Vehicle-mounted wire harness fault diagnosis method and device and storage medium

Abstract

The invention relates to the technical field of wire harness fault diagnosis, and discloses a vehicle-mounted wire harness fault diagnosis method and device and a storage medium. The method comprises the following steps: carrying out equivalent circuit modeling on a measurement node and a wire harness section of the vehicle-mounted wire harness to obtain a node connection relation matrix; performing feature analysis on the voltage data and the current data of the measurement nodes to obtain a fault feature vector and a fault type discrimination rule base; inputting the voltage data, the current data and the node connection relation matrix into a recursive variational graph automatic encoder for data reconstruction to obtain time sequence feature reconstruction data and topological relation reconstruction data; carrying out distributed sub-model division and model fusion to obtain a standardized feature vector; and performing matching analysis on the standardized feature vector and the fault type judgment rule base to obtain fault position information and fault type information, so that automatic optimization and screening of fault features are realized, and the reliability and interpretability of a diagnosis result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wiring harness fault diagnosis, and in particular to a vehicle-mounted wiring harness fault diagnosis method, device and storage medium. Background Art

[0002] Traditional on-board wiring harness fault diagnosis methods mainly rely on manual experience and simple electrical measurements, which are difficult to cope with the multi-type and multi-location fault diagnosis needs in complex wiring harness systems. Although the existing automated diagnosis methods have introduced some intelligent algorithms, they still have the problem of insufficient diagnostic accuracy when dealing with the nonlinear fault characteristics of large-scale wiring harness systems.

[0003] At present, vehicle wiring harness fault diagnosis faces three major challenges: first, the diversity and complexity of fault characteristics. Fault types such as open circuit, short circuit, insulation aging and poor contact in the wiring harness system have different electrical characteristics, and these characteristics are often coupled with each other, which increases the difficulty of fault identification; second, the accuracy of fault location. Due to the wide spatial distribution of the wiring harness system, traditional positioning methods are difficult to accurately locate the fault point, especially in the presence of multiple faults; finally, there is the reliability of the diagnostic results. Existing methods have weak resistance to noise and interference and are easily affected by environmental factors and lead to misjudgment. Summary of the invention

[0004] The present invention provides a vehicle-mounted wiring harness fault diagnosis method, device and storage medium, which realizes automatic optimization and screening of fault characteristics and improves the reliability and interpretability of diagnosis results.

[0005] In a first aspect, the present invention provides a method for diagnosing a vehicle wiring harness fault, the method comprising: Conduct equivalent circuit modeling on the measurement nodes and wiring harness sections of the vehicle wiring harness to obtain the node connection relationship matrix; Performing feature analysis on the voltage data and current data of the measurement node according to the node connection relationship matrix to obtain a fault feature vector and a fault type discrimination rule base; Inputting the voltage data, the current data and the node connection relationship matrix into a recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topological relationship reconstruction data; Performing distributed sub-model division and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector; The standardized feature vector is matched and analyzed with the fault type discrimination rule base to obtain fault location information and fault type information.

[0006] In a second aspect, the present invention provides a vehicle-mounted wiring harness fault diagnosis device, the vehicle-mounted wiring harness fault diagnosis device comprising: A modeling module is used to perform equivalent circuit modeling on the measurement nodes and wiring harness sections of the vehicle wiring harness to obtain a node connection relationship matrix; An analysis module, used to perform feature analysis on the voltage data and current data of the measurement node according to the node connection relationship matrix to obtain a fault feature vector and a fault type discrimination rule base; A reconstruction module, used for inputting the voltage data, the current data and the node connection relationship matrix into a recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topological relationship reconstruction data; A partitioning module, used for performing distributed sub-model partitioning and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector; The matching module is used to match and analyze the standardized feature vector with the fault type discrimination rule base to obtain fault location information and fault type information.

[0007] The third aspect of the present invention provides a vehicle-mounted wiring harness fault diagnosis device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the vehicle-mounted wiring harness fault diagnosis device executes the above-mentioned vehicle-mounted wiring harness fault diagnosis method.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned vehicle wiring harness fault diagnosis method.

[0009] In the technical solution provided by the present invention, by constructing a multi-node equivalent circuit model and combining it with Kirchhoff's law for modeling, the physical characteristics and electrical laws of the wiring harness system are effectively reflected, and the accuracy of fault feature extraction is improved. The wiring harness data is reconstructed by using a recursive variational graph autoencoder. The structure has a dual action mechanism, which can not only reconstruct the timing measurement value and topological relationship, but also serve as a noise filtering module, which significantly enhances the anti-interference ability of fault diagnosis. The model fusion distributed fault diagnosis algorithm and the expectation maximization algorithm are introduced. Through the collaborative work of multiple sub-models, the accurate capture of complex fault features is achieved, and the system's recognition ability for multi-type and multi-location faults is improved. Based on the feature extraction method of gradient descent, the automatic optimization and screening of fault features are achieved through adaptive learning rate and feature importance evaluation, and the adaptability of the diagnosis process is enhanced. A multi-layer verification and cross-analysis mechanism is adopted, combined with fault severity evaluation, a complete fault diagnosis decision chain is established, and the reliability and interpretability of the diagnosis results are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0011] Figure 1 A schematic diagram of a process flow of a vehicle wiring harness fault diagnosis method provided in an embodiment of the present application; Figure 2 A schematic block diagram of the structure of a vehicle wiring harness fault diagnosis device provided in an embodiment of the present application; Figure 3 A schematic block diagram of the structure of a vehicle wiring harness fault diagnosis device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0013] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change based on actual conditions.

[0014] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0015] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0016] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0017] See also Figure 1 , Figure 1A schematic diagram of a process flow of a vehicle wiring harness fault diagnosis method provided in an embodiment of the present application, such as Figure 1 As shown, the vehicle wiring harness fault diagnosis method provided in the embodiment of the present application includes steps S100 to S600.

[0018] Step S100, performing equivalent circuit modeling on the measurement nodes and wiring harness sections of the vehicle wiring harness to obtain a node connection relationship matrix; It is understandable that the execution subject of the present invention may be a vehicle-mounted wiring harness fault diagnosis device, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0019] Specifically, the spatial distribution analysis of the physical position of the measurement node is performed, and the distribution data of the measurement node is formed by obtaining the geometric structure of the vehicle's internal wiring harness system and the spatial coordinate information of the measurement node. In this process, the spatial position data of the measurement node is digitized with the help of three-dimensional scanning, CAD models or existing wiring harness design drawings so that it can reflect the actual physical distribution characteristics. According to the topological structure of the wiring harness, the measurement node distribution data is divided into multiple segments according to the connection relationship. Each segment consists of a pair of physically connected measurement nodes and the wiring harness between them, forming wiring harness section data. Electrical parameter modeling is performed on the wiring harness section data. By combining actual tests or existing design data, the resistance, inductance and capacitance parameters of each wiring harness section are modeled separately. These parameters are calculated through the physical properties of the wiring harness (such as material, conductivity, length, cross-sectional area, etc.). For example, the resistance of the wiring harness is calculated by the formula l / A calculation, where is the resistivity of the material, is the harness length, is the cross-sectional area; the inductance of the harness is estimated by combining its winding characteristics and the distribution of the surrounding magnetic flux, and the capacitance is calculated by the insulating medium and spatial distribution between the harnesses. In this way, the equivalent circuit parameters of each harness section are obtained to form an equivalent circuit model of the harness as a whole. Based on the above harness equivalent circuit parameters, the node voltage equation is constructed according to Kirchhoff's voltage law (KVL) and Kirchhoff's current law (KCL). By defining the voltage and current of each measurement node and the electrical parameters of each harness section, the electrical connection relationship between the nodes in the system is converted into a set of node voltage equations in mathematical form. In the specific construction process, KVL is used to describe the voltage relationship in each closed loop, and KCL is used to characterize the current conservation relationship of each node, forming a set of linear equations containing unknown voltage values ​​and known circuit parameters, which systematically describes the electrical behavior of the harness system. In order to improve the accuracy of the model, the data collected by the voltage sensor and current sensor of the measurement node are standardized. By normalizing or denoising the original data, the influence of measurement errors and environmental interference on the model accuracy is eliminated, and standardized measurement data is obtained. Solve the node voltage equations according to the standardized measurement data to obtain the electrical characteristic parameters between nodes, including the voltage value, current distribution, and power dissipation of each node, which can reflect the real-time electrical state of the harness. Based on the electrical characteristic parameters between nodes, analyze the physical connection relationship of nodes. By analyzing the electrical characteristics between each pair of measurement nodes, such as the conduction state, resistance, and relative changes in voltage and current, deduce the actual physical connection relationship between nodes, form a physical connection matrix, and characterize the connection state of the measurement nodes in physical space. By performing matrix synthesis operations on the physical connection matrix and the electrical characteristic parameters between nodes, the node connection relationship matrix is ​​obtained.

