New energy vehicle fault diagnosis method based on multi-memory fusion and related equipment

Abstract

This application discloses a method and related equipment for fault diagnosis of new energy vehicles based on multi-memory fusion, which can be applied to the field of new energy vehicle technology. This application constructs a multi-layered structured memory system, and after obtaining the target diagnosis stage, it retrieves valid rule text, valid work order information, and valid triplet information by combining the corresponding mapping function and the multi-layered structured memory system. These are then structurally concatenated with spoken natural language query information to obtain parsable prompt text, which is input into a preset large language model and combined with spoken natural language query information and the target diagnosis stage to predict the root cause of the fault. This effectively improves the accuracy and adaptability of fault diagnosis for new energy vehicles. Subsequently, based on the actual root cause determined by the user according to the predicted root cause, the contextual memory bank and semantic memory bank are updated, thereby effectively improving the continuous evolution capability of the fault diagnosis process for new energy vehicles.

Description

Technical Field

[0001] This application relates to the field of new energy vehicle technology, and in particular to a new energy vehicle fault diagnosis method and related equipment based on multi-memory fusion. Background Technology

[0002] In related technologies, new energy vehicles have become the mainstream of the automotive industry. Their core components and technical architecture differ significantly from traditional fuel vehicles, making the causes of failures more complex and leading to frequent new failure scenarios. User descriptions of new energy vehicle failures are often in colloquial, non-standardized natural language, resulting in low interpretability of existing fault diagnosis methods that rely on static knowledge bases or general "black box" models. Furthermore, existing fault diagnosis methods lack effective structured organization of heterogeneous repair knowledge, have rigid retrieval strategies, and lack the ability to accumulate experience, leading to low diagnostic accuracy and poor scenario adaptability, thus failing to meet the precise diagnostic needs of complex new energy vehicle failures.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this application is to propose a fault diagnosis method and related equipment for new energy vehicles based on multi-memory fusion, which can effectively improve the accuracy, adaptability and continuous evolution capability of fault diagnosis for new energy vehicles.

[0005] To achieve the above objectives, one aspect of this application proposes a fault diagnosis method for new energy vehicles based on multi-memory fusion, the method comprising the following steps: A multi-layered structured memory system is constructed, which includes a procedural memory bank, a contextual memory bank, and a semantic memory bank. The procedural memory bank is used to store standardized maintenance rules and process knowledge, the contextual memory bank is used to store historical maintenance work orders, and the semantic memory bank stores historical fault entities and historical triplet information formed by historical semantic relationships between historical fault entities through a knowledge graph. Obtain spoken natural language query information for the current fault scenario; Based on the semantic memory, the colloquial natural language query information is semantically parsed to obtain the target diagnosis stage; Based on the target diagnostic stage and the procedural usage mapping function, a procedural retrieval is performed in the procedural memory to obtain valid rule text; Based on the target diagnostic stage and the contextual usage mapping function, a contextual retrieval is performed in the contextual memory to obtain valid work order information; Based on the target diagnostic stage and semantic usage mapping function, semantic retrieval is performed in the semantic memory to obtain effective triplet information; The valid rule text, the valid work order information, and the valid triplet information are structurally concatenated with the colloquial natural language query information to obtain a parsable prompt text. The parsable prompt text is input into a preset large language model, and the fault prediction is performed by combining the colloquial natural language query information and the target diagnosis stage to obtain the predicted fault root cause. The contextual memory is updated based on the actual root causes of failures determined by the user according to the predicted root causes, and the semantic memory is updated based on the actual root causes of failures determined by the user according to the predicted root causes.

[0006] In some embodiments, the step of semantically parsing the colloquial natural language query information based on the semantic memory to obtain the target diagnosis stage includes: The colloquial natural language query information is converted into a query semantic vector; Calculate the first cosine similarity between the query semantic vector and all historical fault entities in the semantic memory; The target diagnostic stage is determined based on the query semantic vector or the first cosine similarity, combined with multi-dimensional logical judgment conditions.

[0007] In some embodiments, the step of performing a program retrieval in the procedural memory based on the target diagnostic stage and the procedural usage mapping function to obtain valid rule text includes: Calculate the second cosine similarity between the query semantic vector and the rule vectors in the procedural memory; The number of effective rules for the target is determined based on the target diagnostic stage and the procedural usage mapping function; Based on the second cosine similarity, a number of rules equal to the target number of valid rules are retrieved from the procedural memory as the valid rule text.

