Railway-locomotive diesel-engine fault diagnosis system based on knowledge graph

Abstract

The present invention relates to the technical field of railway-locomotive diesel-engine fault diagnosis systems, and particularly relates to a railway-locomotive diesel-engine fault diagnosis system based on a knowledge graph. The system comprises a data collection module, a transmission module, a preprocessing module, an algorithm application module, a storage module and a front-end display module, wherein one end of the data collection module is used for performing signal collection by means of a sensor, and the other end thereof is connected to the preprocessing module by means of the transmission module; the preprocessing module performs diagnosis and positioning on a specific faulty part in combination with a fault classification algorithm of the algorithm application module; and an original signal collected by the data collection module and algorithm diagnosis information are scheduled and stored in the storage module by means of back-end management software, and display interaction materials are then provided for the front-end display module. The present invention retains the previous operation mode of an expert knowledge base in terms of fault diagnosis so as to ensure the basic accuracy of cause analysis, and a neural network intelligent algorithm is introduced to reduce the manual workload and individual difference of people in terms of professionalism.

Description

A railway locomotive diesel engine fault diagnosis system based on knowledge graph Technical Field

[0001] The present invention relates to the technical field of railway locomotive diesel engine fault diagnosis systems, and in particular to a railway locomotive diesel engine fault diagnosis system based on a knowledge graph. Background Art

[0002] As railway users place increasing demands on intelligent operation and maintenance of locomotive diesel engines, simply developing health management systems based on anomaly detection may not be able to meet the increasingly demanding application assurance needs in the future. Users require diagnostic systems that can accurately and quickly locate or classify faults when they occur.

[0003] In actual daily use, individual products often have only a small number of similar failure cases due to manufacturing variations. While traditional fault diagnosis trees based on knowledge graphs and rules can effectively handle fault isolation and classification, they rely heavily on domain knowledge, and domain knowledge struggles to cover all fault types. Field technicians, influenced by their expertise and experience, inevitably make subjective or routine judgments on representative or confusing cases during cause analysis. Therefore, the development of diagnostic systems that can reduce subjective influences while improving efficiency and requiring less manual effort through digital development and machine assistance is urgently needed by engineers and technicians.

[0004] The patent number is CN112596495B, and the patent name is "A method and system for diagnosing industrial equipment faults based on a knowledge graph." This patent proposes a method and system for diagnosing industrial equipment faults based on a knowledge graph. This includes: collecting maintenance data and corresponding fault information from industrial equipment, and constructing a knowledge graph based on the maintenance data and corresponding fault data of the industrial equipment; obtaining current operating data of the industrial equipment to be diagnosed, and performing similarity measurement on the current operating data through the knowledge graph to obtain a similarity value; setting a similarity threshold, and comparing the similarity value with the similarity threshold, and generating corresponding fault information based on the comparison result. The present invention collects maintenance data and fault information from various industrial equipment and combines it with other knowledge information in this field to construct a comprehensive and complete knowledge graph, and then judges the status of the industrial equipment to be diagnosed based on the knowledge graph. This not only allows for timely judgment of equipment faults to avoid dangerous accidents, but also allows for tracing equipment faults and enhancing equipment management security.

[0005] The patent number is CN114579875B, and the patent name is "Equipment Fault Diagnosis and Maintenance Knowledge Recommendation System Based on Knowledge Graph". A equipment fault diagnosis and maintenance knowledge recommendation system based on knowledge graph is proposed. The system includes: a data layer, which is used to collect health management field knowledge of monitoring equipment; an analysis layer, which is used to extract health management field knowledge and construct a knowledge graph, use link prediction to infer and complete the knowledge graph, and based on the completed and inferred knowledge graph, use the knowledge graph feature learning combined with the collaborative filtering recommendation algorithm to calculate the similarity and perform equipment diagnosis and maintenance knowledge prediction and ranking; an application layer, which is used to provide fault diagnosis under abnormal conditions and trend prediction analysis under normal operating conditions for monitoring equipment based on the equipment diagnosis and maintenance knowledge prediction and ranking, so as to recommend equipment failure causes and maintenance plans.

