Maintenance Training System with Intelligent Fault Diagnosis Function and Its Method

Abstract

The present application discloses a maintenance training system with an intelligent fault diagnosis function and its method. By collecting the event text descriptions distributed by the maintenance training system to multiple users and the event data generated by these multiple users, and introducing data processing and semantic analysis algorithms at the backend to perform semantic analysis and correlation comparison of these data, concurrent conflicts are detected in this way. In this way, concurrent conflicts in collaborative virtual maintenance can be automatically detected, and a fault warning prompt can be issued in a timely manner, improving the quality and safety of maintenance training.

Description

Technical Field

[0001] This application relates to the technical field of intelligent maintenance training, and particularly to a maintenance training system with a fault intelligent diagnosis function and its method. Background Art

[0002] Collaborative virtual maintenance is a method that uses virtual reality technology to simulate real maintenance scenarios, enabling multiple users to share the same virtual world in a network environment for collaborative maintenance training. Collaborative virtual maintenance has the advantages of high efficiency, safety, economy, etc., and can be applied to fields such as the maintenance training, fault diagnosis, and fault troubleshooting of complex equipment. However, collaborative virtual maintenance also faces some challenges, one of which is how to handle concurrent conflicts. Concurrent conflicts refer to the inconsistencies and conflicts that may occur when multiple users operate on the same or different objects in the same virtual environment, such as operation conflicts, state conflicts, resource conflicts, etc. Concurrent conflicts will affect the effect and user experience of collaborative virtual maintenance, and may even lead to the collapse of the virtual environment or the failure of the system.

[0003] Therefore, a maintenance training system with a fault intelligent diagnosis function is desired. Summary of the Invention

[0004] This application provides a maintenance training system with a fault intelligent diagnosis function and its method. By collecting the event text descriptions distributed by the maintenance training system for multiple users and the event data generated by these multiple users, and introducing data processing and semantic analysis algorithms at the backend to perform semantic analysis and correlation comparison on these data, concurrent conflicts can be detected. In this way, concurrent conflicts in collaborative virtual maintenance can be automatically detected, and a fault warning prompt can be issued in a timely manner, improving the quality and safety of maintenance training.

[0005] This application also provides a maintenance training system with a fault intelligent diagnosis function, which includes:

[0006] An event text description collection module, configured to obtain a set of event text descriptions distributed by the maintenance training system for multiple users;

[0007] An event data collection module, configured to obtain a set of event data generated by the multiple users;

[0008] An event text description semantic feature analysis module, configured to perform context semantic association analysis based on word granularity on each event text description in the set of event text descriptions to obtain a sequence of event text description semantic feature vectors;

[0009] An event data semantic encoding module, configured to perform semantic encoding on each event data in the set of event data generated by the multiple users respectively to obtain a sequence of event data semantic encoding feature vectors;

[0010] An event semantic difference measurement module, configured to calculate the event semantic difference measurement coefficients between each pair of corresponding event text description semantic feature vectors and event data semantic encoding feature vectors in the sequence of event text description semantic feature vectors and the sequence of event data semantic encoding feature vectors respectively to obtain an event semantic difference measurement vector composed of multiple said event semantic difference measurement coefficients;

[0011] An event semantic difference time series feature encoding module, configured to perform feature extraction on the event semantic difference measurement vector through an event semantic difference time series correlation feature extractor based on a deep neural network model to obtain event semantic difference time series features;

[0012] A conflict detection module, configured to determine whether there is a concurrent conflict based on the event semantic difference time series features;

[0013] An early warning module, configured to issue a fault early warning prompt in response to the classification result indicating that there is a concurrent conflict.

[0014] This application also provides a maintenance training method with a fault intelligent diagnosis function, which includes:

[0015] Obtaining a set of event text descriptions distributed by a maintenance training system to multiple users;

[0016] Obtaining a set of event data generated by the multiple users;

[0017] Performing context semantic association analysis based on word granularity on each event text description in the set of event text descriptions respectively to obtain a sequence of event text description semantic feature vectors;

[0018] Performing semantic encoding on each event data in the set of event data generated by the multiple users respectively to obtain a sequence of event data semantic encoding feature vectors;

[0019] Calculating the event semantic difference measurement coefficients between each pair of corresponding event text description semantic feature vectors and event data semantic encoding feature vectors in the sequence of event text description semantic feature vectors and the sequence of event data semantic encoding feature vectors respectively to obtain an event semantic difference measurement vector composed of multiple said event semantic difference measurement coefficients;

[0020] Performing feature extraction on the event semantic difference measurement vector through an event semantic difference time series correlation feature extractor based on a deep neural network model to obtain event semantic difference time series features;

[0021] Based on the temporal characteristics of the event semantic differences, determine whether there is a concurrent conflict;

[0022] In response to the classification result indicating the existence of a concurrent conflict, issue a fault warning prompt.

