmental variables was constructed based on historical data. When selecting environmental conditions, the characteristic values ​​of wind speed were first determined, and then the characteristic values ​​of wave height and period were selected based on these. Four characteristic values ​​for wind speed were selected, namely the 1 / 3, 1 / 2, 2 / 3, and 5 / 6 quantiles of the wind speed distribution function. The 1-hour average wind speed characteristic values ​​are 5.64, 7.29, 9.09, and 11.49 m / s, respectively. Assuming the anemometer is installed at a height of 60 m above sea level, the 1-minute average wind speed characteristic values ​​at a height of 10 m are 5.56, 6.71, 7.85, and 9.23 m / s, respectively. Four characteristic values ​​for meaningful wave height were selected, namely the 1 / 3, 1 / 2, 2 / 3, and 5 / 6 quantiles of the wave height distribution function, each corresponding to a different average wind speed characteristic value. Four eigenvalues ​​were selected across the zero period, namely the 20%, 40%, 60%, and 80% quantiles of the periodic distribution function. Figure 20 and Figure 21 The diagrams show the fitting of measured wind speed data distribution to a two-parameter Weibull distribution and the fitting of wave distribution parameters to a three-parameter Weibull distribution, respectively.

[0479] Significant Wave Height and Zero-Cycle (Wave Period): Each wind speed corresponds to four wave height characteristic values, and each wave height corresponds to four wave period characteristic values, as shown in Table 8. The wave height and wave period characteristic values ​​correspond to their positions in the table. For example, the wave height of 1.24m in the first row and first column of Table 8 corresponds to the wave period characteristic values ​​of 5.03 s, 5.24 s, 5.46 s, and 5.75 s in the first row and first column of Table 9; the wave height of 1.45m in the first row and second column corresponds to the wave period characteristic values ​​of 5.16 s, 5.36 s, 5.57 s, and 5.84 s in the first row and second column of Table 9; and so on.

[0480] Table 8

[0481]

[0482] Table 9

[0483]

[0484] In this embodiment, the total number of pre-set load conditions is: The system calculates and stores the internal force (including axial force and bending moment) response of members under various sea conditions using SACS simulation software.

[0485] Identify the members in the jacket platform that require attention, obtain the strain of the members that require attention, and then obtain the maximum axial force and maximum bending moment of the members that require attention during the current monitoring period.

[0486] Specifically, this embodiment uses fiber optic sensors to acquire wavelength data to obtain the strain of the member of interest, and then obtains the maximum axial force of the member of interest during the current monitoring period. and the maximum bending moment during the current monitoring period ;

[0487] Specifically, in this embodiment, based on the strain calculation of the members of interest, the axial force and bending moment of the members of interest are considered. In this embodiment, some diagonal braces of different height horizontal layers of the jacket structure and main leg sections of different heights are considered as members of interest. The method for determining the members of interest is a conventional technique in the field and will not be described in detail here.

[0488] In this embodiment, on the rods requiring attention, the sensors are arranged according to the following rules: fiber optic sensors are arranged axially on the structural surface, and four groups are arranged clockwise circumferentially according to numbers 1, 2, 3, and 4, for a total of four groups. Figure 22 As shown in the figure, LC(X) is the fiber optic sensor, and JACKET LEG is the jacket support member. The fiber optic sensor obtains strain information of the structural surface through scattering principles (such as Raman scattering or Brillouin scattering). When the surface strain changes, the optical signal in the fiber changes accordingly. These changing optical signals are converted into electrical signals and calculated to determine the axial force and bending moment values ​​of the member of interest.

[0489] The following calculation steps can be taken: based on the strain of the member, the axial force of the member that needs to be considered can be obtained, and the following formula can be used for calculation:

[0490] ,

[0491] In the formula, This refers to the elastic modulus, with units of Pa. This refers to the cross-sectional area of ​​the jacket support member, expressed in meters (m²). 2 , ; This is strain data obtained without a protective shield; This indicates the axial force correction factor for the sensor housing;

[0492] Specifically, the jacket platform is subjected to cyclic loads for a long time, and changes in axial force can lead to fatigue damage. Since axial force cannot be directly measured by actual sensors, the axial force is initially calculated by the strain of the rods.

[0493] The bending moment of the member requiring attention is calculated using the following formula:

[0494] , ,

[0495] In the formula, The bending moment of the rod around the neutral plane where sensors 2 and 4 are located; The bending moment of the rod around the neutral plane where sensors 1 and 3 are located; The moment of inertia of the tube cross section is expressed in units of . ; This is the correction factor for the bending moment of the sensor housing; D The diameter of the jacket support member; For wall thickness;

[0496] Specifically, the magnitudes of each axial force and bending moment within the current monitoring period are compared to obtain the maximum axial force and maximum bending moment of the members that need to be monitored within the current monitoring period.

[0497] Based on the environmental load information during the current monitoring period, the maximum axial force and maximum bending moment of the jacket platform under the prior working conditions, and the maximum axial force and maximum bending moment of the members that need to be monitored during the current monitoring period, the inertial force index (i.e., twin inversion factor) of the members that need to be monitored is obtained.

[0498] In a specific embodiment, the step of obtaining the inertial force index of the member to be monitored based on the environmental load information during the current monitoring period, the maximum axial force and maximum bending moment of the jacket platform under prior working conditions, and the maximum axial force and maximum bending moment of the member to be monitored during the current monitoring period includes:

[0499] When the underwater service time of the jacket is less than the set service time threshold (2 months in this embodiment), and at this time, before enough environmental data and measured axial force and bending moment data have been collected, the inertial force index of the members that need to be focused on is calculated by offline calculation method.

[0500] In a specific embodiment, the step of calculating the inertial force index of the member of interest using an offline calculation method includes:

[0501] S401: Based on the wave height, wave period, and the first-order twin model, call the existing SACS static calculation module in the SACS software to obtain the maximum horizontal force of the members of the jacket platform that need to be considered under wave load.

[0502] S402: Based on the wave height, wave period and the first-order twin model, call the existing SACS dynamic calculation module in the SACS software to obtain the dynamic response curve of the horizontal force of the member of the jacket under wave load, and then obtain the maximum absolute value of the horizontal force of the member of the jacket, that is, the maximum absolute value of the horizontal force in the dynamic response curve of the horizontal force.

[0503] S403: Based on the maximum horizontal force of the member of the jacket structure that needs to be monitored under the wave load and the maximum absolute value of the horizontal force of the member that needs to be monitored, the wave amplification factor under the prior working condition is obtained.

[0504] Specifically, the formula used for the wave amplification factor is as follows:

[0505] ,

[0506] In the formula: Indicates the wave amplification factor; This indicates the maximum absolute value of the horizontal force on the member that requires attention. This indicates the maximum horizontal force on the members of the jacket structure that require attention under wave loads.

[0507] Specifically, firstly, using the SACS static calculation module, the maximum horizontal force on the jacket members of interest under the action of wind direction, wave height, and wave period in the prior working condition is calculated, denoted as... It should be noted that, since only wave loads are unsteady forces in the SACS static calculation module, only the wave height and wave period from the prior working condition are used here to calculate the maximum horizontal force under wave loads.

[0508] Subsequently, the SACS dynamics module was invoked to calculate the dynamic response curve of the horizontal force under the action of wind direction, wave height, and wave period in the prior working condition. The maximum absolute value of the horizontal force in the dynamic response curve was extracted and denoted as . Therefore, the wave amplification factor calculated by SACS quasi-static method can be obtained.

[0509] S404: Based on the wave amplification factor, and using the SACS static calculation module, the axial force and bending moment of the members that need to be considered, taking into account inertial effects, are obtained.

[0510] Specifically, in the SACS static calculation, the prior load condition is input, and the wave load is assigned a corresponding wave amplification factor in the SACS simulation software. Quasi-static calculations were performed to obtain the axial forces of the members that needed to be considered under the prior working conditions, taking into account inertial effects. With bending moment .

[0511] S405: The inertial force index of the member to be considered is obtained by taking into account the axial force and bending moment of the member to be considered.

[0512] In a specific embodiment, the inertial force index of the member of interest is obtained by considering the axial force and bending moment of the member to be of interest based on the inertial effect. The formula used is as follows:

[0513] ,

[0514] In the formula, The index of axial force and inertia of the member. For the axial force of the member considering inertial effects under prior operating conditions, This represents the maximum axial force of the members in the jacket under the prior operating conditions.

[0515] The formula for calculating the bending moment inertia index is:

[0516] ,

[0517] ,

[0518] In the formula, The bending moment and inertial force indices of the members about the principal inertial axes in the y and z directions, respectively; These are the bending moments of the members about the principal inertial axes in the y and z directions, respectively, considering the inertial effect; These represent the maximum bending moments of the member about the principal axes of inertia in the y and z directions, respectively.

[0519] Specifically, the inertial effect in this embodiment refers to the resistance of the mass of the jacket platform to the motion response when it is accelerated by external forces. The inertial effect significantly alters the mechanical response of the jacket platform by affecting its vibration, dynamic amplification, and energy dissipation, potentially leading to greater displacement and stress, and even resonance, especially under dynamic loads.

[0520] When the time the jacket is in underwater service is greater than or equal to the set service time threshold (2 months in this embodiment), the inertial force index of the members that need to be monitored is calculated online.

[0521] In a specific embodiment, the step of calculating the inertial force index of the member of interest using an online calculation method includes:

[0522] S411: Obtain the maximum static axial force and maximum static bending moment of the member to be monitored during the current monitoring period, as follows:

[0523] When the environmental load information during the current monitoring period of the jacket platform belongs to the prior working condition, the static axial force and static bending moment of the member that needs to be monitored during the current monitoring period are determined to be the maximum value of the member's axial force and the maximum value of its bending moment under the prior working condition.

[0524] When the environmental load information of the current monitoring period of the jacket platform does not belong to the prior working condition, the maximum static axial force and the maximum static bending moment of the member to be monitored during the current monitoring period are obtained based on the environmental load information of the current monitoring period of the jacket platform and the first-level twin model.

[0525] S412: Based on the required parameters, the maximum static axial force and maximum static bending moment of the member under observation during the current monitoring period are obtained, along with the maximum axial force and maximum bending moment of the member under observation during the current monitoring period, to obtain the inertial force index of the member under observation.

[0526] In a specific embodiment, the formula used to obtain the inertial force index of the member of interest is as follows:

[0527] ,

[0528] ,

[0529] ,

[0530] In the formula: The index of axial force and inertia of the member. The bending moment and inertial force indices of the members about the principal inertial axes in the y and z directions, respectively; This represents the maximum axial force of the member requiring monitoring during the current monitoring period. , These are the maximum bending moments of the members requiring attention around the principal inertial axes in the y and z directions during the current monitoring period. The maximum static axial force of the members that require attention during the current monitoring period; , These represent the maximum static bending moments of the members requiring attention within the current monitoring period, about the principal inertial axes in the y and z directions, respectively.

[0531] Based on the inertial force index of the members that need to be of interest, and using a machine learning method based on decision tree (DT), the interpolated inertial force index of the members of the jacket platform is obtained.

[0532] The real-time environmental load information of the jacket platform is obtained, and based on the real-time environmental load information of the jacket platform and the first-level twin model, the real-time static axial force and static bending moment of the members that need to be monitored are obtained.

