A self-checking method and system for integrated ship bridge system
By constructing a self-testing model through multi-source signal processing and data-driven modeling, the problem of automatic anomaly detection in the integrated ship-bridge system was solved, enabling automatic fault location and early warning, thus ensuring stable system operation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA WATERBORNE TRANSPORT RES INST
- Filing Date
- 2025-06-26
- Publication Date
- 2026-05-22
AI Technical Summary
The existing integrated bridge system lacks automatic anomaly detection capabilities, resulting in untimely equipment fault detection and affecting the stable operation of the system.
By using multi-source signal processing, data acquisition, clustering, feature extraction, and data-driven modeling to build a self-checking model, automatic anomaly detection and early warning can be achieved.
It enables automatic fault location and early warning for the integrated bridge system, ensuring long-term stable operation of the system and maintaining the normal operation of other modules when a module malfunctions.
Smart Images
Figure CN120704288B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ship intelligent management technology, and more specifically, to a self-inspection method and system for an integrated bridge system. Background Technology
[0002] The integrated bridge system is an integrated system for ship navigation and control. Through interconnection, it centrally utilizes sensor information, commands, and controls from workstations to improve the safety and efficiency of ship management for operators. Built on the ship's bridge, the integrated bridge system is a highly information-based and automated integrated system that combines navigation, steering, radar collision avoidance, and navigation management. It utilizes modern technologies from multiple disciplines, including computers, automatic control, networks, and information fusion, to centrally acquire sensor information, centrally monitor information, sensitively detect and rapidly respond to events, and promptly issue various control commands to implement effective navigation and control. It serves as the ship's information center and command and control center.
[0003] The integrated bridge system can consist of one or more multi-functional workstations. Each workstation provides users with a multi-tasking interface comprised of ECDIS, radar, and conning, and these interfaces can be switched between each other. The primary function of ECDIS is to provide route planning and navigation monitoring based on electronic charts, AIS data, and sensor data. The primary function of radar is to provide collision avoidance based on radar echo data, AIS data, and sensor data. Conning primarily provides navigation information, navigation control information, central alarm management, and system status monitoring. Since ships contain a large number of mechanical and electronic devices, it is necessary to monitor and promptly eliminate faults in these devices to ensure their normal operation. Therefore, the system possesses a certain degree of self-checking and automatic repair capabilities. When an anomaly is detected, it can issue an early system alarm, alerting relevant personnel for manual intervention. This achieves statistical monitoring and early warning functions, ensuring the long-term stable operation of the entire system. Summary of the Invention
[0004] To address the aforementioned technical problems, this invention proposes a self-inspection method and system for an integrated bridge system. The aim is to achieve automatic anomaly detection of the integrated bridge system through multi-source signal processing, thereby enabling early warning of ship anomalies.
[0005] The first aspect of this invention provides a self-inspection method for an integrated bridge system, comprising the following steps:
[0006] Data on the bridge equipment in the integrated bridge system is collected by a data acquisition device, the bridge equipment data is matched with data functions, and the matched data is preprocessed.
[0007] The data of ship-bridge equipment with data function tags are clustered, and the integrated ship-bridge system is modularized according to the clustering results. Feature extraction is used to obtain the feature factors corresponding to different modules.
[0008] Extract normal operation data of the integrated bridge system, expand the dimensionality of the normal operation data, construct training data based on feature factors, construct self-testing models of different modules of the integrated bridge system based on the data-driven model, and use the training data to train the self-testing models.
[0009] Obtain actual ship-bridge equipment data of the integrated ship-bridge system, use data function labels to match the corresponding self-inspection model, construct an anomaly classifier based on binary logic in the self-inspection model, and output the self-inspection result by estimating the residual between the predicted ship-bridge equipment data and the actual ship-bridge equipment data.