[0020] Step S200, performing feature analysis on the voltage data and current data of the measurement node according to the node connection relationship matrix to obtain a fault feature vector and a fault type discrimination rule base; Specifically, the voltage data and current data of the measurement node are classified and marked according to the fault type. Through data analysis, for common fault types, such as open circuit fault, short circuit fault, insulation aging fault and poor contact fault, the fault type is marked according to the differences in voltage and current performance, and a marked data set containing fault types is generated. The time series features of the voltage data and current data of the measurement node are extracted based on the fault type marking data. The time series analysis method is used to extract the characteristics of the voltage and current changes of each node in the time dimension, and the time series feature data of voltage and current are obtained, including the voltage fluctuation amplitude, the frequency component of the signal, the change rate of the current, etc., which can effectively capture the electrical characteristics in a dynamic environment. In order to mine the fault characteristics, the voltage difference and phase difference between the adjacent nodes of the measurement node are calculated and processed to obtain the difference feature data between the nodes. These data reveal the potential abnormal mode in the electrical connection by quantifying the electrical differences between adjacent nodes. The time series feature data of voltage and current and the difference feature data between nodes are subjected to feature fusion operation to integrate the feature information of various types of data and generate a fault feature vector to reflect the comprehensive electrical characteristics of the measurement node and its adjacent nodes. On this basis, the impedance parameters of the wiring harness system are calculated using the node connection relationship matrix and the fault feature vector to obtain the impedance feature data of the node. By combining the physical parameters of the actual circuit, the equivalent impedance between nodes is calculated to reflect the impact of the fault on the line impedance characteristics. The node impedance feature data is compared and analyzed with the standard impedance parameters in the system to generate a fault feature deviation matrix to describe the difference between the actual impedance characteristics and the normal impedance parameters. Based on the fault feature deviation matrix, a set of fault discrimination conditions is constructed. By analyzing the distribution law of different parameters in the deviation matrix and associating them with the characteristics of different fault types, a series of discrimination conditions are extracted. These conditions are expressed in the form of rules to form a set of fault discrimination rules that can cover various common fault modes. The fault discrimination rules are mapped and associated with the fault type labeling data to establish a corresponding relationship between the fault type and the rule conditions, and a fault type discrimination rule base is generated.

[0021] Step S300, inputting voltage data, current data and node connection relationship matrix into a recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topology relationship reconstruction data; Specifically, the voltage data and current data are input into the time series feature encoding layer of the recursive variational graph autoencoder for dimensionality reduction. The time series feature encoding layer gradually extracts the features of the input data through three graph convolution layers. The graph convolution layer uses the topological information of the wiring harness network and the electrical characteristics of the nodes to capture the local and global correlation of the data. Each graph convolution layer uses the ReLU activation function to enhance the model's ability to express nonlinear features, and combines the batch normalization operation to effectively reduce the problem of gradient disappearance or explosion, while ensuring the stability of the feature distribution. After this step, the voltage and current data are reduced to compact and information-dense initial latent variables. The initial latent variables and the node connection relationship matrix are input into the variational inference layer of the recursive variational graph autoencoder for parameter calculation. In the variational inference layer, the nonlinear transformation of the features is realized through two fully connected layers, and a graph attention layer is used to perform weighted analysis on the relationship between nodes to capture important node interaction characteristics, and the mean vector and variance vector are output to describe the center position and distribution range of the latent variable distribution, respectively. Reparameterized sampling operations are performed based on the mean vector and variance vector to generate random latent variables. The reparameterization technique ensures the differentiability of the sampling process by introducing the randomness of the standard normal distribution, so that the model can learn effectively during the optimization process. The random latent variables are input into the topological relationship reconstruction layer of the recursive variational graph autoencoder to restore the topological features. The topological relationship reconstruction layer contains three graph deconvolution layers and a graph residual connection. The graph deconvolution layer gradually restores the latent variables to high-dimensional data consistent with the original topological structure through the reverse convolution operation, and combines the residual connection mechanism to ensure that important topological information is not lost during the reconstruction process. In this way, the generated topological relationship reconstruction data can truly reflect the physical connection structure between the wiring harness nodes while retaining the implicit characteristics of the input data. At the same time, the random latent variables are decoded by time series features to restore the time series features of voltage and current. In the process of time series feature decoding, three transposed convolution layers are used to gradually restore the latent variables to the resolution of the original data by reverse convolution. Each transposed convolution layer uses the LeakyReLU activation function, which introduces a smaller negative slope in the nonlinear region, making the model more efficient in capturing negative characteristics and gradient flow, and improving the stability and accuracy of the decoding process. Through this step, the time series feature reconstruction data is generated, which can reflect the dynamic change characteristics of the input voltage and current data, and contain the potential information learned by the model.

[0022] Step S400, performing distributed sub-model division and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector; Specifically, the time series feature reconstruction data and the topological relationship reconstruction data are input into the feature classification module of the model fusion distributed fault diagnosis algorithm. In this module, the data is distributedly divided into fault features by using the time series characteristics and topological relationship characteristics of the data, and multiple fault feature sub-models are generated. These sub-models represent the fault features of specific types, such as the manifestations of open circuit, short circuit or poor contact in the wiring harness system. According to multiple fault feature sub-models, the fault state of the vehicle wiring harness is subspace modeled to obtain a fault subspace data set. By analyzing the feature distribution of each sub-model, it is mapped to the corresponding multidimensional subspace so as to describe the distribution of different fault states in the high-dimensional space. The probability distribution analysis of the fault subspace data set is performed, and the feature probability distribution matrix is ​​constructed by using statistical methods and clustering technology. The feature probability distribution matrix and multiple fault feature sub-models are input into the expectation maximization (EM) algorithm for iterative optimization. Through the EM algorithm, the fault probability parameter, fault duration parameter and fault impact range parameter are optimized respectively. These parameters describe the possibility of the fault under different conditions, the distribution law of the duration and the impact range of the fault on the overall performance of the wiring harness. The fault transfer probability of the diagnosis results of multiple fault feature sub-models is calculated to generate a fault transfer probability matrix. This matrix reflects the propagation law of faults between different sub-models. For example, after a sub-model detects a short circuit fault, other sub-models detect associated faults such as voltage fluctuation or poor contact. The fault propagation law is quantified by the probability matrix. Based on the fault occurrence probability parameter, fault duration parameter and fault impact range parameter, multiple fault feature sub-models are combined and optimized to generate the weight coefficient of the fusion model. Using the optimization algorithm, combined with the diagnostic ability and feature expression efficiency of the sub-model, the weight of each sub-model in the overall fusion model is dynamically adjusted to improve the accuracy and robustness of the final diagnosis result. According to the weight coefficient of the fusion model and the fault transfer probability matrix, the output results of multiple fault feature sub-models are weighted and combined to generate the fusion fault diagnosis result. The fusion fault diagnosis result is evaluated globally according to the time series and spatial distribution to generate the global state evaluation index. On this basis, the fusion fault diagnosis result is mapped globally to generate the global fault state data. Based on the global fault state data, the gradient descent method is used to extract features, gradually optimize the representation of the data, and generate a standardized feature vector. Gradient descent feature extraction continuously adjusts the feature representation by minimizing the error objective function, so that the standardized feature vector can accurately reflect the fault characteristics of the wiring harness at the global and local levels.