[0533] Based on the interpolated inertial force index of the jacket platform members and the real-time static axial force and static bending moment of the members that need to be focused on, the inverse values ​​of axial force and bending moment of the jacket platform members are obtained, so as to establish an application database including environmental load information, axial force, bending moment and inertial force index, and form a second-level twin model through the application database.

[0534] Specifically, since only a small number of data acquisition sensors are actually installed on a limited number of members, and most members lack measured data, it is necessary to obtain the axial force, bending moment, and other data for all members through inversion. Furthermore, since the environmental loads are discrete points distributed in the load type space, only a very small number of members requiring attention can be calculated for their inertial force indices. The inertial force indices of other members are obtained through linear interpolation along the height of the jacket structure. This method uses a decision tree-based machine learning approach to interpolate the inertial force indices, described as follows:

[0535] ,

[0536] In the formula, To use the inertial force exponential interpolation function trained by DT, These are the actual monitored wind direction, flow direction, flow velocity, wind speed, wave height, and wave period data obtained by the environmental data monitoring module.

[0537] In a specific embodiment, the inverse values ​​of the axial force and bending moment of the jacket platform members are obtained based on the interpolated inertial force index of the members and the real-time static axial force and static bending moment of the members to be considered. Specifically, after obtaining the inertial force index of the members in the actual environment, since most of them are small values ​​less than 0.2, the first-order term can be taken after Taylor expansion of the inertial force index, and the inverse values ​​of the axial force and bending moment of the jacket platform members can be obtained using the following formula:

[0538] ,

[0539] ,

[0540] ,

[0541] In the formula, These are the inverse values ​​of the axial force of the jacket structure members and the bending moments of the members about the principal inertial axes in the y and z directions, respectively. For the real-time static axial force of the member that needs to be monitored; These are the real-time static bending moments about the principal inertial axes in the y and z directions of the members that need to be monitored; The interpolated axial force inertia index of the member. and These are the interpolated bending moment inertial force indices about the principal inertial axes in the y and z directions, respectively.

[0542] Specifically, the design scheme presented in this embodiment introduces real-time monitoring data and combines it with advanced machine learning algorithms to analyze and learn from large-scale data, thereby dynamically updating and optimizing internal force calculations. This effectively overcomes the limitations of traditional methods and makes internal force calculations more accurate and reliable.

[0543] Specifically, the design scheme presented in this embodiment can adaptively consider various complex working conditions and environmental influences, thereby achieving more scientific structural health monitoring and risk assessment. The inertial force index can consider the structural dynamic effects neglected in the SACS simulation calculation, more accurately reflecting the stress state of each member, and the updating of the inertial force index ensures the practicality of the inversion method under extreme sea conditions. The machine learning-based interpolation method ensures the accuracy of database expansion. Especially in large-span structures and high-rise buildings, the design scheme presented in this embodiment can effectively improve the calculation accuracy of the internal forces of the jacket structure members, providing a scientific database for engineering design, maintenance, and safety assessment, and has significant social and economic value. By combining real-time monitoring data and machine learning algorithms, the data in the database structured in this embodiment can provide more reliable decision support for subsequent intelligent structural health monitoring and risk early warning.

[0544] S3: Construct a tertiary twin model based on the aforementioned secondary twin model. The tertiary twin model includes a force prediction model for the entire jacket structure and its members.

[0545] The overall and component stress prediction model of the jacket platform is used to predict the overall horizontal force and axial force on the components of the jacket platform under corresponding environmental loads.

[0546] In a specific embodiment, the steps for constructing a stress prediction model for the overall jacket structure and its members include:

[0547] An axial force response dataset is established based on the axial force data in the aforementioned two-dimensional twin model;

[0548] In this embodiment, an axial force response-overall horizontal force-environmental load database is formed based on the axial force data obtained from the inversion of the second-level twin model to train the overall and member stress prediction model of the jacket structure, thereby improving the prediction accuracy of the overall and member stress prediction model of the jacket structure.

[0549] The overall horizontal force of the jacket platform is calculated based on the axial force response dataset, forming a one-to-one database of axial force response, overall horizontal force and environmental load. The database of axial force response, overall horizontal force and environmental load is preprocessed to obtain the processed database.

[0550] In a specific embodiment, the overall horizontal force of the jacket platform is calculated based on the axial force response dataset, forming a one-to-one database of axial force response-overall horizontal force-environmental load, including:

[0551] S601: Obtain the two-dimensional model of the jacket platform. Let any two support points at the bottom of the jacket in the two-dimensional model be... Dot and A point, subjected to force as Figure 24 As shown, the overall horizontal force is calculated using the following formula:

[0552] ,

[0553] In the formula, This indicates the external force acting on the catheter support. and They are respectively Dot and The reaction force of the point, and They are respectively Dot and The axial force of the outrigger at the point. and for Dot and The axial force of the diagonal bracing member between points;

[0554] Specifically, the legs and diagonal braces are both members of the jacket platform. The axial force of the legs refers to the axial force borne by the legs (columns) of the jacket platform. In this embodiment, it refers to the axial force borne by the legs at points A and B. The axial force of the diagonal braces refers to the axial force borne by the diagonal braces (diagonal support members) of the jacket platform. In this embodiment, it refers to the axial force borne by the diagonal members connecting the legs at points A and B.

[0555] Simplifying the above equation, we obtain the expression for the external force acting on the jacket:

[0556] ,

[0557] according to The formula for calculating the overall horizontal force of the duct stent 2D model is:

[0558] ,

[0559] In the formula, It is a vector in the horizontal direction; This represents the overall horizontal force in the two-dimensional model of the jacket structure.

[0560] S602: Based on the first-order twin model, the direction vector of each sensor monitoring point along the diagonal brace of the jacket support member is defined as follows: Meanwhile, the unit vector in the x-direction is defined as... The unit vector in the y-direction is ;

[0561] according to The horizontal forces in the x and y directions are obtained from the direction vectors of the jacket structure members, the unit vector in the x direction, and the unit vector in the y direction, as shown in the following formula.

[0562] ,

[0563] ,

[0564] In the formula, For the numbering of the jacket support members, ;

[0565] The overall horizontal force characteristic dataset of the jacket platform is obtained based on the above two formulas. Together with the corresponding axial force response dataset and the environmental load data corresponding to the axial force response dataset, the axial force response-overall horizontal force-environmental load database is formed.

[0566] In this embodiment, the unit vector of the duct frame member is determined based on the actual 3D model of the duct frame as follows:

[0567] ,

[0568] , ,

[0569] ,

[0570] ,

[0571] Specifically, a force prediction model for the overall jacket structure and its members is constructed based on CNN neural networks and KAN neural networks;

[0572] The overall and member stress prediction model of the jacket structure is used to predict the overall horizontal force and member axial force of the jacket structure based on environmental load data and the secondary twin model.

[0573] Specifically, the long-term action of horizontal forces may lead to cumulative structural damage and affect the service life of the platform. In this embodiment, by obtaining the axial force of each rod, the overall horizontal force distribution of the jacket platform can be accurately derived, and the overall horizontal force is decomposed into the axial force of the rods for calculation, which simplifies the stress analysis of complex spatial structures.

[0574] Specifically, by calculating the axial force of each member, the overall horizontal force distribution of the structure can be accurately derived. The calculation of the overall horizontal force by decomposing it into member axial forces simplifies the stress analysis of complex spatial structures. A stress prediction model for the overall jacket and its members is constructed based on CNN and KAN neural networks. This model is used to predict the overall horizontal force and member axial force of the jacket based on environmental load data and the secondary twin model. Two KAN network layers are added between the convolutional and fully connected layers of the original CNN. Introducing the KAN network further enhances the network's theoretical support, efficiency, and robustness. Based on the Kolmogorov-Arnold representation theorem, the KAN neural network has a stronger theoretical foundation and structural simplicity. It can accurately approximate multivariate functions with a fixed number of nodes, has high training efficiency, and avoids the gradient vanishing / exploding problem. The resulting original stress prediction model improves the accuracy and effectiveness of predictions, helping to reduce unnecessary on-site monitoring and maintenance.

[0575] Specifically, the overall and member stress prediction model of the jacket structure can output the predicted overall horizontal force and member axial force of the jacket structure based on the input environmental load, helping personnel to maintain and adjust the jacket structure in a timely manner, ensuring the safety of the platform in extreme weather and other emergencies, and protecting the safety of personnel and equipment.

[0576] Specifically, underwater monitoring equipment (such as fiber optic cables) inevitably ages and deteriorates due to long-term operation in harsh and complex marine environments, leading to system-wide shutdowns of digital twin systems. Therefore, the three-level twin model, based on data accumulated from the database of the two-level twin model, trains and establishes a high-precision intelligent model of the marine environment and the stress state of key components. This model directly reflects the mapping relationship between load and response, improving the accuracy of the inversion algorithm while avoiding system-wide shutdowns caused by the inability of underwater sensors to operate in-situ for extended periods.

[0577] In a specific embodiment, a stress prediction model for the overall jacket structure and its members is constructed based on CNN neural networks and KAN neural networks, including:

[0578] like Figure 25 As shown, the structure of the KAN-CNN network model includes an input layer, a first convolutional layer, a second convolutional layer, a third convolutional layer, a first KAN network layer, a second KAN network layer, a fully connected layer, and an output layer connected in sequence; the two KAN network layers have the same structure and both use B-spline as the learning function.

[0579] Specifically, in this embodiment, two KAN networks are added between the convolutional and fully connected layers of the original CNN. Introducing the KAN network further enhances the network's theoretical support, efficiency, and robustness. Based on the Kolmogorov-Arnold representation theorem, the KAN neural network has a stronger theoretical foundation and structural simplicity, capable of accurately approximating multivariate functions with a fixed number of nodes, resulting in high training efficiency and avoiding the gradient vanishing / exploding problem. The introduction of the KAN network significantly enhances the ability to capture the nonlinear correlation between environmental loads and the jacket platform's mechanical response. This allows the model to more accurately predict the overall horizontal force of the jacket platform when processing multidimensional marine environmental load data, reducing prediction errors and providing a solid guarantee for accurate prediction. This improves the accuracy and effectiveness of predictions, reduces unnecessary on-site monitoring and maintenance, lowers overall operating costs, and allows for better design and maintenance strategies that save resources and time. More accurate force prediction enables timely warnings and adjustments, ensuring the platform's safety in extreme weather and other emergencies, and protecting the safety of personnel and equipment.

[0580] Specifically, the overall stress prediction model of the jacket structure and its members is trained based on the processed database to obtain the trained overall stress prediction model of the jacket structure and its members.

[0581] In a specific embodiment, the stress prediction model of the overall jacket structure and its members is trained based on the processed database to obtain the trained stress prediction model of the overall jacket structure and its members, which includes:

[0582] S501: Randomly initialize the weights and bias parameters in the overall and member stress prediction model of the jacket structure;

[0583] Specifically, in this embodiment, the convolutional kernel size is 1*3, and both the convolutional layer and the fully connected layer use the ReLU activation function;

[0584] S502: Input the processed database into the overall and member stress prediction model of the jacket structure:

[0585] The processed database is input into the first convolutional layer through the input layer, and convolution and activation operations are performed on the data in the processed database to obtain the first feature vector;

[0586] The first feature vector is input into the second convolutional layer, and convolution and activation operations are performed on the first feature vector to obtain the second feature vector;

[0587] The second feature vector is input into the third convolutional layer, where convolution and activation operations are performed on the second feature vector to output the third feature vector.