[0008] In some embodiments, the step of performing scenario retrieval in the scenario memory based on the target diagnostic stage and scenario-based usage mapping function to obtain valid work order information includes: The target number of valid work orders is determined based on the target diagnostic stage and the contextual usage mapping function. Based on the query semantic vector, a preliminary search is performed in the contextual memory to obtain a preliminary candidate set of historical work order information with a number equal to the target number of valid work orders; Calculate the third cosine similarity between the query semantic vector and each historical work order vector in the initial candidate set; Obtain the vehicle meta-feature information corresponding to the query semantic vector, wherein the vehicle meta-feature information includes vehicle model information and current vehicle mileage; The evaluation score for each historical work order in the initial inspection candidate set is calculated based on the third cosine similarity and the vehicle meta-feature information. Valid work order information is determined from the initial candidate set based on the evaluation score.

[0009] In some embodiments, the step of performing semantic retrieval in the semantic memory based on the target diagnostic stage and the semantic usage mapping function to obtain valid triplet information includes: The target associated entity is determined from the semantic memory based on the first cosine similarity. The historical triplet information corresponding to the target associated entity is obtained from the semantic memory as the initial semantic retrieval result; The number of effective triples for the target is determined based on the target diagnostic stage and semantic usage mapping function; Valid triple information is determined from the initial semantic retrieval results based on the target number of valid triples.

[0010] In some embodiments, updating the contextual memory based on the actual root cause of the failure determined by the user according to the predicted root cause includes: The predicted fault root cause is combined with the colloquial natural language query information corresponding to the vehicle meta-feature information to form a new scenario memory work order information. The semantic embedding vector corresponding to the new scenario memory work order information is generated by the text encoder. The new scenario memory work order information is added to the scenario memory library, the semantic embedding vector corresponding to the new scenario memory work order information is added to the vector index, and the mapping relationship between the new scenario memory work order information and the vehicle meta-feature vector is established synchronously in the metadata association table.

[0011] In some embodiments, updating the semantic memory based on the actual root cause of the failure determined by the user according to the predicted root cause includes: If the actual root cause of the failure is determined to belong to a new entity not included in the semantic memory, the actual root cause of the failure is added to the entity set in the semantic memory. New triplet information is constructed based on the actual root causes of the faults; The semantic memory is updated based on the new triplet information.

[0012] To achieve the above objectives, another aspect of this application proposes a fault diagnosis device for new energy vehicles based on multi-memory fusion, the device comprising: The first module is used to construct a multi-layered structured memory system, which includes a procedural memory bank, a contextual memory bank, and a semantic memory bank. The procedural memory bank is used to store standardized maintenance rules and process knowledge, the contextual memory bank is used to store historical maintenance work orders, and the semantic memory bank stores historical fault entities and historical triplet information formed by historical semantic relationships between historical fault entities through a knowledge graph. The second module is used to obtain spoken natural language query information for the current fault scenario; The third module is used to perform semantic parsing on the colloquial natural language query information based on the semantic memory to obtain the target diagnosis stage; The fourth module is used to perform program retrieval in the program memory according to the target diagnostic stage and the program usage mapping function to obtain valid rule text; The fifth module is used to perform scenario retrieval in the scenario memory based on the target diagnostic stage and scenario-based usage mapping function to obtain valid work order information; The sixth module is used to perform semantic retrieval in the semantic memory based on the target diagnostic stage and the semantic usage mapping function to obtain effective triplet information; The seventh module is used to structurally concatenate the valid rule text, the valid work order information, and the valid triplet information with the colloquial natural language query information to obtain parsable prompt text; The eighth module is used to input the parsable prompt text into a preset large language model, and combine it with the colloquial natural language query information and the target diagnosis stage to perform fault prediction and obtain the predicted fault root cause. The ninth module is used to update the contextual memory based on the actual root causes of failures determined by the user according to the predicted root causes of failures, and to update the semantic memory based on the actual root causes of failures determined by the user according to the predicted root causes of failures.

[0013] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0014] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods described above.

[0015] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method.

