Vehicle fault diagnosis methods and related equipment

Abstract

This application provides a vehicle fault diagnosis method and related equipment, belonging to the field of vehicle intelligent diagnosis technology. The method includes: collecting multi-source heterogeneous data from the target vehicle across the entire domain; preprocessing the collected preset multi-source vehicle data to obtain vehicle feature vector data; mapping vehicle risk based on the vehicle feature vector data using a preset multi-dimensional evaluation unit to obtain vehicle risk severity data; predicting fault state classification based on the vehicle risk severity data and the vehicle feature vector data using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data; and performing causal logic encoding based on the vehicle operating state energy level data using a preset large language model to obtain abnormal state causal tracing results. This application embodiment can realize real-time fault diagnosis and root cause tracing of vehicles under multi-physics coupling.

Description

Technical Field

[0001] This application relates to the field of vehicle intelligent diagnostic technology, and in particular to a vehicle fault diagnosis method and related equipment. Background Technology

[0002] The high integration of powertrain and chassis systems in new energy vehicles has led to a surge in massive amounts of nonlinear data, causing traditional monitoring algorithms to face bottlenecks of high latency and high false alarms when processing heterogeneous signals. While deep learning has significantly improved the accuracy of fault identification, it is essentially a mapping based on statistical correlations of data, rather than a deduction of physical causal mechanisms. This "black box" characteristic makes models prone to capturing surface signal features while ignoring underlying failure mechanisms. Consequently, diagnostic results often stop at fault phenomenon classification, making it difficult to accurately locate the root cause of the fault and perform dynamic attribution.

[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 vehicle fault diagnosis method and related equipment, which can realize real-time fault diagnosis and root cause tracing under multi-physical field coupling of vehicles.

[0005] To achieve the above objectives, one aspect of this application proposes a vehicle fault diagnosis method, the method comprising:

[0006] The target vehicle is subjected to full-domain multi-source heterogeneous data collection, and then the collected preset vehicle multi-source data is preprocessed to obtain vehicle feature vector data. Based on the vehicle feature vector data, vehicle risk is mapped through a preset multidimensional evaluation unit to obtain vehicle risk severity data; wherein, the preset multidimensional evaluation unit is constructed based on the feature relative importance index and preset feature factors; Based on the vehicle risk severity data and the vehicle feature vector data, a fault state classification prediction is performed using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data; wherein, the preset random weighted neural network diagnostic model is constructed based on a non-iterative analytical algorithm for model learning; Based on the vehicle operating status energy level data, causal logic encoding is performed using a preset large language model to obtain the causal source tracing results of abnormal states.

[0007] In some embodiments, the step of collecting multi-source heterogeneous data from the target vehicle across the entire domain, and then preprocessing the collected preset multi-source vehicle data to obtain vehicle feature vector data, includes: The vehicle operating status parameter package is dynamically collected by a preset distributed parsing module, and then resampled and aligned by linear interpolation to obtain instantaneous global state vector data. The instantaneous global state vector data is subjected to non-stationary random disturbance discrimination, and then the sampling point data that is determined to be outlier in the instantaneous global state vector data is corrected to obtain the first feature vector data; The first feature vector data is smoothed by a preset low-pass filter to obtain the second feature vector data; The second feature vector data is subjected to normalization correction and standard quantization processing to obtain the vehicle feature vector data.

[0008] In some embodiments, the step of mapping vehicle risk based on the vehicle feature vector data through a preset multi-dimensional evaluation unit to obtain vehicle risk severity data includes: The relative importance index of each monitoring feature is determined by pre-set evaluation parameters, and then the feature significance weight data is calculated by the relative importance index; wherein, the pre-set evaluation parameters include expert system data or historical fault correlation data; The comprehensive vulnerability evaluation data is calculated based on the feature saliency weight data and the preset feature factors; wherein, the preset feature factors include sensitivity factors and influence intensity factors. The feature severity contribution value is calculated based on the vehicle feature vector data and the feature comprehensive vulnerability evaluation data, and then the instantaneous severity value of the whole vehicle is generated based on the feature severity contribution value. The instantaneous severity value of the entire vehicle is compensated and corrected by a dynamic attenuation factor to obtain the vehicle risk severity data.

[0009] In some embodiments, before performing the fault state classification prediction based on the vehicle risk severity data and the vehicle feature vector data using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data, the method further includes: Construct an initial stochastic weighted neural network diagnostic model; wherein the initial stochastic weighted neural network model includes an input layer, multiple hidden layers, and an output layer; The output weight matrix is ​​determined by the pseudo-inverse matrix method, and then the parameters of the initial random weight neural network diagnostic model are updated according to the output weight matrix to obtain the preset random weight neural network diagnostic model.

[0010] In some embodiments, the step of performing causal logic encoding based on the vehicle operating state energy level data using a preset large language model to obtain the causal source tracing result of the abnormal state includes: When the vehicle operating state energy level data is determined to be a non-good energy level, the instantaneous global feature data of the vehicle at the current moment is recorded, and the vehicle control operation parameters are retrieved to construct a global state snapshot data; wherein, the vehicle control operation parameters include the state bit of the vehicle control strategy, navigation environment context data, and driver operation sequence data; The semantic mapping operator is used to convert the global state snapshot data and the instantaneous global feature data of the vehicle into the desired text data. Based on the expected text data, combined with the preset vehicle physical architecture causal chain and fault mode library, causal logic encoding is performed through the preset large language model to obtain the causal tracing result of the abnormal state.

