Fault mode library intelligent generation method and system for new energy electric vehicle

Abstract

The invention discloses an intelligent generation method and system for a fault mode library of a new energy electric vehicle, and the method comprises the following steps: constructing a triple knowledge model, and determining product information and a mapping relation between a product function and a fault mode; generating a knowledge graph based on the triple knowledge model; selecting a plurality of graph evaluation factors based on node types in the knowledge graph; performing orthogonal verification according to the plurality of atlas evaluation factors and preset factor level parameters, and outputting a factor level pair corresponding to the optimal signal-to-noise ratio as an optimal test set; performing knowledge graph correction based on the optimal test set; generating a fault mode library based on the corrected knowledge graph; according to the method, based on the triple knowledge model, the product reliability knowledge chain is constructed, the knowledge graph is generated, the coverage rate and correctness of the knowledge graph are calculated by using the Taguchi method, the knowledge graph is corrected according to the sampling result, and repeated iteration is performed until the given correct rate requirement is met, so that the fault mode library of the new energy electric vehicle is generated.

Description

Technical Field

[0001] The present invention relates to the technical field of large language model processing, and more specifically to an intelligent generation method and system for a fault mode library of new energy electric vehicles. Background Art

[0002] At present, as an innovative product with rapid iteration, new energy electric vehicles have been greatly popularized in recent years. Their reliability is crucial for use safety, user experience, and use cost. Based on in-depth analysis of product function descriptions and working principles, a fault mode library can be constructed.

[0003] However, traditional methods rely on manual experience summary and expert knowledge collation, which are not only inefficient but also difficult to comprehensively cover various fault modes of new energy electric vehicles, thus affecting the accuracy and reliability of fault diagnosis.

[0004] Therefore, how to improve the generation efficiency and coverage rate of the fault mode library is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides an intelligent generation method and system for a fault mode library of new energy electric vehicles, which can build a product reliability knowledge chain based on a triple knowledge model, generate a knowledge graph, calculate the coverage rate and correctness of the knowledge graph using the Taguchi method, correct the knowledge graph according to the sampling results, and repeatedly iterate until the given accuracy requirement is met, thereby generating a fault mode library of new energy electric vehicles.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An intelligent generation method for a fault mode library of new energy electric vehicles, comprising the following steps:

[0008] Build a triple knowledge model to clarify product information and the mapping relationship between product functions and fault modes;

[0009] Generate a knowledge graph based on the triple knowledge model;

[0010] Select multiple graph evaluation factors based on the node types in the knowledge graph;

[0011] Perform orthogonal verification according to multiple graph evaluation factors and preset factor level parameters, and output the factor level pair corresponding to the optimal signal-to-noise ratio as the optimal test set; correct the knowledge graph based on the optimal test set;

[0012] Generate a fault mode library based on the corrected knowledge graph.

[0013] Preferably, building a triple knowledge model specifically includes:

[0014] Construct a reliability triple model, and describe the component products, the failure modes of the products, and the failure determination results of the products of new energy electric vehicles through the reliability triple;

[0015] Construct a product triple knowledge model, and describe the product functions and the corresponding product working principles formed by each product in the new energy electric vehicle through the product triple knowledge model;

[0016] Construct a product reliability knowledge chain, and connect the reliability triple and the product triple according to the mapping relationship between functions and failures through the product reliability knowledge chain.

[0017] Preferably, the reliability triple knowledge model is:

[0018] T1 = (h1, r1, t1)

[0019] Wherein, T1 is the reliability triple knowledge model, h1 is the product name, r1 is the failure mode, and t1 is the failure determination.

[0020] Preferably, the product triple knowledge model is:

[0021] T2 = (h2, r2, t2)

[0022] Wherein, T2 is the product triple knowledge model, h2 is the product composition, r2 is the product function, and t2 is the product working principle.

[0023] Preferably, the product reliability knowledge chain is:

[0024] L = [T1, T2] = [r1, t1, h1(h2), r2, t2]

[0025] Wherein, L is the product reliability knowledge chain.

