A marine wind power blade fault diagnosis system based on multi-source sensor fusion

Abstract

The application relates to the field of offshore wind power intelligent operation and maintenance and industrial control technology, in particular to an offshore wind power blade fault diagnosis system based on multi-source sensor fusion, which comprises the following: a multi-source data acquisition unit which acquires blade vibration, acoustic, strain, infrared data and wind speed, wave height, transport ship attitude and operation state feedback data; an adaptive noise reduction processing unit which generates dynamic noise reduction parameters according to the original sea state data and outputs multi-source clean feature data; a fault diagnosis unit which outputs fault type, severity and fault location information and generates maintenance and hoisting work orders and fault disposal requests; and a closed-loop control unit which generates cooperative control instructions and performs deviation dynamic correction and issues suspension, locking or follow-up holding instructions when the sea state is abnormal. The application improves the fault recognition stability under complex sea conditions and makes the maintenance operation consistent with the diagnosis result.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance and industrial control technology for offshore wind power, specifically to an offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion. Background Technology

[0002] During the operation and maintenance of offshore wind turbines, the blades are exposed to high humidity, strong corrosion and large waves for a long time. Blade fault diagnosis usually requires the combination of multiple status information such as vibration, acoustics, strain and temperature. At the same time, maintenance operations are also affected by wind speed, wave height and changes in the attitude of the work vessel. Therefore, the accuracy of blade status identification and the stability of subsequent hoisting and maintenance operations are of great importance to the safe operation and maintenance of offshore wind turbines. In traditional methods, blade fault monitoring often relies on single sensor data or distributed processing. Signals obtained under environmental disturbances have weak noise interference resistance, and the fault type, fault degree, and fault location are difficult to stably reflect the true state. Moreover, fault judgment, work order arrangement, and hoisting execution are usually independent of each other. When sea state fluctuates or multiple control requests occur concurrently, the coordination of operations and the continuity of control are still low. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a fault diagnosis system for offshore wind turbine blades based on multi-source sensor fusion. Specifically, the technical solution of this invention includes: The multi-source data acquisition unit is used to collect raw data on the vibration, acoustics, strain and infrared status of the blades, raw data on the environmental wind speed, wave height and the attitude of the transport ship, and operational status feedback data of the external hoisting equipment. An adaptive noise reduction processing unit is used to extract the wind and wave features of the raw sea state data, combine them with a preset mapping relationship to reduce noise in the raw blade state data, and output multi-source clean feature data. The fault diagnosis unit is used to generate fault diagnosis results based on multi-source cleaning feature data, and generate maintenance and hoisting work orders and fault handling requests when preset conditions are met. The human-machine collaboration unit is used to receive user input commands, generate hoisting control requests, and output system status to the user; The closed-loop control unit is used to generate collaborative control commands to external hoisting execution equipment based on fault diagnosis results, maintenance and hoisting work orders, raw sea state data, target docking position input by the human-machine collaboration unit, and operation status feedback data; and to calculate deviations and dynamically correct collaborative control commands; generate sea state anomaly requests when sea state is abnormal; and adjudicate conflicts of multiple requests according to preset rules, and issue collaborative control commands when sea state restrictions are met. A standardized interactive interface is used to transmit data and instructions.

[0004] Optionally, the specific methods by which the closed-loop control unit adjudicates conflicts among multiple types of requests include: When generating a request, each unit attaches the target device identifier, priority identifier, and update time information, and synchronizes them to the standardized interaction interface; The closed-loop control unit acquires each request and compares it with a pre-built control attribution table; When a sea state anomaly request meets the preset sea state restrictions, its priority is raised to the highest level and control is updated. When both the fault handling request and the hoisting control request are valid and there is no abnormal sea condition, the priority identifiers and update time information of the two are compared in turn, and the control of the target equipment is switched to the request with higher priority or later update. The standardized interaction interface will send the updated request results for execution.

[0005] Optionally, the system also includes a scheduling management unit for managing the entire lifecycle of each processing unit, specifically including: When a unit is loaded, a unique identifier is assigned and its processing capabilities are registered to the service directory. Data types for sending and receiving are synchronized to complete the initialization. When execution is triggered, a task request is sent to the corresponding unit, the running state is switched, and the business data is notified to be distributed. When a pause or unload command is received, the corresponding flag unit status is checked, task distribution is stopped, or resources are released and the flag is deregistered.

[0006] Optionally, when the target unit is paused, abnormally switched, or unloaded, the scheduling and management unit may maintain the execution of the last valid collaborative control command by the external hoisting execution equipment, or maintain the last valid fault result, until the control ownership is reassigned or a new command is issued.

[0007] Optionally, the system adopts a ship-shore collaborative distributed architecture; the multi-source data acquisition unit is set on the wind turbine body, nacelle, transport ship or auxiliary platform side; the adaptive noise reduction processing unit, fault diagnosis unit and closed-loop control unit are deployed on the shipborne computing node, shore-based control center or edge computing node; the human-machine collaboration unit is deployed on the user client.

[0008] Optionally, the fault diagnosis unit includes: The feature fusion module is used to receive multi-source clean feature data and generate fused feature data; The fault identification module is used to generate blade fault type and severity results based on fused feature data; The fault location module is used to generate fault location information based on strain and acoustic data; The work order generation module is used to generate a maintenance and hoisting work order containing the above information by combining the pre-acquired blade parameters when the fault status meets the preset conditions.

[0009] Optionally, the closed-loop control unit includes: The strategy matching module is used to match the collaborative operation mode of external hoisting execution equipment based on work order information; The pose compensation module is used to generate pose compensation amounts based on the transport ship's attitude data and wind speed data. The control generation module is used to generate collaborative control commands based on the collaborative operation mode, pose compensation amount, and target docking position. The safety cutoff module is used to determine sea state limitations and trigger pause, lock, or follow-up hold commands.