[0023] The global fault state data is input into the objective function construction module, in which the feature optimization objective function is established based on the principle of dual-objective optimization. The feature optimization objective function consists of two core items, of which the reconstruction error minimization item is used to ensure that the model minimizes the deviation from the original data during the feature reconstruction process, thereby enhancing the fidelity of the feature, while the fault diagnosis accuracy maximization item optimizes the feature parameters to achieve the best accuracy of the diagnosis result. The partial derivatives of the feature parameters in the feature optimization objective function are calculated to obtain the gradient direction data of each feature parameter. The gradient direction data represents the change trend of the objective function in the parameter space and is the key basis for guiding the parameter adjustment direction in the optimization process. In the calculation process, the sensitivity of the feature parameters under the optimization objective is determined by deriving the feature parameters one by one and combining the distribution characteristics in the global fault state data. At the same time, based on the gradient direction data, the change amplitude of the feature parameters is dynamically adjusted to adapt to the nonlinear distribution characteristics of the complex fault state. The adaptive learning rate method is used to dynamically adjust the learning rate according to the amplitude and change rate of the gradient direction, thereby avoiding the problem of slow convergence or oscillation caused by the fixed learning rate. The feature parameters are iteratively updated and calculated according to the adaptive learning rate and gradient direction data. Through multiple rounds of iterations, the feature parameters are gradually adjusted to the local optimal point of the feature optimization objective function. In each iteration, the parameter value is updated according to the gradient direction and learning rate of the current feature parameter, and the change trend of the objective function value during the optimization process is monitored in real time. After the iterative update is completed, the feature parameter update results are analyzed, focusing on the contribution of each feature component to the diagnosis result. Through sorting analysis, the importance of each feature component in the fault diagnosis process is quantified, and the feature importance ranking result is generated. According to the feature importance ranking result, the feature parameter update results are threshold screened. According to the set diagnostic performance threshold, the feature components that contribute significantly to the diagnostic results are screened to form a target feature parameter set. Each feature parameter in the target feature parameter set is subjected to mean-variance normalization. The normalization operation eliminates the dimensional differences between different feature parameters, so that the feature vectors are consistent in numerical scale, and a standardized feature vector is obtained.

[0024] Step S500: Match and analyze the standardized feature vector with the fault type discrimination rule base to obtain fault location information and fault type information.

[0025] Specifically, the standardized feature vector is input into the fault diagnosis processing module, and the probability of failure in each section of the wiring harness system is calculated to generate the fault probability distribution data. This data is combined with the physical structure and electrical characteristics of the wiring harness, and the probability of each section of the wiring harness is predicted by the distribution characteristics of the feature vector. The matching degree of the fault probability distribution data of each section of the vehicle wiring harness is calculated with the fault feature rules in the fault type discrimination rule base. By analyzing the fault characteristics of the wiring harness section by section and matching the similarity with the predefined fault feature patterns in the rule base, the matching degree data of each fault type is obtained, indicating the possibility of a specific fault type occurring in different wiring harness sections. The fault type matching degree data is input into the feature consistency verification module for multiple verification and analysis. By performing consistency verification on multiple feature matching results of the same wiring harness section, the erroneous matching caused by noise or abnormal fluctuations is effectively screened out, and a more reliable fault feature verification result is generated. The standardized feature vector is mapped to the measurement node space to generate the physical positioning data of each measurement node. The electrical characteristics of the wiring harness are converted into its physical position through the mapping algorithm to form a calibration result in the spatial dimension. Based on the physical positioning data, the fault position is calibrated to obtain accurate fault position coordinate data. Through the positioning algorithm, the electrical characteristic anomalies can be mapped to specific wiring harness sections and nodes to form a clear distribution in physical space. The fault location coordinate data is cross-validated with the fault probability distribution data of each section of the vehicle wiring harness. Through cross-validation, the consistency between the positioning data and the probability prediction results is checked, and potential errors are corrected to obtain the fault location verification results. On this basis, according to the fault feature verification results and the fault location verification results, the fault repair priority is calculated to generate fault severity level data. Combined with the probability of occurrence of the fault, the scope of impact and the possible fault type, the priority of maintenance is determined through comprehensive evaluation. For faults with higher severity, repair operations will be arranged first to reduce the overall risk of the wiring harness system. The fault severity level data is associated with the repair suggestion rules in the fault type discrimination rule base for analysis to generate fault diagnosis report data. Based on the diagnosis report data, the fault location and fault type information are extracted to generate the final fault location information and fault type information.

[0026] In the embodiment of the present invention, by constructing a multi-node equivalent circuit model and combining it with Kirchhoff's law for modeling, the physical characteristics and electrical laws of the wiring harness system are effectively reflected, and the accuracy of fault feature extraction is improved. The wiring harness data is reconstructed by using a recursive variational graph autoencoder. The structure has a dual action mechanism, which can not only reconstruct the timing measurement value and topological relationship, but also serve as a noise filtering module, which significantly enhances the anti-interference ability of fault diagnosis. The model fusion distributed fault diagnosis algorithm and the expectation maximization algorithm are introduced. Through the collaborative work of multiple sub-models, the accurate capture of complex fault features is achieved, and the system's recognition ability for multi-type and multi-location faults is improved. Based on the feature extraction method of gradient descent, the automatic optimization and screening of fault features are achieved through adaptive learning rate and feature importance evaluation, and the adaptability of the diagnosis process is enhanced. A multi-layer verification and cross-analysis mechanism is adopted, combined with fault severity evaluation, a complete fault diagnosis decision chain is established, and the reliability and interpretability of the diagnosis results are improved.

[0027] In a specific embodiment, the process of executing step S100 may specifically include the following steps: Perform spatial distribution analysis on the physical positions of the measurement nodes to obtain measurement node distribution data, and divide the measurement node distribution data into line segments according to the wire harness topology structure to obtain wire harness segment data; Modeling the resistance, inductance and capacitance parameters of each wiring harness section in the wiring harness section data to obtain the wiring harness equivalent circuit parameters; Based on Kirchhoff's voltage law and current law, node voltage equations are constructed for the harness equivalent circuit parameters to obtain a node voltage equation group; Standardize the data collected by the voltage sensor and current sensor of the measurement node to obtain standardized measurement data, and solve the node voltage equation group according to the standardized measurement data to obtain the electrical characteristic parameters between the nodes; The physical connection relationship of nodes is analyzed based on the electrical characteristic parameters between nodes to obtain a physical connection matrix, and the physical connection matrix is ​​subjected to matrix synthesis operation with the electrical characteristic parameters between nodes to obtain a node connection relationship matrix.