[0026] Preferably, the graph evaluation factors include one or more of product type, product failure mode, product function, product composition, and failure determination.

[0027] Preferably, orthogonal verification is performed according to the multiple graph evaluation factors and the preset factor level parameters, specifically including:

[0028] Configure multiple factor levels for each selected graph evaluation factor respectively;

[0029] Let different factor levels of different graph evaluation factors cross-pair and perform verification, and calculate the signal-to-noise ratio;

[0030] After traversing all the cross-pairs, output the factor level pair with the highest signal-to-noise ratio as the test set.

[0031] Preferably, the calculation process of the signal-to-noise ratio includes:

[0032]

[0033] where X i is the coverage rate of the i-th sample, μ1 is the average coverage rate, and n is the number of samples;

[0034]

[0035] where Y i is the correct rate of the i-th sample, μ2 is the average correct rate, and n is the number of samples.

[0036] An intelligent generation system for a fault mode library of a new energy electric vehicle, comprising:

[0037] A triple extraction module, configured to construct triples to clarify product information and the mapping relationship between product functions and fault modes;

[0038] A graph generation module, configured to construct a knowledge graph according to the product information and the mapping relationship between product functions and fault modes;

[0039] A graph correction module, configured to sample according to the knowledge graph to confirm a test set; and correct the graph according to the test results of the test set;

[0040] A fault mode generation module, configured to generate a fault mode library according to the knowledge graph.

[0041] Through the above technical solutions, compared with the prior art, the present invention discloses an intelligent generation method and system for a fault mode library of a new energy electric vehicle, constructs a knowledge model including product triples and reliability triples, forms a complete product reliability knowledge chain, automatically generates a knowledge graph of faults of a new energy electric vehicle, and innovatively uses the Taguchi method for orthogonal experimental design, quantitatively analyzes the coverage rate and correctness index of the knowledge graph, and then continuously optimizes the knowledge graph through an iterative correction mechanism, and finally generates a structured fault mode library with high coverage rate and accuracy. The present invention realizes the full-process intelligence from data modeling, knowledge graph generation to quality evaluation, significantly improves the construction efficiency and quality of the fault mode library, and provides reliable technical support for the intelligent diagnosis and maintenance of new energy electric vehicles. Description of the Drawings

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on the provided drawings.

[0043] Figure 1 Schematic diagram of an intelligent generation method for a fault mode library of a new energy electric vehicle provided by the present invention. Detailed implementation manners

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0045] Embodiment 1

[0046] The embodiment of the present invention discloses an intelligent generation method for a fault mode library of a new energy electric vehicle. Refer to Figure 1 As shown, to illustrate the above method, two products are taken as examples below, which is an incomplete data set. The method of this embodiment includes the following steps:

[0047] S1: Construct a triple knowledge model

[0048] S11: Construct a reliability triple knowledge model T1=(h1, r1, t1). The composition products, fault modes of products, fault determination results of products, etc. of the new energy electric vehicle are described through the reliability triple. The fault modes of products include functional fault modes and potential fault modes, and the fault determination results of products include functional fault results and potential fault results.

[0049]

[0050] Among them, represents the battery cell, represents that the fault mode is that the voltage output of the 38th and 39th battery cells of the 2nd CSC is 0, represents that the fault determination is a battery cell fault; represents the vehicle control unit, represents that the fault mode is no control signal output, represents that the fault determination is a control unit fault.

[0051] S12: Construct the product triple knowledge model T2 = (h2, r2, t2). Describe information such as the product composition, product functions, and product working principles of new energy electric vehicles through product triples. The functions of the product provide a functional benchmark for subsequent failure mode analysis, and the working principles of the product explain the mechanisms for the product to achieve its functions, which are used to support the logical derivation of fault determination.

[0052]

[0053] Among them, represents the battery cell, represents that the product function is to output voltage, represents that the working principle is to charge and discharge through electrochemical reactions; represents the vehicle control unit, represents that the product function is to output control signals, represents that the working principle is to communicate with each module through the CAN bus.