[0010] Optionally, standardized interactive interfaces include data acquisition interfaces, processing call interfaces, device control interfaces, work order and result interfaces, and rule adjustment interfaces, which are used to match the targeted transmission of corresponding types of data, tasks, instructions, and parameters.

[0011] Optionally, the human-machine collaboration unit receives control commands containing target equipment identifiers and sea state threshold parameters through a visual interactive interface. After completing standardization verification and authorization verification, it transmits the commands to the corresponding processing unit through a standardized interactive interface to adjust the underlying control logic or operation strategy. It also supports restoring the default configuration through a reset command.

[0012] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention simultaneously acquires vibration, acoustic, strain, and infrared data of the blades, as well as wind speed, wave height, transport vessel attitude, and operational status data of external hoisting equipment through a multi-source data acquisition unit. This enables the simultaneous acquisition of the blade's physical state and external sea condition boundary conditions, providing consistent data input for subsequent noise reduction processing, fault diagnosis, and hoisting control. This improves the problem in traditional methods where single-sensor data is prone to feature overload and cannot stably reflect the true fault state. 2. This invention extracts wind speed and wave height features from raw sea state data and generates dynamic noise reduction parameters based on a preset wind-wave vibration coupling mapping relationship. It performs adaptive noise reduction on the raw blade state data, which can introduce the influence of sea state changes on sensor background noise into the threshold adjustment process, reduce the coverage of fault features by background noise in the context of high wind and waves, and improve the availability of multi-source features and the stability of subsequent fault identification under complex sea conditions. 3. This invention achieves unified interaction for data acquisition, processing and invocation, equipment control, work order and result transmission, and rule adjustment through a standardized interactive interface. It also receives user-input blade parameters, target docking positions, and control instructions for fault diagnosis parameters, hoisting operation parameters, and control rules through a human-machine collaboration unit. Before issuing control instructions, it performs format standardization verification, permission scope verification, and object correlation checks, which can improve the consistency of information transmission between various units of the system and the controllability of parameter adjustment. This makes it easier for users to view results, process work orders, and control operations within the same interactive system. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of a closed-loop control system for fault diagnosis and hoisting maintenance of offshore wind turbine blades based on multi-source sensor fusion, provided in an embodiment of this application. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. Example

[0015] Please see Figure 1 A fault diagnosis system for offshore wind turbine blades based on multi-source sensor fusion, the system comprising: The multi-source data acquisition unit is used to collect raw data on the vibration, acoustics, strain and infrared status of the blades, raw data on the environmental wind speed, wave height and the attitude of the transport ship, and operational status feedback data of the external hoisting equipment. An adaptive noise reduction processing unit is used to extract the wind and wave features of the raw sea state data, combine them with a preset mapping relationship to reduce noise in the raw blade state data, and output multi-source clean feature data. The fault diagnosis unit is used to generate fault diagnosis results based on multi-source cleaning feature data, and generate maintenance and hoisting work orders and fault handling requests when preset conditions are met. The human-machine collaboration unit is used to receive user input commands, generate hoisting control requests, and output system status to the user; The closed-loop control unit is used to generate collaborative control commands to external hoisting execution equipment based on fault diagnosis results, maintenance and hoisting work orders, raw sea state data, target docking position input by the human-machine collaboration unit, and operation status feedback data; and to calculate deviations and dynamically correct collaborative control commands; generate sea state anomaly requests when sea state is abnormal; and adjudicate conflicts of multiple requests according to preset rules, and issue collaborative control commands when sea state restrictions are met. Standardized interaction interfaces are used to transmit data and instructions; The specific methods by which the closed-loop control unit adjudicates conflicts among multiple types of requests include: When generating a request, each unit attaches the target device identifier, priority identifier, and update time information, and synchronizes them to the standardized interaction interface; The closed-loop control unit acquires each request and compares it with a pre-built control attribution table; When a sea state anomaly request meets the preset sea state restrictions, its priority is raised to the highest level and control is updated. When both the fault handling request and the hoisting control request are valid and there is no abnormal sea condition, the priority identifiers and update time information of the two are compared in turn, and the control of the target equipment is switched to the request with higher priority or later update. The standardized interaction interface will send the updated request results for execution.