[0588] The third feature vector is input into the first KAN network layer. A learnable combination of activation functions, namely the B-spline function, is applied to each node in the first KAN network layer to linearly aggregate the third feature vector, outputting the fourth feature vector.

[0589] The fourth feature vector is input into the second KAN network layer. The fourth feature vector is linearly aggregated by applying a learnable combination of activation functions, namely the B-spline function, to each node in the second KAN network layer, and the fifth feature vector is output.

[0590] The fifth feature vector is input into the fully connected layer, and the fifth feature vector is linearly transformed by the fully connected layer to output a variable containing the predicted overall horizontal force and axial force of the jacket.

[0591] The predicted overall horizontal and axial forces of the guide frame are output through the output layer;

[0592] S503: Compare the overall horizontal force and axial force of the jacket structure predicted by the overall and member force prediction model with the actual data, and calculate the MSE loss function between the prediction results and the actual labels.

[0593] S504: The gradient of the loss function with respect to each parameter is calculated through the backpropagation algorithm. The backpropagation process propagates the gradient from the output layer to the input layer.

[0594] S505: Based on the calculated gradient information, the gradient descent optimization algorithm is used to update the parameters of each network layer in the overall and rod stress prediction model of the jacket structure. That is, the parameters used for calculation, such as the convolution kernel, bias term, stride, and padding of each convolutional layer, as well as the parameters used for calculation in each connection layer of each KAN network layer and the weights and bias parameters in the fully connected layer, are updated according to the actual optimization results to achieve the prediction effect.

[0595] S506: Repeat the process of calculating loss, backpropagation, and parameter update until the set stopping condition is met (such as reaching the maximum number of iterations or the loss function converges) to obtain the trained overall and member stress prediction model of the jacket.

[0596] In this embodiment, the overall stress prediction model of the jacket and its members can improve the accuracy and effectiveness of prediction, help reduce unnecessary on-site monitoring and maintenance, reduce overall operating costs, and make timely warnings and adjustments through more accurate stress prediction, ensuring the safety of the platform in extreme weather and other emergencies, and protecting the safety of personnel and equipment.

[0597] In a specific embodiment, the axial force response-overall horizontal force-environmental load database is divided into a training set, a test set, and a validation set. The training set, which comprises 80% of the total database, is used to train the neural network model. The validation set, comprising 10% of the total database, is used to fine-tune the model parameters and improve the model's generalization ability. The test set, comprising 10% of the total database, is used to finally evaluate the model's prediction accuracy.

[0598] Specifically, the test set is input into the overall stress prediction model of the jacket structure and its members, the overall stress prediction model of the jacket structure and its members is evaluated, and the overall stress prediction model of the jacket structure and its members is optimized based on the evaluation results, including:

[0599] The test set is input into the overall and member stress prediction model of the jacket structure to obtain the prediction results;

[0600] Calculate the loss of the overall and member stress prediction model of the jacket structure: Compare the prediction results of the overall and member stress prediction model of the jacket structure with the actual data, and calculate the MSE loss value between the prediction results and the actual labels;

[0601] Determine whether the calculated loss value has converged. If it has, no further updates are needed. If it has not, readjust the hyperparameters and the connection structure of the network layers in the model. If the current model does not meet the prediction requirements, change the number or connection relationship of convolutional layers, KAN layers and fully connected layers in the model, and retrain the model.

[0602] In a specific embodiment, the validation set is input into the overall and member stress prediction model of the jacket structure to obtain the overall horizontal force on the jacket platform and the axial force on the jacket members under the environmental load.

[0603] The environmental load data from the validation set is input into the overall and member stress prediction model of the jacket structure to obtain the predicted overall horizontal force and the axial force of the jacket structure members under the environmental load. Compared with the actual data, the deviation is very small, which proves that the overall and member stress prediction model of the jacket structure can output the predicted overall horizontal force and member axial force of the jacket structure according to the input environmental load. This helps personnel to maintain and adjust the jacket structure in a timely manner, ensure the safety of the platform in extreme weather and other emergencies, and protect the safety of personnel and equipment.

[0604] In a specific embodiment, the data interaction management module includes: a data interaction module, a basic data management module, an operation and maintenance management module, a digital twin management module, a 3D digital twin display module, a menu management module, and a user management module;

[0605] The data interaction module includes: a structural stress-strain interaction module, a weight and center of gravity interaction module, a dynamic response interaction module, a tilt angle interaction module, a displacement interaction module, an acceleration interaction module, a wave radar interaction module, a wind speed and direction interaction module, and a cathodic protection interaction module. The structural stress-strain interaction module displays the strain, axial force, and bending moment values ​​calculated from the data collected by the grating fiber optic sensor in various ways, such as data lists and line graphs, and includes several units:

[0606] (a) Real-time monitoring data display unit

[0607] Real-time monitoring data is displayed in the form of a data list, including the measuring point number, full name of the measuring point, member number, elevation, axial force, bending moment MY, bending moment MZ, and data acquisition time. It also supports fuzzy or precise searches using a combination of measuring point number and full name.

[0608] (II) Real-time monitoring data chart display unit

[0609] The fiber optic grating hardware device performs calculations on the data collected by sensors, determining strain, axial force, and bending moment values, and displays the results as a line graph. This graph is linked to the data list; when a measurement point is selected in the data list, the real-time graph dynamically switches between displaying the calculated results from the data collected by the corresponding sensor at that measurement point.

[0610] (III) Historical Monitoring Data Chart Display Unit

[0611] The historical monitoring data chart display is an accumulation of real-time monitoring data charts. Users can dynamically switch between historical data of the last 10 minutes, the last 1 day, and the last 7 days, and display them in a line chart format. Similarly, the charts and data lists have a linkage effect. When a measuring point is selected in the data list, the historical chart will dynamically switch to display the results of calculations on the historical data collected by the sensors under that measuring point.

[0612] (iv) Historical chart data display unit

[0613] Historical chart data display presents charts in a structured data format, displayed as a data list. This includes statistical summaries of historical data for strain, axial force, and bending moment.

[0614] (v) Monitoring historical data file unit

[0615] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0616] (vi) Monitoring Configuration Management Unit

[0617] Monitoring configuration management is the unified configuration and management of structural stress and strain measurement points, sensors, member location information, and key parameters required for calculation.

[0618] (vii) Early warning configuration management unit

[0619] Axial force early warning configuration management sets early warning values ​​for axial force-tensile (KN), axial force-compression (KN), axial force-tensile (KN), and axial force-compression (KN) under the operating environment, extreme environment, and extreme environment.

[0620] (viii) Early Warning Historical Units

[0621] The historical records of axial force warnings are displayed in list format.

[0622] Specifically, the weight and center of gravity interaction module performs calculations on the data collected by the hardware device through sensors, calculating the strain value, average strain value, weight, and center of gravity, and displays the results in various ways such as data lists and line graphs, including several units:

[0623] (a) Real-time monitoring data display unit

[0624] Real-time monitoring data is displayed in the form of a data list, including the measuring point number, full name of the measuring point, member number, elevation, pipe diameter D, wall thickness t, axial force correction coefficient, and data acquisition time. It supports fuzzy or precise queries using a combination of measuring point number and full name.

[0625] (II) Real-time monitoring data chart display unit

[0626] The hardware device performs calculations based on data collected by sensors, determining strain, average strain, weight, and center of gravity, and displays the results as a line graph. This graph is linked to the data list; when a measurement point is selected in the data list, the real-time graph dynamically switches between displaying the calculated results from the data collected by the corresponding sensor at that point.

[0627] (III) Historical Monitoring Data Chart Display Unit

[0628] The historical monitoring data chart display is an accumulation of real-time monitoring data charts. Users can dynamically switch between historical data of the last 10 minutes, the last 1 day, and the last 7 days, and display them in a line chart format. Similarly, the charts and data lists have a linkage effect. When a measuring point is selected in the data list, the historical chart will dynamically switch to display the results of calculations on the historical data collected by the sensors under that measuring point.

[0629] (iv) Historical chart data display unit

[0630] Historical chart data display presents the data in a structured manner, displayed as a data list. This includes statistical summaries of historical data for stress value 1 and stress value 2.

[0631] (v) Monitoring historical data file unit

[0632] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0633] (vi) Monitoring Configuration Management Unit

[0634] Monitoring configuration management involves the unified configuration and management of weight center of gravity measurement points, sensors, rod position information, and key parameter indicators required for calculation.

[0635] Specifically, the dynamic response interaction module will perform calculations on the data collected by the dynamic acquisition device through the sensors, calculating acceleration, frequency domain, and fundamental frequency, and displaying the results in various ways such as data lists and line graphs, including several units:

[0636] (a) Real-time monitoring data display unit

[0637] Real-time monitoring data is displayed in a data list format, including sensor number, pole number, elevation, X-axis acceleration, Y-axis acceleration, Z-axis acceleration, and data acquisition time. Fuzzy or precise search by sensor number is supported.

[0638] (II) Real-time monitoring data chart display unit

[0639] The hardware device performs calculations on the data collected by the sensors, determining acceleration, frequency domain, and fundamental frequency, and displays the results as a line graph. This graph is linked to the data list; when a sensor is selected in the data list, the real-time graph dynamically switches between displaying the calculated results from the data collected by that sensor.

[0640] (III) Historical Monitoring Data Chart Display Unit

[0641] The historical monitoring data chart display is an accumulation of real-time monitoring data charts. Users can dynamically switch between historical data from the last 10 minutes, the last 1 day, and the last 7 days, and display them in a line chart format. Similarly, the charts and data lists have a linked effect. When a sensor is selected in the data list, the historical chart will dynamically switch to display the results of calculations on the historical data collected by that sensor.

[0642] (iv) Historical chart data display unit

[0643] Historical chart data display presents the data in a structured manner, displayed as a data list. This includes statistical summaries of historical data for maximum acceleration, frequency domain, and fundamental frequency.

[0644] (v) Monitoring historical data file unit

[0645] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0646] (vi) Monitoring Configuration Management Unit

[0647] Monitoring configuration management is the unified configuration and management of sensors, pole position information, and key parameters required for calculation that enable dynamic responses.

[0648] (vii) Early warning configuration management unit

[0649] Dynamic response early warning configuration management sets early warning thresholds for the Y-axis maximum, Y-axis minimum, X-axis maximum, and X-axis minimum of the base frequency.

[0650] (viii) Early Warning Historical Units

[0651] The historical records of baseband warnings are displayed in list format.

[0652] Specifically, the tilt angle interaction module performs calculations on the data collected by the dynamic acquisition device through sensors, calculates the maximum values ​​in the X and Y directions, and displays the results in various ways such as data lists and line graphs. It includes several units:

[0653] (a) Real-time monitoring data display unit

[0654] Real-time monitoring data is displayed in a data list format, including sensor number, pole number, elevation, X-direction tilt angle, Y-direction tilt angle, and data acquisition time. Fuzzy or precise search by sensor number is supported.