[0016] The embodiments of this application include at least the following beneficial effects: This application provides a new energy vehicle fault diagnosis method and related equipment based on multi-memory fusion. This solution constructs a multi-layered structured memory system including a procedural memory, a contextual memory, and a semantic memory. Based on the semantic memory, it performs semantic parsing on the spoken natural language query information of the current fault scenario to obtain the target diagnosis stage. Then, based on the target diagnosis stage, it performs retrieval using the corresponding mapping function and the multi-layered structured memory system to obtain valid rule text, valid work order information, and valid triplet information. The valid rule text, valid work order information, and valid triplet information are then structurally concatenated with the spoken natural language query information to obtain a parsable prompt text. This parsable prompt text is input into a preset large language model and combined with the spoken natural language query information and the target diagnosis stage to predict the root cause of the fault. This effectively improves the accuracy and adaptability of new energy vehicle fault diagnosis. Then, based on the actual root cause determined by the user according to the predicted root cause, the contextual memory and semantic memory are updated, thereby effectively improving the continuous evolution capability of the new energy vehicle fault diagnosis process and meeting the precise diagnosis needs of complex faults in new energy vehicles. Attached Figure Description

[0017] Figure 1 This is a flowchart of a new energy vehicle fault diagnosis method based on multi-memory fusion provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of the new energy vehicle fault diagnosis device based on multi-memory fusion provided in the embodiments of this application; Figure 3 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0019] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0020] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0022] In related technologies, new energy vehicles have become the mainstream of the automotive industry. Their core components and technical architecture differ significantly from traditional fuel vehicles, making the causes of failures more complex and leading to frequent new failure scenarios. User descriptions of new energy vehicle failures are often in colloquial, non-standardized natural language, resulting in low interpretability of existing fault diagnosis methods that rely on static knowledge bases or general "black box" models. Furthermore, existing fault diagnosis methods lack effective structured organization of heterogeneous repair knowledge, have rigid retrieval strategies, and lack the ability to accumulate experience, leading to low diagnostic accuracy and poor scenario adaptability, thus failing to meet the precise diagnostic needs of complex new energy vehicle failures.

[0023] In view of this, this application provides a new energy vehicle fault diagnosis method and related equipment based on multi-memory fusion, which can effectively improve the accuracy, adaptability and continuous evolution capability of new energy vehicle fault diagnosis, and meet the needs of accurate diagnosis of complex faults in new energy vehicles.

[0024] The multi-memory fusion-based fault diagnosis method for new energy vehicles provided in this application relates to the field of new energy vehicle technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the multi-memory fusion-based fault diagnosis method for new energy vehicles, but is not limited to the above forms.

[0025] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0026] The embodiments of this application will be described in detail below with reference to the accompanying drawings: Figure 1 This is an optional flowchart of the new energy vehicle fault diagnosis method based on multi-memory fusion provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S190: Step S110: Construct a multi-layer structured memory system, which includes a procedural memory bank, a contextual memory bank, and a semantic memory bank. The procedural memory bank is used to store standardized maintenance rules and process knowledge, the contextual memory bank is used to store historical maintenance work orders, and the semantic memory bank stores historical fault entities and historical triplet information formed by historical semantic relationships between historical fault entities through a knowledge graph. Step S120: Obtain spoken natural language query information for the current fault scenario; Step S130: Perform semantic parsing on the colloquial natural language query information based on the semantic memory to obtain the target diagnosis stage; Step S140: Based on the target diagnostic stage and the procedural usage mapping function, perform a program retrieval in the procedural memory to obtain the effective rule text; Step S150: Based on the target diagnostic stage and the contextual usage mapping function, perform contextual retrieval in the contextual memory to obtain valid work order information; Step S160: Based on the target diagnostic stage and semantic usage mapping function, perform semantic retrieval in the semantic memory to obtain effective triplet information; Step S170: Concatenate the valid rule text, valid work order information, and valid triplet information with the colloquial natural language query information to obtain parsable prompt text; Step S180: Input the parsable prompt text into the preset large language model, and combine it with the colloquial natural language query information and the target diagnosis stage to perform fault prediction and obtain the predicted fault root cause; Step S190: Update the contextual memory and semantic memory based on the actual root causes of failures determined by the user according to the predicted root causes of failures.