[0011] In some embodiments, after performing causal logic encoding based on the vehicle operating state energy level data using a preset large language model to obtain the causal tracing result of the abnormal state, the method further includes: Based on the causal tracing results of the abnormal state, driving intensity labels and environmental load coefficients are extracted; The dynamic attenuation factor is dynamically compensated and corrected based on the driving intensity label and the environmental load coefficient.

[0012] To achieve the above objectives, another aspect of this application provides a vehicle fault diagnosis device, the device comprising: The first module is used to collect multi-source heterogeneous data of the target vehicle across the entire domain, and then preprocess the collected preset multi-source vehicle data to obtain vehicle feature vector data. The second module is used to perform vehicle risk mapping based on the vehicle feature vector data through a preset multi-dimensional evaluation unit to obtain vehicle risk severity data; wherein, the preset multi-dimensional evaluation unit is constructed based on the feature relative importance index and preset feature factors; The third module is used to predict the fault state level by using a preset random weighted neural network diagnostic model based on the vehicle risk severity data and the vehicle feature vector data, thereby obtaining vehicle operating state energy level data; wherein, the preset random weighted neural network diagnostic model is constructed by model learning based on a non-iterative analytical algorithm; The fourth module is used to perform causal logic encoding based on the vehicle operating status energy level data through a preset large language model to obtain the causal source tracing results of abnormal states.

[0013] To achieve the above objectives, another aspect of this application provides an electronic device, the electronic device comprising: 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 performs 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 above-described method.

[0015] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method. The embodiments of this application include at least the following beneficial effects: This application provides a vehicle fault diagnosis method, device, electronic device, storage medium, and program product. This solution involves collecting multi-source heterogeneous data from the target vehicle across the entire domain, and then preprocessing the collected preset multi-source vehicle data to obtain vehicle feature vector data. Next, based on the vehicle feature vector data, the embodiments of this invention perform vehicle risk mapping using a preset multi-dimensional evaluation unit constructed based on the feature relative importance index and preset feature factors to obtain vehicle risk severity data. Further, based on the vehicle risk severity data and the vehicle feature vector data, the embodiments of this invention perform fault state classification prediction using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data. The preset random weighted neural network diagnostic model is constructed based on a non-iterative analytical algorithm. Finally, based on the vehicle operating state energy level data, the embodiments of this invention perform causal logic encoding using a preset large language model to obtain abnormal state causal tracing results, thereby achieving root cause tracing of vehicle faults. It is easy to understand that the embodiments of the present invention perform vehicle risk mapping on vehicle feature vector data, and then perform fault state classification prediction based on the obtained vehicle risk level data and vehicle feature vector data through a preset random weighted neural network diagnostic model, thereby alleviating the problem of rapid response under complex dynamic working conditions, realizing real-time fault diagnosis under multi-physical field coupling of vehicles, and performing causal logic encoding based on vehicle operating state energy level data through a preset large language model to trace the root cause of vehicle faults, thereby achieving accurate tracing of the underlying cause of faults. Attached Figure Description

[0016] Figure 1 This is a flowchart of the vehicle fault diagnosis method provided in the embodiments of this application; Figure 2 This is a schematic diagram of the vehicle fault diagnosis device 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

[0017] 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.

[0018] 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.”

[0019] 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.

[0020] 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.

[0021] Before providing a detailed description of the embodiments of this application, some of the nouns and terms involved in the embodiments of this application will be explained first. The nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0022] Large Language Model (LLM) is an artificial intelligence model based on deep learning. It is trained on massive amounts of text data to capture the complex structure, semantics and contextual relationships of language, and can achieve natural language understanding and generation.

[0023] The high integration of powertrain and chassis systems in new energy vehicles has led to a surge in massive amounts of nonlinear data, causing traditional monitoring algorithms to face bottlenecks of high latency and high false alarms when processing heterogeneous signals. While deep learning has significantly improved the accuracy of fault identification, it is essentially a mapping based on statistical correlations of data, rather than a deduction of physical causal mechanisms. This "black box" characteristic makes models prone to capturing surface signal features while ignoring underlying failure mechanisms. Consequently, diagnostic results often stop at fault phenomenon classification, making it difficult to accurately locate the root cause of the fault and perform dynamic attribution.

[0024] In view of this, this application provides a vehicle fault diagnosis method, device, electronic device, storage medium, and program product. This solution involves collecting multi-source heterogeneous data from the target vehicle across the entire domain, and then preprocessing the collected preset multi-source vehicle data to obtain vehicle feature vector data. Next, based on the vehicle feature vector data, this embodiment performs vehicle risk mapping using a preset multi-dimensional evaluation unit constructed based on the feature relative importance index and preset feature factors to obtain vehicle risk severity data. Further, based on the vehicle risk severity data and the vehicle feature vector data, this embodiment performs fault state classification prediction using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data. The preset random weighted neural network diagnostic model is constructed based on a non-iterative analytical algorithm. Finally, based on the vehicle operating state energy level data, this embodiment performs causal logic encoding using a preset large language model to obtain abnormal state causal tracing results, realizing real-time fault diagnosis and root cause tracing under multi-physics coupling of the vehicle.