[0016] In this embodiment, the system is used for fault diagnosis and hoisting maintenance linkage operations of offshore wind turbine blades; it acquires vibration, acoustic, strain and infrared data of the blades, and combines wind speed, wave height, transport vessel attitude and external hoisting equipment operation status data to perform noise reduction processing on the raw blade status data to generate multi-source clean feature data; then, based on the multi-source clean feature data, it generates blade fault type results, fault severity results and fault location information, and generates maintenance and hoisting work orders and fault handling requests when conditions are met; By combining the target docking position, raw sea state data, and operational status feedback data, collaborative control commands are generated. In the event of abnormal sea conditions or conflicting requests, pause, lock, or follow-up holding commands are issued after adjudicating according to preset rules. Through this process, the diagnostic results can be directly involved in the lifting control, ensuring that the diagnostic results provide real-time data support for the lifting control. Acquire raw data on blade status and sea state. Based on the sensor deployment location, synchronously collect data on blade operation status and environmental status to generate raw data on blade status, raw data on sea state, and operational status feedback data. Among them, blade status raw data refers to the original monitoring quantities generated during blade operation or maintenance, such as the local vibration abrupt change, acoustic emission pulse, strain anomaly and local temperature rise corresponding to the expansion of cracks on the blade surface; sea state raw data refers to the boundary input of the marine operating environment, such as environmental disturbances when the wind speed reaches 25m / s and the wave height reaches 1.5m; and operation status feedback data refers to the real-time operation feedback information of external hoisting execution equipment, such as the crane's operation status, changes in the attitude of the transport ship, and deviations in execution position. Unlike traditional single-sensor independent processing methods, this invention extracts the correlation features of the same fault in different channels through multi-source synchronous acquisition and unified interface transmission. By cross-verifying the features extracted from multiple channels, the signal-to-noise ratio of fault features is improved. In summary, the embodiments of this invention can achieve consistency of diagnostic input based on multi-source synchronous acquisition, avoid the misalignment of fault features caused by asynchronous sampling, and improve the reliability and consistency of fault diagnosis. The raw sea state data is acquired, and the blade state data is adaptively denoised based on wind speed and wave height characteristics to generate multi-source clean feature data. Among them, the wind-wave-vibration coupling mapping relationship refers to the correspondence between changes in wind speed and wave height and changes in the background noise of the blade monitoring signal. For example, when the wind speed increases, the background energy of the vibration channel increases, and when the wave height increases, the hull attitude fluctuations create additional disturbances to the acoustic and strain channels. In practical implementation, to quantify the impact of sea state on monitoring signals, the system combines the current operating time with the data. Below, the current wind speed that directly affects the wind load input to the blades and hoisting equipment during offshore operations. And the current wave height, which represents the sea surface fluctuation input that causes the transport ship's attitude fluctuations. And introduce the critical safe wind speed used in system design. and critical safety wave height The coupling weights used to dynamically balance the contributions of steady-state wind speed, steady-state wave height, and sudden wind speed changes to noise intensity are utilized. , and Examples of values ​​are 0.4, 0.4, and 0.2, and a reference standard for the rate of change of wind speed for dimensional normalization is introduced. Its dimensions are The dynamic change factor of sea state was calculated as follows: Furthermore, after obtaining the dynamic mutation factor of sea state... Then, wavelet decomposition was performed on the original blade state data to obtain the scale parameter. and time translation parameters Determined wavelet coefficients for each layer This is combined with the initial noise standard deviation, which represents the noise floor level of the current channel when there are no valid fault characteristics. The sampling length represents the size of the time window covered by this noise reduction process. And the sea state coupling gain coefficient used to adjust the sensitivity of different sensors to sea state disturbances. For example, the vibration channel value can be set to 1.5, and the infrared channel value can be set to 0.1, thus constructing a dynamic threshold: The absolute values ​​of the wavelet coefficients at each layer are compared with the dynamic threshold. After removing the noise coefficients using the threshold function, the inverse wavelet transform is performed to reconstruct and output multi-source clean feature data. Unlike traditional single noise reduction methods with fixed thresholds, this invention uses sea state feedforward to participate in threshold adjustment, extracts the pattern of coordinated changes in wind and wave and monitoring noise, and achieves adaptive noise suppression based on operating conditions under both high and low sea states. Comparative analysis shows that in the field of marine signal processing, commonly used evaluation benchmarks include signal-to-noise ratio changes, the degree of fault feature retention, and real-time stability during sudden changes in sea state. When the wind speed reaches 25 m / s and the wave height reaches 1.5 m, if the fixed threshold method is still used, the early characteristics of cracks in the vibration channel are easily masked by background vibration. The dynamic threshold in this embodiment will change with... The adjustment is improved, making it more suitable for handling non-stationary noise under complex sea conditions. In summary, the embodiments of the present invention can achieve multi-source signal cleaning based on threshold adjustment involving sea conditions, avoiding excessive filtering of effective fault features, thereby improving the stability of noise reduction effect. Multi-source cleaning feature data is acquired, and the multi-source cleaning feature data is fused and analyzed according to the fault identification and localization logic to generate blade fault type results, fault severity results, and fault location information. Among them, the fault type result refers to the identification result of blade fault categories such as cracks and corrosion, the fault severity result refers to the quantitative result of the degree of fault development, and the fault location information refers to the spatial location of the fault on the blade. In practical implementation, multi-source cleaning feature data is input into a pre-trained one-dimensional and two-dimensional hybrid convolutional neural network, and the network outputs a fault type probability vector. and Fault Severity Index ,in, The range of values ​​is Its physical meaning is to reflect the degree of impact of the current fault on the safety of the blade's continued operation; when Or, when an expanding crack is identified, a maintenance and hoisting work order and a fault handling request are generated; when Furthermore, if no extended cracks are detected, the system will output the normal operating status and maintain daily monitoring. The hybrid convolutional neural network specifically includes a one-dimensional convolutional branch for independently extracting strain and acoustic temporal features, and a two-dimensional convolutional branch for extracting vibration and infrared temporal spectrogram spatial features. The high-dimensional feature tensors output by each branch are concatenated and fused by a fully connected layer and then input into the classifier for joint computation. When locating the fault, considering that strain reflects low-frequency macroscopic deformation and acoustic emission reflects high-frequency microscopic fracture, and that the two have different propagation characteristics, a layered location logic is adopted: based on the location of sensor nodes where strain data shows abnormal changes, the initial macroscopic region where the fault is located is determined; the arrival time difference of multiple acoustic sensors arranged around this region receiving the same acoustic pulse is extracted, and this time difference is combined with the acoustic wave velocity. Calculate fault space coordinates ; Unlike traditional rule-setting methods that rely on a single threshold or human experience, this invention uses fusion features and location information to participate in the judgment, extracting consistent features of different faults in multiple source channels, and realizing the effective fusion of multi-channel diagnostic information and the simultaneous classification and location. In summary, the embodiments of this invention can achieve the joint output of fault category, degree and location based on multi-source fusion diagnosis, providing complete data support for subsequent maintenance decisions. The system acquires blade fault type results, fault severity results, fault location information, maintenance and hoisting work orders, raw sea state data, target docking position and operation status feedback data. Based on the current status of the target equipment, it coordinates the control of external hoisting equipment, generates coordinated control commands, and dynamically corrects the commands based on the operation status feedback data. To correct equipment actions under fluctuating sea conditions, attitude compensation parameters can be generated based on the transport ship's attitude data and wind speed data, and these parameters can be added to the variable at the current operating moment. Below, based on the current joint angle vector of the crane's robotic arm. Calculate the inverse Jacobian matrix Combined with the target location With actual location Determined pose tracking error: Through proportional gain and differential gain Adjust the error components and simultaneously introduce a transformation matrix from the transport ship's body coordinate system to the crane's base coordinate system. Transformation matrix of coordinate system Vector of the transport ship's swaying speed Calculate attitude disturbances and incorporate wind load compensation terms: in, The air drag coefficient, air density, This refers to the wind-received projected area of ​​the equipment and blades. The wind direction is a unit vector; simultaneously, the admittance control coefficient matrix is ​​introduced. This is used to convert wind load physical forces into equivalent pose compensation velocities, generating compensation control quantities: Unlike traditional independent control methods that separate ship motion and wind load effects, this invention incorporates docking deviation, ship sway, and wind load effects into the control simultaneously through the same compensation relationship, thereby improving the control accuracy and response speed of the equipment in complex sea conditions. In the field of offshore lifting control, commonly used evaluation benchmarks include target docking deviation, available working window duration, and safety stability under critical sea state fluctuations. Under the conditions of Beaufort scale 5 winds and wave height of 1.5m, if the control system only issues fixed actions based on the target position, changes in the attitude of the transport ship will cause the lifting point to drift continuously. This embodiment introduces attitude compensation and wind load compensation, which can correct equipment movements before the target position changes, making it more suitable for offshore operations. In summary, this embodiment can achieve coordinated equipment control based on operational status feedback and sea state compensation, avoiding docking deviations exceeding preset safety tolerances due to target deviation accumulation, and improving the precision of offshore hoisting. The system acquires fault handling requests, hoisting control requests, and sea state anomaly requests. Based on pre-set safety priority rules and operation priority rules, it performs conflict resolution on multiple requests to determine the control ownership of the target equipment. Each unit sends the request via a standardized interactive interface after attaching a request identifier, target equipment identifier, priority identifier, and update time information when generating the request. The closed-loop control unit synchronously acquires each request through a standardized interactive interface and verifies the target device identifier, current control affiliation, and priority matching relationship according to the control affiliation table. When a sea state anomaly request meets the preset sea state restriction conditions, the priority of the sea state anomaly request is raised to the highest level, and the control affiliation table is updated. When both the fault handling request and the hoisting control request are valid and neither triggers an abnormal sea state, first compare their priority indicators, then compare their update time information if the priorities are the same, and switch control to the request with the later update time. Unlike the traditional method of fixed equipment occupation by a single control terminal, this invention uses a control attribution table and dual comparison rules to analyze the competition relationship between different requests on the same device, thereby achieving orderly adjudication and smooth switching of multiple command sources. In summary, the embodiments of this invention can achieve control adjudication based on request identifier, priority, and update time, avoiding external hoisting execution equipment from receiving conflicting commands simultaneously, and ensuring the safety of equipment collaborative operation. In this embodiment, the human-machine collaboration unit receives blade parameters, target docking position, and adjustment instructions for fault diagnosis parameters, hoisting operation parameters, and control rules input by the user, and interacts with each unit through a standardized interactive interface. The role of this part is to put on-site operation experience and system automatic results on the same business link, so that users can adjust diagnostic and operation parameters when sea conditions change or faults escalate, while obtaining diagnostic results, work orders, and execution status. Example