[0027] It is understood that the multi-layered structured memory system in this embodiment is a structured knowledge storage architecture that separates, organizes, and indexes heterogeneous automotive fault diagnosis knowledge according to cognitive types, laying the foundation for subsequent accurate and efficient retrieval.

[0028] Specifically, procedural memory This is used to store standardized maintenance rules and process knowledge, such as Standard Operating Procedures (SOPs), Fault Tree Analysis (FTA) flowcharts, and Failure Mode and Effects Analysis (FMEA) entries. The corresponding formulas are as follows: ; In the formula, For the i-th rule data, The fault is uniquely identified by ID. For specific faults (by The detailed text of the specific maintenance rules or standardized procedures corresponding to ) This represents the total number of rules in the procedural memory.

[0029] This embodiment uses a domain-fine-tuned text encoder. Generate the rule vector corresponding to the i-th standardized maintenance rule or standardized process knowledge text: ; In the formula, A text encoder for domain fine-tuning, which will convert the text... Convert to a semantic embedding vector of dimension d .

[0030] Simultaneously, an inner product vector index is constructed using the FAISS library to semantically embed vectors of all standardized maintenance rules or standardized process knowledge texts. Store in the index for subsequent fast semantic matching retrieval: ; In the formula, This indicates a completed procedural memory index. This represents a vector indexing operation based on the inner product metric. This represents the input data, i.e., all A set of semantic embedding vectors corresponding to the knowledge text of a standardized maintenance rule or standardized process.

[0031] Episodic memory bank The formula for storing historical maintenance work orders is as follows: ; In the formula, For the j-th historical maintenance work order data, This refers to the fault description text provided by the user or technician in the work order. This is the root cause of the fault in this work order. The meta-feature vector... It is constructed as a two-dimensional combined feature containing two key attributes, specifically composed of the vehicle model identifier corresponding to the historical repair work order and the vehicle mileage value at the time of the fault.

[0032] Encoding vector of fault description text for: ; Build an index of the fault description text and associate it with metadata: ; In the formula, This indicates the completed contextual memory index. This represents a vector indexing operation based on the inner product metric. This represents the input data, i.e., all A set of semantic embedding vectors corresponding to each historical maintenance work order.

[0033] Simultaneously construct a metadata association mapping table Each work order index identifier in the context memory index library Its corresponding meta-feature vector (Including vehicle model and mileage information) Establish a one-to-one anchoring relationship to support hybrid filtering and reordering based on physical features in the subsequent retrieval process.

[0034] Semantic memory The relationships between entities describing faults, components, etc., in fault diagnosis are stored in the form of a knowledge graph. ; In the formula, For the "head entity-relation-tail entity" triple information of a knowledge graph, entity set Includes information such as fault symptoms and components; relationship set It contains semantic relationships such as "cause" and "related components".

[0035] It is understandable that, after constructing a multi-layered structured memory system for the corresponding field of new energy vehicles, this embodiment obtains colloquial natural language query information for the current fault scenario, and performs semantic parsing on the colloquial natural language query information based on a semantic memory to obtain the target diagnostic stage. Specifically, this embodiment can convert the colloquial natural language query information into a query semantic vector, calculate the first cosine similarity between the query semantic vector and all historical fault entities in the semantic memory, and determine the target diagnostic stage based on the query semantic vector or the first cosine similarity combined with multi-dimensional logical judgment conditions, thereby providing a basis for subsequent hierarchical adaptive retrieval. In this embodiment, the query semantic vector corresponding to the input colloquial natural language query information... as follows: ; The first cosine similarity is calculated by comparing the query semantic vector with the knowledge image in the semantic memory, using the maximum cosine similarity. ; In the formula, For faulty entities in a knowledge graph, this formula calculates the dot product of the query vector and the entity vector. They are then divided by their respective L2 norms to quantify the semantic association between the query semantic vector and the historical fault entity.

[0036] This embodiment combines multi-dimensional logical judgment conditions to determine the target diagnostic stage. The process is as follows: (1) If a user’s clear feedback confirmation behavior is detected based on the query semantic vector, it is directly determined as a fine-grained investigation as the target diagnosis stage; (2) If the number of characters in the colloquial natural language query information is less than 20 and there is a dialogue history, it is considered as a non-substantive interaction and is judged as the preliminary positioning stage as the target diagnosis stage. (3) If the calculated maximum semantic similarity (first cosine similarity) is lower than the preset threshold ( This indicates that the problem has not been included in the existing knowledge base, and it is identified as a new type of fault stage as the target diagnostic stage. (4) If none of the above conditions are met, the system will default to the initial positioning stage and perform a regular breadth search.