[0028] Specifically, the physical location of the measurement node is spatially analyzed. With the help of the design drawings of the vehicle wiring harness or the actual measurement data, the three-dimensional coordinates (e.g. ) calibrate the physical position of each measurement node and generate measurement node distribution data. Assume that there is measurement nodes, and the position of each node is Indicates that Is a node Coordinates in three-dimensional space. These distribution data constitute the topological basis of the measurement node in space. According to the topological structure of the wire harness, the measurement node distribution data is divided into line segments according to the physical connection relationship. Each pair of physically connected measurement nodes and Form a line segment, expressed as , its physical length Calculated by the Euclidean distance formula: ; in, It is a wiring harness paragraph The physical length of and The measurement nodes are and The harness segment data is composed of The collection of , which reflects the basic connection structure of the entire harness system. After obtaining the harness segment data, Model resistance, inductance, and capacitance parameters. Calculated by the following formula: ; in, is the resistivity of the harness material; is the cross-sectional area of ​​the harness; is the length of the harness segment. The inductance of the harness It is estimated by the following formula: ; in, is the magnetic permeability of the wire harness material. Capacitance Modeling through dielectric properties between wire bundles: ; in, is the dielectric constant. Using these formulas, equivalent circuit parameters are generated for each harness segment, including , forming an equivalent circuit model of the entire wiring harness system. Based on these parameters, the node voltage equations are constructed according to Kirchhoff's voltage law (KVL) and Kirchhoff's current law (KCL). For any closed loop, according to KVL, there is: ; in, Is the flow through the harness segment For each node, according to KCL, we have: ; By combining these equations, a system of voltage equations for all nodes is established, which has the form: ; in, is the coefficient matrix containing resistance, inductance and capacitance, is the node voltage vector, is the input vector of the external current source. After the node voltage equations are established, the data collected by the voltage sensor and current sensor of the measurement node are standardized. The normalized value of each data point is calculated: ; in, is the standardized value, is the original sample value, is the mean, is the standard deviation. Remove the deviation caused by external factors in the data to obtain standardized measurement data. Based on the standardized data, solve the node voltage equations to obtain the electrical characteristic parameters between nodes, including the voltage of each node. and the current of each segment Based on the electrical characteristic parameters between nodes, the physical connection relationship of nodes is analyzed. By comparing the current, voltage and impedance relationship between nodes, it is determined whether there is an actual physical connection between nodes and a physical connection matrix is ​​formed. For example, if the node and Voltage difference Within a certain threshold range, the current If there is a significant value, then ,otherwise . Connect the physical matrix Perform matrix synthesis operation with the electrical characteristic parameters between nodes to obtain the node connection relationship matrix For example, if Representation Node and The equivalent impedance relationship between It contains both the topological information of the connection and reflects the electrical characteristics.

[0029] In a specific embodiment, the process of executing step S200 may specifically include the following steps: Classify and mark the voltage data and current data of the measurement node into open circuit fault, short circuit fault, insulation aging fault and poor contact fault to obtain fault type marking data; Based on the fault type label data, time series feature extraction is performed on the voltage data and current data of the measurement node to obtain voltage and current time series feature data; The voltage difference and phase difference between adjacent nodes of the measurement node are calculated and processed to obtain the difference feature data between nodes, and the voltage and current time series feature data and the difference feature data between nodes are subjected to feature fusion operation to obtain the fault feature vector; The impedance parameters of the node connection relationship matrix and the fault characteristic vector are calculated to obtain the node impedance characteristic data, and the standard parameter comparison analysis is performed based on the node impedance characteristic data to obtain the fault characteristic deviation matrix; A set of fault discrimination conditions is constructed according to the fault feature deviation matrix to obtain fault discrimination rules, and the fault discrimination rules are mapped and associated with the fault type label data to obtain a fault type discrimination rule base.

[0030] Specifically, the voltage and current data of the measurement nodes are classified and marked. By analyzing the characteristic effects of different fault types (open circuit, short circuit, insulation aging, poor contact) on voltage and current, classification rules are established for each fault type. For example, a circuit breaker fault is manifested as an abnormal increase in node voltage or current interruption. By setting a voltage threshold and current threshold For identification, the specific rules are: and , then mark the node There is an open circuit fault. Similarly, a short circuit fault is manifested as a sudden drop in node voltage and a sharp increase in current. According to the rule: and , then mark the node There is a short circuit fault. Insulation aging faults are manifested as leakage in the node current and increased voltage fluctuations. The rules are defined by measuring the leakage current and noise level between nodes. For example, if the leakage current , it is marked as insulation aging. Poor contact faults cause irregular voltage fluctuations, and the rules are obtained by analyzing the characteristic values ​​of the fluctuation amplitude and frequency. These classification and marking rules are automatically applied to the data of the measurement nodes to generate a marked data set containing the above fault types. Based on the fault type marking data, the voltage and current data of the measurement nodes are extracted for time series features to capture the dynamic characteristics of the data in the time dimension, such as the voltage fluctuation amplitude, frequency components, and current mutation behavior. Assume Represents the time series of node voltage. The time series features are extracted through Fourier transform or wavelet transform to extract frequency domain features, such as harmonic components ,in It is The amplitude of the harmonics, is the fundamental frequency. Similarly, the current data The instantaneous rate of change of Description is used to capture mutation characteristics. The results of time series feature extraction constitute voltage and current time series feature data, which describe the dynamic behavior of the node. After obtaining the time series feature data, the voltage difference and phase difference between adjacent nodes of the measurement node are calculated and processed to obtain the difference feature data between the nodes. For any pair of adjacent nodes and , the voltage difference By formula Calculate, where and Is a node and The instantaneous voltage value of It is determined by the following formula: ; in represents the inner product of two node voltages, and is the modulus length of each node. These difference feature data can capture the electrical characteristic relationship between nodes and provide an additional source of information for diagnosis. The voltage and current time series feature data are fused with the difference feature data between nodes to generate a fault feature vector. Feature fusion is achieved through weighted combination, for example: ; in is the final fault feature vector, is the time series feature data, is the difference feature data, and is a weighting coefficient used to balance the contribution of different features. and the fault feature vector Input impedance calculation module to calculate the impedance characteristic data of the node. and its adjacent node set ,impedance It is expressed as:

[0031] in is the voltage difference, is the current flowing through the wire bundle between nodes. By calculating the impedance characteristics of each node, a global impedance characteristic data set is generated. Based on the impedance characteristic data, a comparative analysis is performed with the standard parameters to construct a fault characteristic deviation matrix. Assume that the standard impedance is , the deviation matrix element It is expressed as: ; By analyzing the deviation matrix, the fault degree of each node is quantified. According to the fault feature deviation matrix, a set of fault discrimination conditions is constructed. For example, when When , it is judged as a circuit breaker fault; when , it is determined to be a short circuit fault. These conditions are mapped and associated with the fault type tag data to form a fault type discrimination rule base. The rule base can quickly match the fault feature vector with the known rules, so as to accurately diagnose the fault type and provide a basis for maintenance. For example, if it is detected that the voltage difference of a certain section of the harness is too large and the impedance deviation exceeds the standard, it is determined to be a circuit breaker fault according to the rule base, and specific maintenance suggestions are provided.