[0017] The system also includes a scheduling and management unit, which is used to manage each processing unit throughout its entire lifecycle, specifically including: When a unit is loaded, a unique identifier is assigned and its processing capabilities are registered to the service directory. Data types for sending and receiving are synchronized to complete the initialization. When execution is triggered, a task request is sent to the corresponding unit, the running state is switched, and the business data is notified to be distributed. When a pause or unload command is received, the corresponding flag unit status is checked, task distribution is stopped, or resources are released and the flag is deregistered. When the target unit is paused, abnormally switched or unloaded, the scheduling and management unit maintains the execution of the last valid collaborative control command by the external hoisting execution equipment, or maintains the last valid fault result, until the control ownership is reassigned or a new command is issued. The system adopts a ship-shore collaborative distributed architecture; multi-source data acquisition units are set on the wind turbine body, nacelle, transport ship or auxiliary platform side; adaptive noise reduction processing unit, fault diagnosis unit and closed-loop control unit are deployed on shipborne computing nodes, shore-based control center or edge computing nodes; human-machine collaboration unit is deployed on user client.

[0018] The system acquires loading requests from the adaptive noise reduction processing unit, fault diagnosis unit, and closed-loop control unit. Based on the unit's processing capabilities, it performs initialization management on each unit and generates registered unit identifiers and service catalog information. Among these, the unique identifier refers to the unique number of each processing unit in the system, used to distinguish different processing units; the processing capability metadata refers to the description of the processing content that the unit has. The service catalog refers to the collection of records of registered units and their corresponding capabilities in the system. In specific implementation, after a unit is loaded, the scheduling and management unit assigns it a unique identifier, registers the processing capability metadata to the service catalog, sets the unit status to idle, and synchronizes the data types it can receive and the result types it can output to the standardized interaction interface. In specific registration business logic and quantization dictionary instances, the system assigns an identifier number to the adaptive noise reduction processing unit, such as node_0x01, and encapsulates its capability element information into a standardized key-value pair structure; the scheduling management unit persists this key-value pair structure to the service directory to build a global addressing dictionary that supports fast matching; Unlike the traditional static approach of fixing the processing program in a single device, this invention establishes the input-output correspondence between different processing units through unified scheduling and directory registration, enabling manageable and stable invocation of each module throughout its entire lifecycle. In summary, the embodiments of this invention can initialize the processing unit based on the scheduling management unit, ensuring accurate routing and delivery of underlying data communication and scheduling. The system acquires raw blade status data, raw sea state data, or maintenance and hoisting work orders. Based on preset trigger conditions, it sends task requests to the target processing unit, determines the unit's operating status, and distributes the corresponding data. The preset trigger conditions refer to the conditions that are met to start the corresponding processing flow. The operating status refers to the state in which the target unit has started processing the current task. In specific implementation, the scheduling and management unit sends a task request to the corresponding unit, switches the unit's status to running, and notifies the standardized interaction interface to distribute the corresponding data. Unlike traditional methods that involve independent processing and manual switching, this invention synchronizes task request-triggered state switching and data distribution, analyzes the dependencies in the task chain, and ensures automatic sequential connection of the processing chain. In summary, the embodiments of this invention can achieve sequential operation of each processing unit based on unified task triggering, avoiding task delays and data waiting caused by improper connection. Obtain a pause command or unload command, and based on the current status of the target unit, pause or terminate the target unit, generating a task recovery notification and a control ownership update notification. The pause command is a control command that temporarily stops the target unit from receiving new tasks, the unload command is a control command that terminates the operation of the target unit and releases resources, and the control ownership update notification is used to update the information of the control source of the external hoisting execution equipment after the unit exits. In practice, upon receiving a pause command, the scheduling management unit marks the target unit's status as paused and stops distributing new tasks, while simultaneously sending a pause notification to the relevant units; upon receiving an unload command, the scheduling management unit deregisters the target unit's identifier, releases the occupied resources, switches the target unit's status to terminated, and sends a task recycling notification and a control ownership update notification to the standardized interactive interface. Unlike the traditional forced interruption method of directly stopping the program, this invention analyzes the dependencies between various links in the system operation by first updating the state, then stopping task distribution, and then notifying the relevant units of the processing order, so as to realize the smooth and orderly exit of the processing unit during the task. In summary, the embodiments of this invention can realize the suspension and unloading of units based on state-based management, avoiding data loss caused by the sudden interruption of processing tasks. When a target unit is received as paused, abnormally switched or unloaded, the last valid collaborative control command or the last valid fault result is maintained according to the current effective result until the scheduling management unit reallocates control ownership, restores the target unit's operating status or issues a new command. Among them, the last valid collaborative control command refers to the equipment control command that has been confirmed to be valid before the target unit exits, and the last valid fault result refers to the fault type result, fault severity result and fault location information that has been confirmed and output before the fault diagnosis unit exits. Unlike traditional equipment that directly loses reference or clears results when control is interrupted, this invention maintains the last valid result, thus meeting the continuous time requirements of hoisting operations and fault diagnosis, and ensuring a smooth transition of the control and diagnosis process during system scheduling. In offshore equipment control and fault monitoring scenarios, common risk conditions include transient switching of the control unit when the attitude of the transport ship fluctuates at high frequency, and the suspension of fault diagnosis tasks under high load. If the system clears the current control or diagnosis results at this time, the external hoisting equipment may lose the basis for action, and the on-site personnel will not be able to confirm the current fault status. This implementation maintains the last valid result, which can reduce the operational risks caused by such interruptions. In summary, the embodiments of the present invention can maintain the continuity of the process based on the last valid result, avoid abrupt changes in control commands and diagnostic results during unit switching, and improve the continuity of on-site operations. Obtain system deployment requirements, and based on the ship-shore collaborative distributed architecture, deploy and configure the multi-source data acquisition unit, adaptive noise reduction processing unit, fault diagnosis unit, closed-loop control unit, and human-machine collaboration unit, and determine the bidirectional transmission path of results and instructions; whereby, the ship-shore collaborative distributed architecture refers to an architecture in which some units are deployed on the offshore equipment side and some units are deployed on the shore-based control center or edge computing node side. In practice, the multi-source data acquisition unit can be set up on the wind turbine body, nacelle, transport ship or auxiliary platform to collect on-site data nearby; the adaptive noise reduction processing unit, fault diagnosis unit and closed-loop control unit can be deployed on shipborne computing nodes, shore-based control centers or edge computing nodes to allocate processing tasks according to computing power and communication conditions; the human-machine collaboration unit is deployed on the user client. Unlike traditional methods that concentrate all functions on a single device, this invention allocates deployment according to data collection location, computing power, and operational needs. This addresses the need for both local data collection and remote processing in maritime operations, balancing the system's communication burden and processing power to achieve distributed collaborative computing. In summary, embodiments of this invention can achieve a division of labor for data collection, processing, and control based on ship-shore collaborative deployment, improving the system's adaptability and response efficiency in complex maritime environments. Example

[0019] The fault diagnosis unit includes: The feature fusion module is used to receive multi-source clean feature data and generate fused feature data; The fault identification module is used to generate blade fault type and severity results based on fused feature data; The fault location module is used to generate fault location information based on strain and acoustic data; The work order generation module is used to generate a maintenance and hoisting work order containing the above information by combining the pre-acquired blade parameters when the fault status meets the preset conditions.

[0020] The closed-loop control unit includes: The strategy matching module is used to match the collaborative operation mode of external hoisting execution equipment based on work order information; The pose compensation module is used to generate pose compensation amounts based on the transport ship's attitude data and wind speed data. The control generation module is used to generate collaborative control commands based on the collaborative operation mode, pose compensation amount, and target docking position. The safety cutoff module is used to determine sea state limitations and trigger pause, lock, or follow-up hold commands; The standardized interactive interfaces include a data acquisition interface, a processing call interface, a device control interface, a work order and result interface, and a rule adjustment interface, which are used to match the targeted transmission of corresponding types of data, tasks, instructions, and parameters. The human-machine collaboration unit receives control commands containing target equipment identifiers and sea state threshold parameters through a visual interactive interface. After completing standardization verification and authorization verification, it transmits the commands to the corresponding processing unit through a standardized interactive interface to adjust the underlying control logic or operation strategy. It also supports restoring the default configuration through a reset command.