[0037] It is understood that this embodiment can perform semantic matching retrieval on the procedural memory and perform "vector initial detection + meta-feature weighted ranking" retrieval on the contextual memory according to the target diagnosis stage, thereby achieving hierarchical adaptive knowledge retrieval. Specifically, when performing semantic matching retrieval on the procedural memory, this embodiment can calculate the second cosine similarity between the query semantic vector and the rule vectors in the procedural memory, and after determining the number of effective rules for the target based on the target diagnosis stage and the procedural usage mapping function, retrieve the number of rules equal to the number of effective rules for the target from the procedural memory based on the second cosine similarity as effective rule text. This embodiment targets the procedural memory. FMEA rules and standard maintenance procedures are used to accurately retrieve standardized operational knowledge related to the current fault through semantic matching. This includes information retrieval in conversational natural language. The domain-adjusted text encoder converts the text into a query semantic vector: ; The second cosine similarity is calculated by comparing the query semantic vector with the rule vectors in the procedural memory. The formula is: ; In the formula, This is the embedding vector for the rule.

[0038] Simultaneously, based on the target diagnostic stage Determining the target effective rule number using a programmatic usage mapping function : ; Select in descending order of second cosine similarity. The programmatic retrieval results are obtained by combining the rules as valid rule texts. : ; In the formula, .

[0039] Understandably, when performing contextual retrieval in the contextual memory, this embodiment can determine the target number of valid work orders based on the target diagnostic stage and the contextual usage mapping function. A preliminary search is then performed in the contextual memory based on the query semantic vector to obtain a preliminary candidate set of historical work order information equal to the target number of valid work orders. The third cosine similarity between the query semantic vector and each historical work order vector in the preliminary candidate set is then calculated. After obtaining the vehicle meta-feature information corresponding to the query semantic vector, including vehicle model information and current vehicle mileage, an evaluation score is calculated for each historical work order information in the preliminary candidate set based on the third cosine similarity and the vehicle meta-feature information. Valid work order information is then determined from the preliminary candidate set based on the evaluation score. Specifically, the target number of valid work orders... The formula for determining it is as follows: ; Initial Candidate Set The expression is as follows: ; In the formula, , This represents the total number of work orders in the contextual memory bank.

[0040] The evaluation score of the j-th historical work order information The following formula: ; In the formula, The third cosine similarity; Binary factors are matched to vehicle models, and the models are compared with those in historical work orders. Meta-feature vector of the currently diagnosed vehicle Vehicle information Assign values ​​based on whether they match; Defined as a mileage proximity coefficient, this coefficient is a normalized value whose magnitude is inversely proportional to the absolute value of the difference between "historical work order mileage" and "current vehicle mileage". For the initial inspection candidate set... Calculate the evaluation score for each historical work order. , The historical work order information with the highest evaluation score is selected as the valid work order information to form the contextual search results. : ; Understandably, when performing semantic retrieval in the semantic memory, this embodiment can determine the target associated entity from the semantic memory based on the first cosine similarity, then retrieve the historical triplet information corresponding to the target associated entity from the semantic memory as the initial semantic retrieval result. After determining the number of effective triplets based on the target diagnostic stage and the semantic usage mapping function, effective triplet information is then determined from the initial semantic retrieval result based on the number of effective triplets. Specifically, this embodiment can retrieve "entity-relationship" triplets related to the current fault based on the determined target associated entity, providing a semantic association basis of "fault phenomenon-component-root cause" for diagnosis. The target associated entity can be the entity with the highest first cosine similarity, as shown in the following formula: ; In the formula, For the head entity, retrieve the semantic memory. All of the above triples , This constitutes the initial semantic search results.

