A method and system for mining analysis of vehicle wiring harness failure data

Abstract

The present application relates to the technical field of data mining, in particular to a kind of vehicle wiring harness fault data mining analysis method and system, comprising the following steps, extract maintenance time stamp and path segment index to build two-dimensional coding matrix, section division is carried out to vehicle running mileage, generate section fault total number and probability sequence, according to node main circuit and branch line number execution logic comparison, configure path weight parameter, probability sequence and weight parameter are executed tensor operation, obtain aggregation degree parameter and build mining analysis result.In the present application, by reshaping data multidimensional space topology structure, and calculating dynamic recession mapping to capture deep derivative relationship, probability sequence and weight parameter are input into perception network to execute feature extraction, obtain aggregation degree index, according to feature collection operation, deeply mine the space associated conduction path of the deterioration of wiring harness state in different positions, form the fault data mining analysis system with network deep topology feature analysis ability.

Description

Technical Field

[0001] This invention relates to the field of data mining technology, and in particular to a method and system for mining and analyzing vehicle wiring harness fault data. Background Technology

[0002] Data mining technology involves extracting potential patterns, relationships, and statistical features from large amounts of structured or unstructured data. Its core aspects include data collection, data processing, feature extraction, statistical analysis, and pattern recognition. In practical applications, data from different devices or business systems is typically summarized and recorded, categorized and stored according to time sequence or event type, and then the inherent relationships between data are identified through statistical calculations, correlation comparisons, and trend analysis. This technology is widely used in scenarios such as industrial equipment operation record analysis, quality inspection record statistics, and fault record classification. By summarizing and analyzing historical data samples, patterns hidden in the data can be systematically organized and presented, thereby forming a data pattern system that can be used for further research and decision-making reference.

[0003] Traditional vehicle wiring harness fault data mining and analysis methods refer to the statistical organization and pattern analysis of wiring harness fault records generated during vehicle operation or testing. This typically involves collecting wiring harness fault codes, fault occurrence times, vehicle models, and maintenance record information from vehicle fault diagnostic boxes or maintenance records. Fault records from different vehicles are categorized and stored according to vehicle type, wiring harness location, and fault codes. The frequency of occurrence of fault records of the same category is then statistically summarized and sorted chronologically. The frequency of faults occurring in the same wiring harness location across different vehicle models is then compared and analyzed based on the statistical results. Simultaneously, records of wiring harness breakage, poor contact, and insulation damage are categorized and organized in the maintenance records. Finally, vehicle wiring harness fault data is summarized and analyzed through statistical tables and historical record comparisons.

[0004] Traditional analysis methods rely on manual review and basic tabular statistical classification when processing vehicle wiring harness fault records. This conventional frequency accumulation and time sorting comparison mode cannot deeply explore the inherent spatial topological relationships between nodes. Because it only performs shallow feature classification based on vehicle model and single fault code, it is difficult to accurately capture the hidden chain of related faults caused by the evolution of specific mileage. The coarse-grained state comparison method causes the feature dimensions of complex related faults to be isolated from each other, resulting in the inability to complete effective data reconstruction and feature mapping when facing deep fault derivative patterns. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose a method and system for mining and analyzing vehicle wiring harness fault data.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for mining and analyzing vehicle wiring harness fault data, comprising the following steps:

[0007] S1: Extract vehicle code data and wiring harness node numbers from vehicle fault diagnosis box and maintenance records, parse wiring harness node numbers to construct path segment index sequence, cluster code data time difference to generate event chain segment number, and combine path segment index sequence and event chain segment number to construct two-dimensional coding matrix;

[0008] S2: Divide the mileage records to generate mileage segment numbers, calculate the number of node fault triggers under a single mileage segment number in the two-dimensional coding matrix, accumulate the number of node fault triggers, and generate the total number of segment faults;

[0009] S3: Call the coordinate position set by the two-dimensional encoding matrix, compare the number of node fault triggers with the total number of faults in the section, generate a single node ratio result, arrange the single node ratio results, and construct a node fault probability sequence;

[0010] S4: Parse the harness node number, extract the main loop number and branch line number, extract the target node combination according to the node failure probability sequence, perform logic comparison, if the main loop number and branch line number are the same, configure the first constant, if only the main loop number is the same, configure the second constant, otherwise configure the third constant, and generate path weight parameters;

[0011] S5: Integrate the node failure probability sequence and path weight parameters, input them into the multilayer perceptron model to perform calculations, generate node product values, merge the node product values ​​into the clustering parameter, and arrange the clustering parameter to construct the failure data mining and analysis results.