[0054] S13: Construct the product reliability knowledge chain L. Connect the product triple and the reliability triple through the function-fault correspondence relationship to form the product reliability knowledge chain.

[0055]

[0056] Among them, h1 is the merged node battery cell (product name) of the reliability triple and the product triple, and h2 is the merged node vehicle control unit (product name).

[0057] S2: Generate the fault knowledge graph of new energy vehicles.

[0058] S21: Prepare the sliced text of the fault knowledge graph of new energy vehicles; the text includes: customer feedback information, product fault mode information, product repair and handling information, and product fault cause analysis information.

[0059] An example of the sliced text of the fault knowledge graph is as follows:

[0060] Sliced text 1: "The customer feedback that the vehicle reports a battery system fault and the driving speed is limited. Previously, the collected driving data was fed back to the technical staff for analysis, and it was found that the 38th and 39th battery cells of the 2nd CSC were faulty. After the vehicle information was reported to the technical staff, it was required to replace the new battery box for handling, and the fault was eliminated after replacement. Further detection found that the battery cell fault was caused by long-term overloading. The technical staff suggested optimizing the battery management algorithm and regularly maintaining the battery cell status. In subsequent tests, the new battery box ran stably and no speed limit problems occurred again."

[0061] Sliced text 2: "The customer reported that the warning light of 'Electronic Stability Control System (ESC) failure' came on on the instrument panel during vehicle driving. At the same time, the ABS function was not triggered during emergency braking, and the braking distance was significantly extended. The technician checked and found that the right front wheel speed sensor was oxidized due to water ingress, resulting in intermittent signal transmission and triggering the protective lock of the ESC system. Further detection found that the sensor plug was severely oxidized and the internal metal contacts were corroded. A new sensor needed to be replaced and the wiring harness cleaned. After the repair, the test showed that the ABS function returned to normal and the ESC warning light went out. The summary of the failure reason was that the sensor plug was not cleaned in time after water ingress, resulting in the accumulation of oxidation problems. It is recommended that the customer avoid long-term parking of the vehicle after water ingress and regularly check the protection status of the chassis electronic components. In the subsequent upgrade plan, it is planned to add a waterproof rubber sleeve to the sensor plug to improve the sealing performance."

[0062] S22: Generate text slices, divide the long text into different small texts with a fixed byte size of 512 plus an overlapping block size of 50. The fixed byte refers to a data unit with a predetermined size used in data processing, and the overlapping block refers to dividing the data into blocks with partially overlapping content in data processing.

[0063] Generated text slice 1: "The customer reported that the vehicle reported a battery system failure and speed limit during driving. Previously, the driving data was collected and fed back to the technician for analysis, and it was found that the 38th and 39th battery cells of the No. 2 CSC were faulty. After the vehicle information was reported to the technician, it was required to replace the new battery box for processing, and the failure was eliminated after replacement."

[0064] Generated text slice 2: "The customer reported that the warning light of 'Electronic Stability Control System (ESC) failure' came on on the instrument panel during vehicle driving. At the same time, the ABS function was not triggered during emergency braking, and the braking distance was significantly extended. The technician checked and found that the right front wheel speed sensor was oxidized due to water ingress, resulting in signal transmission interruption and triggering the protective lock of the ESC system."

[0065] S23: Knowledge graph generation. According to the definition of the product reliability knowledge chain, set the product name, failure mode, and product function prompt words. The prompt words include instructions and references to some samples. The large language model generates reliability triples and product triples respectively based on the text slices, connects the reliability triples and product triples with corresponding relationships to form a product reliability knowledge chain, and splices the knowledge chain to generate a knowledge graph.

[0066] Prompt word settings:

[0067] Product name: battery cell, vehicle control unit

[0068] Failure mode: no voltage output, no control signal output

[0069] Product function: output voltage, output control signal

[0070] Generate an example of a knowledge graph:

[0071]

[0072] S3: Taguchi method sampling test

[0073] The graph generation given in this case is partial results. To further illustrate step S3, the number of products included in the graph is given below: 400 types, and the number of failure modes is: 1500 types.