[0041] Target number of effective triplets The following formula: ; This embodiment can retrieve results from the initial semantic retrieval results according to relation priority. Each triplet is considered as valid triplet information. The final semantic retrieval results are composed of: ; As can be seen from the above, the number of targets for each library varies during the target diagnosis phase. The specific configuration is as follows: (1) Preliminary localization stage: The system focuses on breadth-first search, with the following configuration parameters: , , That is, it mainly relies on general maintenance rules and semantic associations of knowledge graphs, supplemented by a small number of historical cases; (2) Detailed screening stage: The system focuses on in-depth comparison, and the configuration parameters are as follows: , , At this point, the number of historical work orders retrieved should be increased significantly to take advantage of detailed troubleshooting experience for similar faults. (3) Novel Fault Phase: The system focuses on exploratory reasoning, with the following configuration parameters: , , At this point, the reliance on existing fixed rules is reduced, and instead, potential fault clues are found by increasing the weight of knowledge graph triples and similar cases.

[0042] It is understandable that this embodiment obtains the search results. Valid rule text , A valid work order containing a fault description and the actual root cause. and Group of effective triplet information Then, combine the information with colloquial natural language queries. The structure is spliced ​​to obtain a structured prompt stream that combines rule logic, historical experience and semantic association as parsable prompt text, so that the Large Language Model (LLM) can read it and generate parsable diagnostic conclusions.

[0043] Specifically, this embodiment uses an LLM model that is domain-adapted to parsable prompt text input, leveraging the model's natural language understanding and reasoning capabilities to generate diagnostic results that fit the current fault scenario, and finally outputs parsable diagnostic conclusions as a prediction of the root cause of the fault. for: ; In the formula, Represents LLM, Represents colloquial natural language query information, for , This indicates that the prompt text can be parsed.

[0044] It is understandable that, after obtaining the predicted root cause of the fault, this embodiment can prompt relevant personnel to determine the actual root cause of the fault at the corresponding location of the new energy vehicle, and then incremental writing can be performed. , , The closed-loop update and knowledge evolution continuously improve the accuracy and coverage of subsequent diagnoses. Specifically, this embodiment updates the contextual memory bank. At that time, the colloquial natural language query information corresponding to the predicted root cause of the failure can be used. Vehicle meta-feature information Combined with the actual root cause of the failure Create new scenario memory work order information And add the new contextual memory work order information to the contextual memory bank. Simultaneously, a semantic embedding vector corresponding to the new context memory work order information is generated through a text encoder. Then, the semantic embedding vector corresponding to the new context memory work order information is used. Add to vector index In the meantime, new scenario memory work order information and vehicle meta-feature vectors are simultaneously established in the metadata association table. A one-to-one mapping relationship ensures that the new case can be accurately retrieved in the future.

[0045] It is understood that, in updating the semantic memory, this embodiment can automatically detect the true root cause of the fault. Does it exist in the current semantic entity set? In the middle, if the true root cause This refers to a new entity not yet included in the semantic memory, which will reveal the true root cause of the failure. Entity sets added to the semantic memory And based on the actual root cause of the failure, construct an entity containing "related header entities". —The root cause After obtaining the new triplet information, update the semantic memory based on the new triplet information. This expands the causal relationship between the failure phenomenon and the new root cause at the semantic level.

[0046] As can be seen from the above, the method of this application embodiment integrates procedural memory, contextual memory and semantic memory in cognitive science to construct an intelligent reasoning mechanism that dynamically integrates multiple memories, thereby realizing intelligent identification, root cause reasoning and operation and maintenance decision support for key system failures of new energy vehicles.

[0047] Please see Figure 2 This application also provides a fault diagnosis device for new energy vehicles based on multi-memory fusion, the device comprising: The first module is used to construct a multi-layered structured memory system, which includes a procedural memory bank, a contextual memory bank, and a semantic memory bank. The procedural memory bank is used to store standardized maintenance rules and process knowledge, the contextual memory bank is used to store historical maintenance work orders, and the semantic memory bank stores historical fault entities and historical triplet information formed by historical semantic relationships between historical fault entities through a knowledge graph. The second module is used to obtain spoken natural language query information for the current fault scenario; The third module is used to perform semantic parsing of colloquial natural language query information based on the semantic memory to obtain the target diagnosis stage; The fourth module is used to perform program retrieval in the program memory based on the target diagnostic stage and the program usage mapping function to obtain the effective rule text; The fifth module is used to perform scenario retrieval in the scenario memory based on the target diagnostic stage and scenario-based usage mapping function to obtain valid work order information; The sixth module is used to perform semantic retrieval in the semantic memory based on the target diagnostic stage and semantic usage mapping function to obtain effective triplet information; The seventh module is used to structurally concatenate the valid rule text, valid work order information, and valid triplet information with the colloquial natural language query information to obtain parsable prompt text; The eighth module is used to input parsable prompt text into a preset large language model, and combine it with the spoken natural language query information and the target diagnosis stage to perform fault prediction and obtain the predicted fault root cause. The ninth module is used to update the contextual memory and semantic memory based on the actual root causes of failures determined by the user according to the predicted root causes of failures.