[0025] The vehicle fault diagnosis method provided in this application relates to the field of intelligent vehicle diagnosis technology. The vehicle fault diagnosis method provided in this application can be applied to a terminal, a server, or software running on a terminal or 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 vehicle fault diagnosis method, but is not limited to the above forms.

[0026] 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.

[0027] Figure 1 This is an optional flowchart of the vehicle fault diagnosis method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S110 to S140.

[0028] Step S110: Collect multi-source heterogeneous data of the target vehicle across the entire domain, and then preprocess the collected preset multi-source vehicle data to obtain vehicle feature vector data.

[0029] Step S120: Based on the vehicle feature vector data, perform vehicle risk mapping through a preset multi-dimensional evaluation unit to obtain vehicle risk severity data. The preset multi-dimensional evaluation unit is constructed based on the feature relative importance index and preset feature factors.

[0030] Step S130: Based on vehicle risk severity data and vehicle feature vector data, a fault state classification prediction is performed using a preset stochastic weighted neural network diagnostic model to obtain vehicle operating state energy level data. The preset stochastic weighted neural network diagnostic model is constructed using a non-iterative analytical algorithm.

[0031] Step S140: Based on the vehicle operating status energy level data, perform causal logic encoding through a preset large language model to obtain the causal source tracing results of abnormal states.

[0032] In this specific embodiment, the present invention first performs full-domain multi-source heterogeneous data collection on the target vehicle to obtain raw, preset multi-source vehicle data. This preset multi-source data is then preprocessed to obtain vehicle feature vector data. Specifically, in this embodiment, the target vehicle refers to a vehicle requiring fault diagnosis. Correspondingly, the present invention accesses various heterogeneous bus protocols in parallel through a central gateway to collect corresponding vehicle operating status data in real time, thus obtaining preset multi-source vehicle data. Since the collected raw multi-source data suffers from issues such as latency jitter and inconsistent sampling frequencies, this embodiment preprocesses the preset multi-source vehicle data to obtain vehicle feature vector data, providing reliable and high-fidelity input feature vectors for subsequent vehicle fault diagnosis and identification. Next, the present invention maps vehicle risk based on the vehicle feature vector data using a preset multi-dimensional evaluation unit to obtain vehicle risk level data. Then, it combines the vehicle feature vector data with a preset random weighted neural network diagnostic model to predict fault state classification, obtaining vehicle operating state energy level data. Specifically, in this embodiment, the preset multi-dimensional evaluation unit refers to a multi-dimensional evaluation module used to map multi-source vehicle feature vectors to the severity of real-time vehicle health risks. The preset multidimensional evaluation unit is constructed using a feature relative importance index and preset feature factors. Accordingly, in this embodiment, the preset multidimensional evaluation unit maps the preprocessed vehicle feature vector data to the real-time health risk severity of the vehicle, obtaining a vehicle health risk severity value, i.e., vehicle health risk severity data. Furthermore, the preset random weighted neural network diagnostic model in this embodiment refers to a predictive model built based on a random mapping mechanism for rapid classification and grading of vehicle feature vectors at the vehicle edge. In this embodiment, the preset random weighted neural network diagnostic model is constructed using a non-iterative analytical algorithm for model learning. Accordingly, this embodiment first inputs the vehicle feature vector data into the preset random weighted neural network diagnostic model for random feature mapping, thereby mapping the input features to a high-order random feature space. Then, it combines the vehicle risk severity data to classify the fault state, obtaining vehicle operating state energy level data, such as good energy level, slight energy level, moderate energy level, or critical energy level. In this embodiment of the invention, a pre-defined random weighted neural network model is constructed. This model, through a random mapping mechanism, enables the capture of complex nonlinear coupling relationships between sensor signals in new energy vehicles, mitigating the problem of getting trapped in local optima in traditional backpropagation algorithms. Finally, based on vehicle operating state energy level data, this embodiment of the invention performs causal logic encoding using a pre-defined large language model to obtain causal tracing results for abnormal states.Specifically, in this embodiment of the invention, after predicting the energy level data of the vehicle's operating state, the abnormal energy level is causally logically encoded based on the semantic reasoning capability of a preset large language model, thereby identifying the true root cause of the abnormal vehicle state and obtaining the causal source tracing result of the abnormal state.

[0033] In some embodiments of the present invention, multi-source heterogeneous data is collected from the target vehicle across the entire domain, and then the collected preset multi-source vehicle data is preprocessed to obtain vehicle feature vector data, including but not limited to the following steps: The vehicle operating status parameter package is dynamically collected by a pre-set distributed parsing module, and then resampled and aligned by linear interpolation to obtain instantaneous global state vector data.

[0034] Non-stationary random disturbances are detected in the instantaneous global state vector data, and then the sampling point data that is determined to be outlier in the instantaneous global state vector data is corrected to obtain the first feature vector data.