[0006] The technologies in the above two patents reflect two different research focuses of knowledge graph-based fault diagnosis systems. One is to compare the similarity between operating data and faults with relatively clear characterizations, and generate fault information based on the results; the knowledge graph here can basically be understood as the equipment fault data representation domain. The other is to push inspection and maintenance knowledge when the equipment triggers a fault; the knowledge graph here can basically be understood as a continuously improved fault inspection manual, and the implementation of fault location is still based on the subsequent inspection result feedback. Although the above two focus technologies are both helpful for fault diagnosis, they have the following shortcomings:

[0007] 1. A large number of clear fault samples are required as a basis for comparative diagnosis, which is very difficult to achieve on high-value equipment;

[0008] 2. The diagnostic model is a decision tree-like model with clear graphical mechanisms, which makes it difficult to construct some multidimensional data sets with complex working mechanisms;

[0009] 3. The diagnosis object is equipment operation failure, and the diagnosis of operation weakening or deterioration is less involved;

[0010] 4. The diagnostic carrier is a multi-dimensional fixed device, and its adaptability to railway equipment with large changes in operating environment remains to be discussed. Technical issues

[0011] In order to overcome the existing deficiencies, the present invention provides a railway locomotive diesel engine fault diagnosis system based on knowledge graph. Technical Solutions

[0012] The technical solution adopted by the present invention to solve its technical problems is: a railway locomotive diesel engine fault diagnosis system based on knowledge graph, including a data acquisition module, a transmission module, a preprocessing module, an algorithm application module, a storage module and a front-end display module; one end of the data acquisition module performs signal acquisition through various monitoring sensors arranged on the locomotive diesel engine, and the other end is connected to the preprocessing module through the transmission module; the preprocessing module includes data cleaning, duplicate checking, denoising and other tasks, and the signal transmitted by the transmission module is cleaned and sliced, and the signal feature vector is extracted by the feature vector extractor, and the abnormal working signal feature vector is identified by the abnormal recognition and diagnosis algorithm of the algorithm application module, and it is combined with the fault classification algorithm to diagnose and locate the specific faulty part; the original signal and algorithm diagnosis information collected by the data acquisition module are stored in the storage module through the back-end management software, and then the display interaction material is provided to the front-end display module.

[0013] According to another embodiment of the present invention, the algorithm application module further includes a fault classification algorithm, which builds a failure analysis fault tree for a railway locomotive diesel engine and imports it into a system knowledge base, retrieves the corresponding failure mode in the failure analysis fault tree by identifying the working abnormality signal feature vector using an abnormality recognition and diagnosis algorithm, and generates a pre-classification of all faults based on domain knowledge.

[0014] According to another embodiment of the present invention, it further includes sorting out the pre-classification of abnormal faults associated with each signal based on the mechanistic relationship between the failure analysis fault tree and the detection signal, and visualizing the correlation relationship to generate a correlation graph, and injecting the correlation graph into the graph convolution network as prior knowledge. The graph convolution network combines the signal feature vector as input, dynamically updates the adjacency matrix weight of the correlation graph, and finally obtains a fault location or classification algorithm.

[0015] According to another embodiment of the present invention, the data acquisition module further includes various sensors and data acquisition equipment XIOS arranged on the diesel engine and its supercharger; one end of the data acquisition equipment XIOS collects sensor signals installed on the machine, and the other end is bidirectionally connected to the LCS locomotive control system via Ethernet, thereby forming a railway locomotive diesel engine operating status monitoring network.

[0016] According to another embodiment of the present invention, it further includes that the sensor signals on the machine include a diesel engine data acquisition unit and a supercharger VTG control unit; the diesel engine data acquisition unit and the supercharger VTG control unit transmit the collected basic control data to the MVC basic control protection unit and the data acquisition device XIOS; the MVC basic control protection unit communicates with the LCS locomotive control system and receives the control signal sent by the latter, the LCS locomotive control system communicates with the data acquisition device XIOS in both directions via Ethernet, and can directly communicate with the supercharger VTG control unit when the data acquisition device XIOS fails; the MVC basic control protection unit communicates with the data acquisition device XIOS via the CAN bus, thereby forming a railway locomotive diesel engine control network.

[0017] According to another embodiment of the present invention, it further includes that the data acquisition device XIOS communicates with the edge computing device in a two-way manner to realize on-board real-time diagnosis and adaptation, and the LCS locomotive control system realizes ground diagnostic algorithm training iteration and big data storage by connecting to the locomotive ground intelligent operation and maintenance platform.

[0018] According to another embodiment of the present invention, the edge computing device further includes a CPU board, a network communication board, a power board, a switch board and a backplane; the backplane connects the CPU board, the network communication board and the power board to the switch board for communication, and supplies power through the power board; the network communication board communicates with the data acquisition device XIOS network. Beneficial effects

[0019] Beneficial effects of the present invention:

[0020] This invention not only retains the previous operating mode of the expert knowledge base in fault diagnosis to ensure the basic accuracy of cause analysis, but also introduces neural network intelligent algorithms to reduce manual workload and individual professional differences. As the number of fault cases in the knowledge graph increases and users iterate on the system's usage feedback, the accuracy of its recommended diagnostic results will also increase.