[0032] Among them, before performing feature analysis on the voltage data and current data of the measurement nodes according to the node connection relationship matrix, it also includes: inputting the voltage data and current data of the measurement nodes and the node connection relationship matrix into the auxiliary variable introduction module for nonlinear equation conversion to obtain a pseudo-linear measurement equation group describing the electrical relationship between the measurement nodes, the pseudo-linear measurement equation group includes voltage relationship terms, current relationship terms and impedance relationship terms between the nodes; performing Lp norm minimization optimization on each relationship parameter of the pseudo-linear measurement equation group to obtain an initial weight coefficient matrix of each measurement item, the initial weight coefficient matrix reflects the relative importance of different measurement items; performing weighted least squares iterative calculation on each parameter in the pseudo-linear measurement equation group according to the initial weight coefficient matrix to obtain the first-stage pseudo-linear least squares estimation result, the weighted least squares iterative calculation adopts an adaptive step size to update the parameters; comparing and analyzing the first-stage pseudo-linear least squares estimation result with the actual sampling data of the measurement node, the result is obtained. to the measurement error vector and the estimation accuracy matrix, which describes the estimation reliability of each measurement point; based on the measurement error vector and the estimation accuracy matrix, a parameter dependency function is constructed to obtain a dependency matrix reflecting the mutual influence between the measurement parameters, and the dependency matrix contains direct dependencies and indirect dependencies; the dependency matrix and the first-stage pseudo-linear least squares estimation result are input into the second-stage iterative reweighting module for Lp norm minimization calculation to obtain the second-stage weight coefficient considering the parameter dependency, and the second-stage weight coefficient reflects the importance of the parameter after considering the mutual dependence; the measurement error vector is weighted and optimized according to the second-stage weight coefficient to obtain the optimized estimation result including the electrical characteristics of the measurement node, and the optimized estimation result reduces the influence of measurement noise and interference; the iterative update amplitude of the optimized estimation result is calculated and analyzed, and when the parameter change amount of two consecutive iterations is less than the preset threshold, the current optimized estimation result is used as the final voltage data and current data.

[0033] In a specific embodiment, the process of executing step S300 may specifically include the following steps: The voltage data and current data are input into the time series feature encoding layer of the recursive variational graph autoencoder for dimensionality reduction to obtain the initial latent variables. The time series feature encoding layer includes three graph convolution layers, each of which uses the ReLU activation function and batch normalization. The initial latent variables and node connection relationship matrix are input into the variational inference layer of the recursive variational graph autoencoder for parameter calculation to obtain the mean vector and variance vector. The variational inference layer includes two fully connected layers and one graph attention layer. Perform reparameterized sampling operations based on the mean vector and variance vector to obtain random latent variables; The random latent variables are input into the topological relationship reconstruction layer of the recursive variational graph autoencoder to restore the topological features and obtain the topological relationship reconstruction data. The topological relationship reconstruction layer includes three graph deconvolution layers and one graph residual connection. The time series feature decoding operation is performed on the random latent variables to obtain the time series feature reconstruction data. The time series feature decoding operation uses three transposed convolutional layers, and each transposed convolutional layer uses the LeakyReLU activation function.

[0034] Specifically, the input voltage data and current data are preprocessed. These data are represented by nodes, where the voltage and current of each node are used as feature vectors. ,in Is a node The voltage value, Is a node Assume that the entire network has nodes, the initial input data is expressed as , and combined with the node connection relationship matrix The voltage data and current data are input into the temporal feature encoding layer of the recursive variational graph autoencoder. This layer consists of three graph convolution layers, which are used to extract the spatial correlation and temporal dynamic characteristics of node features. For the calculation of each layer, the graph convolution formula is: ; It is The node feature matrix of the layer, the initial input is is the normalized adjacency matrix, is the degree matrix; It is The weight matrix of the layer; is the activation function, using the ReLU (rectified linear unit) function, defined as After each graph convolution layer, batch normalization is used to normalize the feature distribution to accelerate training convergence and improve model stability. After three graph convolution layers, the output feature matrix Reduced to the initial latent variable ,in is the dimension of the latent variable. and node connection matrix Input the variational inference layer of the recursive variational graph autoencoder to calculate the mean vector of the latent variable and variance vector This layer consists of two fully connected layers and one graph attention layer. The fully connected layer captures the global characteristics of the latent variables through linear transformation and activation function. The calculation formula is: ; in is the weight matrix, is the bias vector, is the activation function. The graph attention layer uses the attention mechanism to assign different weights to each pair of nodes. The calculation formula is:

[0035] in, Is a node and The attention coefficient, is the attention weight vector, represents a vector concatenation operation, Is a node The set of neighbor nodes. By calculation, we get the mean vector and variance vector Based on the mean vector and variance vector, a reparameterized sampling operation is performed to generate random latent variables The sampling process is implemented through the reparameterization technique, and its formula is: ; in represents element-wise product, is a noise vector sampled from a standard normal distribution. This process ensures the differentiability of sampling, thus supporting gradient descent optimization. Input the topological relationship reconstruction layer of the recursive variational graph autoencoder to restore the topological features. This layer consists of three graph deconvolution layers and a graph residual connection. The graph deconvolution layer gradually restores the topological features of the nodes through the inverse convolution operation. The formula is: ; in It is The residual connection adds the initial topological information to the reconstruction result to ensure that no information is lost, and finally generates the topological relationship reconstruction data. At the same time, for the random hidden variable Perform time series feature decoding operations to restore time series features. The decoding process consists of three transposed convolution layers, each of which uses the LeakyReLU activation function, and its formula is: ; in, It is the reconstructed voltage and current timing characteristic data.

[0036] In a specific embodiment, the process of executing step S400 may specifically include the following steps: The time series feature reconstruction data and the topology relationship reconstruction data are input into the model and integrated into the feature classification module of the distributed fault diagnosis algorithm to divide the fault feature distribution and obtain multiple fault feature sub-models; The fault state of the vehicle wiring harness is modeled in subspace according to multiple fault characteristic submodels to obtain a fault subspace data set, and a probability distribution analysis is performed on the fault subspace data set to obtain a characteristic probability distribution matrix; The characteristic probability distribution matrix and multiple fault characteristic sub-models are input into the expectation maximization algorithm for iterative optimization to obtain the fault probability parameter, fault duration parameter and fault impact range parameter; The fault transfer probability is calculated for the diagnosis results of multiple fault feature sub-models to obtain a fault transfer probability matrix, which describes the propagation law of faults between different fault feature sub-models. Based on the fault probability parameter, fault duration parameter and fault impact range parameter, multiple fault feature sub-models are combined and optimized to obtain the weight coefficient of the fusion model; The output results of multiple fault feature sub-models are weighted and combined according to the fusion model weight coefficient and the fault transfer probability matrix to obtain the fusion fault diagnosis result; Performing a global state evaluation on the fused fault diagnosis results according to the time series and spatial distribution to obtain a global state evaluation index, and performing a global state mapping on the fused fault diagnosis results based on the global state evaluation index to obtain global fault state data; Gradient descent feature extraction is performed based on the global fault status data to obtain a standardized feature vector.