[0035] The first feature vector data is smoothed by a preset low-pass filter to obtain the second feature vector data.

[0036] The second feature vector data is subjected to normalization correction and standard quantization to obtain vehicle feature vector data.

[0037] In this specific embodiment, the present invention first dynamically collects vehicle operating status parameter packets through a preset distributed parsing module, and then performs resampling and alignment using linear interpolation to obtain instantaneous global state vector data. Specifically, the present invention accesses heterogeneous bus protocols such as vehicle Ethernet, CAN-FD, and LIN in parallel through a central gateway, and then extracts key operating status parameter packets covering the power domain, energy domain, and chassis domain in real time through the preset distributed parsing module, i.e., vehicle operating status parameter packets. To alleviate the inherent transmission delay jitter problem of different bus links, the present invention uses the PTP protocol to achieve vehicle node clock synchronization and injects a high-precision hardware timestamp at the moment of signal triggering. Meanwhile, to alleviate the problem of inconsistent sampling frequencies of sensors in different control domains, this embodiment of the invention employs linear interpolation for resampling alignment. The PTP protocol provides reliable input variables for linear interpolation, ensuring that the physical meaning of the interpolation calculation is accurate. For example, if the synchronization reference time is... And this moment falls at two original sample collection points. and Between these, the corresponding signal eigenvalues ​​are aligned. The calculation formula is as follows:

[0038] Where, in the formula and The original time series sequences are respectively in and The sampled value at time.

[0039] Accordingly, this embodiment of the invention uses this mechanism to uniformly project heterogeneous signal streams with unequal frequencies and asynchronous operation onto a standardized synchronous time axis, constructing an instantaneous global state vector. As shown in the following formula:

[0040] Where, in the formula For the first The sensor signals are synchronized at the same time. eigenvalues; This refers to the total feature dimension being monitored.

[0041] Next, this embodiment of the invention performs non-stationary random disturbance judgment on the instantaneous global state vector data, and then corrects the sampling point data that is determined to be outlier in the instantaneous global state vector data to obtain the first feature vector data. Specifically, this embodiment of the invention first performs non-stationary random disturbance judgment on the instantaneous global state vector data. Non-stationary random interference is identified. For example, if the sampled data deviates from the local statistical mean by more than three times the standard deviation threshold, it is identified as an outlier caused by electromagnetic transient interference, and the effective value from the previous moment is used to logically overwrite it, thereby obtaining the pre-calibrated feature vector, i.e., the first feature vector data. Further, in this embodiment of the invention, the first feature vector data is smoothed using a preset low-pass filter to obtain the second feature vector data. Specifically, to eliminate the interference of electromagnetic noise generated by the high-frequency switching of the drive motor and mechanical vibration during high-speed vehicle operation on performance loss characteristics, this embodiment of the invention uses a preset low-pass filter, such as introducing... A Butterworth low-pass filter of order 1, applied to the eigenvectors after initial calibration ( The key dimensions in the filter are smoothed to obtain the second feature vector data. This filter has the characteristic of having the flattest amplitude-frequency response within the passband, and its squared amplitude function is defined as follows:

[0042] in, The amplitude response of the filter. Let the filter order be . The cutoff frequency, This represents the signal angular frequency.

[0043] Furthermore, this embodiment of the invention performs normalization correction and standard quantization processing on the second feature vector data to obtain vehicle feature vector data. Specifically, after completing the physical-level data synchronization alignment and noise reduction cleaning through the above steps, this embodiment of the invention, based on this, transforms the original signal into a standardized feature vector most suitable for efficient recognition by the neural network through statistical regularity correction and dimensional unification, i.e., vehicle feature vector data. Among these features, the sensor signals of new energy vehicles often do not conform to a standard normal distribution, but rather exhibit a nonlinear skewed distribution, which makes it difficult for subsequent neural networks to capture patterns. Therefore, in response to the nonlinear skewed distribution characteristics often exhibited by new energy vehicle sensor signals, this embodiment of the invention performs a Box-Cox transformation to achieve normalization of the features, the transformation logic of which is as follows:

[0044] Where, in the formula The original input feature values ​​represent the original sensor signal data after the above initial calibration and filtering processes, i.e., the second feature vector data. The transformed output eigenvalues ​​represent the new eigenvalues ​​after normalization correction. Additionally, the parameters... The features are determined online via maximum likelihood estimation. Subsequently, embodiments of the present invention perform standard scalarization processing on the transformed features, compressing the heterogeneous features to a uniform scaling space, as shown in the following equation:

[0045] Where, in the formula The mean, The standard deviation is used. Through the above mathematical transformation, the scalar values ​​of each dimension processed in this embodiment of the invention are... Encapsulation constitutes instantaneous global normalized feature vector That is, the vehicle feature vector data, as shown in the following formula:

[0046] In some embodiments of the present invention, vehicle risk severity data is obtained by mapping vehicle risk based on vehicle feature vector data through a preset multi-dimensional evaluation unit, including but not limited to the following steps: The relative importance index of each monitoring feature is determined by pre-set evaluation parameters, and then the feature significance weight data is calculated using the feature relative importance index. The pre-set evaluation parameters include expert system data or historical fault correlation data.