[0012] The present invention improves upon this invention by including the following: the two-dimensional encoding matrix includes a topological mapping vector, span evolution coordinates, and spatial location identifiers; the total number of segment faults includes an overall loss benchmark, cumulative fatigue frequency, and global degradation index; the node fault probability sequence includes local risk propensity, life cycle decay curve, and spatial vulnerability characteristics; the path weight parameters include structural coupling coefficients, hierarchical correlation factors, and spatial linkage influence; and the fault data mining and analysis results include a core high-risk map, a state evolution benchmark, and a health assessment baseline.

[0013] The present invention is improved in that the specific steps for obtaining the two-dimensional encoding matrix are as follows:

[0014] S111: Extract vehicle code data and wiring harness node numbers from vehicle fault diagnosis box and maintenance records. Perform character segmentation on wiring harness node numbers to extract branch path segment numbers. Calculate the hierarchical position parameters attached to the branch path segment numbers. Sort the branch path segment numbers according to the hierarchical position parameters, establish hierarchical mapping relationship, and generate path segment index sequence.

[0015] S112: Extract maintenance timestamp values ​​based on the vehicle reporting data. Sort the extracted multiple sets of maintenance timestamp values ​​in ascending order according to their value size. Calculate the correlation difference between adjacent maintenance timestamp values. Call a preset time threshold and compare the difference with the time threshold. When the difference is lower than the time threshold, aggregate the corresponding vehicle reporting data into the same set. When the difference is higher than the time threshold, establish an independent set. Assign a corresponding identifier to each set to obtain the event chain segment number.

[0016] S113: Call the path segment index sequence and event chain segment number to construct a two-dimensional projection surface for the intersection of the path segment index sequence and event chain segment number, extract all distributed coordinates within the two-dimensional projection surface, assign status values ​​to the distributed coordinates, change the status value of the distributed coordinates with existing fault records to one, change the status value of the distributed coordinates with missing fault records to zero, establish a topological association structure based on the arrangement pattern of status values, and obtain a two-dimensional encoding matrix.

[0017] The present invention is improved in that the step of obtaining the total number of faults in the section is specifically as follows:

[0018] S211: Extract vehicle mileage records, sort the feature values ​​in the vehicle mileage records in ascending order, obtain the preset segment length parameter, perform modular operation and boundary interval division on the sorted vehicle mileage records according to the segment length parameter, assign incremental identifiers according to the division sequence, and establish mileage segment numbers.

[0019] S212: Call the two-dimensional encoding matrix, extract the elements of the abnormal state mapping matrix inside the two-dimensional encoding matrix, delineate the boundary of the numerical interval according to the mileage segment number, perform slicing and truncation on the matrix elements, obtain segmented sub-matrix feature data, perform non-zero numerical distribution frequency statistical operation on the segmented sub-matrix feature data within a single mileage segment number, and generate the node fault trigger count.

[0020] S213: Based on the same mileage segment number range, lock the internal branch node sequence of the interval, call the fault trigger count of the corresponding index position node to construct the interval node fault frequency accumulation array, perform linear summation arithmetic operation on all scalar element values ​​inside the interval node fault frequency accumulation array, map the summation output value to the vertical variable position of the corresponding interval coordinate system, and obtain the total number of segment faults.

