AI-fused cloud platform ship intelligent management method and system

Abstract

The invention discloses an AI-fused cloud platform ship intelligent management method and system, belongs to the technical field of ship management, and aims to solve the problems of data islanding, weak real-time decision-making ability and lack of adaptive optimization of a traditional system. The system adopts a data acquisition layer, an edge intelligent layer, a cloud decision-making layer and an application interaction layer architecture. The data acquisition layer acquires ship navigation, equipment, energy and environment data through a sensor; the edge intelligent layer preprocesses the data, diagnoses faults through threshold detection and lightweight AI, and transmits the data in a grading manner; the cloud decision-making layer generates an optimal plan by means of deep learning, reinforcement learning and digital twinning; the interaction layer is applied to visualize information, and operation information and actual effect data are fed back to optimize the AI model. According to the invention, data intercommunication of the whole ship is realized, decision real-time performance and comprehensive operation efficiency are improved, and communication and operation and maintenance costs are reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ship management, and particularly relates to a cloud platform ship intelligent management method and system fusing AI. BACKGROUND

[0002] Traditional ship intelligent management systems have realized a certain degree of function integration and application expansion at present. With the aid of various sensors and communication technologies, the system can collect and transmit the basic navigation parameters of the ship position, speed, heading, etc. in real time, and provide the ship operator with real-time dynamic information of the ship to assist him / her in making navigation decisions. In terms of equipment management, some systems can monitor the running state of key equipment of the ship, such as engine, generator, etc., and issue an alarm when the equipment parameters are abnormal, to a certain extent, to ensure the safe operation of the equipment. The cargo transportation management function has also been developed, which can monitor and record the cargo loading and unloading process and the cargo storage state, and improve the standardization of cargo transportation management. Moreover, the system can also realize the collection and simple analysis of ship fuel consumption data, and provide data support for ship energy saving management. However, such traditional systems have many problems to be solved. From the perspective of system integration, since the various subsystems of the ship are often provided by different manufacturers, the communication protocols and data formats lack unified standards, which makes it difficult for data to flow smoothly between subsystems, forming an information island, and greatly limiting the comprehensive analysis and collaborative management of the overall operation of the ship. In terms of data quality, the ship operating environment is complex, and the sensors are easily affected by vibration, electromagnetic interference, etc., resulting in problems such as noise, drift and even loss of collected data, which reduces the accuracy and reliability of the data, and the accuracy of analysis and decision-making based on such data is greatly reduced. In terms of network communication, the communication signal is restricted by factors such as ocean environment and geographical location when the ship is sailing, the satellite communication bandwidth is limited and the signal is unstable, and the ground communication network coverage is insufficient, resulting in frequent data transmission delays and interruptions, affecting the timeliness and effectiveness of real-time monitoring and remote control. Moreover, the traditional system relies on pre-set rules for decision-making, and lacks the self-adaptive ability to complex and variable ship operating conditions and environment, and is difficult to dynamically optimize the ship operating strategy according to real-time conditions, which has obvious shortcomings in improving the efficiency of ship operation and energy saving and emission reduction.

[0003] Meanwhile, the traditional ship intelligent management system has significant shortcomings in the application of AI technology, which limits its management efficiency and intelligent level. In terms of data processing, the system lacks deep analysis capability of AI algorithms and can only perform simple filtering and alarm on sensor data through preset thresholds, which cannot mine potential correlations between data and is difficult to discover early abnormal features of equipment or predict failure trends. At the decision-making level, the traditional system relies on manual experience or fixed rules for energy scheduling, route planning and other operations, and lacks self-adaptive ability. It cannot dynamically optimize strategies like AI algorithms in the face of complex and variable sea conditions, weather and shipping demand, resulting in fuel waste and low sailing efficiency. In addition, the system lacks the self-learning and iterative mechanism of AI and cannot continuously improve the management strategy according to the changes of ship operating conditions and new data accumulation, making it difficult to meet the increasingly upgraded energy efficiency, safety and environmental protection requirements of the shipping industry.