[0021] The present invention can realize machine anomaly identification and identify the weakening and degradation of the operating state before the failure. It does not require a large number of fault mark samples and is very friendly to the anomaly identification of high-value equipment.

[0022] 3. The fault location diagnostic method is a graph convolutional network that combines knowledge graphs and prior knowledge, which is more suitable for equipment failures with complex working mechanisms;

[0023] 4. The system integrates an adaptive algorithm of the diagnostic model, which is more user-friendly to the special situation of China's railways where the operating environment changes greatly. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present invention will be further described below with reference to the accompanying drawings and examples.

[0025] FIG1 is a block diagram of the present invention;

[0026] Figure 2 is a block diagram of the data interaction structure of the ground link;

[0027] FIG3 is a block diagram of the data interaction of the vehicle link;

[0028] FIG4 is a block diagram of the internal connection structure of the preprocessing module and the algorithm module.

[0029] In the figure, 1. data acquisition module, 2. transmission module, 3. preprocessing module, 4. algorithm application module, 5. storage module, and 6. front-end display module. Modes for Carrying Out the Invention

[0030] As shown in Figure 1, a structural diagram of the present invention is a railway locomotive diesel engine fault diagnosis system based on a knowledge graph, comprising a data acquisition module 1, a transmission module 2, a preprocessing module 3, an algorithm application module 4, a storage module 5 and a front-end display module 6; one end of the data acquisition module 1 performs signal acquisition through various monitoring sensors arranged on the locomotive diesel engine, and the other end is connected to the preprocessing module 3 through the transmission module 2; the preprocessing module 3 includes data cleaning, duplicate checking, denoising and other tasks, and the signal transmitted by the transmission module 2 is cleaned and sliced, and the signal feature vector is extracted by the feature vector extractor, and the abnormal working signal feature vector is identified by the abnormal recognition and diagnosis algorithm of the algorithm application module 4, and it is combined with the fault classification algorithm to diagnose and locate the specific faulty part; the original signal and algorithm diagnosis information collected by the data acquisition module 1 are stored in the storage module 5 through the back-end management software scheduling, and then the front-end display module 6 is provided with display interactive materials.

[0031] Specifically, the data acquisition module 1 collects data from the diesel engine and supercharger, imports data including altitude, atmospheric pressure, and data based on fault mechanism and data correlation analysis results, and performs fault identification based on a given fault domain; the transmission module 2 is used to encrypt and temporarily store the original data, and transmit it back to the ground through the network; the preprocessing module 3 parses, cleans, checks for duplicates, denoises, slices, and performs time-frequency conversion on the data received from the transmission module 2 to prepare for the subsequent algorithm application module 4; the algorithm application module 4 executes an anomaly recognition algorithm on the data processed by the preprocessing module 3 to find abnormal signals, and proposes fault classification and positioning through internal algorithms.

[0032] According to another embodiment of the present invention, the algorithm application module 4 further includes a fault classification algorithm, which builds a failure analysis fault tree for the railway locomotive diesel engine and imports it into the system knowledge base, retrieves the corresponding failure mode in the failure analysis fault tree by identifying the working abnormality signal feature vector through the abnormality recognition and diagnosis algorithm, and generates all fault pre-classifications based on domain knowledge.

[0033] According to another embodiment of the present invention, it further includes sorting out the pre-classification of abnormal faults associated with each signal based on the mechanistic relationship between the failure analysis fault tree and the detection signal, and visualizing the correlation relationship to generate a correlation graph, and injecting the correlation graph into the graph convolution network as prior knowledge. The graph convolution network combines the signal feature vector as input, dynamically updates the adjacency matrix weight of the correlation graph, and finally obtains a fault location or classification algorithm.

[0034] According to another embodiment of the present invention, the data acquisition module 1 further includes various sensors and data acquisition equipment XIOS arranged on the diesel engine and its supercharger; one end of the data acquisition equipment XIOS collects sensor signals installed on the machine, and the other end is bidirectionally connected to the LCS locomotive control system via Ethernet, thereby forming a railway locomotive diesel engine operating status monitoring network.