[0021] The present invention is improved in that the step of obtaining the node failure probability sequence is specifically as follows:

[0022] S311: Call the coordinate position set by the two-dimensional coding matrix, extract the number of node fault triggers and the total number of section faults, collect the node compressive strength and real-time stress value, monitor the insulation layer wear and initial insulation layer thickness, detect the local vibration frequency and reference resonance frequency, and calculate the single node ratio result.

[0023] S312: For the single node ratio result, obtain the harness node number and mileage segment number, lock the cross positioning point according to the horizontal axis associated with the harness node number and the vertical axis associated with the mileage segment number, write the single node ratio result into the storage area mapped by the cross positioning point, and establish a state mapping feature set.

[0024] S313: Call the mileage segment number to extract the increasing sorting pattern, and perform linear extraction and recombination on the single node ratio result recorded in the state mapping feature set according to the increasing sorting pattern, merge discrete numerical terms, and construct a node failure probability sequence.

[0025] The present invention is improved in that the formula for obtaining the ratio result of a single node is specifically as follows:

[0026] ;

[0027] in, This represents the ratio result of a single node. This represents the number of times a node failure has been triggered. The total number of faults in the representative section. The normalized value representing the real-time stress. The normalized value representing the compressive strength of the node. Normalized value representing the amount of insulation wear. The normalized value representing the initial thickness of the insulation layer. The normalized value representing the local vibration frequency. This represents the normalized value of the reference resonant frequency.

[0028] The present invention is improved in that the step of obtaining the path weight parameter is specifically as follows:

[0029] S411: Call the harness node number, perform cutting and isolation on its internal character sequence, extract the independent characters of the first segment to form the main loop number, cut off the remaining characters of the second segment to form the branch line number, obtain the preset topology rule, perform key-value alignment on the node mapping of the main loop number and the branch line number according to the topology rule, and establish a node topology identifier set;

[0030] S412: Based on the node failure probability sequence, obtain a preset proportional coefficient, sort the values ​​within the node failure probability sequence in descending order, call a preset risk threshold value to truncate and filter the sorted values, lock the corresponding nodes above the threshold, perform cross-pairing operation on the nodes, and obtain the target node combination.

[0031] S413: For the target node combination, call the node topology identifier set to extract the main loop number and branch line number corresponding to the paired node, perform equality judgment comparison on the main loop number and branch line number within the paired node respectively, read the first constant when the main loop number and branch line number are the same, read the second constant when the main loop number is the same and the branch line number is different, read the third constant when the main loop number is different, assign the extracted constant to the corresponding link, and generate path weight parameters.

[0032] The present invention is improved in that the steps for obtaining the fault data mining and analysis results are specifically as follows:

[0033] S511: Call the node failure probability sequence, extract the sequence element values, set the network hidden layer connection weights based on the path weight parameters, perform matrix inner product operation on the sequence element values ​​and connection weights to extract linear transformation feature quantities, collect preset scalars as bias term parameters, perform arithmetic addition operation on the bias term parameters and linear transformation feature quantities to obtain intermediate network tensors, compare the tensor values ​​with zero values, remove negative values, and obtain the node product values;

[0034] S512: For the target node combination, extract the internal node mapping coordinates, lock the feature temporary storage interval according to the mapping coordinates, call the node product value to extract the scalar term, perform aggregation extraction on all node scalar terms within the coverage of the same target node combination, load the aggregation term into the feature temporary storage interval and perform linear accumulation and summation arithmetic calculation to generate the clustering degree parameter.

[0035] S513: Obtain the node traversal search directory, extract the clustering parameter corresponding to all nodes based on the search directory, sort the parameter values ​​in descending order according to the size rule, obtain the position number, assign the feature dimension mapping index to the node according to the position number, align and concatenate the feature dimension mapping index with the clustering parameter, construct the data tuple array, and establish the fault data mining analysis results.

[0036] A system for mining and analyzing vehicle wiring harness fault data, the system being used to implement the aforementioned method for mining and analyzing vehicle wiring harness fault data, the system comprising:

[0037] The two-dimensional coding processing module extracts vehicle code data and wiring harness node numbers from the vehicle fault diagnosis box and maintenance records, parses the wiring harness node numbers to construct a path segment index sequence, clusters the time difference of the code data, generates event chain segment numbers, and combines the path segment index sequence and event chain segment numbers to construct a two-dimensional coding matrix.