[0074] S31: Define evaluation metrics

[0075] Randomly select product types at different levels. Among the selected product types, randomly select product functions at different levels as the total set under this factor level combination. Through the graph query function, search for all the selected product types and their related product functions in the knowledge graph, count their quantities, and judge their correctness.

[0076] Each factor level combination is randomly sampled 10 times, and the average value of the calculated coverage rate and correctness rate is taken as the coverage rate and correctness rate under this factor level combination.

[0077] Coverage rate: The ratio of the correctly extracted product types and product functions in the knowledge graph to the total product types and product functions.

[0078] Correctness: The accuracy of the extracted information in the knowledge graph.

[0079] It is stipulated that the knowledge graph is valid when the coverage rate reaches more than 85% and the correctness rate reaches more than 92%.

[0080] S32: Select factors and levels.

[0081] Factors: Product type, product failure mode

[0082] Product types: 150, 170, 190, 210, 230

[0083] Product failure modes: 600, 650, 700, 750, 800

[0084] S33: Orthogonal design method

[0085] Select 2 factors, each factor has 5 levels, and use the L25(5^2) orthogonal table for orthogonal verification.

[0086] S34: Calculate the signal-to-noise ratio

[0087] For each experimental combination, calculate the signal-to-noise ratio (S / N ratio) of coverage and correctness: The maximum signal-to-noise ratio is obtained when the number of product types is 230 and the number of product failure modes is 700. At this time, the signal-to-noise ratios are -1.09116 and -1.01472 respectively.

[0088] S35: Analysis Results

[0089] By comparing and analyzing the larger-the-better signal-to-noise ratios obtained for each experimental combination, the results are as follows: The combination of 230 product types and 700 product failure modes is the optimal factor level combination. At this time, the volatility of the test results is small and the most stable. This factor level combination can better represent the complete knowledge graph and is determined as the test set. The calculated coverage rate of the knowledge graph reaches 90.4151%, and the correctness reaches 86.0812%. It is determined that the correctness of the knowledge graph is insufficient.

[0090] S4: Knowledge Graph Correction.

[0091] 1. Construct a question-and-answer training set according to the triple knowledge model.

[0092] 2. Use a script to automatically compare the system output with the standard answer and identify incorrect triple information in the input data.

[0093] 3. Correct the incorrect triple entity information in the training set to form correct triple entity information; generate a knowledge graph test set again according to step 3, and iteratively test the knowledge graph until the given correctness requirement is met.

[0094] During the text slicing process, the large language model only infers the failure mode as a short circuit in the sensor internal circuit based on "internal metal contact corrosion", rather than the actual cause of poor contact due to plug oxidation.

[0095] Before correction: For internal circuit short circuit, no control signal output, ESC system mis-triggered, ABS function failure

[0096] After correction: For poor contact caused by plug oxidation, For signal transmission interruption, ESC system protection lock

[0097] Generate a knowledge graph test set again according to step 3, and iteratively test the knowledge graph until the given correctness requirement is met.

[0098] S5: Generation of Failure Mode Library

[0099] Input the prompt into the large language model to obtain the output of the failure mode library in tabular form.

[0100] Structured Prompt: According to the knowledge graph, list the fault mode library in tabular form, including the content product name, product function, and fault mode

[0101] The large language model generates a fault mode library for new energy electric vehicles based on the current input, as shown in the following table:

[0102] Table 1: Fault Mode Library of New Energy Electric Vehicles

[0103] Product Name Product Function Fault Mode Battery System Output Voltage No. 2 CSC Cell Fault Vehicle Control System Output Control Signal Signal Transmission Interruption, ESC System Protection Lockout

[0104] Example 2

[0105] Based on the same inventive concept, an embodiment of the present invention discloses an intelligent generation system for a fault mode library of a new energy electric vehicle, including:

[0106] A triple extraction module for constructing triples to clarify the product information and the mapping relationship between the product function and the fault mode;

[0107] A graph generation module for constructing a knowledge graph according to the product information and the mapping relationship between the product function and the fault mode;

[0108] A graph correction module for sampling according to the knowledge graph to confirm the test set; and correcting the graph according to the test results of the test set;

[0109] A fault mode generation module for generating a fault mode library according to the knowledge graph.