[0035] According to another embodiment of the present invention, it further includes that the sensor signals on the machine include a diesel engine data acquisition unit and a supercharger VTG control unit; the diesel engine data acquisition unit and the supercharger VTG control unit transmit the collected basic control data to the MVC basic control protection unit and the data acquisition device XIOS; the MVC basic control protection unit communicates with the LCS locomotive control system and receives the control signal sent by the latter, the LCS locomotive control system communicates with the data acquisition device XIOS in both directions via Ethernet, and can directly communicate with the supercharger VTG control unit when the data acquisition device XIOS fails; the MVC basic control protection unit communicates with the data acquisition device XIOS via the CAN bus, thereby forming a railway locomotive diesel engine control network.

[0036] According to another embodiment of the present invention, it further includes that the data acquisition device XIOS communicates with the edge computing device in a two-way manner to realize on-board real-time diagnosis and adaptation, and the LCS locomotive control system realizes ground diagnostic algorithm training iteration and big data storage by connecting to the locomotive ground intelligent operation and maintenance platform.

[0037] Specifically, since the data acquisition device XIOS participates in the protective control of the locomotive, the carrier running the diagnostic algorithm is deployed on an independent edge computing device to further reduce the impact on the locomotive control system. After the data acquisition device XIOS performs necessary protective diagnosis on the machine operating parameters, it forwards the data to the edge computing device. After the diagnostic algorithm deployed on it performs fault diagnosis, the edge computing device will feedback the identified abnormal events to the data acquisition device XIOS, and send it back to the LCS locomotive control system together with the operating parameters, providing the driver with visual data and operation and maintenance suggestions.

[0038] According to another embodiment of the present invention, the edge computing device further includes a CPU board, a network communication board, a power board, a switch board and a backplane; the backplane connects the CPU board, the network communication board and the power board to the switch board for communication, and supplies power through the power board; the network communication board communicates with the data acquisition device XIOS network.

[0039] Specifically, the CPU board is used for data flow processing, control, data storage, and control software carrier of the load device; the network communication board is responsible for communicating with the XIOS network and receiving collected data; the power board is used to power the chassis, supporting 110V (+-40%) input and 5V output (including high temperature derating); the switch board is used to connect the device content data exchange; the backplane is used to provide Gigabit Ethernet circuits to connect each sub-board to the switch board for communication.

[0040] The first step is to collect data through the data acquisition module 1, encrypt and temporarily store the data through the transmission module 2, and then transmit it back to the ground through the network. The data return architecture of the diagnostic system of the data acquisition module 1 is divided into two links: on-board and ground. The on-board link is composed of various monitoring sensors arranged on the locomotive diesel engine, which aggregate the operating equipment into the data acquisition device XIOS. The data acquisition device XIOS is used as a physical separation between the monitoring network and the control network MVC to minimize interference; the data acquisition device XIOS performs necessary protective diagnosis on the machine operating parameters and forwards the data to the edge computing device, and the diagnostic algorithm deployed on it performs fault diagnosis and feeds back the results to the data acquisition device XIOS; because the data acquisition device XIOS participates in the protective control of the locomotive, the carrier for the operation of the diagnostic algorithm is deployed on an independent edge computing device to further reduce the impact on the locomotive control system. The independent device includes a separate backplane, power board, CPU board, network communication board, and data exchange board. Edge computing devices are conducive to expanding computing power when subsequent diagnostic system items are improved. Independent power supply boards are conducive to avoiding the problem of poor power supply quality of on-board power supply. Data exchange boards can realize the download and reading of data and diagnostic logs. Edge computing devices will feed back identified abnormal events to the data acquisition equipment XIOS, and return them together with the operating parameters to the LCS locomotive control system to provide the driver with visual data and operation and maintenance suggestions. After the locomotive arrives at the section, the data and event log packages compiled in the locomotive control system are imported into the diesel engine PHM ground server through manual download or wireless transmission. The operating data is diagnosed and algorithm training and optimization are carried out on the ground server with stronger computing power. The application information and fault logs are fed back to the locomotive ground intelligent operation and maintenance platform.

[0041] In the second step, the pre-processing module 3 receives the data received by the transmission module 2 and performs operations such as parsing, cleaning, duplicate checking, denoising, slicing, and time-frequency conversion to prepare for the subsequent algorithm application module 4;

[0042] The third step is to build a fault tree based on prior knowledge. Based on the reverse process of the fault tree, different signals are sorted out to pre-classify possible faults. Based on the fault pre-classification, software is used to generate a high-dimensional correlation diagram between various signals and faults. The correlation diagram is programmed based on the graph convolutional neural network to generate a fault classification algorithm. The abnormal identification algorithm is applied to the data of the algorithm application module 4 after the pre-processing module 3 to find the abnormal signal and execute the fault classification algorithm to finally complete the fault location or classification.