[0037] Specifically, the time series feature reconstruction data and the topology relationship reconstruction data are input into the feature classification module of the model fusion distributed fault diagnosis algorithm. Assume that the time series feature reconstruction data is ,in is the number of nodes, is the time series feature dimension, and the topological relationship reconstruction data is , describing the connection weights between nodes. In the feature classification module, combined with and Use clustering algorithms (such as K-Means or graph-based clustering methods) to distribute fault features. By partitioning, all node features are divided into sub-models, each sub-model Contains a set of features related to a specific fault mode. The fault state of the vehicle wiring harness is modeled in subspace according to the multiple fault feature submodels, and a fault subspace dataset is generated. The node set of the sub-model is , and its corresponding subspace feature matrix is ​​defined as By performing probability distribution analysis on these subspace data sets, the probability distribution of each feature in the subspace is calculated. It is Nodes in the submodel The characteristic vector of , then its probability distribution is obtained by kernel density estimation :

[0038] in is a kernel function (such as a Gaussian kernel), is the smoothing parameter. By calculating the probability distribution of all nodes, the characteristic probability distribution matrix is ​​formed , whose elements Representation Node and nodes In subspace The joint probability within. The feature probability distribution matrix and multiple fault signature sub-models Input into the expectation maximization (EM) algorithm for iterative optimization to obtain the failure probability parameter , Fault duration parameters and fault impact range parameters The EM algorithm initializes these parameters and defines the expected function ,in is a set of parameters, It is The parameter value of the iteration. The expected function is of the form:

[0039] By maximizing Update the parameters until convergence and obtain the optimized , and After obtaining these parameters, the fault transfer probability is calculated for the diagnosis results of multiple fault feature sub-models to generate the fault transfer probability matrix . Among them, the matrix elements Indicates that the fault is from the sub-model Transfer to The probability is calculated as:

[0040] in Is a node and The distance between them reflects the propagation law of faults in space. , Fault duration and the scope of the fault , combine and optimize multiple fault feature sub-models to obtain the fusion model weight coefficient :

[0041] Combination and the failover probability matrix , weighted combination of the output results of multiple fault feature sub-models:

[0042] in It is The diagnostic results of each sub-model are is the fusion fault diagnosis result. After obtaining the fusion fault diagnosis result, the global state evaluation is performed according to the time series and spatial distribution, and the global state evaluation index is calculated. The evaluation index includes fault coverage :

[0043] and fault concentration :

[0044] The fusion results are mapped to the global state through evaluation indicators to generate global fault state data Based on the global fault status data, the gradient descent method is used to extract features and optimize the feature vector The objective function is: ; The standardized feature vector is obtained by iterative optimization .

[0045] In a specific embodiment, the execution step performs gradient descent feature extraction based on the global fault state data to obtain a standardized feature vector, which may specifically include the following steps: The global fault state data is input into the objective function construction module to construct a dual-objective optimization function to obtain a feature optimization objective function, which includes a reconstruction error minimization term and a fault diagnosis accuracy maximization term; The partial derivatives of the feature parameters in the feature optimization objective function are calculated to obtain the gradient direction data of the feature parameters, and the change amplitude of the feature parameters is dynamically adjusted based on the gradient direction data to obtain the adaptive learning rate; Iteratively update and calculate the feature parameters according to the adaptive learning rate and gradient direction data to obtain the feature parameter update result, and rank and analyze the diagnostic contribution of each feature component in the feature parameter update result to obtain the feature importance ranking result; The feature parameter update results are threshold screened according to the feature importance ranking results to obtain a target feature parameter set, and the feature parameters in the target feature parameter set are normalized by mean variance to obtain a standardized feature vector.

[0046] Specifically, the global fault state data is input into the objective function building module to generate an optimization function containing two objectives. The global fault state data is represented as ,in is the number of failed nodes, is the feature dimension of each node. The first optimization objective is to minimize the reconstruction error, which is used to ensure the accurate reconstruction of the original data during the feature extraction process; the second is to maximize the fault diagnosis accuracy, which is used to optimize the diagnostic performance. These two objectives are jointly expressed by the following formula:

[0047] in, is the characteristic parameter matrix in the optimization process, is the projection matrix used for reconstruction, is the classification weight vector, Representation Node The fault category, and is a hyperparameter used to balance the two objectives. The gradient of the feature optimization objective function is calculated to obtain the gradient direction data of the feature parameters. about The partial derivative of is calculated as follows:

[0048] Among them, the first term represents the gradient contribution of the reconstruction error to the characteristic parameter, and the second term represents the gradient contribution of the fault diagnosis accuracy to the characteristic parameter. The gradient direction data can reflect the direction and amplitude that should be adjusted during the characteristic parameter optimization process. In order to improve the optimization efficiency, the change amplitude of the characteristic parameter is dynamically adjusted based on the gradient direction data to obtain the adaptive learning rate. The adaptive learning rate is calculated using the RMSProp or Adam algorithm. Taking RMSProp as an example, its update formula is:

[0049] in, It is The node at the iteration The learning rate, is the initial learning rate, is the exponentially weighted average of the squared gradients, is a small constant that prevents the denominator from being zero, is the weight decay factor. According to the adaptive learning rate And gradient direction data, iteratively update the feature parameters, the formula is:

[0050] This update process can converge to the local optimal solution of the optimization target after multiple iterations. The diagnostic contribution of each feature component in the feature parameter update result is sorted and analyzed to quantify the impact of each feature on fault diagnosis. The contribution is defined as the cumulative amplitude of the gradient change during the optimization process, for example: ; in It is a feature The diagnostic contribution of is the total number of iterations. Sort all features according to their contribution to obtain the feature importance ranking result. Based on the feature importance ranking result, perform threshold screening on the feature parameter update results and retain the features with a diagnostic contribution higher than the set threshold. The characteristic components of the target feature parameter set are formed The characteristic parameters in this set are normalized by mean and variance to eliminate dimension differences. The normalization formula is:

[0051] in, and Characteristics After standardization, the final standardized feature vector is obtained. .

[0052] In a specific embodiment, the process of executing step S500 may specifically include the following steps: The standardized feature vector is input into the fault diagnosis processing module to calculate the fault probability, and the fault probability distribution data of each section of the vehicle wiring harness is obtained; Calculate the matching degree of the fault probability distribution data of each section of the vehicle wiring harness and the fault characteristic rules in the fault type discrimination rule library to obtain the fault type matching degree data; The fault type matching degree data is input into the feature consistency verification module for multiple verification and analysis to obtain the fault feature verification result; The standardized feature vector is spatially mapped to the measurement node to obtain the physical positioning data of the measurement node, and the fault position is calibrated based on the physical positioning data to obtain the coordinate data of the fault position; Cross-validate the fault location coordinate data with the fault probability distribution data of each section of the vehicle wiring harness to obtain the fault location verification result; Calculate the fault repair priority based on the fault feature verification results and the fault location verification results to obtain the fault severity level data; The fault severity level data is correlated with the repair suggestion rules in the fault type discrimination rule base to obtain fault diagnosis report data, and the fault location and fault type information is extracted based on the fault diagnosis report data to obtain fault location information and fault type information.