[0047] The comprehensive vulnerability evaluation data is calculated by combining the feature saliency weight data with preset feature factors. These preset feature factors include sensitivity factors and impact intensity factors.

[0048] The severity contribution value of a feature is calculated based on the vehicle feature vector data and the feature comprehensive vulnerability evaluation data. Then, the instantaneous severity value of the entire vehicle is generated based on the feature severity contribution value.

[0049] By compensating and correcting the instantaneous severity value of the entire vehicle using a dynamic attenuation factor, the severity data of vehicle risk is obtained.

[0050] In this specific embodiment, the present invention first determines the relative importance index of each monitoring feature through preset evaluation parameters, and then calculates the feature significance weight data through the feature relative importance index. Specifically, the preset evaluation parameters in this embodiment include expert system data or historical correlation data. The present invention constructs a multi-dimensional evaluation system, i.e., preset multi-dimensional evaluation units, to determine the relative importance and sensitivity of each monitoring feature in fault diagnosis. The present invention determines the feature relative importance index of each monitoring feature based on an expert system or historical fault correlation. Then, the feature saliency weight data is calculated. Accordingly, in the embodiments of the present invention, the features Significance weight The calculation formula is as follows:

[0051] in, This represents the total number of features monitored across the entire region.

[0052] Furthermore, in this embodiment of the invention, feature comprehensive vulnerability evaluation data is calculated by combining the calculated feature saliency weight data with preset feature factors. Specifically, this embodiment of the invention predefines preset feature factors for each feature, including sensitivity factors. With influence intensity factor Then, features are constructed using the parameters mentioned above. Comprehensive vulnerability evaluation function The feature synthesis vulnerability evaluation data is calculated as shown in the following formula:

[0053] In this embodiment of the invention, the comprehensive vulnerability evaluation function is a static function used to characterize the inherent vulnerability potential of each monitoring feature, which is determined by the feature's importance, sensitivity, and impact intensity.

[0054] Furthermore, in this embodiment of the invention, a feature severity contribution value is calculated based on vehicle feature vector data and feature comprehensive vulnerability evaluation data, and then a vehicle-wide instantaneous severity value is generated based on the feature severity contribution value. Specifically, this embodiment of the invention uses the standardized real-time feature components output in the above steps... That is, vehicle feature vector data, used as the excitation function. It is then nonlinearly coupled with the aforementioned comprehensive vulnerability evaluation function to calculate this feature in... The contribution value of the severity at any given moment As shown in the following formula:

[0055] In this embodiment of the invention, the severity contribution value is a dynamic value, representing the actual anomaly level of the feature at the current moment, obtained by multiplying its real-time data by the static vulnerability. Accordingly, based on this, this embodiment of the invention generates a vehicle-wide instantaneous severity value by weighted aggregation of the functional domain components. As shown in the following formula:

[0056] In this embodiment of the invention, the instantaneous severity value of the entire vehicle is the weighted aggregation result of all feature contribution values, reflecting the original comprehensive risk score of the vehicle at the current moment.

[0057] Finally, this embodiment of the invention compensates and corrects the instantaneous severity value of the entire vehicle using a dynamic attenuation factor to obtain vehicle risk severity data. Specifically, to mitigate the problem of irreversible physical degradation of core components as accumulated mileage increases, this embodiment of the invention introduces a dynamic attenuation factor. The severity value across the entire vehicle is compensated and corrected. The corrected real-time severity value is shown below. The calculation formula is as follows:

[0058] Where, in the formula The quantization order for the vehicle's cumulative mileage. The base decay rate is the dynamic decay factor. This invention, through the introduction of dynamic deviation correction of the time-scale decay factor, effectively mitigates false alarms caused by differences in diagnostic standards between older and newer vehicles.

[0059] In some embodiments of the present invention, before performing fault state classification and prediction based on vehicle risk severity data and vehicle feature vector data through a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data, the vehicle fault diagnosis method provided in the embodiments of the present invention further includes, but is not limited to, the following steps: An initial stochastic weighted neural network diagnostic model is constructed. This model includes an input layer, multiple hidden layers, and an output layer.

[0060] The output weight matrix is ​​determined by the pseudo-inverse matrix method, and then the parameters of the initial random weight neural network diagnostic model are updated according to the output weight matrix to obtain the preset random weight neural network diagnostic model.