[0021] Acquire multi-source cleaning feature data, and according to the internal processing flow of the fault diagnosis unit, fuse, identify, locate and generate work orders for the multi-source cleaning feature data to form a complete set of results required for maintenance and hoisting; Among them, the feature fusion module is used to organize the clean features of vibration, acoustic, strain and infrared channels into a unified analysis input; the fault identification module is used to output the blade fault type and fault severity results; the fault location module is used to output the fault location information; and the work order generation module is used to integrate the fault conclusion and blade parameters into a maintenance and hoisting work order. Multi-source cleaning feature data is acquired, and the multi-source cleaning feature data is fused based on the complementary relationship of fault characterization in each channel to generate fused feature data. The fused feature data refers to the analysis data after combining feature information from different channels that points to the same fault. For example, periodic anomalies in the vibration channel, sudden pulses in the acoustic channel, local deformation in the strain channel, and temperature rise areas in the infrared channel all point to a certain crack fault. The specific data flow and feature alignment rules are as follows: For one-dimensional time-series data such as vibration, acoustics, and strain, a short-time Fourier transform is used to divide it into multiple fixed-length data windows, which are then mapped into a two-dimensional time-frequency spectrum. Based on the spatial resolution of infrared thermal imaging, the generated two-dimensional time-frequency spectrum is resampled and interpolated to ensure that the pixel size and resolution of each channel feature map remain consistent. The vibration time-frequency spectrum, acoustic time-frequency spectrum, strain time-frequency spectrum, and infrared image are then tensor-stitched according to the channel dimension to form fused feature data. Unlike the traditional method of processing the results of various sensors independently and then comparing them manually, this invention first forms a unified input through a feature fusion module, extracts the correspondence of the same fault on multiple channels, and realizes mutual support and effective supplementation of feature information from different sensors. In summary, the embodiments of this invention can realize the centralized expression of fault information based on feature fusion, avoid the dispersion of subsequent identification and positioning basis, and improve the robustness of multi-source data joint diagnosis. The process involves acquiring fused feature data, classifying and determining its severity based on a fault identification model, and generating blade fault type and severity results. The blade fault type result refers to the identified fault category, while the severity result refers to the quantified severity of the fault, such as a severity index. It can be between 0 and 1, and is used to indicate the level of impact of the current fault on the operational safety of the blade; In practice, a pre-trained one-dimensional and two-dimensional hybrid convolutional neural network can be used to identify the fused feature data and output a fault type probability vector. and Fault Severity Index ;when If the fault type result indicates an extended crack, it can be considered that the preset conditions are met; Unlike traditional low-resolution identification methods that rely solely on a single threshold to determine anomalies, this invention uses both classification and severity results to analyze the correlation between fault type and severity, enabling precise differentiation and joint determination of fault categories and safety impact levels. In summary, embodiments of this invention can achieve joint output of category and severity based on a fault identification model, providing comprehensive and reliable technical support for maintenance decisions. The process involves acquiring strain and acoustic data from the fused feature data, calculating the fault location based on a hierarchical localization mechanism, and generating fault location information. Specifically, strain data is analyzed to determine the macroscopic structural region where deformation characteristics exceed a set threshold. At least three acoustic channel signals from the surrounding area are extracted, and precise coordinates are calculated based on the relationship between time of arrival (TOA) and wave velocity. TOA refers to the time difference between the arrival of the same high-frequency fracture wave at different acoustic sensors, and wave velocity... The actual speed at which sound waves propagate in the blade material; In a specific quantitative simulation example, assuming the coordinates of the three activated acoustic sensors around the macroscopic region where the strain anomaly occurs are as follows: , and ;set up The moment the acoustic pulse is received is , for , for ;by As a reference benchmark, calculate and Time difference between Converted into distance difference ; Based on this, multiple sets of spatial distance difference constraint models are established, with the constraint condition that the difference in Euclidean distance from the target coordinates to each acoustic sensor position is equal to the corresponding time difference conversion distance; by simultaneously solving the system... Multiple distance difference equations are used as a reference, and Newton's iteration algorithm or grid search method is employed for three-dimensional spatial optimization iteration to obtain the optimal analytical solution. ; This inference rule based on strict geometric distance constraints transforms fuzzy area determination into definite spatial coordinate system calculations. Unlike the traditional method of roughly estimating the fault area based solely on the amplitude of a single channel, this invention uses both strain and acoustic channels to participate in the localization process, extracting the time difference pattern in the fault wave propagation path, thereby improving the accuracy of fault location coordinate calculations in three-dimensional space. Comparative analysis shows that in blade fault location scenarios, common evaluation criteria include location error and location stability. When sea state disturbances exceed the calibration threshold, relying solely on vibration amplitude distribution can easily misidentify near-end noise as a fault source. This embodiment uses the time difference of arrival of strain data and acoustic data, which can further eliminate the influence of background disturbances on the basis of multi-source cleaning feature data, making it more suitable for offshore blade positioning scenarios. In summary, the embodiments of the present invention can realize fault location information output based on time difference positioning, providing an accurate spatial coordinate basis for the generation of maintenance work orders. The system obtains the severity of the fault, the type of the blade fault, the location of the fault, and the blade parameters. Based on preset conditions, it integrates the maintenance requirements and generates a maintenance and hoisting work order that includes the location of the fault and the blade parameters. The blade parameters refer to the model parameters, blade length, or blade weight information input by the human-machine collaboration unit. Their function is to provide object constraints for subsequent hoisting strategy matching. Unlike the traditional method of generating diagnostic reports and hoisting work orders separately, this invention directly organizes diagnostic results into maintenance and hoisting work orders through a work order generation module, establishing a correspondence between fault locations and maintenance actions, and realizing direct data connection between diagnostic results and work execution stages. In summary, the embodiments of this invention can generate work orders based on fault results and blade parameters, avoiding manual information processing on-site and improving work preparation efficiency. Obtain fault location information and blade parameters from maintenance and hoisting work orders, and match the collaborative operation mode of the crane and transport vessel according to the operating capacity of the external hoisting equipment to generate a collaborative operation mode; where the collaborative operation mode refers to the coordination method between the crane and the transport vessel under the current operation task. Unlike the traditional approach of using uniform operating procedures for different fault locations, this invention uses a strategy matching module to directly input fault location information and blade parameters as operating mode inputs. It analyzes the impact of differences in maintenance objects on hoisting paths and action arrangements, and realizes the function of dynamically allocating operating methods according to differentiated task requirements. In summary, the embodiments of this invention can achieve collaborative operating mode matching based on work order content, ensuring that the hoisting strategy matches the actual maintenance location constraints and improving overall operating performance. Acquire the transport ship's