[0655] (II) Real-time monitoring data chart display unit

[0656] The hardware device performs calculations on the data collected by the sensors, determining the maximum values ​​in the X and Y directions and displaying the results as a line graph. This graph is linked to the data list; when a sensor is selected in the data list, the real-time graph dynamically switches between displaying the calculated results from the data collected by that sensor.

[0657] (III) Historical Monitoring Data Chart Display Unit

[0658] The historical monitoring data chart display is an accumulation of real-time monitoring data charts. Users can dynamically switch between historical data from the last 10 minutes, the last 1 day, and the last 7 days, and display them in a line chart format. Similarly, the charts and data lists have a linked effect. When a sensor is selected in the data list, the historical chart will dynamically switch to display the results of calculations on the historical data collected by that sensor.

[0659] (iv) Historical chart data display unit

[0660] Historical chart data display presents charts with structured data in a list format. This includes statistical summaries of historical maximum values.

[0661] (v) Monitoring historical data file unit

[0662] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0663] (vi) Monitoring Configuration Management Unit

[0664] Monitoring configuration management is the unified configuration and management of sensors, pole position information, and key parameters required for calculation that enable dynamic responses.

[0665] (vii) Early warning configuration management unit

[0666] Tilt angle warning configuration management sets warning values ​​for the angle of tilt caused by operation and extreme environments;

[0667] (viii) Early Warning Historical Units

[0668] The historical records of tilt warnings are displayed in list format.

[0669] Specifically, the displacement interaction module performs calculations on the data collected by the dynamic acquisition device through sensors, calculates the maximum values ​​in the X and Y directions, and displays the results in various ways, such as data lists and line graphs. It includes several units:

[0670] (a) Real-time monitoring data display unit

[0671] Real-time monitoring data is displayed in a data list format, showing sensor number, pole number, elevation, X-direction displacement, Y-direction displacement, and data acquisition time. Fuzzy or precise search by sensor number is supported.

[0672] (II) Real-time monitoring data chart display unit

[0673] The hardware device performs calculations on the data collected by the sensors, determining the maximum values ​​in the X and Y directions and displaying the results as a line graph. This graph is linked to the data list; when a sensor is selected in the data list, the real-time graph dynamically switches between displaying the calculated results from the data collected by that sensor.

[0674] (III) Historical Monitoring Data Chart Display Unit

[0675] The historical monitoring data chart display is an accumulation of real-time monitoring data charts. Users can dynamically switch between historical data from the last 10 minutes, the last 1 day, and the last 7 days, and display them in a line chart format. Similarly, the charts and data lists have a linked effect. When a sensor is selected in the data list, the historical chart will dynamically switch to display the results of calculations on the historical data collected by that sensor.

[0676] (iv) Historical chart data display unit

[0677] Historical chart data display presents charts with structured data in a list format. This includes statistical summaries of historical maximum values.

[0678] (v) Monitoring historical data file unit

[0679] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0680] (vi) Monitoring Configuration Management Unit

[0681] Monitoring configuration management is the unified configuration and management of sensors, pole position information, and key parameters required for calculation that enable dynamic responses.

[0682] (vii) Early warning configuration management unit

[0683] Displacement warning configuration management sets warning values ​​for the angle of displacement during operation and under extreme environments;

[0684] (viii) Early Warning Historical Units

[0685] The historical displacement warning records are displayed in list format.

[0686] Specifically, the anemometer interaction module performs calculations on the data collected by the hardware device through sensors, calculates the wind speed and direction, and displays the results in various ways, such as data lists and line graphs. It includes several units:

[0687] (a) Real-time monitoring data display unit

[0688] Real-time monitoring data is displayed in the form of a data list, which includes angle, wind speed, and data collection time.

[0689] (II) Real-time monitoring data chart display unit

[0690] The hardware device performs a calculation based on the data collected by the sensors, calculates the wind speed and wind direction, and displays the effect in the form of a line graph.

[0691] (III) Historical Monitoring Data Chart Display Unit

[0692] The historical monitoring data charts are an accumulation of real-time monitoring data. Users can dynamically switch between historical data from the last 10 minutes, the last 1 day, and the last 7 days, and display the data in a line chart format.

[0693] (iv) Historical chart data display unit

[0694] Historical chart data display presents charts in a structured format, displayed as a data list. This includes statistical summaries of historical wind speed and direction data.

[0695] (v) Monitoring historical data file unit

[0696] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0697] (vi) Monitoring Configuration Management Unit

[0698] Monitoring configuration management refers to the unified configuration and management of environmental monitoring.

[0699] (vii) Early warning configuration management unit

[0700] Wind speed warning configuration management allows you to set warning values ​​for wind speed op and wind speed ex;

[0701] (viii) Early Warning Historical Units

[0702] Historical wind speed warning records are displayed in list format.

[0703] Specifically, the wave interaction module displays data collected by the hardware device through sensors, showing significant wave height, maximum wave height, TM2 period, wave crest direction, wave crest period, wave crest wavelength, ocean current direction, ocean current speed, and time. It presents the data in various ways, including data lists and line graphs, and includes several units:

[0704] (a) Real-time monitoring data display unit

[0705] Real-time monitoring data is displayed in the form of a data list, which includes significant wave height, maximum wave height, TM2 period, wave crest direction, wave crest period, wave crest wavelength, ocean current direction, ocean current speed, and time.

[0706] (II) Real-time monitoring data chart display unit

[0707] The hardware device displays data collected by sensors, including significant wave height, maximum wave height, TM2 period, wave crest direction, wave crest period, wave crest wavelength, ocean current direction, ocean current speed, and time, and presents the results in the form of text and line graphs.

[0708] (III) Historical Monitoring Data Chart Display Unit

[0709] The historical monitoring data chart display is an accumulation of real-time monitoring data charts. Users can dynamically switch between historical data of the last 10 minutes, the last 1 day, and the last 7 days, and display them in the form of line charts.

[0710] (iv) Historical chart data display unit

[0711] Historical chart data display presents the charts in a structured data format, in the form of a data list, including statistical summaries of data such as significant wave height, maximum wave height, TM2 period, wave crest direction, wave crest period, wave crest wavelength, ocean current direction, ocean current speed, and time.

[0712] (v) Monitoring historical data file unit

[0713] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0714] (vi) Monitoring Configuration Management Unit

[0715] Monitoring configuration management refers to the unified configuration and management of environmental monitoring.

[0716] (vii) Early warning configuration management unit

[0717] Early warning configuration management involves setting warning values ​​for flow velocity op, flow velocity ex, wave height op, and wave height ex.

[0718] (viii) Early Warning Historical Units

[0719] Historical records of flow velocity and wave height warnings are displayed in list format.

[0720] Specifically, the cathodic protection interactive module displays the data collected by the hardware device through sensors, showing the Ag reference electrode potential value, Zn reference electrode potential value, and sacrificial anode release current value, and presenting the effects in various ways such as data lists and line graphs. It includes several units:

[0721] (a) Real-time monitoring data display unit

[0722] The real-time monitoring data is displayed in the form of a data list, which includes the number, potential value, and data acquisition time of the Ag reference electrode potential, the number, potential value, and data acquisition time of the Zn reference electrode potential, and the number, current value, and data acquisition time of the sacrificial anode release current.

[0723] (II) Real-time monitoring data chart display unit

[0724] The hardware device displays the values ​​of Ag reference electrode potential, Zn reference electrode potential, and sacrificial anode release current through data collected by sensors, using line graphs to illustrate the effects.

[0725] (III) Historical Monitoring Data Chart Display Unit

[0726] The historical monitoring data charts are an accumulation of real-time monitoring data. Users can dynamically switch between historical data from the last 10 minutes, the last 1 day, and the last 7 days, and display the data in a line chart format.

[0727] (iv) Historical chart data display unit

[0728] Historical chart data display presents the data in a structured manner, in the form of a data list. This includes a statistical summary of the Ag reference electrode potential number, potential value, and data acquisition time; the Zn reference electrode potential number, potential value, and data acquisition time; and the sacrificial anode release current number, current value, and data acquisition time.

[0729] (v) Monitoring historical data file unit

[0730] Historical monitoring data files are created by summarizing, packaging, and compressing real-time monitoring data, and saving it in CSV file format to a file server for archiving. The historical CSV file information is displayed as a data list, allowing users to easily download the historical CSV files and providing support for subsequent business operations.

[0731] Specifically, in this embodiment, the basic data management module includes: a design data management unit, an engineering data management unit, a file management unit, and a SACS model management unit.

[0732] Specifically, the design data management unit is responsible for the unified management and maintenance of platform basic information data, structured data and unstructured data of detailed design calculation results on the design data management page. This includes:

[0733] (a) Platform Basic Data Unit

[0734] The platform's basic data includes overall platform information, structural calculation summaries, basic structural information, sea conditions, pile and soil conditions, and weight control information. This information is managed and maintained uniformly within the jacket platform health management system of this embodiment and presented in the form of data forms.

[0735] (ii) Detailed design calculation result data unit

[0736] The detailed design calculation results include four parts of data: static analysis, fatigue analysis, seismic analysis, and collapse analysis. Static analysis is further divided into four parts: load, members, nodes, and pile foundation data. Fatigue analysis includes fatigue safety factors, grinding control nodes, support additions, and inner ring additions. Seismic analysis includes foundation data, toughness analysis results, and strength analysis results. All this data is managed and maintained within the jacket platform health management system and presented in the form of data forms.

[0737] Specifically, the engineering data management unit includes a material tracking unit and an NDT tracking unit:

[0738] (a) Material Tracking Unit

[0739] Material tracking unifies the management and maintenance of the relationship between structural members, material tracking, and material certificates, and displays the relationships between structural members, material tracking, and material certificates in a data list. In addition, it supports adding, deleting, modifying, and querying data, file downloading, and importing and exporting data.

[0740] (ii) NDT tracking unit

[0741] NDT tracking includes NDT structural overview, weld information, and NDT data. This data is managed and maintained uniformly within the jacket platform health management system. It also supports adding, deleting, modifying, querying data, downloading files, and importing and exporting data.

[0742] Specifically, file management involves the unified management and maintenance of unstructured file types involved in the project, and supports functions such as file upload, download, addition, modification, deletion, and query.

[0743] Specifically, the SACS model management unit supports the import of SACS model data files and displays the imported data in a list format. The duct stent platform health management system integrates 3D modeling capabilities, and the maintenance and management of SACS model data within this system is the fundamental support for generating 3D models.

[0744] Specifically, in this embodiment, the operation and maintenance management module includes: an overall assessment management unit, a partial assessment management unit, a testing project management unit, an inspection plan management unit, and an implementation record management unit. The overall assessment management unit and the partial assessment management unit conduct overall or partial assessments of the offshore jacket through assessment items and rule settings. The jacket platform health management system filters risk levels through assessment scores, providing data support for subsequent testing plans and implementation.

[0745] Specifically, the overall assessment management unit includes platform risk assessment, overall scoring item management, and overall risk rule management; the partial assessment management unit includes partial assessment, original assessment, and partial assessment risk rules, specifically including:

[0746] (a) Assessment Item Setting Unit

[0747] The catheter stent platform health management system supports custom assessment items and the maintenance and management of these items, including functions such as adding, deleting, modifying, and querying assessment items.