[0004] Therefore, how to provide a ship intelligent management method and system of an AI integrated cloud platform to realize efficient, accurate and sustainable optimization of ship management is a problem that needs to be solved by those skilled in the art. SUMMARY

[0005] Therefore, the present application provides a ship intelligent management method and system of an AI integrated cloud platform to solve the above problems. The technical scheme adopted by the present application is as follows: In a first aspect, the present application provides a ship intelligent management method of an AI integrated cloud platform, comprising the following steps: S1, receiving ship operating physical signals and converting the physical signals into ship operating data; encoding the ship operating data and adding time stamps and equipment identifiers, and then transmitting the ship operating data to an edge computing unit; S2, the edge computing unit receives the ship operating data, pre-processes the ship operating data to obtain real-time monitoring data, executes fault diagnosis on the real-time monitoring data, and outputs fault diagnosis results; encrypts the real-time monitoring data and fault diagnosis results, and then transmits them to a cloud decision layer; S3, the cloud decision layer receives the encrypted real-time monitoring data and fault diagnosis results, analyzes and fuses them to generate a full fleet operation database; executes AI analysis on the full fleet operation database, and outputs fault prediction results, energy scheduling strategies and alternative route planning schemes; simulates the energy scheduling strategies and alternative route planning schemes through a digital twin model, and selects and outputs the optimal plan; S4, the application interaction layer receives the real-time monitoring data output by the edge computing unit and the optimal plan output by the cloud decision layer, and converts them into visual information; converts the operation instructions of the operation personnel combined with the optimal plan into operation information, and feeds back the actual effect data after the execution of the optimal plan to the cloud decision layer.

[0006] Preferably, in step S1, the ship operation data includes navigation state data, equipment parameter data, energy consumption data and environmental data.

[0007] Preferably, in step S2, the real-time fault diagnosis includes: inputting the real-time monitoring data, outputting explicit emergency fault or no fault through threshold detection; if the no fault is outputted, analyzing the real-time monitoring data through a machine learning model to output an implicit early fault diagnosis result; the machine learning model receives the real-time monitoring data, outputs data space features and time sequence features, and generates the fault diagnosis result containing fault location, type and severity level after fusion.

[0008] Preferably, in step S2, it further includes a transmission strategy according to priority, the priority specifically includes fault emergency priority and data real-time priority, the fault emergency priority is sorted according to the emergency type of the fault diagnosis result, the data real-time priority is sorted according to the monitoring frequency of the real-time monitoring data, and the transmission strategy of encrypting the real-time monitoring data and fault diagnosis result and transmitting them to the cloud decision layer is determined based on the fault emergency priority and / or the data real-time priority.

[0009] Preferably, the fault diagnosis result located in the first priority order of the fault emergency degree and / or the real-time monitoring data located in the first priority order of the data real-time priority is transmitted by using exclusive bandwidth and retransmission mechanism; the fault diagnosis result located in the second priority order of the fault emergency degree and / or the real-time monitoring data located in the second priority order of the data real-time priority is transmitted after batch compression; the fault diagnosis result located in the third priority order of the fault emergency degree and / or the real-time monitoring data located in the third priority order of the data real-time priority is transmitted with delay; wherein the priority orders of the first priority, the second priority and the third priority are arranged in order from front to back.

[0010] Preferably, in step S3, the AI analysis includes: receiving the whole fleet operation database, outputting the fault prediction result of potential fault probability and influence range through a deep learning algorithm; outputting the optimized energy scheduling strategy through a reinforcement learning algorithm; outputting a plurality of the alternative route planning schemes in combination with historical navigation records and real-time weather data; the digital twin model receives the energy scheduling strategy and the alternative route planning scheme, outputs energy consumption, equipment load and risk assessment, and selects the optimal plan based on the risk assessment.

[0011] Preferably, in step S4, the visualization information is generated by the real-time monitoring data and the optimal plan conversion, including charts, maps, three-dimensional models; the operation information includes touch screen operation, voice instruction or physical key control, which is input into the cloud decision layer together with the actual effect data to drive the AI algorithm model parameter optimization.

[0012] In another aspect, the present application also discloses a cloud platform ship intelligent management system integrating AI, which is used to realize the above-mentioned cloud platform ship intelligent management method integrating AI, and comprises a data acquisition layer, an edge intelligent layer, a cloud decision layer and an application interaction layer which are coupled in sequence. The data acquisition layer is located on the ship and is used to receive ship running physical signals and convert them into the ship running data, and then output them to the edge intelligent layer. The edge intelligent layer is located on the ship and is used to receive the ship running data, output real-time monitoring data and fault diagnosis results to the cloud decision layer, and receive the control instructions output by the cloud decision layer. The cloud decision layer is used to receive the real-time monitoring data and the fault diagnosis results, output the control instructions to the edge intelligent layer, and receive the operation information and the actual effect data output by the application interaction layer. The application interaction layer is used to receive the real-time monitoring data output by the edge intelligent layer and the optimal plan output by the cloud decision layer, convert them into visualization information, and convert the operation instructions of the operator into the operation information, and then output them to the cloud decision layer together with the actual effect data.