attitude and wind speed data. Based on the impact of sea state disturbances on external lifting equipment, compensate for the equipment's movements to generate attitude compensation values. These attitude compensation values ​​are used to correct deviations in the crane and transport ship's movements, effectively counteracting target drift caused by ship sway and wind load. In practice, the aforementioned... and The calculation is incorporated to ensure that the attitude compensation amount simultaneously reflects changes in the transport ship's attitude and wind speed. Unlike traditional methods that only remedy deviations after they occur at the end of the equipment, this invention incorporates the effects of ship attitude and wind speed into the control in advance through a pose compensation module. It calculates the continuous disturbance of sea state changes to the target docking position, thereby achieving feedforward prediction and dynamic compensation of operational deviations. In summary, the embodiments of this invention can achieve pose compensation based on the transport ship's attitude data and wind speed data, avoiding pose divergence of external hoisting equipment under continuous disturbances and improving the tracking accuracy of the control. The system acquires the collaborative operation mode, pose compensation amount, and target docking position. Based on the current operation target, it generates collaborative control instructions for external hoisting equipment. The target docking position refers to the operation position that the hoisting equipment should reach under the current work order, and the collaborative control instructions refer to the set of execution instructions sent to equipment such as cranes and transport ships. Unlike traditional methods that separate the selection of operation mode and the generation of equipment actions, this invention uses a control generation module to take the collaborative operation mode, pose compensation amount, and target docking position as inputs, analyzes the coupling relationship between the operation intention and the on-site disturbance, and realizes a close coupling between the command generation process and the dynamic sea state fluctuations on-site. In summary, the embodiments of this invention can generate control commands based on the target position and compensation results, enhancing the stability of the collaborative actions of the execution equipment under dynamic sea conditions. The system acquires raw sea state data, determines whether the current operation can continue based on preset sea state constraints, and issues pause, lock, or follow-up commands when the conditions are met. The preset sea state constraints refer to the operational boundary conditions determined based on parameters such as wind speed and wave height. The pause command is used to temporarily stop active docking, the lock command is used to keep the equipment in a safe and fixed state, and the servo-hold command is used to maintain the equipment's ability to follow changes in sea state while stopping active docking; in practice, the dynamic change factor of sea state can continue to be utilized. Make a judgment when When a safety cutoff is triggered; At the same time, coordinated control commands are continuously issued; among them, The sea state change threshold value is pre-calibrated based on historical safe lifting operation data and the mechanical threshold of the crane. Its value range can be configured, for example, from 0.65 to 0.85. Unlike traditional methods that only manually terminate operations after the equipment exhibits obvious dangerous behavior, this invention uses a safety cut-off module to directly cut off the operation when the sea conditions reach the boundary conditions, extracting early warning features before the formation of dangerous conditions, and realizing proactive prevention and cut-off based on sea condition boundaries. In summary, the embodiments of this invention can achieve safe cut-off based on sea condition constraints, avoiding equipment operation in sea conditions that exceed design boundaries and reducing the safety risks of offshore hoisting. Acquire raw blade status data, raw sea state data, task requests, processing results, collaborative control commands, status feedback information, maintenance and hoisting work orders, and rule adjustment commands, and transmit them through standardized interaction interfaces depending on the interaction object; Unlike the traditional approach of using different temporary links for different data, this invention fixes the paths for acquisition, processing, control, and parameter tuning through interface division of labor, analyzes the differences in the flow of various types of data in the system, and realizes standardized routing and standardized interaction of various heterogeneous data. In summary, the embodiments of this invention can realize data and command transmission based on standardized interaction interfaces, enhancing interface compatibility and data transmission reliability when the system is expanded. The system acquires control commands input by the user and verifies, issues, and provides feedback on the control commands based on the target blade identifier, target equipment identifier, sea state threshold parameters, fault judgment parameters, and operation preference parameters. In practice, the human-machine collaboration unit receives control commands through a visual interactive interface and performs format standardization verification, permission scope verification, and object association check in sequence. After the verification is passed, the commands are sent to the adaptive noise reduction processing unit, fault diagnosis unit, or closed-loop control unit through a standardized interactive interface. Upon receiving the command, the corresponding processing unit updates the relevant parameter configuration, determines the subordinate relationship between the target object and the existing rules, and adjusts the diagnostic logic, hoisting strategy, or control priority. When a reset command is received, the target parameter configuration or control rules are restored to the default state, and the parameter update result, rule update result, or reset result is fed back. Unlike traditional methods of manually adjusting or distributing parameters on-site, this invention standardizes the adjustment objects, adjustment ranges, and feedback results through a human-machine collaboration unit and rule adjustment interface. It establishes a correspondence between the modification subject, the modification content, and the effective status, achieving traceability and closed-loop verification of the entire control rule adjustment process. In summary, the embodiments of this invention can realize parameter and rule adjustment based on a human-machine collaboration unit, avoiding malfunctions caused by parameter adjustments without strict authorization verification, and improving the standardization of system operation. In practical applications, multi-source data acquisition units are deployed on the wind turbine body, nacelle, transport vessel and auxiliary platform to continuously collect vibration, acoustic, strain and infrared data of the blades, as well as wind speed, wave height and transport vessel attitude data; the collected raw data of blade status and sea state are sent to the adaptive noise reduction processing unit through the data acquisition interface. The adaptive noise reduction processing unit adjusts according to the dynamic change factor of sea state. and dynamic threshold After noise reduction, the multi-source cleaning feature data is sent to the fault diagnosis unit; the fault diagnosis unit outputs the blade fault type result, fault severity result, and fault location information, and, if the conditions are met... Or, if the crack is identified as an extended crack, a repair and hoisting work order will be generated; The maintenance and hoisting work order, along with the target docking position input by the human-machine collaboration unit, is sent to the closed-loop control unit. The closed-loop control unit generates collaborative control commands based on the collaborative operation mode, posture compensation amount, and target docking position, and sends them to the external hoisting execution equipment via the equipment control interface. If the sea state reaches the preset sea state limit, the safety cutoff module will directly issue a pause, lock, or follow-up hold command; if a fault handling request, a hoisting control request, and a sea state anomaly request exist at the same time, the command will be issued after the priority identifier and update time information in the control attribution table are used for adjudication. During the application process, users can view diagnostic results, work orders, and execution status through the human-machine collaboration unit, and can also input sea state threshold parameters, fault judgment parameters, and operation preference parameters for adjustment; newly collected data continues to enter the data acquisition interface, new control commands continue to enter the rule adjustment interface, and new processing results and status feedback continue to be transmitted back to the user client, so that blade fault diagnosis, maintenance preparation, and hoisting operations can be carried out continuously within the same system.