[0043] In the fourth step, the original signal collected by the data acquisition module 1 and the algorithm diagnosis information of the algorithm application module are stored in the storage module 5 through the back-end management software scheduling, and then provide display interaction materials to the front-end display module 6.

[0044] The above description is only illustrative of the present invention and not restrictive. Those skilled in the art will understand that many modifications, changes or equivalents may be made without departing from the spirit and scope defined by the appended claims, but all of them will fall within the scope of protection of the present invention.

Claims

1. A fault diagnosis system for railway locomotive diesel engines based on a knowledge graph, characterized in that, it includes a data acquisition module (1), a transmission module (2), a preprocessing module (3), an algorithm application module (4), a storage module (5) and a front-end display module (6); one end of the data acquisition module (1) collects signals through various monitoring sensors arranged on the locomotive diesel engine, and the other end is connected to the preprocessing module (3) through the transmission module (2); the preprocessing module (3) includes operations such as data cleaning, duplicate checking, and noise removal. The signals transmitted by the transmission module (2) are cleaned and sliced, and then signal feature vectors are extracted through a feature vector extractor. The abnormal working signal feature vectors are identified through the abnormal recognition and diagnosis algorithm of the algorithm application module (4), and the specific faulty parts are diagnosed and located by combining the fault classification algorithm; the original signals and algorithm diagnosis information collected by the data acquisition module (1) are scheduled and stored in the storage module (5) through the back-end management software, and then display and interaction materials are provided to the front-end display module (6).

2. The fault diagnosis system for railway locomotive diesel engines based on a knowledge graph according to claim 1, characterized in that, the fault classification algorithm of the algorithm application module (4) constructs a failure analysis fault tree for the railway locomotive diesel engine and imports it into the system knowledge base. By retrieving the corresponding failure modes in the failure analysis fault tree for the abnormal working signal feature vectors identified by the abnormal recognition and diagnosis algorithm, all fault pre-classifications based on domain knowledge are generated.

3. The fault diagnosis system for railway locomotive diesel engines based on a knowledge graph according to claim 2, characterized in that, Based on the mechanism relationship between the failure analysis fault tree and the detection signals, each signal-related abnormal fault pre-classification is sorted out, and its association relationship is visualized to generate an association graph. The association graph is injected into the graph convolutional network as prior knowledge. The graph convolutional network combines the signal feature vectors as inputs to dynamically update the weights of the adjacency matrix of the association graph, and finally obtains a fault location or classification algorithm.

4. The fault diagnosis system for railway locomotive diesel engines based on a knowledge graph according to claim 1, characterized in that, the data acquisition module (1) includes various sensors and the data acquisition device XIOS arranged on the diesel engine and its supercharger; one end of the data acquisition device XIOS collects the sensor signals installed on the machine, and the other end is bidirectionally connected to the LCS locomotive control system through Ethernet, thereby forming a monitoring network for the operating status of the railway locomotive diesel engine.

5. The fault diagnosis system for railway locomotive diesel engines based on a knowledge graph according to claim 4, characterized in that, The sensor signals on the machine include a diesel engine data acquisition unit and a supercharger VTG control unit; the diesel engine data acquisition unit and the supercharger VTG control unit transmit the collected basic control data to the MVC basic control protection unit and the data acquisition device XIOS; the MVC basic control protection unit communicates with the LCS locomotive control system and receives the control signals sent by the latter. The LCS locomotive control system communicates bidirectionally with the data acquisition device XIOS through Ethernet and can directly communicate with the supercharger VTG control unit in case of a failure of the data acquisition device XIOS; the MVC basic control protection unit communicates with the data acquisition device XIOS through the CAN bus, thereby forming a railway locomotive diesel engine control network.

6. A railway locomotive diesel engine fault diagnosis system based on a knowledge graph according to claim 4, characterized in that the data acquisition device XIOS communicates bidirectionally with the edge computing device to achieve on-vehicle real-time diagnosis and adaptation, and the LCS locomotive control system realizes ground diagnosis algorithm training iteration and big data storage by connecting to the locomotive ground intelligent operation and maintenance platform.

7. A railway locomotive diesel engine fault diagnosis system based on a knowledge graph according to claim 6, characterized in that the edge computing device internally includes a CPU board, a network communication board, a power supply board, a switching board and a backplane; the backplane connects the CPU board, the network communication board and the power supply board to the switching board for communication and is powered by the power supply board; the network communication board communicates with the data acquisition device XIOS through the network.

Images