[0748] (II) Evaluation Rule Setting Unit

[0749] The catheter stent platform health management system supports custom assessment rules for assessment items and maintains the relationship between assessment items and rules;

[0750] (III) Scoring Unit

[0751] Users can select assessment items themselves. The system calculates the total score based on the assessment items selected by the user. The final score is linked to the risk rules. After the user selects the risk rules, the system automatically determines the risk level, risk recommendations, handling requirements, and risk matrix diagram based on the assessment score.

[0752] Specifically, risk rules involve customizing rules, levels, and matrices, and linking them to risk assessments. These rules include:

[0753] (a) Risk rule setting unit

[0754] Users can customize risk rules, including adding, deleting, modifying, and querying them. Creating risk rules includes specifying the rule name and sorting order.

[0755] (II) Risk Level Setting Unit

[0756] Users can add, delete, modify, and query risk levels;

[0757] (III) Risk Matrix Unit

[0758] Users can customize the risk matrix settings and select the risk level within the risk matrix.

[0759] Specifically, this system supports custom detection items, including operations such as adding, modifying, deleting, querying, and viewing details, specifically including:

[0760] (a) New additions

[0761] The function to add new testing items includes: testing item name, selection of testing record template (structural thickness measurement, marine organism measurement, mechanical damage bending, anode, potential, circular pipe thickness measurement, beam thickness measurement, water ingress detection, ACFM), uploading of requirement documents, and requirement descriptions;

[0762] (II) Amendments

[0763] Specifically, this system supports modification of newly added inspection items. Users can modify the inspection item name, reselect the inspection record template (structural thickness measurement, marine organism measurement, mechanical damage bending, anode, potential, circular pipe thickness measurement, beam thickness measurement, water ingress detection, ACFM), re-upload the upload requirement file, and modify the requirement description.

[0764] (iii) Deletion

[0765] This system supports the deletion of existing test items, which can be done individually or in batches.

[0766] (iv) Inquiry

[0767] This system supports precise or fuzzy matching queries by combining the detection item number and item name.

[0768] (V) Details

[0769] The testing details page displays the created testing project records. The details include the testing project name, testing implementation record template, requirement documents, and requirement descriptions.

[0770] Specifically, this system supports custom detection plans, including operations such as adding, modifying, deleting, querying, and viewing details:

[0771] (a) New additions

[0772] The newly added inspection plan function includes inspection plan name, plan start time, plan end time, selection of inspection items (structural thickness measurement, marine organism measurement, mechanical damage bending, anode, potential, circular pipe thickness measurement, beam thickness measurement, water ingress detection, ACFM) and calls 3D model service. By operating the visual model, the member to be inspected is selected, effectively combining the plan, inspection items, and member in this system;

[0773] (II) Amendments

[0774] This system supports modification of newly added inspection plans, including modification of the inspection plan name, plan start time, plan end time, reselection of inspection items (structural thickness measurement, marine organism measurement, mechanical damage bending, anode, potential, circular pipe thickness measurement, beam thickness measurement, water ingress detection, ACFM), and reselection of rods;

[0775] (iii) Deletion

[0776] This system supports the deletion of existing test plans, which can be done individually or in batches.

[0777] (iv) Inquiry

[0778] This system supports precise or fuzzy matching queries by combining the detection plan number, plan name, start time, and end time.

[0779] (V) Details

[0780] The inspection plan details page displays the created inspection plan records. Details include the inspection plan name, start time, end time, items to be inspected, rods, and 3D models.

[0781] Specifically, the implementation record management is for implementation and maintenance personnel. By accessing the system menu and selecting "Implementation Record Management," implementation personnel can view the testing plans and items pre-defined by the administrator, perform specific testing items based on the test records, and upload the test results back to the system for unified maintenance and management. Implementation personnel can also view past testing history records within this system. Specific functions are as follows:

[0782] (a) Download the testing implementation template

[0783] This system supports the download of inspection project implementation templates. Maintenance personnel can download them to their local machines, view the projects and pole information that need to be maintained, and fill in the relevant inspection content for later transmission back to this system for preparation.

[0784] (ii) Implement template data import

[0785] This system supports the implementation of template data import function. After users fill in the template data offline, they can use this function to upload the data to this system and display it in the form of a data list.

[0786] (III) New Implementation Records

[0787] If you do not want to use the template data import function, this system also supports the function of adding new detection implementation records;

[0788] (iv) Historical Testing Records

[0789] This system allows maintenance personnel to view past implementation and testing records. It supports keyword searches by plan name, start time, and end time.

[0790] In this embodiment, the twin management module includes: a working condition management unit and a digital twin calculation result management unit;

[0791] Specifically, the operating condition management unit is used to manage the operating condition data required to establish the digital twin model;

[0792] Specifically, the digital twin calculation result management unit is used to manage the simulation calculation results of the jacket platform obtained based on the modified SACS simulation model, the secondary digital twin model, and the overall health status of the jacket platform obtained based on the tertiary digital twin model.

[0793] In this embodiment, the 3D digital twin display module includes an information overview unit and a digital twin display unit;

[0794] The information overview unit includes:

[0795] (a) Sensor status statistics

[0796] Statistics show the activity and total number of sensors, and the online and offline ratios of sensors are displayed in the form of charts and pie charts;

[0797] (ii) Monitoring peak data

[0798] Peak data from the monitoring sensors is displayed as a line graph;

[0799] (iii) Fundamental frequency

[0800] The fundamental frequency data of the display structure is presented in the form of a line chart;

[0801] (iv) Axial force

[0802] The axial force data of the monitored rods is displayed in the form of a bar chart;

[0803] (v) Risk warning statistics within 24 hours

[0804] Displays risk warning information within 24 hours, showing its percentage distribution in pie chart format;

[0805] (vi) Marine environment

[0806] Displays marine environmental data such as wind direction, wind speed, and waves, presented in instrument charts;

[0807] (vii) Cathodic protection

[0808] The data, including Ag reference electrode potential, Zn reference electrode potential, and sacrificial anode release current, are displayed as line graphs.

[0809] Specifically, the digital twin display unit includes:

[0810] (a) Overview Map

[0811] The platform's geographical location is displayed on an overview map;

[0812] (II) Basic Information

[0813] Display basic platform information, including the company to which it belongs, water depth, platform type, number of wells, and jacket type;

[0814] (III) Monitoring and Twins

[0815] Displays monitoring and twin data, including data on acceleration, displacement, tilt angle, weight center of gravity, stress and strain, and environmental conditions;

[0816] (iv) Assessment and Testing

[0817] Displays assessment and testing data, including the most recent assessment time, the maximum UC value of the main leg in place, and the minimum lifespan of fatigue analysis.

[0818] Specifically, in this embodiment, a system management module is also included, which specifically includes: user management, role management, menu management, and department management;

[0819] In this embodiment, the user management module is used for centralized management and maintenance of user information, including:

[0820] (a) New users

[0821] New user information includes user nickname, department, mobile phone number, email address, username, password, gender, status, job title, role, and remarks;

[0822] (ii) Modify user

[0823] Modifying user information includes user nickname, department, mobile phone number, email address, user password, user gender, status, job title, role, and remarks;

[0824] (iii) Deleting a user

[0825] This system supports deleting created users;

[0826] (iv) User Inquiry

[0827] Users can search for users by entering their username, mobile phone number, status, or creation time.

[0828] (v) Import

[0829] This system supports importing user data from Excel spreadsheets and displays the imported user data as a data list.

[0830] (vi) Export

[0831] This system supports exporting user data lists in Excel format.

[0832] Specifically, this system includes a role management module for assigning and managing roles to users who wish to use the system, including:

[0833] (a) New characters

[0834] The information for newly added roles includes role name, assigned permissions, role order, status, assignment menu, and remarks;

[0835] (ii) Modify the character

[0836] Modifying role information includes role name, assigned permissions, role order, status, assignment menu, and remarks;

[0837] (iii) Deleting a character

[0838] This system supports deleting created roles;

[0839] (iv) Query Role

[0840] You can query roles by entering the role name, permission characters, status, and creation time keywords;

[0841] (v) Export

[0842] This system supports exporting character data lists in Excel format.

[0843] Specifically, this system includes a department management module for assigning and managing departments to users who wish to use the system, including:

[0844] (a) New additions

[0845] The newly added information includes the superior department, department name, display sort, person in charge, contact number, email address, and department status;

[0846] (II) Amendments

[0847] The modifications include the superior department, department name, display sorting, person in charge, contact number, email address, and department status;

[0848] (iii) Deletion

[0849] This system supports deleting departments that have already been created;

[0850] (iv) Inquiry

[0851] You can search for departments by entering the department name or status keyword.

[0852] In this embodiment, the menu management module is used to assign and manage menu permissions for users who want to use the system, including:

[0853] (a) New additions

[0854] The new information includes parent menu, menu type, menu icon, menu name, display sort, whether it is an external link (whether it is an external page reference, which is not referenced in this system), route address (front-end page path), display status (whether the menu is hidden), and menu status (status: disabled, enabled).

[0855] (II) Amendments

[0856] Modifications include parent menu, menu type, menu icon, menu name, display sorting, whether it is an external link (whether it is an external page reference, which is not referenced in this system), route address (front-end page path), display status (whether the menu is hidden), and menu status (status: disabled, enabled).

[0857] (iii) Deletion

[0858] This system supports deleting created menu items;

[0859] (iv) Inquiry

[0860] You can search for the menu by entering the menu name or status keyword.

[0861] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A jacket platform health management system based on digital twin technology, characterized in that, The method comprises the following steps: The data acquisition module is used for high-frequency real-time acquisition of jacket platform monitoring data from different types of data acquisition devices through a combination of nested multithreading and asynchronous multithreading, wherein the jacket platform monitoring data comprises jacket platform structure data and marine environmental load monitoring data. The data acquisition module is used for high-frequency real-time acquisition of jacket platform monitoring data from different types of data acquisition devices through a combination of nested multithreading and asynchronous multithreading, wherein the jacket platform monitoring data comprises jacket platform structure data and marine environmental load monitoring data. A plurality of marine data acquisition main threads are created on an industrial computer, and the plurality of marine data acquisition main threads are used for starting the industrial computer to perform a task of high-frequency real-time acquisition of jacket platform monitoring data from different types of data acquisition devices. A media-level time controller sub-thread is nested in each marine data acquisition main thread, and the industrial computer performs the task of high-frequency real-time acquisition of jacket platform monitoring data from data acquisition devices based on an interval time set by the media-level time controller sub-thread, wherein each data acquisition device stores monitoring data transmitted from a plurality of sensors. The media-level time controller sub-thread transmits the acquired monitoring data to a marine data transmission module. The data acquisition module is also used for obtaining jacket platform model data. The marine data transmission module is used for transmitting the data acquired by the data acquisition module to a land data receiving and processing module through a set sea-land data transmission mode. The land data receiving module is used for receiving the data transmitted by the marine data transmission module and performing an import storage operation on the received data according to a set import storage rule. The data preprocessing module is used for performing a preprocessing operation on the stored data, wherein the preprocessing operation comprises data noise reduction processing and data missing supplement processing. The data distributed computing module comprises a plurality of data computing sub-modules, and each data computing sub-module is used for performing any one or more of data solving, data querying and early warning computing on the preprocessed data according to the data type of the preprocessed data. The digital twin module comprises: A first-level digital twin model, which is a SACS simulation model corrected based on preprocessed data and jacket platform model data, and is used for jacket platform simulation calculation and jacket platform health index early warning. A second-level digital twin model, which is formed by an application database, and is used for further providing data support for a third-level digital twin model, wherein the application database is formed based on the data processed by the data distributed computing module and the first-level digital twin model. A third-level digital twin model, which is a deep learning model trained based on the second-level digital twin model, and is used for predicting the overall health status of the current jacket platform. The data interaction management module is used for providing a jacket platform related data display interface to enable a user to obtain and manage jacket platform information in real time.

2. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The set sea-land data transmission mode comprises: According to the data type of the jacket platform monitoring data, a corresponding raw data file is formed, according to the number of the raw data file, the corresponding time span and the corresponding processing thread, the jacket platform monitoring data is split according to the corresponding time span in time sequence and a compressed file and corresponding check information, a time stamp and id information are formed; A signal is sent to the land data receiving and processing module at a fixed time interval, and whether the current network state is normal is judged according to whether the feedback signal sent by the land data receiving and processing module can be received, and when the network state is normal, the compressed file and the corresponding check information are sent to the land data receiving and processing module, and an information table representing the sent compressed file and the real-time receiving state of the compressed file is generated; When the network state is abnormal, the operation process of collecting offshore data and generating a compressed file is performed until the network state is normal, and the generated compressed file and the corresponding check information are continuously sent to the land data receiving and processing module.

3. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The step of receiving the data transmitted by the offshore data sending module and importing and storing the received data according to the set import storage rule by the land data receiving module includes: The received compressed file and the corresponding check information are checked according to the check information, it is judged whether the transmitted compressed file has been received, whether the compressed file is correct and complete, and the correct and complete compressed file is saved to form a data record table; the data record table is used to represent the information of the received compressed file and the real-time saving state of the corresponding compressed file; According to the data record table, the saved compressed file is decompressed, the decompressed data is imported into the database, and the saving state of the corresponding compressed file in the data record table is updated.

4. The jacket platform health management system based on digital twin technology according to claim 3, wherein, The specific steps of importing the decompressed data into the database by the land data receiving module include: According to the data type coding of the decompressed data, a data partition unit for data storage is obtained; At least three time level split sub-modules are set to split the monitoring time series data according to the time level data splitting strategy and obtain minute-level sequence data, hour-level sequence data and day-level sequence data, and then obtain the split sequence data of each time level; The split sequence data of each time level is written into the data partition unit according to the set writing rule to store the monitoring time series data according to the data type, and then the storage data of different time levels is obtained; A timing task mode is adopted to poll and obtain the split progress of the decompressed data of the at least three time level split sub-modules, obtain a progress query signal, and according to the progress query signal, control the time level split sub-module to stop running after confirming that the split sequence data of the corresponding time level is completed.

5. The jacket platform health management system based on digital twin technology according to claim 4, wherein, The time level data splitting strategy is: The current progress time stamp of each type of data in the corresponding decompressed data is obtained; The current progress timestamp is taken as a starting time point, and a timestamp node serving as an ending time point is confirmed according to split time levels of the split sub-modules corresponding to the at least three time levels, so as to realize split sequence data acquisition of each time level.

6. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The method for creating a plurality of marine data collection main threads comprises: According to the data type of the marine data, the corresponding data collection path, data storage path and time span are configured, and the number of corresponding processing threads is calculated according to the file quantity of different data types of the marine data, and the number of processing threads required to be allocated for each file is calculated according to the time span of different data types; According to the file quantity of different data types of the marine data, the number of corresponding processing threads is calculated, and the number of processing threads required to be allocated for each file is calculated according to the time span of different data types, comprising: Set the configuration field of the file quantity, if the file quantity of a certain data type is greater than the configuration field of the file quantity, then allocate a basic thread for each file in the data type for processing, if the file quantity of a certain data type is less than the configuration field of the file quantity, then allocate a basic thread for all files in the data type for processing; Set the configuration field of the upper limit of the time interval, if the time interval of the data in a certain data file of a certain data type is greater than the set configuration field, then split the data in the file according to the time span to obtain the number of allocated threads after time span calculation, and process the split data in parallel; if the time interval of the data in a certain data file of a certain data type is less than the set configuration field, then allocate a thread for the data in the file for processing.

7. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The step of performing noise reduction processing on the stored data based on the data preprocessing module comprises: Set the corresponding noise reduction method for the stored data to obtain analysis data; An adaptive noise reduction selection method is constructed, the analysis data is analyzed according to the adaptive noise reduction selection method, it is judged whether the noise reduction method corresponding to the monitoring data is replaced, and the stored data is noise reduced according to the noise reduction method obtained after the judgment, and the analysis data together constitute a noise reduction data set, comprising: Set the number of analysis data and the noise point criterion, the noise point criterion includes a standard deviation criterion and a noise point number criterion: Set the standard deviation criterion to n' times of the standard deviation, and the data outside the range of mean plus or minus n' times of the standard deviation is noise point data; Set the specific value of the noise point number criterion, if the number of noise points in the analysis data quantity is lower than the number of noise point data, then use the adjacent difference value algorithm for data noise reduction; if the number of noise points in the analysis data quantity is higher than the number of noise point data, then use the band-pass filtering algorithm for data noise reduction; Set the analysis period, process the monitoring data in the next analysis period according to the replaced noise reduction method to obtain new noise reduction data, and the analysis data together constitute a noise reduction data set.

8. The jacket platform health management system based on digital twin technology according to claim 7, wherein, The step of performing data missing supplement processing on the noise reduction data set based on the data preprocessing module comprises: Based on the noise reduction data set, a target sensor data set to be supplemented is obtained; The sensor data missing judgment strategy is constructed, and the target sensor data set is preprocessed. The preprocessed target sensor data set is divided into training sets and test sets of short-term data missing types and long-term data missing types according to the sensor data missing judgment strategy. The target sensor data set includes non-failed data of the target sensor itself and non-failed data of other sensors; The bidirectional recurrent neural network and the GRU unit are introduced to obtain the time relationship of the sensor, the full connection layer is introduced to obtain the spatial relationship between the sensor data, and a double-branch data processing model based on the bidirectional recurrent neural network and the GRU unit is constructed. The double-branch data processing model is used to judge different data missing types according to the sensor data missing type judgment strategy, and select the corresponding branch to process the input data; The training sets of the two data missing types are input into the double-branch data processing model for training to obtain a sensor data recovery model. The sensor data recovery model is used to recover the missing data according to different data missing types using the corresponding sub-model; The test set is input into the sensor data recovery model to obtain the predicted missing data. The sensor data recovery model is evaluated and optimized according to the error value of the predicted value and the true value, and an optimized sensor data recovery model is obtained. The non-failed data of the target sensor itself and the non-failed data of other sensors are input into the optimized sensor data recovery model to judge the missing data type, and the missing data of the target sensor is output to supplement the missing data in the target sensor data set to be supplemented, and a preprocessed data set is obtained.

9. The jacket platform health management system based on digital twin technology according to claim 8, wherein, The double-branch data processing model includes an input layer, a missing data type discriminator, a network calling module, a short-term missing data recovery model, a long-term missing data recovery model, and an output layer. The missing data type discriminator is used to judge whether the data missing type is a short-term data missing type or a long-term data missing type according to the sensor data missing judgment strategy, and the judgment result is input into the network calling module to call the corresponding branch for training. The network calling module is used to call the short-term missing data recovery model or the long-term missing data recovery model to process the input data according to the judgment result of the missing data type discriminator; The short-term missing data recovery model includes m layers of full connection layers, 1 layer of GRU layers, and r layers of Bi-GRU layers connected in sequence, and is used to recover the short-term missing data. The long-term missing data recovery model includes m layers of full connection layers and n layers of GRU layers connected in sequence, and is used to recover the long-term missing data.

10. The jacket platform health management system based on digital twin technology according to claim 9, wherein, The training sets of the two data missing types are input into the double-branch data processing model for training to obtain a sensor data recovery model, including: S31: A set of hyperparameters, i.e., m, n, and r, are randomly determined, and the weights and biases of each network layer in the double-branch data processing model are randomly initialized under the parameters; S32: The data training set is input to the missing data type discriminator through the input layer for judgment. If it is a short-term data missing type, the judgment result is input to the network calling module, S33 is entered to call the short-term missing data recovery model for training. If it is a long-term data missing type, the judgment result is input to the network calling module, S34 is entered to call the long-term missing data recovery model for training; S33: The short-term missing data training set is input to m fully connected layers, and the input data is reduced in dimension and features are extracted through multi-layer linear and nonlinear transformation to capture the spatial correlation between different sensors; The features extracted through the m fully connected layers are input to a GRU layer to obtain the short-term time dependence in the input features; The features processed through the GRU layer are input to r Bi-GRU layers to obtain the context information of a certain time period in the past and a certain time period in the future; The features extracted through the Bi-GRU layer are integrated and fused through the output layer to predict the missing data corresponding to the current data point as the missing data of the first time point; The missing data of the first time point is added to the data training set, and the training process is repeated to predict the missing data corresponding to the current data point as the missing data of the second time point; The predicted missing data corresponding to the current data point is continuously added to the data training set for training until the missing data of the x*-1 time point is added to the data training set for training, and the missing data corresponding to the current data point is predicted as the missing data of the x* time point; x* represents the time point of the end position of the short-term missing data; S34: The long-term missing data training set is input to m fully connected layers, and the input data is reduced in dimension and features are extracted through multi-layer linear and nonlinear transformation to capture the spatial correlation between different sensors; The features extracted through the m fully connected layers are input to n GRU layers to obtain the long-term time dependence in the input features; The features extracted through the n GRU layers are integrated and fused through the output layer to predict the missing data corresponding to the current data point as the missing data of the first time point; The missing data of the first time point is added to the data training set, and the training process is repeated to predict the missing data corresponding to the current data point as the missing data of the second time point; The predicted missing data corresponding to the current data point is continuously added to the data training set for training until the missing data of the y*-1 time point is added to the data training set for training, and the missing data corresponding to the current data point is predicted as the missing data of the y* time point; y* represents the time point of the end position of the long-term missing data; S35: Set the loss function and the optimizer: set MSE as the loss function, and use the Adam optimizer for optimization; S36: Calculate the loss function of the predicted data and the real data, calculate the gradient according to the loss function, and update the weights and biases of the model using the Adam optimizer according to the calculated gradient; S37: Set the numerical range of the hyperparameters, adjust the number of layers of the full connection layer, GRU layer and Bi-GRU layer of the model based on the set numerical range according to the grid search method, and obtain the sensor data recovery model.

11. The jacket platform health management system based on digital twin technology according to claim 10, wherein, The sensor data missing judgment strategy comprises: Determine the start time point and the end time point of the missing data time period in the training set, and calculate the time length of the missing data time period; Determine the time length of the available data time period after the missing data, which is the time length from the end time point of the missing data time period to the start time point of the next missing data time period; Set the time threshold of the missing data and the time threshold of the available data, when the time length of the missing data time period is less than the time threshold of the missing data, and the time length of the available data time period is greater than the threshold of the available data, the missing data of the target sensor is short-term data missing type; When the time length of the missing data time period is greater than the time threshold of the missing data or the time length of the available data time period is less than the threshold of the available data, the missing data of the target sensor is long-term data missing type.

12. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The data distributed computing module comprises a structural stress and strain calculation submodule, which is used to solve the strain data in the preprocessed data, comprising: Solve the strain data in the preprocessed data to obtain the axial force data, and calculate as follows: ， wherein E is the elastic modulus in Pa, A is the cross-sectional area of the jacket leg in m 2 , ; is the strain data obtained without the shroud; is the axial force correction factor for the sensor shroud; Solve the strain data in the preprocessed data to obtain the bending moment data, and calculate as follows: ， ， In the formula, M is the bending moment of the rod around the neutral plane of the No. 2 and No. 4 sensors; M is the bending moment of the rod around the neutral plane of the No. 1 and No. 3 sensors; I is the moment of inertia of the pipe cross section, with the unit of ; M is the bending moment of the rod around the neutral plane of the No. 2 and No. 4 sensors; D D is the diameter of the jacket rod; t is the wall thickness; Obtain the correction coefficient of the axial force and the bending moment, and correct the calculated axial force and bending moment data to obtain the corrected axial force data and bending moment data.

13. The jacket platform health management system based on digital twin technology according to claim 12, wherein, The specific steps of obtaining the correction coefficient of the axial force and the bending moment comprise: S11: Design a corresponding sensor outer shield according to the type of the strain sensor, and establish a simulation model of the strain sensor outer shield; S12: Construct a strain sensor clamping block model, and calculate the strain data of the strain sensor under different load conditions when the two clamping bases of the strain sensor are without a sensor outer shield according to the strain sensor clamping block model; S13: Construct a strain sensor clamping block model with a strain sensor outer shield according to the simulation model of the strain sensor outer shield, and calculate the strain data of the strain sensor under different load conditions when the two clamping bases of the strain sensor have a sensor outer shield according to the strain sensor clamping block model with a strain sensor outer shield; S14: Calculate the correction coefficient of the strain sensor according to the two types of strain data obtained in S12 and S13, wherein the correction coefficient comprises an axial force correction coefficient and a bending moment correction coefficient.

14. The jacket platform health management system based on digital twin technology according to claim 13, wherein, The calculation of the axial force correction coefficient and the bending moment correction coefficient of the strain sensor comprises: S141: Calculate the axial force correction coefficient according to the obtained strain data, and the formula is: ， wherein, represents an axial force correction factor of the sensor shroud, represents strain data acquired in the case of no shroud; represents strain data acquired in the case of a shroud. S142: Calculate the bending moment correction coefficient according to the obtained strain data, and the formula is: * ， wherein, represents a bending moment correction coefficient, represents a bending moment correction coefficient of different points of the sensor, ; represents strain data acquired in the case of no shield under a bending moment load, represents strain data acquired in the case of a shield under a bending moment load.

15. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The structural stress and strain calculation submodule for solving the strain data in the preprocessed data further comprises: The actual strain value of the main leg of the multi-leg support structure of the jacket platform under the action of the upper block is calculated to obtain the weight gravity center of the upper block, including: Obtaining the strain value of the main leg of the multi-leg support structure of the jacket platform under the action of the upper block in the preprocessed data; According to the two strain values, the axial force of the main leg of the multi-leg support structure of the jacket platform is obtained through the main leg axial force calculation formula; According to the axial force of the main leg of the multi-leg support structure of the jacket platform, the weight of the upper block is obtained to obtain the gravity center coordinates of the upper block.

16. The jacket platform health management system based on digital twin technology according to claim 15, wherein, The main leg axial force calculation formula is as follows: ， ， ， In the formula: This indicates the axial force of the main leg of the multi-leg support structure of the jacket platform; Indicates the elastic modulus; This represents the cross-sectional area of ​​the main leg of the multi-leg support structure of the jacket platform; This indicates the serial number of the strain sensor deployed at the monitoring point, that is, the serial number of the strain value of the main leg of the multi-leg support structure of the jacket platform under no-load condition obtained at the location of the monitoring point. This indicates the total number of strain sensors deployed at the monitoring point, which is the total number of strain values ​​of the main legs of the multi-leg support structure of the jacket platform under no-load conditions obtained at the location of the monitoring point. Indicates the influence coefficient of the sensor housing; The first monitoring point was deployed at the location of the monitoring point. i Actual monitoring data from each strain sensor; The first monitoring point was deployed at the location of the monitoring point. i No-load monitoring data from a strain sensor; This indicates the diameter of the main leg of the multi-leg support structure of the jacket platform; This indicates the wall thickness of the main leg of the multi-leg support structure of the jacket platform; Represents a counting function; Indicates actual monitoring data The total number of strain sensors that are not zero.

17. The jacket platform health management system based on digital twin technology according to claim 16, wherein, According to the axial force of the main leg of the multi-leg support structure of the jacket platform, the weight of the upper block is obtained to obtain the gravity center coordinates of the upper block, including: The calculation formula of the weight of the upper block is as follows: ， In the formula: This indicates the weight of the upper component, in kg. The first leg support structure of the jacket platform Axial force of the main leg; This indicates the sequence number of the main leg of the multi-leg support structure of the jacket platform. This indicates the total number of main legs of the multi-leg support structure of the jacket platform; It is the acceleration due to gravity; The calculation formula of the gravity center coordinates of the upper block is as follows: ， ， In the formula: The X-axis coordinate representing the centroid of the upper block; The Y-axis coordinate representing the centroid of the upper block; The first leg support structure of the jacket platform k The X-axis coordinates of each main leg; The first leg support structure of the jacket platform k The Y-axis coordinate of each main leg.

18. The jacket platform health management system based on digital twin technology according to claim 17, wherein, The data distributed calculation module further includes a jacket platform model calculation submodule, which is used to solve the pile-soil p-y curve in the preprocessed data, including: The pile-soil p-y curve of the jacket installation position is corrected, and the method used is as follows: S121: Construct an implicit function based on the cyclic load coefficient, standard soil resistance and initial modulus of internal friction angle to obtain the difference format of the implicit function: ， ， ， wherein is the implicit function for the 1st order natural frequency; is the implicit function for the 2nd order natural frequency; is the implicit function for the 3rd order natural frequency; is the 1st order natural frequency increment; is the 2nd order natural frequency increment; is the 3rd order natural frequency component; denotes the difference step size of A denotes the difference step size of denotes the difference step size of denotes the difference step size of denotes the difference step size of denotes the difference step size of A denotes the cyclic load coefficient; denotes the standard soil resistance; denotes the internal friction angle initial modulus; Then, the discrete partial derivative of the implicit function is: , , ， , , ， , , ， wherein: , , , , , , , , are intermediate calculation parameters; denotes the difference step of the discretized A ; denotes the difference step of the discretized ; denotes the difference step of the discretized ; Therefore, the discrete difference format is: ， ， ， Wherein, ， ， ， ， ， ， ， ， ， where: Implicit function representing the 1st order natural frequency An increment in the direction of A ; Implicit function representing the 1st order natural frequency An increment in the direction of ; Implicit function representing the 1st order natural frequency An increment in the direction of ; Implicit function representing the 2nd order natural frequency An increment in the direction of A ; Implicit function representing the 2nd order natural frequency An increment in the direction of ; Implicit function representing the 2nd order natural frequency An increment in the direction of ; Implicit function representing the 3rd order natural frequency An increment in the direction of A ; Implicit function representing the 3rd order natural frequency An increment in the direction of ; Implicit function representing the 3rd order natural frequency An increment in the direction of ; Then the cyclic load coefficient, standard soil resistance and initial modulus of internal friction angle are obtained, and the formula used is: ， ， ， ， In the formulae: represents a set iteration threshold value; S122: Based on the differential step size of the cyclic load coefficient, the differential step size of the standard soil resistance, and... The differential step size is used to obtain the corrected cyclic load coefficient, corrected standard soil resistance, and corrected initial modulus of internal friction angle, in order to obtain the corrected pile-soil Py curve at the jacket installation position. The formula used is as follows: ， In the formula: represents the horizontal soil resistance; represents the initial modulus of internal friction; represents the depth of action.

19. The jacket platform health management system based on digital twin technology according to claim 18, wherein, The jacket platform model calculation submodule is also used to solve the wall thickness of the jacket bar in the preprocessed data, including: The wall thickness of the jacket bar is corrected, and the formula used is as follows: ， wherein: tcorrrepresents the corrected wall thickness of the jacket leg; D t1represents the as-delivered pipe diameter of the jacket leg; m1represents the actual mass of the jacket leg after delivery; l1represents the length of the jacket leg.

20. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The data query corresponding to the data provided by the data distributed calculation module includes: Receiving the query request of the query data sent by the preset client; Receiving the timestamp query priority of the query request of the query data sent by the preset client; According to the query request of the query data, the response data corresponding to the query condition of the query request is determined; Wherein, the timestamp query priority is the query matching rule formulated for different time span query conditions, specifically: If the query request of the query data is a data query condition within one minute, the response data corresponding to the second level is preferentially called; If the query request of the query data is a data query condition within one hour, the response data corresponding to the minute level is preferentially called; If the query request of the query data is a data query condition within one day, the response data corresponding to the hour level is preferentially called; If the query request of the query data is a data query condition within one year, the response data corresponding to the day level is preferentially called.

21. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The pre-warning calculation corresponding to the data provided by the data distributed calculation module includes: The pre-warning method selection rule is constructed according to different data characteristics in the pre-processed data set, and a pre-warning method corresponding to the data is selected according to the pre-warning rule, a corresponding safety range and an alarm value are set, and a pre-warning calculation result is obtained; The pre-warning method selection rule is constructed according to different data characteristics in the pre-processed data set, and a pre-warning method corresponding to the data is selected according to the pre-warning rule, a corresponding safety range and an alarm value are set, and a pre-warning calculation result is obtained; S101: When the monitoring data comes from the horizontal direction and needs to be combined with the data value of the horizontal direction for pre-warning, the ring boundary pre-warning method is selected; S102: When the monitoring data is a specific numerical value of a certain data type and exceeds the normal value range, the upper and lower limit pre-warning method is selected; S103: When the monitoring data is a specific numerical value of a certain data type at different positions in the vertical direction, and the difference between the specific numerical values at different positions exceeds the normal range, the settlement difference pre-warning method is selected.

22. The jacket platform health management system based on digital twin technology according to claim 4, wherein, The set writing rule includes: The start time point and the end time point are taken as data query conditions, that is, the split sequence data greater than the start time point and less than the end time point is taken as the query condition of the time span acquisition data; The maximum data value and the minimum data value of the split sequence data and the corresponding query condition are written into the data partition unit; The end time point is updated to the current progress timestamp.