[0048] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0049] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0050] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0051] Please see Figure 3 , Figure 3 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 310 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 320 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 320 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 320 and is called and executed by the processor 310 using the methods described in the embodiments of this application. Input / output interface 330 is used to realize information input and output; The communication interface 340 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 350 transmits information between various components of the device (e.g., processor 310, memory 320, input / output interface 330, and communication interface 340); The processor 310, memory 320, input / output interface 330 and communication interface 340 are connected to each other within the device via bus 350.

[0052] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0053] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0054] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0055] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0056] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0057] This application provides a new energy vehicle fault diagnosis method and related equipment based on multi-memory fusion. Through multi-memory fusion, stage-adaptive retrieval, and closed-loop update mechanisms, it can effectively improve the accuracy, adaptability, and continuous evolution capability of new energy vehicle fault diagnosis. Furthermore, this embodiment can accurately identify the diagnosis stage and dynamically invoke applicable knowledge to generate reliable suggestions traceable to maintenance rules, similar cases, or semantic definitions, effectively reducing the decision-making cost for maintenance personnel. Simultaneously, it automatically accumulates new fault experience, continuously expanding the coverage of complex scenarios, completely overcoming the core defects of traditional methods such as insufficient knowledge adaptability, rigid strategies, and inability to learn.

[0058] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0059] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0060] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0061] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0062] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0063] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0064] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0065] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0066] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0067] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0068] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A fault diagnosis method for new energy vehicles based on multi-memory fusion, characterized in that, The method includes the following steps: A multi-layered structured memory system is constructed, which includes a procedural memory bank, a contextual memory bank, and a semantic memory bank. The procedural memory bank is used to store standardized maintenance rules and process knowledge, the contextual memory bank is used to store historical maintenance work orders, and the semantic memory bank stores historical fault entities and historical triplet information formed by historical semantic relationships between historical fault entities through a knowledge graph. Obtain spoken natural language query information for the current fault scenario; Based on the semantic memory, the colloquial natural language query information is semantically parsed to obtain the target diagnosis stage; Based on the target diagnostic stage and the procedural usage mapping function, a procedural retrieval is performed in the procedural memory to obtain valid rule text; Based on the target diagnostic stage and the contextual usage mapping function, a contextual retrieval is performed in the contextual memory to obtain valid work order information; Based on the target diagnostic stage and semantic usage mapping function, semantic retrieval is performed in the semantic memory to obtain effective triplet information; The valid rule text, the valid work order information, and the valid triplet information are structurally concatenated with the colloquial natural language query information to obtain a parsable prompt text. The parsable prompt text is input into a preset large language model, and the fault prediction is performed by combining the colloquial natural language query information and the target diagnosis stage to obtain the predicted fault root cause. The contextual memory is updated based on the actual root causes of failures determined by the user according to the predicted root causes, and the semantic memory is updated based on the actual root causes of failures determined by the user according to the predicted root causes.

2. The method according to claim 1, characterized in that, The step of semantically parsing the colloquial natural language query information based on the semantic memory to obtain the target diagnosis stage includes: The colloquial natural language query information is converted into a query semantic vector; Calculate the first cosine similarity between the query semantic vector and all historical fault entities in the semantic memory; The target diagnostic stage is determined based on the query semantic vector or the first cosine similarity, combined with multi-dimensional logical judgment conditions.

3. The method according to claim 2, characterized in that, The step of performing a program retrieval in the procedural memory based on the target diagnostic stage and the procedural usage mapping function to obtain valid rule text includes: Calculate the second cosine similarity between the query semantic vector and the rule vectors in the procedural memory; The number of effective rules for the target is determined based on the target diagnostic stage and the procedural usage mapping function; Based on the second cosine similarity, a number of rules equal to the target number of valid rules are retrieved from the procedural memory as the valid rule text.