[0038] The section fault summary module divides mileage records to generate mileage section numbers, calculates the number of node fault triggers under a single mileage section number in the two-dimensional coding matrix, accumulates the number of node fault triggers, and generates the total number of section faults.

[0039] The fault probability analysis module calls the coordinate position set by the two-dimensional coding matrix, compares the number of node fault triggers with the total number of faults in the section, generates a single node ratio result, arranges the single node ratio results, and constructs a node fault probability sequence.

[0040] The path weight allocation module parses the harness node number, extracts the main loop number and branch line number, extracts the target node combination and performs logic comparison according to the node failure probability sequence, if the main loop number and branch line number are the same, configures the first constant, if only the main loop number is the same, configures the second constant, otherwise configures the third constant, and generates path weight parameters.

[0041] The clustering analysis module integrates the node failure probability sequence and path weight parameters, inputs them into the multilayer perceptron model to perform calculations, generates node product values, merges the node product values ​​into the clustering parameters, and arranges the clustering parameters to construct the failure data mining analysis results.

[0042] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0043] In this invention, a two-dimensional coding matrix is ​​constructed by extracting maintenance timestamps and path segment indices to reshape the multidimensional spatial topology of the data. Segmentation is performed based on vehicle mileage, and node failure probability sequences are calculated. A dynamic decay mapping between harness nodes is established to accurately capture deep-level derivative relationships. The main circuit and branch line numbers of the harness nodes are extracted and analyzed. Path weight parameters are configured by logical comparison. The probability sequences and weight parameters are input into the perception network, and tensor operations are performed to extract the clustering index. Based on multidimensional feature aggregation operations, the spatial transmission paths of harness state degradation at different locations are mined, forming a fault data mining and analysis system with deep topology analysis capabilities. Attached Figure Description

[0044] Figure 1 This is a flowchart of the method of the present invention;

[0045] Figure 2 This is a flowchart illustrating how the present invention obtains a two-dimensional encoding matrix;

[0046] Figure 3 This is a flowchart for obtaining the total number of faults in a section according to the present invention;

[0047] Figure 4 This is a flowchart illustrating the process of obtaining the node failure probability sequence according to the present invention;

[0048] Figure 5 This is a flowchart illustrating the process of obtaining path weight parameters according to the present invention;

[0049] Figure 6 This is a flowchart illustrating the process of obtaining fault data mining and analysis results according to the present invention. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0051] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0052] Please see Figure 1 This invention provides a technical solution, a method for mining and analyzing vehicle wiring harness fault data, comprising the following steps:

[0053] S1: Extract vehicle code data and wiring harness node numbers from vehicle fault diagnosis box and maintenance records. Perform parsing operation on wiring harness node numbers to extract branch path segment numbers and construct path segment index sequence. Extract maintenance timestamp values ​​from vehicle code data and perform ascending sort operation. Perform difference comparison calculation on maintenance timestamp values ​​of adjacent positions. Perform clustering operation based on comparison operation results to generate event chain segment numbers. Input path segment index sequence and event chain segment numbers into one-hot coding algorithm to construct two-dimensional coding matrix.

[0054] S2: Collect vehicle mileage records and perform ascending sorting operation. Perform segment division operation on the vehicle mileage records according to the set mileage segment length to generate mileage segment numbers. Extract matrix elements in the two-dimensional coding matrix according to the mileage segment numbers and perform data partitioning operation. Calculate the number of node fault triggers in the two-dimensional coding matrix under a single mileage segment number. Perform cumulative calculation on the number of node fault triggers for all nodes in the same mileage segment number to generate the total number of segment faults.