[0061] In this specific embodiment, the present invention first constructs an initial stochastic weighted neural network diagnostic model. Specifically, the initial stochastic weighted neural network model constructed in this embodiment includes an input layer, multiple hidden layers, and an output layer. The embodiment receives vehicle feature vector data, i.e., instantaneous global standardized feature vectors, through the input layer. Meanwhile, the input weights of the hidden layer nodes and bias Hidden layers are randomly assigned during initialization and remain fixed during inference. Accordingly, in this embodiment of the invention, the hidden layers are activated by the Sigmoid function. The input features are mapped to a high-dimensional random feature space. This embodiment of the invention ensures that the model can capture the complex nonlinear coupling relationships between sensor signals in new energy vehicles through a random mapping mechanism, and effectively alleviates the problem of easily getting trapped in local optima in traditional backpropagation algorithms. Furthermore, this embodiment of the invention determines the output weight matrix using a pseudo-inverse matrix method, and then updates the parameters of the initial random weight neural network diagnostic model based on the output weight matrix to obtain a preset random weight neural network diagnostic model. Specifically, this embodiment of the invention determines the output weight matrix using a non-iterative analytical method, such as the pseudo-inverse matrix method. This is to meet the computational efficiency requirements of in-vehicle terminals for real-time diagnostics. Accordingly, this embodiment of the invention calculates the output weight matrix to complete the model's learning phase, using historical data (the target output matrix and randomly generated hidden layer states) to calculate the most accurate mapping relationship. For example, let the hidden layer output matrix be... The target output matrix is Then the output weight matrix The calculation formula is as follows:

[0062] Where, in the formula Output matrix for hidden layer The Moore-Ponous pseudoinverse. Since this process involves only analytic linear matrix operations, the system can complete the processing of the current vehicle feature vector in a very short time. This classification reasoning method enables high-frequency real-time monitoring of the operating status of new energy vehicles. The formula directly calculates the output weights, avoiding iterative training and meeting the real-time requirements of edge computing.

[0063] It should be noted that, in this embodiment of the invention, the output layer uses the Softmax activation function to convert the abstract features extracted by the hidden layer into probability distributions corresponding to the four health levels, and then combines this with the dynamically adjusted severity value. In this embodiment of the invention, the current operating state of the vehicle is divided into the following four energy levels in real time: Good energy level: If < The system is determined to be operating stably and requires no intervention. Slight energy level: if < If a potential loss or minor fault trend is detected, background recording is initiated. Intermediate energy level: If < If the performance is determined to be significantly degraded, an instrument maintenance warning will be triggered. Critical level: If If a significant security risk is detected, an emergency protection strategy is triggered, and the subsequent LLM precise traceability module is immediately activated.

[0064] in, , and These are the health rating thresholds derived from the statistical distribution of historical big data.

[0065] In some embodiments of the present invention, causal logic encoding is performed on vehicle operating state energy level data using a preset large language model to obtain abnormal state causal tracing results, including but not limited to the following steps: When the vehicle's operating state level data is determined to be a non-ideal level, the instantaneous global characteristic data of the vehicle at the current moment is recorded, and the vehicle control operation parameters are retrieved to construct a global state snapshot data. The vehicle control operation parameters include the state bits of the vehicle control strategy, navigation environment context data, and driver operation sequence data.

[0066] The semantic mapping operator transforms global state snapshot data and instantaneous global feature data of vehicles into desired text data.

[0067] Based on the expected text data, combined with the preset vehicle physical architecture causal chain and fault mode library, causal logic encoding is performed through a preset large language model to obtain the causal tracing results of abnormal states.

[0068] In this specific embodiment, the present invention first determines whether the vehicle operating state energy level triggers the corresponding causal tracing process. When it is determined that the vehicle operating state energy level data is a non-good energy level, the LLM precise tracing step is triggered. Specifically, in this embodiment, a non-good energy level includes a vehicle's current operating state being a critical energy level. Accordingly, when it is determined that the vehicle operating state energy level data is a non-good energy level, the present invention first records the vehicle's instantaneous global feature data at the current moment and retrieves the vehicle control operation parameters, including the state bits of the vehicle control strategy, navigation environment context data, and driver operation sequence data, to construct a global state snapshot data. Next, this embodiment of the invention uses a semantic mapping operator to transform the global state snapshot data and the instantaneous global feature data of the vehicle into the desired text data. Specifically, this embodiment of the invention uses a semantic mapping operator... The system combines quantified instantaneous global feature data of vehicles with discrete state snapshots. The data is transformed into structured descriptive text. A semantic mapping operator performs semantic discrimination on the vehicle's instantaneous global feature data based on a preset feature threshold range, historical statistical benchmarks, and control logic consistency rules. This text not only describes the abnormal amplitude of physical quantities but also reconstructs the macroscopic operational background at the time of the anomaly through logical encapsulation, providing a semantic foundation for subsequent reasoning. Furthermore, in this embodiment, the expected text data is combined with a predefined causal chain of the vehicle's physical architecture in the form of a structured causal relationship graph and a fault mode library stored in the form of standardized fault mode entries. Causal logic encoding is performed using a preset large language model to obtain the causal tracing result of the abnormal state. Specifically, in this embodiment, a preset large language model, such as a cloud-based LLM engine, receives the semantically encapsulated prompt words, i.e., the expected text data, and uses its pre-set causal chain of the vehicle's physical architecture and fault mode library for deep reasoning. Among them, the pre-set large language model identifies the true root cause of abnormal states by analyzing the logical conflicts between cross-domain signals in parallel, and obtains the causal tracing results of abnormal states, including abnormal root cause candidates, corresponding causal paths and associated confidence labels. The causal tracing results of abnormal states are used to assist in diagnostic analysis and model parameter correction, but do not directly participate in the real-time control decision of the vehicle.