[0023] Compared with the prior art, the maintenance training system and method with a fault intelligent diagnosis function provided by the present application collect the event text descriptions distributed by the maintenance training system for multiple users and the event data generated by these multiple users, and introduce data processing and semantic analysis algorithms at the backend to perform semantic analysis and correlation comparison of these data, so as to detect concurrent conflicts. In this way, it can automatically detect concurrent conflicts in collaborative virtual maintenance and issue a fault warning prompt in a timely manner, improving the quality and safety of maintenance training. Brief Description of the Drawings

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In the drawings:

[0025] Figure 1 It is a block diagram of a maintenance training system with a fault intelligent diagnosis function provided in an embodiment of the present application.

[0026] Figure 2 It is a flowchart of a maintenance training method with a fault intelligent diagnosis function provided in an embodiment of the present application.

[0027] Figure 3 It is a schematic diagram of the system architecture of a maintenance training method with a fault intelligent diagnosis function provided in an embodiment of the present application.

[0028] Figure 4 It is an application scenario diagram of a maintenance training system with a fault intelligent diagnosis function provided in an embodiment of the present application. Detailed Description of the Embodiments

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer and more understandable, the following will further elaborate on the embodiments of the present application with reference to the drawings. Herein, the illustrative embodiments of the present application and their descriptions are used to explain the present application, but do not limit the present application.

[0030] Unless otherwise specified, all technical and scientific terms used in the embodiments of this application have the same meanings as those commonly understood by those skilled in the technical field of this application. The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit the scope of this application.

[0031] In the description of the embodiments of this application, it should be noted that, unless otherwise specified and defined, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or the connection inside two components. It can be directly connected or indirectly connected through an intermediate medium. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific situations.

[0032] It should be noted that the terms "first / second / third" involved in the embodiments of this application are only used to distinguish similar objects and do not represent a specific order for the objects. It can be understood that "first / second / third" can be interchanged in a specific order or sequence when permitted. It should be understood that the objects distinguished by "first / second / third" can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here.

[0033] It should be understood that in order to ensure the normal progress of collaborative virtual maintenance, it is necessary to effectively control and solve conflict problems. Conflict detection is an important part of conflict control. It refers to judging whether there are conflicts in the system through a certain algorithm or mechanism and feeding back the conflict information to the user or the system. Conflict detection is a supplementary means adopted when conflict avoidance cannot completely eliminate conflicts. It can timely discover and handle conflicts. In this way, conflicts can be resolved at the primary stage, avoiding the ineffective work and huge rework caused by the spread of conflicts, affecting the quality and efficiency of collaborative work, and thus avoiding greater impacts caused by conflicts.

[0034] The key to collaborative virtual maintenance is to solve three technical problems: concurrent conflict control, consistency implementation, and collaborative awareness.

[0035] 1. Concurrent conflict control: Concurrent conflict control can be solved from the following three aspects.

[0036] Conflict avoidance. Restrict user operations through certain rules so that they do not conflict. The specific methods are: (1) Lock-based concurrency control. Users need to apply before operating. If the other party does not release the lock, the current user will be in a long-term starvation state. This mechanism is not very responsive and has poor real-time interactivity for users. (2) Optimistic concurrency control. Allow transactions to execute unimpeded until all operations are completed, and then verify them when submitting. If they pass the verification, they will be submitted, otherwise they will be restarted. This mechanism tolerates temporary conflicts and handles conflicts in a centralized manner. The cost of restarting is too high. (3) Time-stamp concurrency control. Each transaction is assigned a time stamp, and the order of transaction execution is determined by the size of the time stamp. This mechanism has a good degree of concurrency. (4) Token mechanism. Only one user can obtain an access token at a time, and other collaborators can only wait for the release of the current access token. It is a simple exclusive method that affects the collaboration of users. Conflict is an essential phenomenon in the collaborative process. The adoption of various conflict avoidance technologies and means can only reduce and avoid a certain number and type of conflicts to a certain extent, but cannot completely eliminate conflicts.