4. The method according to claim 2, characterized in that, The step of performing a scenario retrieval in the scenario memory based on the target diagnostic stage and scenario-based usage mapping function to obtain valid work order information includes: The target number of valid work orders is determined based on the target diagnostic stage and the contextual usage mapping function. Based on the query semantic vector, a preliminary search is performed in the contextual memory to obtain a preliminary candidate set of historical work order information with a number equal to the target number of valid work orders; Calculate the third cosine similarity between the query semantic vector and each historical work order vector in the initial candidate set; Obtain the vehicle meta-feature information corresponding to the query semantic vector, wherein the vehicle meta-feature information includes vehicle model information and current vehicle mileage; The evaluation score for each historical work order in the initial inspection candidate set is calculated based on the third cosine similarity and the vehicle meta-feature information. Valid work order information is determined from the initial candidate set based on the evaluation score.

5. The method according to claim 2, characterized in that, The step of performing semantic retrieval in the semantic memory based on the target diagnostic stage and semantic usage mapping function to obtain effective triplet information includes: The target associated entity is determined from the semantic memory based on the first cosine similarity. The historical triplet information corresponding to the target associated entity is obtained from the semantic memory as the initial semantic retrieval result; The number of effective triples for the target is determined based on the target diagnostic stage and semantic usage mapping function; Valid triple information is determined from the initial semantic retrieval results based on the target number of valid triples.

6. The method according to claim 4, characterized in that, The step of updating the contextual memory bank based on the actual root causes of failures determined by the user according to the predicted root causes includes: The predicted fault root cause is combined with the colloquial natural language query information corresponding to the vehicle meta-feature information to form a new scenario memory work order information. The semantic embedding vector corresponding to the new scenario memory work order information is generated by the text encoder. The new scenario memory work order information is added to the scenario memory library, the semantic embedding vector corresponding to the new scenario memory work order information is added to the vector index, and the mapping relationship between the new scenario memory work order information and the vehicle meta-feature vector is established synchronously in the metadata association table.

7. The method according to claim 1, characterized in that, The step of updating the semantic memory based on the actual root cause of the fault determined by the user according to the predicted root cause includes: If the actual root cause of the failure is determined to belong to a new entity not included in the semantic memory, the actual root cause of the failure is added to the entity set in the semantic memory. New triplet information is constructed based on the actual root causes of the faults; The semantic memory is updated based on the new triplet information.

8. A fault diagnosis device for new energy vehicles based on multi-memory fusion, characterized in that, The device includes: The first module is used to construct a multi-layered structured memory system, which includes a procedural memory bank, a contextual memory bank, and a semantic memory bank. The procedural memory bank is used to store standardized maintenance rules and process knowledge, the contextual memory bank is used to store historical maintenance work orders, and the semantic memory bank stores historical fault entities and historical triplet information formed by historical semantic relationships between historical fault entities through a knowledge graph. The second module is used to obtain spoken natural language query information for the current fault scenario; The third module is used to perform semantic parsing on the colloquial natural language query information based on the semantic memory to obtain the target diagnosis stage; The fourth module is used to perform program retrieval in the program memory according to the target diagnostic stage and the program usage mapping function to obtain valid rule text; The fifth module is used to perform scenario retrieval in the scenario memory based on the target diagnostic stage and scenario-based usage mapping function to obtain valid work order information; The sixth module is used to perform semantic retrieval in the semantic memory based on the target diagnostic stage and the semantic usage mapping function to obtain effective triplet information; The seventh module is used to structurally concatenate the valid rule text, the valid work order information, and the valid triplet information with the colloquial natural language query information to obtain parsable prompt text; The eighth module is used to input the parsable prompt text into a preset large language model, and combine it with the colloquial natural language query information and the target diagnosis stage to perform fault prediction and obtain the predicted fault root cause. The ninth module is used to update the contextual memory based on the actual root causes of failures determined by the user according to the predicted root causes of failures, and to update the semantic memory based on the actual root causes of failures determined by the user according to the predicted root causes of failures.

9. An electronic device, characterized in that, include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the method as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

Images