[0069] In some embodiments of the present invention, after performing causal logic encoding based on vehicle operating state energy level data using a preset large language model to obtain the causal source tracing result of the abnormal state, the vehicle fault diagnosis method provided in the examples of the present invention further includes, but is not limited to, the following steps: The driving intensity label and environmental load coefficient are extracted based on the causal tracing results of the abnormal state.

[0070] The dynamic attenuation factor is dynamically compensated and corrected based on the driving intensity label and environmental load coefficient.

[0071] In this specific embodiment, the present invention first extracts driving intensity labels and environmental load coefficients based on the causal tracing results of abnormal states, and then dynamically compensates and corrects dynamic component factors based on the driving intensity labels and environmental load coefficients. Specifically, for the heterogeneous losses generated by new energy vehicles under different driving styles, a pre-set Large Language Model (LLM) extracts driving intensity labels based on the tracing conclusions (causal tracing results of abnormal states). With environmental load factor The base decay rate in the vehicle risk mapping step Dynamic compensation and correction are performed. Specifically, the corrected dynamic degradation factor... The calculation formula is as follows:

[0072] Where, in the formula This is a nonlinear compensation function generated by a pre-defined Large Language Model (LLM) based on the results of logical analysis. Correspondingly, the calculated... The onboard OTA (Over-The-Air) system provides real-time feedback to the edge diagnostic engine to update the severity calculation benchmark and the judgment threshold in the health grading thresholds mentioned above. This formula adaptively adjusts the aging coefficient based on driving style and environmental load determined by LLM analysis, achieving full lifecycle evolution of the diagnostic model. This embodiment of the invention improves the system's diagnostic sensitivity and robustness throughout the vehicle's entire lifecycle through this parameter evolution closed-loop.

[0073] It is readily understood that the embodiments of this invention can alleviate the problem of real-time fault warning and root cause tracing under multi-physics coupling in the chassis system of new energy vehicles. The embodiments of this invention first utilize the non-iterative learning characteristics of the random weighted neural network (NNRW) to achieve millisecond-level anomaly capture and energy level determination of multi-source heterogeneous signals at the edge, ensuring the real-time response of the system under complex operating conditions. This triggers the cloud-based Large Language Model (LLM) to retrieve multi-dimensional state snapshots. By introducing logical reasoning guided by physical knowledge, it effectively compensates for the shortcomings of traditional deep learning models that rely solely on statistical correlation and lack physical interpretability, achieving a deep leap from "signal appearance classification" to "fault root cause localization." Simultaneously, by constructing a dynamic parameter feedback closed loop based on driving style and environmental load, the edge diagnostic model is endowed with adaptive evolution capabilities throughout its entire lifecycle, significantly improving the system's robustness and operational transparency. Specifically, the embodiments of this invention achieve a closed loop from online diagnosis to deep root cause tracing through the deep collaboration of the random weighted neural network (NNRW) and the Large Language Model (LLM). Correspondingly, by leveraging the analytical solution characteristics of the Randomized Weighted Neural Network (NNRW), online real-time classification is achieved, alleviating the challenge of rapid response under complex dynamic operating conditions. Simultaneously, through the semantic mapping and causal reasoning mechanism of the Large Language Model (LLM), accurate tracing of the underlying causes of faults is realized, facilitating precise and thorough vehicle repair and maintenance. Furthermore, this embodiment of the invention incorporates a dynamic parameter calibration loop based on feedback from the Large Language Model (LLM), enabling the diagnostic model to adaptively evolve throughout the vehicle's entire lifecycle and under diverse operating conditions. This allows for accurate differentiation between strategic protection restrictions and actual hardware damage, significantly reducing false alarm rates while providing efficient and transparent technical support for the operation and maintenance of new energy vehicles.

[0074] Please see Figure 2 This application also provides a vehicle fault diagnosis device that can implement the above-described method. The device includes: The first module 210 is used to collect multi-source heterogeneous data of the target vehicle across the entire domain, and then preprocess the collected preset multi-source vehicle data to obtain vehicle feature vector data.

[0075] The second module 220 is used to map vehicle risk based on vehicle feature vector data through a preset multi-dimensional evaluation unit to obtain vehicle risk severity data. The preset multi-dimensional evaluation unit is constructed based on the feature relative importance index and preset feature factors.

[0076] The third module 230 is used to predict the fault state level by using a preset random weighted neural network diagnostic model based on vehicle risk severity data and vehicle feature vector data, thereby obtaining vehicle operating state energy level data. The preset random weighted neural network diagnostic model is constructed using a non-iterative analytical algorithm.

[0077] The fourth module 240 is used to perform causal logic encoding based on the vehicle's operating status energy level data through a preset large language model to obtain the causal source tracing results of abnormal states.

[0078] 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.

[0079] 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.

[0080] 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.

[0081] 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.

[0082] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0083] 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.

[0084] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0085] 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.

[0086] 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.

[0087] 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.

[0088] 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.

[0089] 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.

[0090] 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.

[0091] 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.

[0092] 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.

[0093] 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.

[0094] 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.

[0095] 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.

[0096] 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.