[0013] Preferably, the data acquisition layer comprises: A navigation state sensor is used to receive ship position, heading, speed physical signals. A device state sensor is used to receive engine, generator, pump vibration, temperature, pressure, speed physical signals, and output device parameter data. An energy consumption sensor is used to receive fuel flow, power consumption, and output energy consumption data. An environmental parameter sensor is used to receive cargo hold temperature and humidity, atmospheric parameters, and output environmental data.

[0014] Preferably, the edge intelligent layer comprises: A data preprocessing module is used to receive the ship running data and output real-time monitoring data. A fault diagnosis module is used to receive the real-time monitoring data and output explicit emergency fault or implicit early fault diagnosis results. A data classification transmission module is used to receive the real-time monitoring data and the fault diagnosis results, divide them according to priority and then output them after encryption. A safety protection module is configured to receive the explicit emergency fault, output a local audible and visual alarm, and trigger a safety protection operation.

[0015] Compared with the prior art, the beneficial effects of the present application include: 1. Solving the data island problem: by adapting to the general data exchange protocol and API standard, integrating the heterogeneous subsystem data of the ship, breaking through the data barriers between the modules of the traditional system, and realizing the interconnection and intercommunication of the whole ship data.

[0016] 2. Improving real-time decision-making ability: the edge-cloud collaborative architecture classifies data processing, and the edge layer quickly completes local fault diagnosis and emergency response, which greatly shortens the key decision-making time compared with the traditional delay mode relying on the cloud.

[0017] 3. Realizing multi-target dynamic optimization: using mature AI algorithms to optimize the ship energy consumption, equipment health, navigation efficiency and other multi-dimensional data, changing the limitations of traditional single-target management system, and significantly improving the comprehensive operation efficiency.

[0018] 4. Enhancing the adaptive ability of the system: the man-machine collaborative feedback mechanism combined with the machine learning framework enables the system to iteratively adapt to the operation habits of the crew and changes in the shipping environment, overcoming the shortcoming of the lack of intelligent evolution of the traditional system.

[0019] 5. Reducing communication and operation and maintenance costs: the edge layer preprocessing reduces the amount of satellite transmission data, and the cloud unified management of multi-ship data avoids the high communication costs and scattered operation and maintenance problems caused by insufficient bandwidth of the traditional system. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.

[0021] Figure 1 A flow chart of a ship intelligent management method based on an AI-fused cloud platform is provided for the embodiments of the present application. Figure 2 A system architecture diagram of a ship intelligent management system based on an AI-fused cloud platform is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0022] 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, and 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.

[0023] In one embodiment, such as Figure 1 As shown, the present invention includes the following steps: S1. Receive ship operation physical signals, convert the physical signals into ship operation data; encode the ship operation data, add timestamps and device identifiers, and then transmit it to the edge computing unit; S2. The edge computing unit receives the ship operation data, preprocesses the ship operation data to obtain real-time monitoring data, performs fault diagnosis on the real-time monitoring data, outputs fault diagnosis results, and encrypts the real-time monitoring data and fault diagnosis results before transmitting them to the cloud decision layer. S3. The cloud-based decision-making layer receives the encrypted real-time monitoring data and fault diagnosis results, parses and integrates them to generate a full fleet operation database; performs AI analysis on the full fleet operation database, and outputs fault prediction results, energy scheduling strategies and alternative route planning schemes; simulates the energy scheduling strategies and alternative route planning schemes through a digital twin model, and selects and outputs the optimal plan. S4. The application interaction layer receives the real-time monitoring data output by the edge computing unit and the optimal plan output by the cloud decision layer, and converts them into visual information; it converts the operator's operation instructions based on the optimal plan into operation information, and feeds it back to the cloud decision layer along with the actual effect data after the optimal plan is executed.

[0024] In one embodiment, the navigation status data corresponds to the ship's latitude and longitude, heading angle, speed, and roll and pitch angles; the equipment parameter data corresponds to the vibration, temperature, pressure, and speed of the engine, generator, and pump; the energy consumption data corresponds to fuel flow and electrical energy consumption; and the environmental data corresponds to cargo hold temperature and humidity and atmospheric parameters. For analog signals such as vibration and temperature, the sensor's built-in analog-to-digital converter converts them into digital signals; while digital signals such as speed and switching signals are directly encoded and transmitted. Before transmission, all data undergoes preliminary encapsulation, adding metadata such as timestamps and device IDs. Subsequently, industrial communication protocols such as Modbus RTU and CANopen are used to transmit the data to the edge computing unit at a pre-set sampling frequency. To ensure the accuracy and integrity of data transmission, mechanisms such as CRC checksums are also used to verify the data during transmission.