[0110] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the description in the method part.

[0111] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligently generating a fault mode library of a new energy electric vehicle, characterized in that, It includes the following steps: Construct a triple knowledge model to clarify product information and the mapping relationship between product functions and failure modes; Generate a knowledge graph based on the triple knowledge model; Select multiple graph evaluation factors based on the node types in the knowledge graph; Perform orthogonal verification according to the multiple graph evaluation factors and preset factor level parameters, and output the factor level pair corresponding to the optimal signal-to-noise ratio as the optimal test set; correct the knowledge graph based on the optimal test set; Generate a failure mode library based on the corrected knowledge graph.

2. The intelligent generation method of the fault mode library of a new energy electric vehicle according to claim 1, wherein, Construct a triple knowledge model, specifically including: Construct a reliability triple model to describe the component products, failure modes of the products, and failure determination results of new energy electric vehicles through the reliability triple; Construct a product triple knowledge model to describe the product functions and corresponding product working principles of each product component in new energy electric vehicles through the product triple knowledge model; Construct a product reliability knowledge chain to connect the reliability triple and the product triple according to the mapping relationship between function and failure through the product reliability knowledge chain.

3. The intelligent generation method of the fault mode library for a new energy electric vehicle according to claim 2, wherein The reliability triple knowledge model is: T1 = (h1, r1, t1) where T1 is the reliability triple knowledge model, h1 is the product name, r1 is the failure mode, and t1 is the failure determination.

4. The intelligent generation method of a fault mode library for a new energy electric vehicle according to claim 2, wherein, The product triple knowledge model is: T2 = (h2, r2, t2) where T2 is the reliability triple knowledge model, h2 is the product name, r2 is the failure mode, and t2 is the failure determination.

5. The intelligent generation method of a fault mode library for a new energy electric vehicle according to claim 2, characterized in that The product reliability knowledge chain is: L = [T1, T2] = [r1, t1, h1(h2), r2, t2] where L is the product reliability knowledge chain.

6. The intelligent generation method of the fault mode library of a new energy electric vehicle according to claim 1, characterized in that, The graph evaluation factors include one or more of product type, product failure mode, product function, product composition, and failure determination.

7. A method for intelligently generating a fault mode library of a new energy electric vehicle according to claim 1, characterized in that, Performing orthogonal verification according to the multiple graph evaluation factors and preset factor level parameters specifically includes: Configure multiple factor levels for each selected graph evaluation factor respectively; Let different factor levels of different graph evaluation factors cross-group and pair for verification, and calculate the signal-to-noise ratio; after traversing all the cross-group pairs, output the factor level pair with the highest signal-to-noise ratio as the training set.

8. A method for intelligently generating a fault mode library of a new energy electric vehicle according to claim 7, characterized in that, The calculation process of the signal-to-noise ratio includes: where X i is the coverage rate of the i-th sample, μ1 is the average coverage rate, and n is the number of samples; Among them, Y i is the correct rate of the i-th sample, μ2 is the average correct rate, and n is the number of samples.

9. An intelligent generation system for a fault mode library of a new energy electric vehicle, characterized in that, Adopt any one of the generation methods described in claims 1-8, including: A triple extraction module for constructing triples to clarify product information and the mapping relationship between product functions and failure modes; A graph generation module for constructing a knowledge graph according to the product information and the mapping relationship between product functions and failure modes; A graph correction module for sampling according to the knowledge graph to confirm the test set; correcting the graph according to the test results of the test set; A failure mode generation module for generating a failure mode library according to the knowledge graph.

Images