[0053] Specifically, the standardized feature vector is input into the fault diagnosis processing module to calculate the fault probability of each section of the vehicle wiring harness. Assume that the standardized feature vector is ,in is the number of nodes, is the feature dimension; assuming that the topological connectivity of the bundle is represented by the adjacency matrix Description. The fault probability calculation is based on node characteristics and topological relationships, and is modeled through a probabilistic graph model (such as a conditional random field). The fault probability of each harness is defined as ,in Represents a connected node and The probability formula is expressed as:

[0054] in, is the characteristic function, describing the node and The feature interaction of is a weight vector. Through normalization, the fault probability distribution data of each section of the vehicle wiring harness is generated . For the failure probability distribution data The matching degree is calculated with the fault characteristic rules in the fault type discrimination rule base. Assume that each fault type defined in the rule base The characteristic rule is , which is related to the harness segment The matching degree is calculated by the following formula:

[0055] in, Is a harness segment Type of fault The matching degree of is in the range of [0,1]. Calculate the matching degree and obtain the matching degree data matrix ,in is the number of fault types. Input feature consistency verification module to perform multiple verification analysis. Feature consistency verification calculates the confidence of verification results by counting the consistency level between multiple matching results. The matching results exist The mean in this analysis is , the variance is , then the confidence is defined as:

[0056] Generate fault signature verification results On this basis, the standardized feature vector Perform spatial mapping of measurement nodes. Assume that the physical coordinates of each node are By combining the eigenvector and topological relationship, the line segment The calibration formula is:

[0057] in, Represents a harness segment Combined with the above calibration process, the fault location coordinate data is generated . The fault location coordinate data Failure probability distribution data Perform cross validation. Verify the accuracy of position calibration by comparing the correspondence between position coordinates and probability distribution. For example, if a harness segment The probability value If the value is significantly higher than the surrounding sections, the position calibration is considered reliable. The cross-validation results form the fault position verification result matrix . Verify results based on fault characteristics Verify the results with the fault location , calculate the fault repair priority of each harness section. The repair priority is defined by the comprehensive probability, matching degree and verification result: ; in, It is a section Maintenance priority, is the weight coefficient, which is used to balance the contribution of the three. , and obtain fault severity level data. Perform correlation analysis on the fault severity level data and the repair suggestion rules in the fault type discrimination rule base. Generate fault diagnosis report data by searching the rule base for suggestions corresponding to severity and fault type. The report data contains detailed information such as the fault type, severity, and repair suggestions of the wiring harness section. Based on the diagnostic report data, the specific fault location information and fault type information are extracted.

[0058] See also Figure 2 , Figure 2 A schematic block diagram of the structure of the vehicle wiring harness fault diagnosis device 200 provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the vehicle-mounted wiring harness fault diagnosis device 200 includes: A modeling module 210 is used to perform equivalent circuit modeling on the measurement nodes and wiring harness sections of the vehicle wiring harness to obtain a node connection relationship matrix; The analysis module 220 is used to perform feature analysis on the voltage data and current data of the measurement node according to the node connection relationship matrix to obtain a fault feature vector and a fault type discrimination rule base; The reconstruction module 230 is used to input the voltage data, the current data and the node connection relationship matrix into the recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topological relationship reconstruction data; A partitioning module 240 is used to perform distributed sub-model partitioning and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector; The matching module 250 is used to match and analyze the standardized feature vector with the fault type discrimination rule base to obtain fault location information and fault type information.

[0059] Through the collaborative cooperation of the above components, a multi-node equivalent circuit model is constructed and modeled in combination with Kirchhoff's law, which effectively reflects the physical characteristics and electrical laws of the wiring harness system and improves the accuracy of fault feature extraction. The recursive variational graph autoencoder is used to reconstruct the wiring harness data. This structure has a dual action mechanism, which can not only reconstruct the timing measurement value and topological relationship, but also serve as a noise filtering module, which significantly enhances the anti-interference ability of fault diagnosis. The model fusion distributed fault diagnosis algorithm and expectation maximization algorithm are introduced. Through the collaborative work of multiple sub-models, the accurate capture of complex fault features is achieved, and the system's recognition ability for multi-type and multi-location faults is improved. The feature extraction method based on gradient descent realizes the automatic optimization and screening of fault features through adaptive learning rate and feature importance evaluation, and enhances the adaptability of the diagnosis process. The multi-layer verification and cross-analysis mechanism is adopted, combined with the fault severity evaluation, to establish a complete fault diagnosis decision chain, which improves the reliability and interpretability of the diagnosis results.

[0060] See also Figure 3 , Figure 3 The present invention provides a schematic block diagram of the structure of an on-board wiring harness fault diagnosis device 300, which includes a processor 301 and a memory 302. The processor 301 and the memory 302 are connected via a device bus 303, wherein the memory 302 may include a non-volatile storage medium and an internal memory.

[0061] The non-volatile storage medium can store a computer program. The computer program includes program instructions. When the program instructions are executed by the processor 301, the processor 301 can execute any of the above-mentioned vehicle wiring harness fault diagnosis methods.

[0062] The processor 301 is used to provide computing and control capabilities to support the operation of the entire vehicle wiring harness fault diagnosis device 300 .

[0063] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor 301, the processor 301 can execute any of the above-mentioned vehicle wiring harness fault diagnosis methods.

[0064] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present application, and does not constitute a limitation on the vehicle-mounted wiring harness fault diagnosis device 300 involved in the scheme of the present application. The specific vehicle-mounted wiring harness fault diagnosis device 300 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0065] It should be understood that the processor 301 may be a central processing unit (CPU), and the processor 301 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0066] It should be noted that technicians in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the vehicle wiring harness fault diagnosis device 300 described above can refer to the corresponding process of the aforementioned vehicle wiring harness fault diagnosis method, and will not be repeated here.

[0067] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, the one or more processors implement the vehicle wiring harness fault diagnosis method provided in the embodiment of the present application.

[0068] The computer-readable storage medium may be an internal storage unit of the vehicle-mounted wiring harness fault diagnosis device 300 of the aforementioned embodiment, such as a hard disk or memory of the vehicle-mounted wiring harness fault diagnosis device 300. The computer-readable storage medium may also be an external storage device of the vehicle-mounted wiring harness fault diagnosis device 300, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped with the vehicle-mounted wiring harness fault diagnosis device 300.

[0069] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0070] If the integrated unit is implemented in the form of 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 the present application is essentially 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. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0071] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for diagnosing a vehicle wiring harness fault, characterized in that: include: Conduct equivalent circuit modeling on the measurement nodes and wiring harness sections of the vehicle wiring harness to obtain the node connection relationship matrix; Performing feature analysis on the voltage data and current data of the measurement node according to the node connection relationship matrix to obtain a fault feature vector and a fault type discrimination rule base; Inputting the voltage data, the current data and the node connection relationship matrix into a recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topological relationship reconstruction data; Performing distributed sub-model division and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector; The standardized feature vector is matched and analyzed with the fault type discrimination rule base to obtain fault location information and fault type information.

2. The vehicle wiring harness fault diagnosis method according to claim 1, characterized in that: The equivalent circuit modeling of the measurement nodes and harness sections of the vehicle harness is performed to obtain a node connection relationship matrix, including: Performing spatial distribution analysis on the physical positions of the measurement nodes to obtain measurement node distribution data, and dividing the measurement node distribution data into line segments according to the line bundle topology structure to obtain line bundle segment data; Modeling the resistance, inductance and capacitance parameters of each wiring harness section in the wiring harness section data to obtain wiring harness equivalent circuit parameters; Based on Kirchhoff's voltage law and current law, node voltage equations are constructed for the wiring harness equivalent circuit parameters to obtain a node voltage equation group; Standardizing the data collected by the voltage sensor and the current sensor of the measurement node to obtain standardized measurement data, and solving the node voltage equation group according to the standardized measurement data to obtain electrical characteristic parameters between nodes; A node physical connection relationship analysis is performed based on the electrical characteristic parameters between the nodes to obtain a physical connection matrix, and a matrix synthesis operation is performed on the physical connection matrix and the electrical characteristic parameters between the nodes to obtain a node connection relationship matrix.