[0025] In one embodiment, in step S2, the real-time fault diagnosis includes: The collected data is cleaned, denoised and feature extracted, effectively eliminating outliers and interference signals. The real-time monitoring data is output through threshold detection to output explicit emergency faults or no faults. If the output is no fault, the real-time monitoring data is analyzed by a machine learning model to output an implicit early fault diagnosis result. The machine learning model receives the real-time monitoring data, outputs data space features and time sequence features, and generates the fault diagnosis result containing fault location, type, and severity level after fusion. Specifically: After data access, filtering is first performed to remove white noise and periodic interference, and then data compression algorithm is used to reduce data transmission bandwidth requirement. The execution steps of the machine learning model are as follows: first, receive the timestamp and device ID standardized data transmitted by the data collection layer, extract the spatial features of device vibration, temperature and other data through CNN, and analyze the time sequence features of the data through LSTM, input the feature data into a pre-set classifier, and output the device health level, such as normal / slight abnormal / severe abnormal; based on the health level, if it is determined to be abnormal, associate with a pre-set rule base to determine whether to trigger a local emergency response. In the fault diagnosis process, a threshold detection method based on statistical analysis and a machine learning algorithm are combined to identify the device operation mode. For the same input data sequence, the execution logic of "threshold detection first, then machine learning algorithm" is adopted: first, the threshold detection method based on statistical analysis is used to complete preliminary abnormal screening in a short time, such as engine temperature ≥ 180℃, vibration acceleration ≥ 5g, to quickly identify "explicit emergency fault"; then, the machine learning algorithm analyzes the data with potential abnormalities that do not exceed the threshold, such as engine temperature fluctuating in the range of 160-170℃ for a long time and vibration acceleration abnormally fluctuating in the range of 2-3g, to identify implicit early faults by mining data space and time sequence features. A progressive diagnosis process of rapid safety backup and deep hidden danger identification is formed In one embodiment, in step S2, a transmission strategy according to priority is also included, the priority specifically includes fault emergency priority and data real-time priority, the fault emergency priority is sorted according to the emergency fault type of the fault diagnosis result, the data real-time priority is sorted according to the monitoring frequency of the real-time monitoring data, and the transmission strategy of encrypting the real-time monitoring data and the fault diagnosis result and transmitting them to the cloud decision layer is determined based on the fault emergency priority and / or the data real-time priority.

[0026] The fault diagnosis result located in the first priority order of the fault emergency degree and / or the real-time monitoring data located in the first priority order of the data real-time priority is transmitted by using exclusive bandwidth and retransmission mechanism; The fault diagnosis result in the second priority order of the fault emergency degree and / or the real-time monitoring data in the second priority order of the data real-time priority is transmitted after batch compression; The fault diagnosis result in the third priority order of the fault emergency degree and / or the real-time monitoring data in the third priority order of the data real-time priority is transmitted with delay; The priority orders of the first priority, the second priority and the third priority are arranged in sequence from front to back.

[0027] Specifically, in the data tag predefinition stage, fixed tags are bound to each sensor data, and the importance is divided into three levels of very high, medium and low according to the degree of influence on ship safety; the real-time is divided into three levels of millisecond, minute and hour according to the decision response time. At the same time, the priority is supported to be corrected according to the equipment state during operation. If a device triggers a slight abnormality, the importance and real-time level of the subsequent data of the device are temporarily increased by one level, and automatically fall back after the device returns to normal.

[0028] For different working condition data, transmission strategy and encryption protocol are automatically selected. For very high importance and millisecond level real-time data, such as emergency fault alarm, exclusive bandwidth and retransmission mechanism are used to preferentially occupy the communication bandwidth, and if the transmission fails, it will be immediately retransmitted. For medium importance and minute level real-time data, such as device health report, batch compression transmission is used, and data is packaged once every 5 minutes. For low importance and hour level real-time data, such as historical energy consumption statistics, delay transmission is used, and the signal is temporarily stored locally when the signal is weak, and uploaded after the signal is restored. At the same time, sensitive data such as fault diagnosis results and operation instructions are encrypted using TLS1.3 protocol; for non-sensitive data such as environmental temperature and humidity, AES-128 lightweight encryption is used. The edge computing unit has a built-in data tag-strategy / protocol mapping table, which automatically matches and executes after receiving data.

[0029] The edge computing unit divides the data into emergency instructions, real-time monitoring, historical records and other categories based on the above mechanism, and uploads them to the cloud decision layer using different transmission strategies and encryption protocols.