23. The jacket platform health management system based on digital twin technology according to claim 1, wherein, The step of constructing the first-level digital twin model based on the pre-processed data and the jacket platform model data includes: Based on the jacket platform model data, a SACS simulation model of the jacket is established, and the jacket platform model data includes the pipe diameter and wall thickness of the jacket member; The first three order natural frequencies of the jacket platform are obtained according to the pre-processed data, and the measured data includes the average acceleration, displacement of the monitoring points of the jacket platform, and the axial force data of the main leg and the horizontal support of the jacket platform; The predicted first three order natural frequencies of the jacket platform are obtained based on the SACS simulation model; The SACS simulation model is corrected based on the measured first three order natural frequencies and the predicted first three order natural frequencies to obtain the first-level twin model.

24. The jacket platform health management system based on digital twin technology according to claim 23, wherein, The construction steps of the second-level digital twin model include: Based on the environmental load information in the current monitoring time period and the first-level twin model in the data processed by the data distributed computing module, the maximum axial force and the maximum bending moment of the jacket platform under the prior condition are calculated; the environmental load information includes: wind speed, wind direction, wave height, wave period, sea water flow direction, and sea water flow speed; The member that needs to be paid attention to in the jacket platform is determined, and the strain of the member is obtained, and then the maximum axial force and the maximum bending moment of the member in the current monitoring time period are obtained; Based on the environmental load information in the current monitoring time period, the maximum axial force and the maximum bending moment of the jacket platform under the prior condition, and the maximum axial force and the maximum bending moment of the member in the current monitoring time period, the inertia force index of the member is obtained; According to the inertia force index of the member, and based on the machine learning method of the decision tree, the interpolated inertia force index of the jacket platform member is obtained; Obtaining real-time environmental load information of the jacket platform, and obtaining real-time static axial force and static bending moment of a rod member needing attention according to the real-time environmental load information of the jacket platform and the primary twin model; Obtaining inversion values of the axial force and the bending moment of the rod member of the jacket platform based on the interpolated inertia force index of the rod member of the jacket platform and the real-time static axial force and the static bending moment of the rod member needing attention, to establish an application database including the environmental load information, the axial force, the bending moment and the inertia force index, and to form a secondary twin model through the application database.

25. The jacket platform health management system based on digital twin technology according to claim 24, wherein, The deep learning model is a jacket overall and rod member stress prediction model, and the steps of constructing the jacket overall and rod member stress prediction model include: Establishing an axial force response data set based on the axial force data in the secondary twin model; Calculating the overall horizontal force of the jacket platform according to the axial force response data set, forming a one-to-one corresponding axial force response-overall horizontal force-environmental load database, and obtaining a processed database by preprocessing the axial force response-overall horizontal force-environmental load database; Constructing the jacket overall and rod member stress prediction model based on a CNN neural network and a KAN neural network; The jacket overall and rod member stress prediction model is used to predict the overall horizontal force of the jacket and the axial force of the rod member according to the environmental load data and the secondary twin model; Training the jacket overall and rod member stress prediction model based on the processed database to obtain a trained jacket overall and rod member stress prediction model.

26. The jacket platform health management system based on digital twin technology according to claim 24, wherein, The specific steps of obtaining the inertia force index of the rod member needing attention based on the environmental load information in the current monitoring time period, the maximum axial force and the maximum bending moment of the jacket platform under the priori working condition, and the maximum axial force and the maximum bending moment of the rod member needing attention in the current monitoring time period include: When the time of the jacket platform serving underwater is less than a set service time threshold, an offline calculation method is adopted to obtain the inertia force index of the rod member needing attention; When the time of the jacket platform serving underwater is greater than or equal to the set service time threshold, an online calculation method is adopted to obtain the inertia force index of the rod member needing attention.

27. The jacket platform health management system based on digital twin technology according to claim 26, wherein, The step of obtaining the inertia force index of the rod member needing attention by the offline calculation method includes: S401: According to the wave height, the wave period and the primary twin model, an existing SACS static force calculation module is called to obtain the maximum horizontal force of the rod member needing attention of the jacket platform under wave load; S402: According to the wave height, the wave period and the primary twin model, an existing SACS dynamic force calculation module is called to obtain the horizontal force dynamic response curve of the rod member needing attention of the jacket under wave load, and then obtain the absolute value maximum of the horizontal force of the rod member needing attention; S403: According to the maximum horizontal force of the rod member needing attention of the jacket under wave load and the absolute value maximum of the horizontal force of the rod member needing attention, a wave amplification coefficient is obtained; S404: According to the wave amplification coefficient, the axial force and the bending moment of the rod member needing attention considering the inertial effect are obtained based on the SACS static calculation module; S405: The inertial force index of the rod member needing attention is obtained according to the axial force and the bending moment of the rod member needing attention considering the inertial effect.

28. The jacket platform health management system based on digital twin technology according to claim 27, wherein, The inertial force index of the rod member needing attention is obtained according to the axial force and the bending moment of the rod member needing attention considering the inertial effect, and the formula used is as follows: ， In the formula, is the axial force inertia force index of the rod, is the axial force of the rod under the prior working condition considering the inertia effect, is the maximum value of the axial force of the rod in the jacket under the prior working condition; ， ， In the formula, bending moment inertia force index of the bar member around the y, z direction principal inertia axis, respectively; bending moment of the bar member around the y, z direction principal inertia axis, respectively, considering the inertia effect; respectively represent the maximum value of the bending moment of the bar member around the y, z direction principal inertia axis.

29. The jacket platform health management system based on digital twin technology according to claim 28, wherein, The formula used for obtaining the inertial force index of the rod member needing attention is as follows: ， ， ， In the formula: is the axial force inertia force index of the bar, are the bending moment inertia force indices of the bar around the main inertia axes y and z respectively; is the maximum axial force of the bar under attention in the current monitoring time interval; , are the maximum bending moments of the bar under attention around the main inertia axes y and z respectively in the current monitoring time interval; is the maximum static axial force of the bar under attention in the current monitoring time interval; , are the maximum static bending moments of the bar under attention around the main inertia axes y and z respectively in the current monitoring time interval.

30. The jacket platform health management system based on digital twin technology according to claim 24, wherein, The inversion values of the axial force and the bending moment of the jacket platform rod member are obtained based on the interpolated inertial force index of the jacket platform rod member and the real-time static axial force and static bending moment of the rod member needing attention, and the formula used is as follows: ， ， ， In the formula, respectively are the inverse values of the axial force of the truss rod, the bending moment of the rod around the main inertia axis of y, z direction, is the real-time static axial force of the rod which needs to be concerned; respectively are the real-time static bending moments of the rod around the main inertia axis of y, z direction which needs to be concerned; is the interpolated axial force inertia force index of the rod, respectively are the interpolated bending moment inertia force indexes of the rod around the main inertia axis of y, z direction.

31. The jacket platform health management system based on digital twin technology according to claim 23, wherein, The SACS simulation model is corrected based on the measured first three order natural frequencies and the predicted first three order natural frequencies, including: S111: Compare the error between the measured first three order natural frequencies of the jacket platform and the predicted first three order natural frequencies, if greater than the set threshold, execute S112, otherwise, execute S114; S112: The SACS simulation model is corrected by using the weight center of gravity of the jacket upper block, the wall thickness of the jacket rod member and the pile-soil p-y curve of the jacket installation position calculated by the structure stress and strain calculation submodule and the jacket platform model calculation submodule, to obtain the corrected SACS simulation model; S113: The predicted first three order natural frequencies of the jacket platform are reacquired based on the corrected SACS simulation model, and the reacquired predicted first three order natural frequencies of the jacket platform are taken as the current predicted first three order natural frequencies, and S111 is executed; S114: The correction of the SACS simulation model is completed.

32. The jacket platform health management system based on digital twin technology according to claim 25, wherein, The jacket overall and rod member stress prediction model is constructed based on CNN neural network and KAN neural network, including: The structure of the jacket overall and rod member stress prediction model includes an input layer, a first convolutional layer, a second convolutional layer, a third convolutional layer, a first KAN network layer, a second KAN network layer, a fully connected layer and an output layer connected in turn; The two KAN network layers have the same structure and use B-spline as the learning function.

33. The jacket platform health management system based on digital twin technology according to claim 32, wherein, The jacket overall and rod member stress prediction model is trained based on the processed database to obtain the trained jacket overall and rod member stress prediction model, including: S501: Randomly initialize the weight and bias parameters in the jacket overall and rod member stress prediction model; S502: Input the processed database into the jacket overall and rod member stress prediction model: The processed database is input into the first convolutional layer through the input layer, and convolution and activation operations are performed on the data in the processed database to obtain the first feature vector; The first feature vector is input into the second convolutional layer, and convolution and activation operations are performed on the first feature vector to obtain the second feature vector; The second feature vector is input into the third convolutional layer, and convolution and activation operations are performed on the second feature vector to output the third feature vector; The third feature vector is input to the first KAN network layer, and a learnable activation function combination, i.e., a B-spline function, is applied to each node in the first KAN network layer to perform linear aggregation on the third feature vector, and a fourth feature vector is output; The fourth feature vector is input to the second KAN network layer, and a learnable activation function combination, i.e., a B-spline function, is applied to each node in the second KAN network layer to perform linear aggregation on the fourth feature vector, and a fifth feature vector is output; The fifth feature vector is input to the fully connected layer, and the fully connected layer performs linear transformation on the fifth feature vector, and outputs a variable, which contains the predicted overall horizontal force of the jacket and the axial force of the rod; The overall horizontal force and the axial force of the jacket predicted by the jacket overall and rod force prediction model are compared with the real data, and the MSE loss function between the prediction result and the real label is calculated; S504: The gradient of the loss function with respect to each parameter is calculated by the back propagation algorithm, and the gradient is propagated from the output layer to the input layer in the process of back propagation; S505: According to the calculated gradient information, the gradient descent optimization algorithm is used to update the parameters of each network layer in the jacket overall and rod force prediction model; S506: Repeat the process of calculating the loss, back propagation and parameter updating until the set stop condition is reached, and obtain the trained jacket overall and rod force prediction model. According to the axial force response data set, the overall horizontal force of the jacket platform is calculated to form a one-to-one corresponding axial force response-overall horizontal force-environment load database, including:

34. The jacket platform health management system based on digital twin technology according to claim 25, wherein, Simplify the above formula to obtain the expression of the external force on the jacket, which is: S601: Obtain a two-dimensional model of the jacket platform, and set any two support points of the bottom of the jacket in the two-dimensional model as point and point, and the overall horizontal force is calculated as shown in the following formula: ， wherein Fextdenotes the external force on the jacket, and are respectively the support reaction forces at the points and are respectively the support leg axial forces of the support legs at the points and and and are respectively the brace axial forces of the brace members between the points and According to the above two formulas, the platform overall horizontal force feature data set is obtained, and the corresponding axial force response data set and the environmental load data corresponding to the axial force response data set are combined to form an axial force response-overall horizontal force-environment load database. ， According to The overall horizontal force of the jacket 2D model is calculated by the formula: ， wherein is the horizontal vector; represents the overall horizontal force of the jacket 2D model; S602: define the jacket member directional vector of each sensor monitoring point along the diagonal bracing direction as , and define the unit vector in the x direction as , and the unit vector in the y direction as ; According to , the horizontal force in the x direction and the horizontal force in the y direction are obtained from the conduit frame bar element direction vector, the x direction unit vector, and the y direction unit vector, as follows: ， ， In the formulae, The numbers of the tubular frame members, ; ​

Images