[0010] In this solution, data from the bridge equipment in the integrated bridge system is collected by a data acquisition device. This data is then matched with data functions, and the matched data undergoes preprocessing, specifically:
[0011] The system uses preset sensors in the integrated bridge system as data collectors to collect multi-sensor signals from the integrated bridge system. The multi-sensor signals are evaluated based on the functional response of the relevant bridge equipment, and multi-sensor signals that meet the preset evaluation criteria are selected as bridge equipment data.
[0012] The selected bridge equipment data is matched with the data function, and the matched bridge equipment data is preprocessed by data noise reduction and data cleaning to remove any missing bridge equipment data.
[0013] Read historical alarm information from the integrated bridge system, extract the data types involved in the historical alarm information, and use the data types to perform data verification in the remaining bridge equipment data. If the remaining bridge equipment data does not contain the given data type, then expand the data according to the data type.
[0014] Once the data verification is successful, the preprocessed ship-bridge equipment data is divided according to the data function, generating ship-bridge equipment data with different data function labels.
[0015] In this solution, the data of bridge equipment with data function tags are clustered, and the integrated bridge system is modularized based on the clustering results, specifically as follows:
[0016] The K-means algorithm was used to cluster the bridge equipment data with data function labels. The K-means algorithm was then optimized using an optimized genetic algorithm. Initial cluster centers were selected, and Mahalanobis distance was used as the metric function to calculate the membership degree between the bridge equipment data sample points and the initial cluster centers.
[0017] The system presets soft clustering thresholds and hard clustering thresholds for the integrated ship-bridge system. When the membership degree between a ship-bridge equipment data sample point and the initial cluster center is not less than the hard clustering threshold, the ship-bridge equipment data sample point is assigned to the corresponding initial cluster center to generate the corresponding cluster.
[0018] When the membership degree between a bridge equipment data sample point and the initial cluster center is greater than or equal to the soft clustering threshold and less than the hard clustering threshold, the bridge equipment data sample point is assigned to all the initial cluster centers that meet the conditions. After all bridge equipment data sample points are assigned, the cluster centers are continuously updated during the iterative clustering process, and the clustering result of the last clustering is output.
[0019] Based on the output clustering results, obtain the corresponding ship-bridge equipment data subsets, and treat the ship-bridge equipment corresponding to each ship-bridge equipment data subset as a comprehensive ship-bridge system module.
[0020] In this scheme, feature extraction is used to obtain the feature factors corresponding to different modules, specifically:
[0021] Obtain subsets of ship bridge equipment data from different integrated ship bridge system modules, extract historical fault data of ship bridge equipment based on big data methods, and use Spearman correlation analysis to obtain ship bridge equipment data variables whose correlation degree meets the preset standard from the subsets of ship bridge equipment data.
[0022] A one-dimensional convolutional neural network is constructed as a feature extraction network. The data variables of the ship bridge equipment are imported into the one-dimensional convolutional neural network. The deep features of the data are extracted by using convolution operations. An attention mechanism is introduced in the fully connected layer to obtain the attention weights of the one-dimensional feature vector.
[0023] The attention weights are used to sort the one-dimensional feature vectors. A preset number of one-dimensional feature vectors are truncated according to the sorting results. If there are redundant features in the truncated one-dimensional feature vectors, the redundant features are removed and a corresponding number of one-dimensional feature vectors are selected from the sorting results. If there are no redundant vectors in the selected one-dimensional feature vectors, they are used as feature factors of the integrated ship-bridge system module.
[0024] In this scheme, normal operation data of the integrated ship-bridge system is extracted, the dimensionality of the normal operation data is expanded, and training data is constructed based on feature factors, specifically as follows:
[0025] Normal operation data is filtered from the data subsets of bridge equipment corresponding to different integrated bridge system modules, and the filtered normal operation data is expanded in dimension using compressed sensing algorithm to improve data quality.