[0030] In one embodiment, in step S3, the AI deep analysis includes: The full fleet operation database is received, and the fault prediction result of the potential fault probability and the influence range is output by a deep learning algorithm; the optimized energy scheduling strategy is output by a reinforcement learning algorithm; combined with historical navigation records and real-time weather data, a plurality of alternative route planning schemes are output; the digital twin technology receives the energy scheduling strategy and the alternative route planning scheme, and outputs energy consumption, device load and risk assessment, and selects the optimal plan based on the risk assessment. Specifically: In the data access stage, three types of specific data uploaded by the edge computing unit are received, one type is filtered and denoised standardized sensor data containing device ID, timestamp and parameter value, such as "engine compartment / 1 / engine / 202405201000 / vibration acceleration / 2.3g" and "cargo hold-2 / cargo hold / 202405201000 / temperature / 25℃"; the second type is local analysis result data, that is, fault diagnosis results containing fault location, type and severity level and data quality reports containing abnormal data proportion and sensor state, such as "cargo hold temperature and humidity sensor-abnormal data proportion 3%-working normally" and "deck wind speed sensor-abnormal data proportion 80%-recommended for maintenance"; the third type is local execution record data, that is, emergency operation records executed by the edge computing unit, such as "202405201005-1-generator-triggered load reduction operation-execution success". After data access, first, format conversion and standardization processing are performed to eliminate the heterogeneity of different ship data; then, real-time online analysis is performed on the data by a stream computing engine, for example, real-time ranking and abnormal detection are performed on the energy consumption data of the entire fleet. The AI model training module continuously optimizes the performance of fault prediction and energy management models based on historical data by comprehensively using supervised learning, unsupervised learning and reinforcement learning methods. The digital twin model synchronizes ship sensor data in real time, drives the dynamic update of the virtual model, and simulates the implementation effect of different plans to evaluate potential risks and benefits. Finally, the optimal plan that has passed safety verification is selected and issued by the cloud decision layer to the edge computing unit for execution.

[0031] In one embodiment, in step S4, the visualization information is generated by converting the real-time monitoring data and the optimal plan, including charts, maps and three-dimensional models; the operation information includes touch screen operation, voice instruction or physical key control, and is input into the cloud decision layer together with the actual effect data to drive AI algorithm model parameter optimization.

[0032] In addition, as Figure 2 shown, the application also discloses an AI-fused cloud platform ship intelligent management system for realizing the AI-fused cloud platform ship intelligent management method, which comprises a data acquisition layer, an edge intelligent layer, a cloud decision layer and an application interaction layer which are coupled in sequence: The data acquisition layer is located on the ship and is used for receiving ship operation physical signals and converting them into the ship operation data, and outputting the ship operation data to the edge intelligent layer; The edge intelligent layer is located on the ship and is used for receiving the ship operation data, outputting real-time monitoring data and fault diagnosis results to the cloud decision layer, and receiving control instructions output by the cloud decision layer; The cloud decision layer is configured to receive the real-time monitoring data and the fault diagnosis result, output the control instruction to the edge intelligent layer, and receive operation information and actual effect data output by the application interaction layer; The application interaction layer is configured to receive the real-time monitoring data output by the edge intelligent layer and the optimal plan output by the cloud decision layer, convert the real-time monitoring data and the optimal plan into visual information, convert an operation instruction of an operator into the operation information, and output the operation information and the actual effect data to the cloud decision layer.

[0033] In an embodiment, the data acquisition layer comprises: a navigation state sensor configured to receive physical signals of a ship position, a heading, and a speed; a device state sensor configured to receive physical signals of vibration, temperature, pressure, and rotation speed of an engine, a generator, and a pump, and output device parameter data; an energy consumption sensor configured to receive physical signals of fuel flow and power consumption, and output energy consumption data; an environmental parameter sensor configured to receive physical signals of cargo hold temperature and humidity and atmospheric parameters, and output environmental data. The data acquisition layer is distributed in key areas of the ship, including an engine room, a cargo hold, a power distribution room, a deck, and a bridge, and serves as a basic link of system data sensing. The data acquisition layer is responsible for real-time acquisition of all-factor information of ship operation. In terms of navigation state monitoring, the layer is responsible for collecting data such as latitude, longitude, heading angle, speed, and roll and pitch angles, to provide necessary basic information for subsequent route planning and navigation safety guarantee. In the field of device health management, the layer continuously monitors parameters such as vibration, temperature, pressure, and rotation speed of core devices such as an engine, a generator, and a pump, to capture subtle changes in the device operation process in a timely manner. Energy consumption data acquisition focuses on indicators such as fuel flow and power consumption, to provide data basis for ship energy efficiency optimization. In addition, the layer also performs real-time monitoring on cargo hold temperature and humidity, atmospheric environmental parameters, and the like, to ensure cargo storage safety and improve ship environmental adaptability.