[0037] Conflict detection. Conflicts in concurrent operations in the system are judged through certain judgment rules. The use of conflict avoidance measures can only reduce but not completely eliminate conflicts. In order to detect conflicts early and resolve them in the early stages as much as possible, avoid ineffective work and huge rework caused by the spread of conflicts, and affect the quality and efficiency of collaborative work, it is necessary to use effective technical means to detect these conflicts in time when the conflicts are potential and about to occur or have occurred but have not yet spread further, and make correct handling accordingly. The specific methods are: (1) Conflict detection based on Petri nets. (2) Conflict detection based on truth values. (3) Conflict detection based on constraints. (4) Conflict detection based on heuristic classification.

[0038] Specifically, conflict detection based on Petri nets. This method uses Petri nets to model objects, operations, and states in collaborative virtual maintenance, and determines whether there is a conflict by analyzing the transitions and labels in the Petri nets. Conflict detection based on truth tables. This method uses truth tables to represent the operations that can be performed by each object in collaborative virtual maintenance in different states, and determines whether there is a conflict by comparing the values ​​of corresponding positions in the truth table. Conflict detection based on constraints. This method uses constraints to describe the conditions that should be met between objects in collaborative virtual maintenance or between objects and the environment, and determines whether there is a conflict by checking whether the constraints are met. Conflict detection based on heuristic classification. This method uses heuristic rules to classify operations in collaborative virtual maintenance into different categories, and determines whether there is a conflict by comparing the categories of operations.

[0039] Conflict resolution. According to certain rules, coordinate and handle the conflicting operations. When the system detects a conflict, it must provide corresponding countermeasures and suggestions for the collaborative members based on the characteristics, forms, and content of the generated conflict to resolve the conflict. The methods are as follows: (1) Backtracking method. After a conflict occurs, the process rolls back forward until the conflicting operation is revoked, but it is difficult to determine a reasonable backtracking span. (2) Constraint relaxation. This method is a mutual compromise, and the conflict resolution method is achieved at the cost of modifying the goal. (3) Arbitration and negotiation resolution. All work must be suspended before the conflict is resolved. Whether a user participates in the current conflicting operation or not, it seriously affects the continuity of collaboration. (4) Multiple versions. Each time a conflict occurs, a new version is generated. This mechanism well preserves the operation intentions of the collaborators, but as the operations gradually deepen, there will be more and more versions, making it difficult to control.

[0040] Based on this, in the technical solution of this application, a maintenance training system with a fault intelligent diagnosis function is proposed. The process of conflict detection includes the following steps: 1. Event definition. An event refers to any operation performed by a user on a device or component in collaborative virtual maintenance, such as selection, movement, rotation, installation, disassembly, etc. Each event has a unique identifier and some related attributes, such as occurrence time, initiating user, target object, operation type, etc. 2. Event recording. During the operation of the system, all events generated by users are captured and recorded in real time, and the event information is stored in an event queue. The event queue is a data structure sorted according to the event occurrence time, which can ensure the timeliness of events. 3. Event distribution. The system distributes the corresponding events to all users participating in collaborative virtual maintenance according to the event information in the event queue. Each user has a local event buffer area for storing the received events and sorting them according to the event occurrence time. 4. Event comparison. Each user compares the events generated by himself / herself with the events generated by other users and the distributed events according to his / her local event buffer area to determine whether there is a conflict. If there are significant differences in the target object and operation type between the generated event and the distributed event, and the event occurrence times of each user overlap, then these two events are considered to have a conflict. 5. Event feedback. If a conflict is detected, the system will provide conflict prompts and solution suggestions to the relevant users, allowing the users to choose whether to accept or modify their operations according to the actual situation to eliminate or reduce the conflict.

[0041] Correspondingly, in the above-mentioned maintenance training system with a fault intelligent diagnosis function, the process of event comparison is of great significance for conflict detection. Therefore, in order to detect and handle conflicts in a timely manner and avoid greater impacts caused by conflicts, the technical concept of this application is to collect the event text descriptions distributed by the maintenance training system for multiple users and the event data generated by these multiple users, and introduce data processing and semantic analysis algorithms in the backend to perform semantic analysis and correlation comparison of these data, so as to detect concurrent conflicts. In this way, it is possible to automatically detect concurrent conflicts in collaborative virtual maintenance and issue fault warning prompts in a timely manner, improving the quality and safety of maintenance training.