[0026] The feature factors of different integrated bridge system modules are obtained as data indicators. The feature parameters are extracted from the expanded normal operation data using the data indicators. During the extraction process, noise is randomly added through additive injection to generate noise-enhanced feature parameters. The feature parameters corresponding to each data indicator and the noise-enhanced feature parameters are aggregated to generate a dataset. The dataset is divided into training data and test data based on a preset ratio.
[0027] In this solution, self-testing models for different modules of the integrated bridge system are constructed based on a data-driven model. The training data is used to train the self-testing models, specifically as follows:
[0028] A self-testing model for different integrated bridge system modules is constructed using an LSTM autoencoder network. The training data of different integrated bridge system modules is imported into the self-testing model for training. The memory unit and gating mechanism of the LSTM network are used to obtain the temporal features of the training data. Data encoding is performed based on the temporal features to achieve data reconstruction.
[0029] The training data reconstructed by the decoder module is decoded, and the decoded training data is mapped to the sample label space for comparison to obtain the reconstruction error. The reconstruction error guides the self-testing model to perform iterative training.
[0030] The self-test model is trained using the generalization ability of the test samples. The residual threshold of the anomaly classifier in the self-test model is output. When the classification performance of the self-test model reaches the preset standard, the self-test model of the current integrated bridge system module is output.
[0031] In this solution, actual bridge equipment data of the integrated bridge system is obtained, and a corresponding self-testing model is matched using data function labels. An anomaly classifier is constructed based on binary logic within the self-testing model. The self-testing result is output by estimating the residual between the equipment data and the actual equipment data. Specifically:
[0032] Obtain actual bridge and dock equipment data of the integrated bridge and dock system; determine the corresponding self-testing model based on the data function tags corresponding to the actual bridge and dock equipment data; after passing the actual bridge and dock equipment data through a feature extraction network, import the output of the feature extraction network into the self-testing model to obtain the estimated bridge and dock equipment data under normal operation.
[0033] The self-testing model obtains residual data between the estimated ship-bridge equipment data and the actual ship-bridge equipment data. In the anomaly classifier, the probability distribution of the integrated ship-bridge system at the current moment is obtained based on the residual data. The probability distribution at the current moment is compared with the probability distribution corresponding to the residual threshold to obtain the deviation value. When the deviation value is greater than the preset deviation threshold, it proves that there is an anomaly in the integrated ship-bridge system and an anomaly warning is generated.
[0034] The second aspect of the present invention provides a self-inspection system for an integrated bridge system, the system comprising a feature factor analysis unit, an integrated bridge system self-inspection unit, a bridge equipment data acquisition unit, and a self-inspection anomaly early warning unit;
[0035] The feature factor analysis unit uses the ship-bridge equipment data to perform clustering to divide the integrated ship-bridge system into modular parts, and uses feature extraction to obtain the feature factors corresponding to different modules.
[0036] The integrated bridge system self-inspection unit extracts the normal operation data of the integrated bridge system, constructs training data based on feature factors, constructs self-inspection models for different modules of the integrated bridge system based on the data-driven model, constructs an anomaly classifier based on binary logic in the self-inspection model, and trains it using the training data.
[0037] The bridge equipment data acquisition unit acquires the actual bridge equipment data of the integrated bridge system and performs data preprocessing.
[0038] The self-inspection anomaly early warning unit imports the preprocessed actual ship-bridge equipment data into the self-inspection model, outputs the self-inspection result by estimating the residual between the predicted ship-bridge equipment data and the actual ship-bridge equipment data, and generates a self-inspection anomaly early warning.
[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0040] This invention features an automatic fault diagnosis function for integrated bridge systems. It can automatically locate system faults and alert on-duty personnel and equipment maintenance staff via alarms. When an anomaly is detected through self-checking, a system alarm can be issued in advance, prompting relevant personnel to intervene manually. This achieves statistical monitoring and early warning functions, ensuring the long-term stable operation of the entire system.