[0034] In an embodiment, the edge intelligent layer comprises: a data preprocessing module configured to receive the ship operation data and output real-time monitoring data; a fault diagnosis module configured to receive the real-time monitoring data and output explicit emergency fault or implicit early fault diagnosis results; a data classification transmission module configured to receive the real-time monitoring data and the fault diagnosis results, divide and encrypt the real-time monitoring data and the fault diagnosis results according to priorities, and output the real-time monitoring data and the fault diagnosis results; a safety protection module configured to receive the explicit emergency fault, output a local audible and visual alarm, and trigger a safety protection operation.

[0035] Specifically, the edge intelligence layer is installed in the central control room of the ship, has perfect electromagnetic shielding and heat dissipation design, serves as the intelligent processing core of the ship, and undertakes key tasks such as data preprocessing, real-time fault diagnosis and emergency response in the system. The overall model architecture includes preset rules and machine learning models. The preset rules are used to process deterministic emergency scenarios such as device parameter exceeding safety threshold and key sensor offline, and quickly trigger emergency response. The lightweight CNN and simplified LSTM machine learning models are used to process non-deterministic potential abnormal scenarios such as slight changes in vibration signals and abnormal energy consumption trends, and to mine data correlation features to identify early fault signs. The two form a complementary mechanism for emergency response and hidden danger identification, and can perform rule verification and model analysis on the same data of the same data acquisition layer. In the data preprocessing stage, the layer cleans, denoises and extracts features from the raw data transmitted by the data acquisition layer, effectively eliminates abnormal values and interference signals, and significantly improves data quality. In the real-time fault diagnosis link, the preset rules and existing machine learning models are used to cooperatively evaluate the device operating state, so that early fault hidden dangers can be identified in a timely manner. Once an emergency fault is detected, the edge intelligence layer will immediately trigger the local sound and light alarm device, and automatically execute safety protection measures such as shutdown and power cutoff according to the preset logic, to ensure the safety of ship equipment and personnel. In addition, the layer also serves as the communication hub between the ship and the cloud, responsible for data caching, priority division and encrypted transmission, to ensure that critical information is uploaded to the cloud decision layer in priority.

[0036] Further, the cloud decision layer is deployed in a professional data center, supporting private cloud, public cloud or hybrid cloud and other deployment modes. As the core processing hub of the system, the cloud decision layer undertakes important tasks such as multi-source data fusion analysis, AI model training, global strategy optimization and digital twin simulation. By integrating massive ship operation data, the layer builds a comprehensive ship operation state knowledge base; by using deep learning, reinforcement learning and other artificial intelligence algorithms, the layer deeply mines the potential laws behind the data, and then realizes intelligent decision-making such as equipment fault prediction, energy efficiency optimization and route planning; the digital twin module builds a highly realistic digital image of the ship in the virtual space based on the ship physical model and real-time data, for strategy rehearsal and risk assessment; in addition, the cloud platform also has multi-ship collaborative management function, supporting unified scheduling and resource optimization of the fleet, to improve overall operation efficiency.

[0037] In one embodiment, the application interaction layer includes: a visual display module for receiving the real-time monitoring data and the optimal plan and outputting visual information; a human-computer interaction module for receiving operation instructions of an operator and outputting operation information; a feedback module for receiving the actual effect data and outputting the operation information to the cloud decision layer after fusion.

[0038] Specifically, the application interaction layer is distributed in the ship control room operation terminal, the shore-based management center workstation, and the mobile terminal device, which serves as the key interface for human-computer interaction, providing a visual operation interface and decision support tool for the crew and management personnel. On the ship side, this layer provides real-time equipment monitoring, alarm processing, local strategy adjustment, and other functions to help the crew master the ship's running state in a timely manner and perform related operations; the shore-based management center focuses on macro management at the fleet level, supporting functions such as energy consumption statistical analysis, carbon footprint tracking, and maintenance plan development, to facilitate overall operational decision-making by management personnel; the mobile terminal device provides convenient remote monitoring and emergency command capabilities, enabling management personnel to view key ship data, receive alarm information, and issue orders at any time and from any location. In addition, the interaction layer also has an AI decision suggestion display and manual intervention interface, enabling human-machine collaborative decision-making and improving the scientific nature and flexibility of decision-making.

[0039] The overall process is described in detail as follows: The sensor network distributed at key positions on the ship collects navigation state, equipment parameters, energy consumption, and environmental data at a preset frequency. The collected data, after preliminary encoding and time stamping, is transmitted to the edge computing unit through the industrial bus; the data acquisition layer inputs raw sensor data to the edge intelligent layer and receives sensor anomaly reports from the edge intelligent layer, and adjusts the sampling frequency according to the reports.