[0022] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A fault diagnosis system for offshore wind turbine blades based on multi-source sensor fusion, characterized in that, The system includes: The multi-source data acquisition unit is used to collect raw data on the vibration, acoustics, strain and infrared status of the blades, raw data on the environmental wind speed, wave height and the attitude of the transport ship, and operational status feedback data of the external hoisting equipment. An adaptive noise reduction processing unit is used to extract the wind and wave features of the raw sea state data, combine them with a preset mapping relationship to reduce noise in the raw blade state data, and output multi-source clean feature data. The fault diagnosis unit is used to generate fault diagnosis results based on multi-source cleaning feature data, and generate maintenance and hoisting work orders and fault handling requests when preset conditions are met. The human-machine collaboration unit is used to receive user input commands, generate hoisting control requests, and output system status to the user; The closed-loop control unit is used to generate collaborative control commands to external hoisting execution equipment based on fault diagnosis results, the maintenance and hoisting work orders, raw sea state data, target docking position input by the human-machine collaboration unit, and operation status feedback data; and to calculate deviations and dynamically correct the collaborative control commands; generate sea state anomaly requests when sea state is abnormal; and adjudicate conflicts of multiple requests according to preset rules, and issue the collaborative control commands when sea state restrictions are met. A standardized interactive interface is used to transmit data and instructions.

2. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 1, characterized in that, The specific methods by which the closed-loop control unit adjudicates conflicts among multiple types of requests include: When generating a request, each unit attaches the target device identifier, priority identifier, and update time information, and synchronizes them to the standardized interaction interface; The closed-loop control unit acquires each request and compares it with a pre-built control attribution table; When a sea state anomaly request meets the preset sea state restrictions, its priority is raised to the highest level and control is updated. When both the fault handling request and the hoisting control request are valid and there is no abnormal sea condition, the priority identifiers and update time information of the two are compared in turn, and the control of the target equipment is switched to the request with higher priority or later update. The standardized interaction interface will send the updated request results for execution.

3. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 1, characterized in that, The system also includes a scheduling and management unit for managing each processing unit throughout its entire lifecycle, specifically including: When a unit is loaded, a unique identifier is assigned and its processing capabilities are registered to the service directory. Data types for sending and receiving are synchronized to complete the initialization. When execution is triggered, a task request is sent to the corresponding unit, the running state is switched, and the business data is notified to be distributed. When a pause or unload command is received, the corresponding flag unit status is checked, task distribution is stopped, or resources are released and the flag is deregistered.

4. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 3, characterized in that, When the target unit is paused, abnormally switched, or unloaded, the scheduling and management unit maintains the execution of the last valid collaborative control command by the external hoisting execution equipment, or maintains the last valid fault result, until the control ownership is reassigned or a new command is issued.

5. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 1, characterized in that, The system adopts a ship-shore collaborative distributed architecture; the multi-source data acquisition unit is set on the wind turbine body, nacelle, transport ship or auxiliary platform side; the adaptive noise reduction processing unit, fault diagnosis unit and closed-loop control unit are deployed on the shipborne computing node, shore-based control center or edge computing node; the human-machine collaboration unit is deployed on the user client.

6. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 1, characterized in that, The fault diagnosis unit includes: The feature fusion module is used to receive multi-source clean feature data and generate fused feature data; The fault identification module is used to generate blade fault type and severity results based on fused feature data; The fault location module is used to generate fault location information based on strain and acoustic data; The work order generation module is used to generate a maintenance and hoisting work order containing the above information by combining the pre-acquired blade parameters when the fault status meets the preset conditions.

7. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 1, characterized in that, The closed-loop control unit includes: The strategy matching module is used to match the collaborative operation mode of the external hoisting execution equipment based on the work order information; The pose compensation module is used to generate pose compensation amounts based on the transport ship's attitude data and wind speed data. The control generation module is used to generate collaborative control commands based on the collaborative operation mode, pose compensation amount, and target docking position. The safety cutoff module is used to determine sea state limitations and trigger pause, lock, or follow-up hold commands.

8. The offshore wind turbine blade fault diagnosis system based on multi-source sensor fusion according to claim 1, characterized in that, The standardized interactive interface includes a data acquisition interface, a processing call interface, a device control interface, a work order and result interface, and a rule adjustment interface, which are used to match the targeted transmission of corresponding types of data, tasks, instructions, and parameters.

9. A fault diagnosis system for offshore wind turbine blades based on multi-source sensor fusion as described in claim 1, characterized in that, The human-machine collaboration unit receives control commands containing target device identifiers and sea state threshold parameters through a visual interactive interface. After completing standardization verification and authorization verification, it transmits the commands to the corresponding processing unit through a standardized interactive interface to adjust the underlying control logic or operation strategy. It also supports restoring the default configuration through a reset command.

Images