[0041] The integrated bridge system's self-inspection system adopts a modular design. When one module of the integrated bridge system malfunctions, other modules can still operate normally. When a normally functioning module detects an anomaly in another related module, it monitors in real time whether the relevant module has recovered and, once it has recovered, restores the system to normal operation as soon as possible. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments or examples of the present invention, the accompanying drawings used in the embodiments or examples will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained according to these drawings without creative effort.
[0043] Figure 1 A flowchart of a self-inspection method for an integrated bridge system is shown.
[0044] Figure 2 The flowchart illustrates the process of obtaining feature factors corresponding to different modules using feature filtering.
[0045] Figure 3 The flowchart illustrates the process of obtaining self-inspection results for an integrated ship-bridge system using a self-inspection model.
[0046] Figure 4 A block diagram of a self-testing system for an integrated ship bridge system is shown. Detailed Implementation
[0047] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.
[0048] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.
[0049] like Figure 1 As shown, the first embodiment of the present invention provides a self-testing method for an integrated bridge system, comprising:
[0050] S102, Collect bridge equipment data in the integrated bridge system through a data acquisition device, match the bridge equipment data with data functions, and perform data preprocessing on the matched data;
[0051] S104, cluster the bridge equipment data with data function tags, divide the integrated bridge system into modules based on the clustering results, and use feature extraction to obtain the feature factors corresponding to different modules.
[0052] S106, extract normal operation data of the integrated bridge system, expand the dimensionality of the normal operation data, construct training data based on feature factors, construct self-testing models of different modules of the integrated bridge system based on the data-driven model, and train the self-testing models using the training data.
[0053] S108: Obtain the actual ship-bridge equipment data of the integrated ship-bridge system, match the corresponding self-inspection model using data function labels, construct an anomaly classifier based on binary logic in the self-inspection model, and output the self-inspection result by estimating the residual between the predicted ship-bridge equipment data and the actual ship-bridge equipment data.
[0054] It should be noted that the system utilizes preset sensors within the integrated bridge system as data collectors to acquire multi-sensor signals. These signals are evaluated based on the functional response of the relevant bridge equipment. Signals meeting preset evaluation criteria are selected as bridge equipment data. Preferably, the evaluation is based on the response of the data function, selecting successfully responding sensor signals as bridge equipment data. For example, successful generation of measurement displays during onboard equipment monitoring indicates a successful response. The selected bridge equipment data is matched with the data function. The matched data undergoes data noise reduction and data cleaning preprocessing to remove missing data. Historical alarm information from the integrated bridge system is read, and the data types involved in these alarms are extracted. This data type is then used to verify the remaining bridge equipment data. If the remaining data does not contain a given data type, data expansion is performed based on that data type. After successful data verification, the preprocessed bridge equipment data is divided according to the data function, generating bridge equipment data with different data function tags.
[0055] It should be noted that the K-means algorithm is used to cluster the bridge equipment data with data function labels. An optimized genetic algorithm is then used to further refine the K-means algorithm. The genetic algorithm is further improved, and a chaotic algorithm is used to create initial population chaos, making the initial population distribution more uniform. Based on the cluster silhouette coefficient as the fitness function, bridge equipment data samples with high fitness values are selected for replication during the iteration of the genetic algorithm. The search direction for the initial cluster centers is obtained, and the search is performed in the direction of high importance to select the initial cluster centers. Mahalanobis distance is used as the metric function to calculate the membership degree between the bridge equipment data sample points and the initial cluster centers. Preset soft clustering thresholds and hard clustering thresholds corresponding to the integrated bridge system are also used. When the membership degree between a bridge equipment data sample point and the initial cluster center is not less than the hard clustering threshold, the bridge equipment data sample point is assigned to the corresponding initial cluster center to generate the corresponding cluster. When the membership degree between a bridge equipment data sample point and the initial cluster center is greater than or equal to the soft clustering threshold and less than the hard clustering threshold, the bridge equipment data sample point is assigned to all initial cluster centers that meet the conditions. After all bridge equipment data sample points are assigned, the cluster centers are continuously updated during the iterative clustering process, and the clustering result of the last clustering is output. Based on the output clustering result, the corresponding bridge equipment data subset is obtained, and the bridge equipment corresponding to each bridge equipment data subset is regarded as a comprehensive bridge system module.