[0040] Further, the edge computing unit receives raw data and performs noise filtering, outlier removal, data compression, and other preprocessing operations in sequence; in the fault diagnosis link, a progressive logic of threshold detection followed by machine learning algorithm is adopted, and a ship working condition adaptation layer is added to the machine learning model, which corrects the raw sensor data through real-time sea state and environmental temperature data for special scenarios such as high sea state and low temperature, improving the accuracy of diagnosis. For emergency faults such as equipment parameters exceeding the safety threshold, local alarms are triggered immediately and preset protection measures are executed. The preprocessed data, local fault diagnosis results, and execution records are classified and packaged, and transmitted to the cloud decision layer through 5G / satellite link, with priority given to critical data. At the same time, the edge layer receives model parameter update instructions from the cloud decision layer to optimize the local diagnosis model; the edge layer relies on the full data from the acquisition layer to ensure diagnostic coverage, and the acquisition layer relies on feedback from the edge layer to adjust the data acquisition strategy.

[0041] Further, the cloud platform receives data uploaded from multiple ships, performs format unification and data fusion, and constructs a full-fleet operation database. Existing AI algorithms such as deep learning and reinforcement learning are used for global analysis, including: Equipment fault prediction: predicting potential fault probability and impact within the next 72 hours; Energy efficiency optimization: dynamically adjust ship energy distribution strategy based on real-time water conditions and shipping demand; Route planning: generate multiple alternative optimal economic routes by combining weather data and historical navigation records.

[0042] Verify the feasibility of each alternative strategy in a virtual environment through digital twin technology, assess potential risks, and select the optimal plan. The verified control instructions and optimization suggestions are issued to the edge computing unit of the corresponding ship. The cloud decision layer inputs global fault reports and optimal plans to the application interaction layer, while receiving feedback from the application interaction layer about manual adjustments to optimize the AI model reward function. The cloud layer relies on the local diagnostic results of the edge layer to narrow the analysis range, while the edge layer relies on the strategy instructions from the cloud layer to perform optimization operations.

[0043] Further, the ship end operation terminal displays the decision-making suggestions issued by the cloud in real time. The crew can choose to automatically execute or manually adjust the strategy, and the adjustment information is synchronized to the cloud. The system continuously collects actual effect data after the execution of the strategy, which is used to optimize the AI model parameters, forming a closed-loop iteration of "data-decision-feedback".

[0044] The above provides a detailed description of the AI-integrated cloud platform ship intelligent management method and system. The specific examples in this embodiment illustrate the principles and implementation methods of the present application. The above examples are used to help understand the method and its core idea; at the same time, for those skilled in the art, according to the idea of the present application, the specific implementation method and application range will be changed; in summary, the content of the specification should not be understood as a limitation of the present application.

[0045] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined in this embodiment can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application should not be limited to the embodiments shown in this embodiment, but should conform to the widest range consistent with the principles and novel features disclosed in this embodiment.

Claims

1. An AI-fused cloud platform ship intelligent management method, characterized in that, The method comprises the following steps: S1, receiving a ship operation physical signal, converting the physical signal into ship operation data, encoding the ship operation data, adding a time stamp and a device identifier, and then transmitting the ship operation data to an edge computing unit; S2, the edge computing unit receives the ship operation data, pre-processes the ship operation data to obtain real-time monitoring data, and outputs a fault diagnosis result; The real-time monitoring data and the fault diagnosis result are encrypted and transmitted to a cloud decision layer; S3, the cloud decision layer receives the encrypted real-time monitoring data and fault diagnosis result, analyzes and fuses them to generate a full fleet operation database, performs AI analysis on the full fleet operation database, and outputs a fault prediction result, an energy scheduling strategy, and an alternative route planning scheme; the energy scheduling strategy and the alternative route planning scheme are simulated by a digital twin model, and the optimal plan is selected and output; S4, the application interaction layer receives the real-time monitoring data output by the edge computing unit and the optimal plan output by the cloud decision layer, converts them into visual information, and converts the operation instructions of the operation personnel combined with the optimal plan into operation information, which is fed back to the cloud decision layer together with the actual effect data after the execution of the optimal plan.

2. The AI-fused cloud platform ship intelligent management method according to claim 1, characterized in that, In step S1, the ship operation data includes navigation state data, device parameter data, energy consumption data, and environmental data.