[0042] In an embodiment of the present application, Figure 1 is a block diagram of a maintenance training system with a fault intelligent diagnosis function provided in an embodiment of the present application. As Figure 1 shown, the maintenance training system 100 with a fault intelligent diagnosis function according to an embodiment of the present application includes: an event text description acquisition module 110, configured to obtain a set of event text descriptions distributed by the maintenance training system for multiple users; an event data acquisition module 120, configured to obtain a set of event data generated by the multiple users; an event text description semantic feature analysis module 130, configured to perform context semantic association analysis based on word granularity on each event text description in the set of event text descriptions to obtain a sequence of event text description semantic feature vectors; an event data semantic encoding module 140, configured to perform semantic encoding on each event data in the set of event data generated by the multiple users to obtain a sequence of event data semantic encoding feature vectors; an event semantic difference measurement module 150, configured to calculate the event semantic difference measurement coefficients between each corresponding event text description semantic feature vector and event data semantic encoding feature vector in the sequence of event text description semantic feature vectors and the sequence of event data semantic encoding feature vectors respectively to obtain an event semantic difference measurement vector composed of multiple event semantic difference measurement coefficients; an event semantic difference time series feature encoding module 160, configured to extract features from the event semantic difference measurement vector through an event semantic difference time series correlation feature extractor based on a deep neural network model to obtain event semantic difference time series features; a conflict detection module 170, configured to determine whether there is a concurrent conflict based on the event semantic difference time series features; and an early warning module 180, configured to issue a fault warning prompt in response to the classification result indicating the existence of a concurrent conflict.

[0043] Specifically, in the technical solution of the present application, first, a set of event text descriptions distributed by the maintenance training system for multiple users is obtained, and a set of event data generated by the multiple users is obtained. Then, considering that there is event semantic information in each of the event text descriptions about the events distributed by the maintenance training system for each user, this event semantic information is crucial for concurrent conflict detection and is the basis for judging whether there is a conflict. Therefore, in order to better capture the semantic information in the event text descriptions, in the technical solution of the present application, each event text description in the set of event text descriptions is respectively subjected to word segmentation processing and then encoded through an event text semantic encoder including a word embedding layer and an LSTM model to respectively extract the context semantic association feature information based on word granularity in each event text description, so as to obtain a sequence of event text description semantic feature vectors.

[0044] In a specific embodiment of the present application, the event text description semantic feature analysis module is used to: respectively perform word segmentation processing on each event text description in the set of event text descriptions and then pass it through an event text semantic encoder including a word embedding layer and an LSTM model to obtain a sequence of event text description semantic feature vectors.

[0045] It should be understood that the event data may include various attributes and features, such as target objects, operation types, timestamps, etc., and there is event semantic information in these data about the events generated by user operations. In order to capture the semantic information in these event data to assist in detecting concurrent conflicts, in the technical solution of the present application, each event data in the set of event data generated by the multiple users is further semantically encoded respectively to extract the semantic encoding feature information in the event data generated by each user respectively, so as to obtain a sequence of event data semantic encoding feature vectors.

[0046] Then, in order to measure the semantic difference degree between the event text description semantic features distributed by the system for each user and the event data encoding features generated by each user to judge whether the operation behavior of each user is consistent with the events distributed by the system. Based on this, in the technical solution of the present application, the event semantic difference measurement coefficients between each corresponding event text description semantic feature vector and event data semantic encoding feature vector in the sequence of event text description semantic feature vectors and the sequence of event data semantic encoding feature vectors are further calculated respectively to obtain an event semantic difference measurement vector composed of multiple event semantic difference measurement coefficients. In particular, here, the event semantic difference measurement vector can reflect the semantic difference degree between the event text description and the event data and is used to further detect the risk of concurrent conflicts. Specifically, when the semantic difference between the event text description and the event data is relatively large, there may be a risk of concurrent conflicts.

[0047] In a specific embodiment of the present application, the event semantic difference measurement module is configured to: calculate the event semantic difference measurement coefficient between each corresponding event text description semantic feature vector and event data semantic coding feature vector in the sequence of event text description semantic feature vectors and the sequence of event data semantic coding feature vectors by using the following semantic difference measurement formula to obtain an event semantic difference measurement vector composed of a plurality of the event semantic difference measurement coefficients; wherein, the semantic difference measurement formula is:

[0048]

[0049] wherein, p (x,y) and q (x,y) are the event semantic difference measurement coefficients between each corresponding event text description semantic feature vector and event data semantic coding feature vector in the sequence of event text description semantic feature vectors and the sequence of event data semantic coding feature vectors, W is the scale of each event text description semantic feature vector in the sequence of event text description semantic feature vectors, H is the scale of each event data semantic coding feature vector in the sequence of event data semantic coding feature vectors, s i are the plurality of event semantic difference measurement coefficients of the event semantic difference measurement vector, and log represents the logarithmic function with base 2.