[0056] According to an embodiment of the present invention, feature factors corresponding to different modules are obtained by feature extraction, specifically as follows:
[0057] S202, obtain a subset of ship bridge equipment data from different integrated ship bridge system modules, extract historical fault data of ship bridge equipment based on big data methods, and use Spearman correlation analysis to obtain ship bridge equipment data variables in the subset of ship bridge equipment data whose correlation degree meets the preset standard;
[0058] S204, construct a one-dimensional convolutional neural network as a feature extraction network, import the data variables of the ship bridge equipment into the one-dimensional convolutional neural network, use convolution operations to extract deep features of the data, and introduce an attention mechanism in the fully connected layer to obtain the attention weights of the one-dimensional feature vector.
[0059] S206, the one-dimensional feature vectors are sorted using the attention weights, and a preset number of one-dimensional feature vectors are truncated according to the sorting results. If there are redundant features in the truncated one-dimensional feature vectors, the redundant features are removed and a corresponding number of one-dimensional feature vectors are selected from the sorting results. If there are no redundant vectors in the selected one-dimensional feature vectors, they are used as feature factors of the integrated ship-bridge system module.
[0060] It should be noted that the feature extraction network consists of an input layer, a convolutional module, and a fully connected layer. The convolutional module includes one convolutional layer, one pooling layer, and one activation layer. The pooling layer uses max pooling to filter features, and the activation layer uses the ReLU activation function to increase the network's non-linearity. After extracting deep features from the data through the convolutional module, the fully connected layer performs attention-weighted operations on the input features, selecting feature factors for the integrated bridge system module.
[0061] Normal operation data is selected from the data subsets of bridge equipment corresponding to different integrated bridge system modules. The selected normal operation data is then expanded using a compressed sensing algorithm. Compressed sensing can recover and reconstruct the original signal from a small amount of normal operation data while retaining all the features of the original signal. Dimension expansion improves data quality, makes it easier to extract signal features, and enhances self-testing accuracy. Feature factors of different integrated bridge system modules are obtained as data indicators. These indicators are used to extract feature parameters from the expanded normal operation data. During the extraction process, noise is randomly added through additive injection to generate noise-enhanced feature parameters. The feature parameters corresponding to each data indicator and the noise-enhanced feature parameters are aggregated to generate a dataset. This dataset is then divided into training and test data based on a preset ratio. The noise injection method enhances the predictive performance of the subsequent self-testing network.
[0062] It should be noted that using LSTM autoencoder networks to construct self-testing models for different integrated bridge system modules, preferably using multi-layer LSTM autoencoders, can better capture, learn, and mine the features contained in the data, obtaining effective data encoding. The LSTM autoencoder network includes an LSTM encoder module and an LSTM decoder module. Training data from different integrated bridge system modules is imported into the self-testing model for training. The memory units and gating mechanism of the LSTM network are used to obtain the temporal features of the training data, and data encoding is performed based on these temporal features to achieve data reconstruction. The reconstructed training data is then decoded by the decoder module, mapped to the sample label space for comparison, and the reconstruction error is obtained. The reconstruction error guides the iterative training of the self-testing model. The generalization ability of the self-testing model is used to train the model, and the residual threshold of the anomaly classifier in the self-testing model is output. When the classification performance of the self-testing model reaches a preset standard, the self-testing model of the current integrated bridge system module is output.
[0063] Figure 3 A flowchart is shown to obtain the self-inspection results of the integrated ship-bridge system using a self-inspection model.