3. The vehicle wiring harness fault diagnosis method according to claim 1, characterized in that: The characteristic analysis of the voltage data and the current data of the measurement node according to the node connection relationship matrix to obtain the fault characteristic vector and the fault type discrimination rule base includes: Classify and mark the voltage data and current data of the measurement node into open circuit fault, short circuit fault, insulation aging fault and poor contact fault to obtain fault type marking data; Extracting time series characteristics of voltage data and current data of the measurement node based on the fault type label data to obtain voltage and current time series characteristic data; Calculating and processing the voltage difference and phase difference between adjacent nodes of the measurement node to obtain inter-node difference feature data, and performing feature fusion operation on the voltage and current time series feature data and the inter-node difference feature data to obtain a fault feature vector; Perform impedance parameter calculation on the node connection relationship matrix and the fault characteristic vector to obtain node impedance characteristic data, and perform standard parameter comparison analysis based on the node impedance characteristic data to obtain a fault characteristic deviation matrix; A fault discrimination condition set is constructed according to the fault feature deviation matrix to obtain a fault discrimination rule, and the fault discrimination rule is mapped and associated with the fault type label data to obtain a fault type discrimination rule base.

4. The vehicle wiring harness fault diagnosis method according to claim 1, characterized in that: The step of inputting the voltage data, the current data and the node connection relationship matrix into a recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topological relationship reconstruction data includes: Inputting the voltage data and the current data into the time series feature encoding layer of the recursive variational graph autoencoder for dimensionality reduction processing to obtain initial latent variables, wherein the time series feature encoding layer includes three graph convolution layers, each of which uses a ReLU activation function and batch normalization; Inputting the initial latent variables and the node connection relationship matrix into the variational inference layer of the recursive variational graph autoencoder for parameter calculation to obtain a mean vector and a variance vector, wherein the variational inference layer includes two fully connected layers and one graph attention layer; Performing a reparameterized sampling operation based on the mean vector and the variance vector to obtain a random latent variable; Inputting the random latent variables into the topological relationship reconstruction layer of the recursive variational graph autoencoder to restore the topological features, and obtaining topological relationship reconstruction data, wherein the topological relationship reconstruction layer includes three graph deconvolution layers and one graph residual connection; A time series feature decoding operation is performed on the random latent variable to obtain time series feature reconstruction data, wherein the time series feature decoding operation uses three transposed convolutional layers, and each transposed convolutional layer uses a LeakyReLU activation function.

5. The vehicle wiring harness fault diagnosis method according to claim 1, characterized in that: The performing distributed sub-model division and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector includes: Inputting the time series feature reconstruction data and the topology relationship reconstruction data into the feature classification module of the model fusion distributed fault diagnosis algorithm to perform fault feature distribution division to obtain multiple fault feature sub-models; Performing subspace modeling on the vehicle wiring harness fault state according to the multiple fault feature submodels to obtain a fault subspace data set, and performing probability distribution analysis on the fault subspace data set to obtain a feature probability distribution matrix; Inputting the characteristic probability distribution matrix and the multiple fault characteristic sub-models into an expectation maximization algorithm for iterative optimization to obtain a fault occurrence probability parameter, a fault duration parameter, and a fault impact range parameter; Performing fault transfer probability calculation on the diagnosis results of the multiple fault feature sub-models to obtain a fault transfer probability matrix, wherein the fault transfer probability matrix describes the propagation law of the fault between different fault feature sub-models; Combining and optimizing the multiple fault characteristic sub-models based on the fault occurrence probability parameter, the fault duration parameter, and the fault impact range parameter to obtain a fusion model weight coefficient; Performing weighted combination on the output results of the plurality of fault feature sub-models according to the fusion model weight coefficient and the fault transfer probability matrix to obtain a fusion fault diagnosis result; Performing a global state evaluation on the fused fault diagnosis result according to the time series and spatial distribution to obtain a global state evaluation index, and performing a global state mapping on the fused fault diagnosis result based on the global state evaluation index to obtain global fault state data; Gradient descent feature extraction is performed based on the global fault state data to obtain a standardized feature vector.

6. The vehicle wiring harness fault diagnosis method according to claim 5, characterized in that: The step of performing gradient descent feature extraction based on the global fault state data to obtain a standardized feature vector includes: Inputting the global fault state data into an objective function construction module to construct a dual-objective optimization function to obtain a feature optimization objective function, wherein the feature optimization objective function includes a reconstruction error minimization term and a fault diagnosis accuracy maximization term; Performing partial derivative calculation on the feature parameters in the feature optimization objective function to obtain gradient direction data of the feature parameters, and dynamically adjusting the variation amplitude of the feature parameters based on the gradient direction data to obtain an adaptive learning rate; Iteratively updating and calculating the feature parameters according to the adaptive learning rate and the gradient direction data to obtain a feature parameter update result, and ranking and analyzing the diagnostic contribution of each feature component in the feature parameter update result to obtain a feature importance ranking result; The feature parameter update result is threshold screened according to the feature importance ranking result to obtain a target feature parameter set, and the feature parameters in the target feature parameter set are normalized by mean variance to obtain a normalized feature vector.

7. The vehicle wiring harness fault diagnosis method according to claim 1, characterized in that: The matching analysis of the standardized feature vector with the fault type discrimination rule base to obtain fault location information and fault type information includes: Inputting the standardized feature vector into a fault diagnosis processing module to calculate the fault probability, and obtaining the fault probability distribution data of each section of the vehicle wiring harness; Calculating the matching degree of the fault probability distribution data of each section of the vehicle wiring harness and the fault characteristic rules in the fault type discrimination rule base to obtain fault type matching degree data; Inputting the fault type matching degree data into a feature consistency verification module for multiple verification and analysis to obtain a fault feature verification result; Performing measurement node space mapping on the standardized feature vector to obtain physical positioning data of the measurement node, and calibrating the fault position based on the physical positioning data to obtain fault position coordinate data; Cross-validating the fault location coordinate data with the fault probability distribution data of each section of the vehicle wiring harness to obtain a fault location verification result; Calculate the fault repair priority according to the fault feature verification result and the fault location verification result to obtain fault severity level data; The fault severity level data is correlated with the repair suggestion rules in the fault type discrimination rule base to obtain fault diagnosis report data, and based on the fault diagnosis report data, information on the fault location and fault type is extracted to obtain fault location information and fault type information.

8. A vehicle-mounted wiring harness fault diagnosis device, characterized in that: The method for diagnosing a vehicle wiring harness fault according to any one of claims 1 to 7 comprises: A modeling module is used to perform equivalent circuit modeling on the measurement nodes and wiring harness sections of the vehicle wiring harness to obtain a node connection relationship matrix; An analysis module, used to perform feature analysis on the voltage data and current data of the measurement node according to the node connection relationship matrix to obtain a fault feature vector and a fault type discrimination rule base; A reconstruction module, used for inputting the voltage data, the current data and the node connection relationship matrix into a recursive variational graph autoencoder for data reconstruction to obtain time series feature reconstruction data and topological relationship reconstruction data; A partitioning module, used for performing distributed sub-model partitioning and model fusion on the time series feature reconstruction data and the topological relationship reconstruction data to obtain a standardized feature vector; The matching module is used to match and analyze the standardized feature vector with the fault type discrimination rule base to obtain fault location information and fault type information.

9. A vehicle-mounted wiring harness fault diagnosis device, characterized in that: The vehicle-mounted wiring harness fault diagnosis device comprises: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory to enable the vehicle wiring harness fault diagnosis device to execute the vehicle wiring harness fault diagnosis method according to any one of claims 1 to 7.

10. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the vehicle wiring harness fault diagnosis method as described in any one of claims 1-7 is implemented.

Images