[0050] Furthermore, considering that in actual concurrent conflict detection, it is not only necessary to detect whether there are semantic features such as target objects and operation types with large differences between the events generated by each user and the distributed events, but also to consider whether the event occurrence times of each user overlap. If the event occurrence times overlap, it is considered that there is a conflict between the events of these two users. Based on this, in the technical solution of the present application, the event semantic difference measurement vector is further subjected to feature mining through an event semantic difference time-series correlation feature extractor based on a one-dimensional convolutional layer to extract the time-series correlation features of the difference feature information between the distributed event semantics and the generated event semantics of each user in the time dimension, so as to obtain an event semantic difference time-series feature vector. It should be understood that by extracting the time-series feature vector of the event semantic difference, the time-series correlation between events can be better analyzed, which is very important for concurrent conflict detection because concurrent conflicts usually involve the time relationship and sequence relationship between events. By analyzing the time-series features of the event semantic difference, the occurrence and evolution of concurrent conflicts can be more accurately judged, and corresponding early warning prompts can be provided.

[0051] In a specific embodiment of the present application, the event semantic difference time-series correlation feature extractor based on the deep neural network model is an event semantic difference time-series correlation feature extractor based on a one-dimensional convolutional layer.

[0052] Further, the event semantic difference temporal feature encoding module is configured to: extract features from the event semantic difference metric vector through the event semantic difference temporal correlation feature extractor based on a one-dimensional convolutional layer to obtain an event semantic difference temporal feature vector as the event semantic difference temporal feature.

[0053] Subsequently, the event semantic difference temporal feature vector is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether there is a concurrent conflict. That is, classification processing is performed using the temporal correlation features in the time dimension of the difference feature information between the distribution event semantics and the generation event semantics of each user, so as to detect concurrent conflicts. In particular, in response to the classification result indicating the existence of a concurrent conflict, a fault warning prompt is issued.

[0054] In a specific embodiment of the present application, the conflict detection module is configured to: pass the event semantic difference temporal feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is a concurrent conflict.

[0055] In an embodiment of the present application, the maintenance training system with a fault intelligent diagnosis function further includes a training module for training the event text semantic encoder including a word embedding layer and an LSTM model, the event semantic difference time series correlation feature extractor based on a one-dimensional convolutional layer, and the classifier. The training module includes: a training event text description acquisition unit for obtaining a set of training event text descriptions distributed by the maintenance training system to multiple users; a training event data acquisition unit for obtaining a set of training event data generated by the multiple users; a training event text description semantic feature analysis unit for performing word segmentation on each training event text description in the set of training event text descriptions and then obtaining a sequence of training event text description semantic feature vectors through the event text semantic encoder including the word embedding layer and the LSTM model; a training event data semantic encoding unit for respectively performing semantic encoding on each training event data in the set of training event data generated by the multiple users to obtain a sequence of training event data semantic encoding feature vectors; a training event semantic difference measurement unit for respectively calculating the training event semantic difference measurement coefficients between each corresponding training event text description semantic feature vector and training event data semantic encoding feature vector in the sequence of training event text description semantic feature vectors and the sequence of training event data semantic encoding feature vectors to obtain a training event semantic difference measurement vector composed of multiple training event semantic difference measurement coefficients; a training event semantic difference time series feature encoding unit for performing feature extraction on the training event semantic difference measurement vector through the event semantic difference time series correlation feature extractor based on a one-dimensional convolutional layer to obtain a training event semantic difference time series feature vector; a training optimization unit for correcting the training event semantic difference time series feature vector to obtain a corrected training event semantic difference time series feature vector; a training classification unit for passing the corrected training event semantic difference time series feature vector through the classifier to obtain a classification loss function value; a training unit for training the event text semantic encoder including the word embedding layer and the LSTM model, the event semantic difference time series correlation feature extractor based on a one-dimensional convolutional layer, and the classifier based on the classification loss function value.