[0064] According to an embodiment of the present invention, actual bridge equipment data of the integrated bridge system is obtained, and a corresponding self-testing model is matched using data function labels. An anomaly classifier is constructed based on binary logic within the self-testing model. The self-testing result is output by estimating the residual between the equipment data and the actual equipment data. Specifically:
[0065] S302, Obtain actual bridge equipment data of the integrated bridge system, determine the corresponding self-test model based on the data function label corresponding to the actual bridge equipment data, pass the actual bridge equipment data through the feature extraction network, import the output of the feature extraction network into the self-test model, and obtain the estimated bridge equipment data under normal operation.
[0066] S304, obtain the residual data between the estimated bridge equipment data and the actual bridge equipment data through the self-testing model, and obtain the probability distribution of the integrated bridge system at the current moment in the anomaly classifier based on the residual data;
[0067] S306, compare the probability distribution at the current moment with the probability distribution corresponding to the residual threshold to obtain the deviation value. When the deviation value is greater than the preset deviation threshold, it proves that there is an anomaly in the integrated ship-bridge system and generates an anomaly warning.
[0068] It should be noted that, based on the bridge equipment data of the online integrated bridge system, the operating status of the integrated bridge system is monitored in real time to achieve automatic detection and accurate diagnosis of system faults. Normally, when the integrated bridge system is in normal operation, the deviation of the residual probability distribution is within a relatively stable and small fluctuation range. When the integrated bridge system is in an abnormal operating state, the deviation of the residual probability distribution gradually increases and fluctuates drastically, gradually deviating from the stable value under normal operating conditions. By monitoring the changing trend of the residual probability distribution deviation in real time, the deterioration state of the integrated bridge system can be automatically detected. Based on the alarm thresholds set in the trained self-testing model and the bridge equipment data of the integrated bridge system, real-time monitoring is performed to promptly and effectively detect abnormal changes in the online integrated bridge system and issue early warning information, achieving early fault warning for the online integrated bridge system.
[0069] The second aspect of the present invention provides a self-inspection system for an integrated bridge system, the system comprising a feature factor analysis unit, an integrated bridge system self-inspection unit, a bridge equipment data acquisition unit, and a self-inspection anomaly early warning unit;
[0070] The feature factor analysis unit uses the ship-bridge equipment data to perform clustering to divide the integrated ship-bridge system into modular parts, and uses feature extraction to obtain the feature factors corresponding to different modules.
[0071] The integrated bridge system self-inspection unit extracts the normal operation data of the integrated bridge system, constructs training data based on feature factors, constructs self-inspection models for different modules of the integrated bridge system based on the data-driven model, constructs an anomaly classifier based on binary logic in the self-inspection model, and trains it using the training data.
[0072] The bridge equipment data acquisition unit acquires the actual bridge equipment data of the integrated bridge system and performs data preprocessing.
[0073] The self-inspection anomaly early warning unit imports the preprocessed actual ship-bridge equipment data into the self-inspection model, outputs the self-inspection result by estimating the residual between the predicted ship-bridge equipment data and the actual ship-bridge equipment data, and generates a self-inspection anomaly early warning.
[0074] The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined, integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling, direct coupling, or communication connection between the various components shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0075] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the various embodiments of the present invention, all functional units can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in a combination of hardware and software functional units.
[0076] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0077] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.