3. The AI-fused cloud platform ship intelligent management method of claim 1, wherein, In step S2, the real-time fault diagnosis includes: Input the real-time monitoring data, output the explicit emergency fault or no fault through threshold detection; if no fault is output, analyze the real-time monitoring data through a machine learning model to output an implicit early fault diagnosis result; the machine learning model receives the real-time monitoring data, outputs data space features and time sequence features, and fuses them to generate the fault diagnosis result containing fault location, type, and severity level.

4. The AI-fused cloud platform ship intelligent management method of claim 1, wherein, In step S2, it also includes a transmission strategy according to priority, which specifically includes fault emergency priority and data real-time priority. The fault emergency priority is sorted according to the emergency fault type of the fault diagnosis result, the data real-time priority is sorted according to the monitoring frequency of the real-time monitoring data, and the transmission strategy of encrypting and transmitting the real-time monitoring data and the fault diagnosis result to the cloud decision layer is determined based on the fault emergency priority and / or the data real-time priority.

5. The AI-fused cloud platform ship intelligent management method according to claim 4, characterized in that, The specific steps are as follows: The fault diagnosis result in the first priority order of the fault emergency degree and / or the real-time monitoring data in the first priority order of the data real-time priority are transmitted using exclusive bandwidth and retransmission mechanism; The fault diagnosis result in the second priority order of the fault emergency degree and / or the real-time monitoring data in the second priority order of the data real-time priority are transmitted after batch compression; The fault diagnosis result in the third priority order of the fault emergency degree and / or the real-time monitoring data in the third priority order of the data real-time priority are transmitted with delay. The priority order of the first priority, the second priority and the third priority is arranged from front to back.

6. The AI-fused cloud platform ship intelligent management method according to claim 1, characterized in that, In step S3, the AI analysis includes: The full fleet operation database is received, the fault prediction result of the potential fault probability and the influence range is output by a deep learning algorithm, the optimized energy scheduling strategy is output by a reinforcement learning algorithm, a plurality of the alternative route planning schemes are output in combination with historical navigation records and real-time weather data, the digital twin model receives the energy scheduling strategy and the alternative route planning scheme, and energy consumption, equipment load and risk assessment are output, and the optimal plan is selected based on the risk assessment.

7. The AI-fused cloud platform ship intelligent management method of claim 1, wherein, In step S4, the visualization information is generated by conversion of the real-time monitoring data and the optimal plan, including charts, maps and three-dimensional models, and the operation information includes touch screen operation, voice instruction or physical key control, and the actual effect data are jointly input into the cloud decision layer to drive AI algorithm model parameter optimization.

8. An AI-fused cloud platform ship intelligent management system, characterized in that, The cloud platform ship intelligent management method for realizing the fusion AI of any one of claims 1-7 comprises a data acquisition layer, an edge intelligent layer, a cloud decision layer and an application interaction layer coupled in sequence: The data acquisition layer is located on the ship and is used for receiving ship operation physical signals, converting the ship operation physical signals into ship operation data, and outputting the ship operation data to the edge intelligent layer; The edge intelligent layer is located on the ship and is used for receiving the ship operation data, outputting real-time monitoring data and fault diagnosis results to the cloud decision layer, and receiving control instructions output by the cloud decision layer; The cloud decision layer is used for receiving the real-time monitoring data and the fault diagnosis results, outputting the control instructions to the edge intelligent layer, and receiving operation information and actual effect data output by the application interaction layer; The application interaction layer is used for receiving real-time monitoring data output by the edge intelligent layer and optimal plans output by the cloud decision layer, converting the real-time monitoring data and the optimal plans into visualization information, and converting operation instructions of an operator into the operation information, and the operation information and the actual effect data are cooperatively output to the cloud decision layer.

9. The AI-fused cloud platform ship intelligent management system according to claim 8, characterized in that, The data acquisition layer comprises: A navigation state sensor is used for receiving ship position, heading and speed physical signals; A device state sensor is used for receiving vibration, temperature, pressure and rotation speed physical signals of an engine, a generator and a pump, and outputting device parameter data; An energy consumption sensor is used for receiving fuel flow and power consumption, and outputting energy consumption data; An environment parameter sensor is used for receiving cargo hold temperature and humidity and atmospheric parameters, and outputting environment data.

10. The AI-fused cloud platform ship intelligent management system according to claim 8, characterized in that, The edge intelligent layer comprises: A data preprocessing module is used for receiving the ship operation data and outputting real-time monitoring data; A fault diagnosis module is used for receiving the real-time monitoring data and outputting explicit emergency fault or implicit early fault diagnosis results; A data classification transmission module is used for receiving the real-time monitoring data and the fault diagnosis results, dividing and encrypting according to priority, and outputting; A safety protection module is used for receiving the explicit emergency fault, outputting local sound and light alarms, and triggering safety protection operations.

Images