[0056] In the technical solution of this application, the sequence of the training event text description semantic feature vectors represents the text semantic features of the bidirectional association between the short-range and long-range source semantic contexts based on word segmentation embedding semantics of each training event text description in the set of training event text descriptions, while the sequence of the training event data semantic encoding feature vectors represents the encoded semantic features of each event data in the set of training event data generated by multiple users. Considering the feature correspondence difference between the training event text description and the training event data, the training event semantic difference metric vector composed of multiple training event semantic difference metric coefficients will have relatively significant distribution discreteness. In this way, after the local association features are extracted from the training event semantic difference metric vector through the event semantic difference time series association feature extractor based on a one-dimensional convolutional layer, the obtained training event semantic difference time series feature vector will have the game discretization of the association distribution information, thus affecting the classification training of the training event semantic difference time series feature vector through the classifier.

[0057] Based on this, in this application, preferably, each time the training event semantic difference time series feature vector passes through the classifier for iterative training, the training event semantic difference time series feature vector is corrected. Specifically, the training optimization unit includes: first, calculating the self-association matrix of the training event semantic difference time series feature vector and its own transpose, and calculating the inner product of the i-th and j-th row vectors of the self-association matrix as the matrix value at the (i, j) position of the weight matrix to obtain the weight matrix. Then, after multiplying the weight matrix by the self-association matrix, further multiplying the result by the training event semantic difference time series feature vector in a matrix-vector multiplication to obtain the correction vector. Finally, multiplying the correction vector by the training event semantic difference time series feature vector in a dot product to obtain the corrected training event semantic difference time series feature vector.

[0058] In this way, based on the self-association dimension of the training event semantic difference time series feature vector as the object to be modulated, the association expansion is performed based on the spatial sub-dimension complexity of the high-dimensional feature space of the feature distribution. Thus, the decomposable association dimension offset of the training event semantic difference time series feature vector is introduced into the heterogeneous association embedding space, and the joint fine-tuning of the decomposable dimension set represented by the heterogeneous association embedding space is used to enhance its association self-consistent relationship, so as to improve the predetermined category harmony of the feature vector of the training event semantic difference time series feature vector in the association target classification domain, thereby improving the training effect of the training event semantic difference time series feature vector through the classifier. In this way, it is possible to automatically detect concurrent conflicts in collaborative virtual maintenance during the maintenance training process and issue a fault warning prompt in a timely manner when a conflict is detected, thereby improving the quality and safety of the maintenance training.

[0059] In summary, the maintenance training system 100 with intelligent fault diagnosis function according to the embodiments of the present application is elucidated, which can automatically detect concurrent conflicts in collaborative virtual maintenance and issue fault warning prompts in a timely manner, improving the quality and safety of maintenance training.

[0060] As described above, the maintenance training system 100 with intelligent fault diagnosis function according to the embodiments of the present application can be implemented in various terminal devices, such as a server for maintenance training with intelligent fault diagnosis function. In one example, the maintenance training system 100 with intelligent fault diagnosis function according to the embodiments of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the maintenance training system 100 with intelligent fault diagnosis function can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the maintenance training system 100 with intelligent fault diagnosis function can also be one of the many hardware modules of the terminal device.

[0061] Alternatively, in another example, the maintenance training system 100 with intelligent fault diagnosis function and the terminal device can also be separate devices, and the maintenance training system 100 with intelligent fault diagnosis function can be connected to the terminal device through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0062] Figure 2 It is a flowchart of a maintenance training method with intelligent fault diagnosis function provided in the embodiments of the present application. Figure 3 It is a schematic diagram of the system architecture of a maintenance training method with intelligent fault diagnosis function provided in the embodiments of the present application. As Figure 2 and Figure 3As shown, a maintenance training method with a fault intelligent diagnosis function includes: 210, obtaining a set of event text descriptions distributed by a maintenance training system to multiple users; 220, obtaining a set of event data generated by the multiple users; 230, respectively performing context semantic association analysis based on word granularity on each event text description in the set of event text descriptions to obtain a sequence of event text description semantic feature vectors; 240, respectively performing semantic encoding on each event data in the set of event data generated by the multiple users to obtain a sequence of event data semantic encoding feature vectors; 250, respectively calculating an event semantic difference metric coefficient between each corresponding event text description semantic feature vector and event data semantic encoding feature vector in the sequence of event text description semantic feature vectors and the sequence of event data semantic encoding feature vectors to obtain an event semantic difference metric vector composed of multiple event semantic difference metric coefficients; 260, performing feature extraction on the event semantic difference metric vector through an event semantic difference time series association feature extractor based on a deep neural network model to obtain event semantic difference time series features; 270, determining whether there is a concurrent conflict based on the event semantic difference time series features; 280, in response to the classification result indicating the existence of a concurrent conflict, issuing a fault warning prompt.