Claims
1. A self-inspection method for an integrated ship-bridge system, characterized in that, Includes the following steps: Data on the bridge equipment in the integrated bridge system is collected by a data acquisition device, the bridge equipment data is matched with data functions, and the matched data is preprocessed. The data of the bridge equipment with data function tags are clustered. Based on the clustering results, the integrated bridge system is divided into several integrated bridge system modules. The optimized genetic algorithm is used to optimize the K-means algorithm to select the initial cluster center. The Mahalanobis distance and preset soft and hard clustering thresholds are used to determine the cluster affiliation of the data sample points. For the segmented integrated bridge system modules, a one-dimensional convolutional neural network combined with an attention mechanism is used to extract feature factors representing the operating status of the module from the corresponding bridge equipment data subset. Based on the feature factors of each integrated bridge system module, normal operation data of the integrated bridge system is extracted and expanded in dimension to construct training data for different modules. An LSTM autoencoder network is used to construct self-testing models for different modules of the integrated bridge system, and the training data is used to train the self-testing models. Obtain actual ship-bridge equipment data of the integrated ship-bridge system, use data function labels to match the corresponding self-inspection model, construct an anomaly classifier based on binary logic in the self-inspection model, and output the self-inspection result by estimating the residual between the ship-bridge equipment data and the actual ship-bridge equipment data. The feature factors corresponding to different modules are obtained as follows: Obtain subsets of ship bridge equipment data from different integrated ship bridge system modules, extract historical fault data of ship bridge equipment based on big data methods, and use Spearman correlation analysis to obtain ship bridge equipment data variables whose correlation degree meets the preset standard from the subsets of ship bridge equipment data. A one-dimensional convolutional neural network is constructed as a feature extraction network. The data variables of the ship bridge equipment are imported into the one-dimensional convolutional neural network. The deep features of the data are extracted by using convolution operations. An attention mechanism is introduced in the fully connected layer to obtain the attention weights of the one-dimensional feature vector. The attention weights are used to sort the one-dimensional feature vectors. A preset number of one-dimensional feature vectors are truncated according to the sorting results. If there are redundant features in the truncated one-dimensional feature vectors, the redundant features are removed and a corresponding number of one-dimensional feature vectors are selected from the sorting results. If there are no redundant vectors in the selected one-dimensional feature vectors, they are used as feature factors of the integrated ship-bridge system module. Construct self-testing models for different modules of the integrated bridge system, and train the self-testing models using the training data, specifically as follows: A self-testing model for different integrated bridge system modules is constructed using an LSTM autoencoder network. The training data of different integrated bridge system modules is imported into the self-testing model for training. The memory unit and gating mechanism of the LSTM network are used to obtain the temporal features of the training data. Data encoding is performed based on the temporal features to achieve data reconstruction. The reconstructed training data is decoded by the decoder module, and the decoded training data is mapped to the sample label space for comparison to obtain the reconstruction error. The reconstruction error guides the self-testing model to perform iterative training. The generalization ability of the self-testing model is trained using test samples, and the residual threshold of the anomaly classifier in the self-testing model is output. When the classification performance of the self-testing model reaches the preset standard, the self-testing model of the current integrated bridge system module is output.
2. The self-inspection method for an integrated ship-bridge system according to claim 1, characterized in that, Data from the bridge equipment in the integrated bridge system is collected via a data acquisition device. This data is then matched with relevant data functions, and the matched data undergoes preprocessing, specifically: The system uses preset sensors in the integrated bridge system as data collectors to collect multi-sensor signals from the integrated bridge system. The multi-sensor signals are evaluated based on the functional response of the relevant bridge equipment, and multi-sensor signals that meet the preset evaluation criteria are selected as bridge equipment data. The selected bridge equipment data is matched with the data function, and the matched bridge equipment data is preprocessed by data noise reduction and data cleaning to remove any missing bridge equipment data. Read historical alarm information from the integrated bridge system, extract the data types involved in the historical alarm information, and use the data types to perform data verification in the remaining bridge equipment data. If the remaining bridge equipment data does not contain the given data type, then expand the data according to the data type. Once the data verification is successful, the preprocessed ship-bridge equipment data is divided according to the data function, generating ship-bridge equipment data with different data function labels.