[0097] 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 vehicle fault diagnosis method, characterized in that, The method includes the following steps: The target vehicle is subjected to full-domain multi-source heterogeneous data collection, and then the collected preset vehicle multi-source data is preprocessed to obtain vehicle feature vector data. Based on the vehicle feature vector data, vehicle risk is mapped through a preset multidimensional evaluation unit to obtain vehicle risk severity data; wherein, the preset multidimensional evaluation unit is constructed based on the feature relative importance index and preset feature factors; Based on the vehicle risk severity data and the vehicle feature vector data, a fault state classification prediction is performed using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data; wherein, the preset random weighted neural network diagnostic model is constructed based on a non-iterative analytical algorithm for model learning; Based on the vehicle operating status energy level data, causal logic encoding is performed using a preset large language model to obtain the causal source tracing results of abnormal states.

2. The method according to claim 1, characterized in that, The process involves collecting multi-source heterogeneous data from the target vehicle across the entire domain, and then preprocessing the collected multi-source vehicle data to obtain vehicle feature vector data, including: The vehicle operating status parameter package is dynamically collected by a preset distributed parsing module, and then resampled and aligned by linear interpolation to obtain instantaneous global state vector data. The instantaneous global state vector data is subjected to non-stationary random disturbance discrimination, and then the sampling point data that is determined to be outlier in the instantaneous global state vector data is corrected to obtain the first feature vector data; The first feature vector data is smoothed by a preset low-pass filter to obtain the second feature vector data; The second feature vector data is subjected to normalization correction and standard quantization processing to obtain the vehicle feature vector data.

3. The method according to claim 1, characterized in that, The process of mapping vehicle risk based on the vehicle feature vector data through a preset multi-dimensional evaluation unit to obtain vehicle risk severity data includes: The relative importance index of each monitoring feature is determined by pre-set evaluation parameters, and then the feature significance weight data is calculated by the relative importance index; wherein, the pre-set evaluation parameters include expert system data or historical fault correlation data; The comprehensive vulnerability evaluation data is calculated based on the feature saliency weight data and the preset feature factors; wherein, the preset feature factors include sensitivity factors and influence intensity factors. The feature severity contribution value is calculated based on the vehicle feature vector data and the feature comprehensive vulnerability evaluation data, and then the instantaneous severity value of the whole vehicle is generated based on the feature severity contribution value. The instantaneous severity value of the entire vehicle is compensated and corrected by a dynamic attenuation factor to obtain the vehicle risk severity data.

4. The method according to claim 1, characterized in that, Before performing the fault state classification and prediction based on the vehicle risk severity data and the vehicle feature vector data using a preset random weighted neural network diagnostic model to obtain vehicle operating state energy level data, the method further includes: Construct an initial stochastic weighted neural network diagnostic model; wherein the initial stochastic weighted neural network model includes an input layer, multiple hidden layers, and an output layer; The output weight matrix is ​​determined by the pseudo-inverse matrix method, and then the parameters of the initial random weight neural network diagnostic model are updated according to the output weight matrix to obtain the preset random weight neural network diagnostic model.

5. The method according to claim 1, characterized in that, The step of performing causal logic encoding based on the vehicle operating state energy level data using a preset large language model to obtain the causal source tracing result of the abnormal state includes: When the vehicle operating state energy level data is determined to be a non-good energy level, the instantaneous global feature data of the vehicle at the current moment is recorded, and the vehicle control operation parameters are retrieved to construct a global state snapshot data; wherein, the vehicle control operation parameters include the state bit of the vehicle control strategy, navigation environment context data, and driver operation sequence data; The semantic mapping operator is used to convert the global state snapshot data and the instantaneous global feature data of the vehicle into the desired text data. Based on the expected text data, combined with the preset vehicle physical architecture causal chain and fault mode library, causal logic encoding is performed through the preset large language model to obtain the causal tracing result of the abnormal state.

6. The method according to claim 3, characterized in that, After performing causal logic encoding based on the vehicle operating state energy level data using a preset large language model to obtain the causal tracing result of the abnormal state, the method further includes: Based on the causal tracing results of the abnormal state, driving intensity labels and environmental load coefficients are extracted; The dynamic attenuation factor is dynamically compensated and corrected based on the driving intensity label and the environmental load coefficient.

7. A vehicle fault diagnosis device, characterized in that, The device includes: The first module is used to collect multi-source heterogeneous data of the target vehicle across the entire domain, and then preprocess the collected preset multi-source vehicle data to obtain vehicle feature vector data. The second module is used to perform vehicle risk mapping based on the vehicle feature vector data through a preset multi-dimensional evaluation unit to obtain vehicle risk severity data; wherein, the preset multi-dimensional evaluation unit is constructed based on the feature relative importance index and preset feature factors; The third module is used to predict the fault state level by using a preset random weighted neural network diagnostic model based on the vehicle risk severity data and the vehicle feature vector data, thereby obtaining vehicle operating state energy level data; wherein, the preset random weighted neural network diagnostic model is constructed by model learning based on a non-iterative analytical algorithm; The fourth module is used to perform causal logic encoding based on the vehicle operating status energy level data through a preset large language model to obtain the causal source tracing results of abnormal states.

8. 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 6.

9. A computer-readable storage medium storing 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 6.

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 6.

Images