[0063] In the maintenance training method with a fault intelligent diagnosis function, respectively performing context semantic association analysis based on word granularity on each event text description in the set of event text descriptions to obtain a sequence of event text description semantic feature vectors includes: respectively performing word segmentation on each event text description in the set of event text descriptions and then passing through an event text semantic encoder including a word embedding layer and an LSTM model to obtain the sequence of event text description semantic feature vectors.

[0064] Those skilled in the art can understand that the specific operations of each step in the above maintenance training method with a fault intelligent diagnosis function have been introduced in detail in the description of the Figure 1 maintenance training system with a fault intelligent diagnosis function above, and therefore, its repeated description will be omitted.

[0065] Figure 4 This is an application scenario diagram of a maintenance training system with a fault intelligent diagnosis function provided in an embodiment of the present application. As Figure 4 shown, in this application scenario, first, obtain a set of event text descriptions distributed by the maintenance training system to multiple users (for example, C1 as shown in Figure 4 ); obtain a set of event data generated by the multiple users (for example, as shown in Figure 4as schematically shown in C2); then, input the obtained set of event text descriptions and the set of event data into a server (e.g., as schematically shown in S) deployed with a maintenance training algorithm having a fault intelligent diagnosis function, where the server can process the set of event text descriptions and the set of event data based on the maintenance training algorithm having a fault intelligent diagnosis function to determine whether there is a concurrent conflict; in response to the classification result indicating the existence of a concurrent conflict, issue a fault warning prompt. Figure 4 as schematically shown in S), wherein the server can process the set of event text descriptions and the set of event data based on the maintenance training algorithm having a fault intelligent diagnosis function to determine whether there is a concurrent conflict; in response to the classification result indicating the existence of a concurrent conflict, issue a fault warning prompt.

[0066] In the above specific embodiments, the purpose, technical solutions, and beneficial effects of the present application have been further described in detail. It should be understood that the above are only specific embodiments of the present application and are not used to limit the protection scope of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A maintenance training system with intelligent fault diagnosis function, characterized in that: include: An event text description collection module is used to obtain a collection of event text descriptions distributed by the maintenance training system to multiple users; An event data collection module, used to obtain a collection of event data generated by the multiple users; An event text description semantic feature analysis module, used to perform contextual semantic association analysis based on word granularity on each event text description in the set of event text descriptions to obtain a sequence of event text description semantic feature vectors; An event data semantic encoding module, used to semantically encode each event data in the set of event data generated by the plurality of users to obtain a sequence of event data semantic encoding feature vectors; An event semantic difference measurement module is used to respectively calculate the event semantic difference measurement coefficients between each group of corresponding event text description semantic feature vectors and event data semantic coding feature vectors in the sequence of the event text description semantic feature vectors and the sequence of the event data semantic coding feature vectors to obtain an event semantic difference measurement vector composed of a plurality of the event semantic difference measurement coefficients; An event semantic difference temporal feature encoding module is used to extract features from the event semantic difference measurement vector through an event semantic difference temporal association feature extractor based on a deep neural network model to obtain event semantic difference temporal features; A conflict detection module, used to determine whether there is a concurrent conflict based on the time sequence characteristics of the semantic differences of the events; An early warning module, for issuing a fault early warning prompt in response to the classification result indicating that there is a concurrent conflict; Wherein, the event semantic difference measurement module is used to: The event semantic difference measurement coefficient between each group of corresponding event text description semantic feature vectors and event data semantic coding feature vectors in the sequence of the event text description semantic feature vectors and the sequence of the event data semantic coding feature vectors is calculated using the following semantic difference measurement formula to obtain an event semantic difference measurement vector composed of a plurality of the event semantic difference measurement coefficients; The semantic difference measurement formula is: Among them, p (x,y) and q (x,y) is the event semantic difference measurement coefficient between each group of corresponding event text description semantic feature vectors and event data semantic coding feature vectors in the sequence of event text description semantic feature vectors and the sequence of event data semantic coding feature vectors, W is the scale of each event text description semantic feature vector in the sequence of event text description semantic feature vectors, H is the scale of each event data semantic coding feature vector in the sequence of event data semantic coding feature vectors, s i are a plurality of event semantic difference measurement coefficients of the event semantic difference measurement vector, and log represents a logarithmic function with base 2; The conflict detection module is used to: pass the event semantic difference time series feature vector through a classifier to obtain a classification result, and the classification result is used to indicate whether there is a concurrency conflict.