3. The self-inspection method for an integrated ship-bridge system according to claim 1, characterized in that, The data on bridge equipment with data function tags are clustered, and the integrated bridge system is modularized based on the clustering results, specifically as follows: The K-means algorithm is used to cluster the bridge equipment data with data function labels. When the membership degree between the bridge equipment data sample point and the initial cluster center is not less than the hard clustering threshold, the bridge equipment data sample point is assigned to the corresponding initial cluster center to generate the corresponding cluster. When the membership degree between a bridge equipment data sample point and the initial cluster center is greater than or equal to the soft clustering threshold and less than the hard clustering threshold, the bridge equipment data sample point is assigned to all the initial cluster centers that meet the conditions. After all bridge equipment data sample points are assigned, the cluster centers are continuously updated during the iterative clustering process, and the clustering result of the last clustering is output. Based on the output clustering results, obtain the corresponding ship-bridge equipment data subsets, and treat the ship-bridge equipment corresponding to each ship-bridge equipment data subset as a comprehensive ship-bridge system module.
4. The self-inspection method for an integrated ship-bridge system according to claim 1, characterized in that, Extract normal operation data from the integrated bridge system, expand the dimensionality of the normal operation data, and construct training data based on feature factors, specifically as follows: Normal operation data is filtered from the data subsets of bridge equipment corresponding to different integrated bridge system modules, and the filtered normal operation data is expanded in dimension using compressed sensing algorithm to improve data quality. The feature factors of different integrated bridge system modules are obtained as data indicators. The feature parameters are extracted from the expanded normal operation data using the data indicators. During the extraction process, noise is randomly added through additive injection to generate noise-enhanced feature parameters. The feature parameters corresponding to each data indicator and the noise-enhanced feature parameters are aggregated to generate a dataset. The dataset is divided into training data and test data based on a preset ratio.
5. The self-inspection method for an integrated ship-bridge system according to claim 1, characterized in that, Obtain actual bridge equipment data from the integrated bridge system, use data function labels to match the corresponding self-testing model, construct an anomaly classifier based on binary logic within the self-testing model, and output the self-testing result by estimating the residual between the equipment data and the actual equipment data. Specifically: Obtain actual bridge and dock equipment data of the integrated bridge and dock system; determine the corresponding self-testing model based on the data function tags corresponding to the actual bridge and dock equipment data; after passing the actual bridge and dock equipment data through a feature extraction network, import the output of the feature extraction network into the self-testing model to obtain the estimated bridge and dock equipment data under normal operation. The self-testing model obtains residual data between the estimated ship-bridge equipment data and the actual ship-bridge equipment data. In the anomaly classifier, the probability distribution of the integrated ship-bridge system at the current moment is obtained based on the residual data. The probability distribution at the current moment is compared with the probability distribution corresponding to the residual threshold to obtain the deviation value. When the deviation value is greater than the preset deviation threshold, it proves that there is an anomaly in the integrated ship-bridge system and an anomaly warning is generated.
6. A self-inspection system for an integrated bridge system, characterized in that, A self-inspection method for an integrated ship-bridge system as described in any one of claims 1-5 is provided, the system comprising a feature factor analysis unit, an integrated ship-bridge system self-inspection unit, a ship-bridge equipment data acquisition unit, and a self-inspection anomaly early warning unit. The feature factor analysis unit uses the ship-bridge equipment data to perform clustering to divide the integrated ship-bridge system into modular parts, and uses feature extraction to obtain the feature factors corresponding to different modules. The integrated bridge system self-inspection unit extracts the normal operation data of the integrated bridge system, constructs training data based on feature factors, constructs self-inspection models for different modules of the integrated bridge system based on the data-driven model, constructs an anomaly classifier based on binary logic in the self-inspection model, and trains it using the training data. The bridge equipment data acquisition unit acquires the actual bridge equipment data of the integrated bridge system and performs data preprocessing. The self-inspection anomaly early warning unit imports the preprocessed actual ship-bridge equipment data into the self-inspection model, outputs the self-inspection result by estimating the residual between the predicted ship-bridge equipment data and the actual ship-bridge equipment data, and generates a self-inspection anomaly early warning.