2. The maintenance training system with intelligent fault diagnosis function according to claim 1 is characterized in that: The event text description semantic feature analysis module is used to: perform word segmentation on each event text description in the set of event text descriptions and then pass it through an event text semantic encoder including a word embedding layer and an LSTM model to obtain a sequence of semantic feature vectors of the event text descriptions.

3. The maintenance training system with intelligent fault diagnosis function according to claim 2 is characterized in that: The event semantic difference temporal association feature extractor based on the deep neural network model is an event semantic difference temporal association feature extractor based on a one-dimensional convolutional layer.

4. The maintenance training system with intelligent fault diagnosis function according to claim 3 is characterized in that: The event semantic difference temporal feature encoding module is used to: extract features from the event semantic difference measurement vector through the event semantic difference temporal association feature extractor based on the one-dimensional convolution layer to obtain an event semantic difference temporal feature vector as the event semantic difference temporal feature.

5. The maintenance training system with intelligent fault diagnosis function according to claim 4 is characterized in that: It also includes a training module for training the event text semantic encoder including the word embedding layer and the LSTM model, the event semantic difference temporal association feature extractor based on the one-dimensional convolutional layer, and the classifier.

6. The maintenance training system with intelligent fault diagnosis function according to claim 5 is characterized in that: The training module comprises: A training event text description acquisition unit, used to obtain a set of training event text descriptions distributed by the maintenance training system to multiple users; A training event data collection unit, used to obtain a set of training event data generated by the multiple users; A training event text description semantic feature analysis unit, used for performing word segmentation on each training event text description in the set of training event text descriptions and then passing the result through the event text semantic encoder including the word embedding layer and the LSTM model to obtain a sequence of training event text description semantic feature vectors; A training event data semantic encoding unit, used to semantically encode each training event data in the set of training event data generated by the plurality of users to obtain a sequence of semantic encoding feature vectors of the training event data; A training event semantic difference measurement unit is used to respectively calculate the training event semantic difference measurement coefficients between each group of corresponding training event text description semantic feature vectors and training event data semantic coding feature vectors in the sequence of the training event text description semantic feature vectors and the sequence of the training event data semantic coding feature vectors to obtain a training event semantic difference measurement vector composed of a plurality of the training event semantic difference measurement coefficients; A training event semantic difference temporal feature encoding unit, used for extracting features from the training event semantic difference measurement vector through the event semantic difference temporal association feature extractor based on the one-dimensional convolution layer to obtain a training event semantic difference temporal feature vector; A training optimization unit, used for correcting the training event semantic difference time series feature vector to obtain a corrected training event semantic difference time series feature vector; A training classification unit, used for passing the corrected training event semantic difference time series feature vector through the classifier to obtain a classification loss function value; A training unit is used to train the event text semantic encoder including the word embedding layer and the LSTM model, the event semantic difference temporal association feature extractor based on the one-dimensional convolutional layer, and the classifier based on the classification loss function value.

7. A maintenance training method with intelligent fault diagnosis function, using the maintenance training system with intelligent fault diagnosis function as claimed in claim 1, characterized in that: include: Obtain a collection of event text descriptions distributed by a maintenance training system to multiple users; Acquire a set of event data generated by the multiple users; Performing a contextual semantic association analysis based on word granularity on each event text description in the set of event text descriptions to obtain a sequence of semantic feature vectors of the event text descriptions; Semantically encoding each event data in the set of event data generated by the multiple users to obtain a sequence of semantically encoded feature vectors of the event data; Respectively calculating the event semantic difference measurement coefficients between each group of corresponding event text description semantic feature vectors and event data semantic coding feature vectors in the sequence of the event text description semantic feature vectors and the sequence of the event data semantic coding feature vectors to obtain an event semantic difference measurement vector composed of a plurality of the event semantic difference measurement coefficients; Extracting features from the event semantic difference measurement vector using an event semantic difference temporal correlation feature extractor based on a deep neural network model to obtain event semantic difference temporal features; Determining whether there is a concurrent conflict based on the semantic difference temporal characteristics of the events; In response to the classification result indicating that there is a concurrent conflict, a fault warning prompt is issued.

8. The maintenance training method with intelligent fault diagnosis function according to claim 7, characterized in that: A contextual semantic association analysis based on word granularity is performed on each event text description in the set of event text descriptions to obtain a sequence of event text description semantic feature vectors, including: each event text description in the set of event text descriptions is subjected to word segmentation processing and then passed through an event text semantic encoder including a word embedding layer and an LSTM model to obtain a sequence of event text description semantic feature vectors.

Images