[0055] S3: Call the coordinate position set by the two-dimensional encoding matrix, perform ratio calculation on the number of node fault triggers to the total number of section faults to generate a single node ratio result, write the single node ratio result into the matrix position where the harness node number and mileage section number are intersected, and perform a permutation operation on the single node ratio result in the matrix position according to the sorting rule of the mileage section number to construct a node fault probability sequence.

[0056] S4: Perform a parsing operation on the harness node number to extract the main circuit number and branch line number. Based on the node failure probability sequence, extract the target node combination and perform a logical comparison operation. When the main circuit number and branch line number associated with the target node combination are the same, configure the first constant as the path weight parameter. When the main circuit number is the same and the branch line number is different, configure the second constant as the path weight parameter. When the main circuit number is different, configure the third constant as the path weight parameter.

[0057] S5: Extract the sequence element values ​​within the node failure probability sequence, input the sequence element values ​​and path weight parameters into the multilayer perceptron model to perform calculation operations to generate node product values, accumulate the node product values ​​to the clustering parameter associated with the target node combination, and perform distribution and arrangement operations on the clustering parameters of all nodes to construct the fault data mining and analysis results.

[0058] The two-dimensional coding matrix includes topological mapping vectors, span evolution coordinates, and spatial location identifiers. The total number of segment faults includes overall loss benchmarks, cumulative fatigue frequency, and global degradation indicators. The node fault probability sequence includes local risk tendency, life cycle decay curves, and spatial vulnerability characteristics. The path weight parameters include structural coupling coefficients, hierarchical correlation factors, and spatial linkage influence. The fault data mining and analysis results include core high-risk maps, state evolution benchmarks, and health assessment baselines.

[0059] Please see Figure 2 The specific steps for obtaining the two-dimensional encoding matrix are as follows:

[0060] S111: Extract vehicle code data and wiring harness node numbers from vehicle fault diagnosis box and maintenance records. Perform character segmentation on wiring harness node numbers to extract branch path segment numbers. Calculate the hierarchical position parameters attached to the branch path segment numbers. Sort the branch path segment numbers according to the hierarchical position parameters, establish hierarchical mapping relationship, and generate path segment index sequence.

[0061] The system continuously listens for and captures diagnostic fault code messages in real time via the vehicle controller area network bus interface at a baud rate of 500kbps, accessing the underlying diagnostic communication protocol. For the captured hexadecimal message data, a pre-defined fault code parsing dictionary is used to perform bitwise masking and offset extraction operations, separating the vehicle code data and the accompanying hexadecimal wiring harness node numbers. The hexadecimal wiring harness node numbers are converted into a unified US Information Interchange Standard (USI) code string format. Fixed delimiters (e.g., hyphens "-") within the string are located, and a string splitting function is called to divide the complete string into multiple independent character substrings. The set of segmented character substrings is read, and the numeric character at the second index position is extracted as the branch path segment number. The alphanumeric character at the third index position is extracted as the hierarchical position parameter. A lexicographical mapping is performed on the hierarchical position parameters, converting hierarchical identifiers such as "L1", "L2", and "L3" into integer scalars of 1, 2, and 3. A structured array is created with integer scalars as the primary key and branch path segment numbers as the value keys. The quicksort algorithm is called, with two pointers pointing to the first and last ends of the array respectively. A pivot value is selected, and the elements within the structured array are sorted in ascending order based on their primary keys. If the primary key values ​​are the same, a second ascending order is performed based on the value keys, completing the hierarchical mapping relationship for all branch path segment numbers. The sorted structured array is traversed, and the branch path segment numbers are sequentially appended to a contiguous memory address space, generating a one-dimensional path segment index sequence.

[0062] S112: Extract maintenance timestamp values ​​based on vehicle reporting data. Sort the extracted multiple sets of maintenance timestamp values ​​in ascending order according to their value size. Calculate the correlation difference between adjacent maintenance timestamp values. Call a preset time threshold and compare the difference with the time threshold. When the difference is lower than the time threshold, aggregate the corresponding vehicle reporting data into the same set. When the difference is higher than the time threshold, establish an independent set. Assign a corresponding identifier to each set to obtain the event chain segment number.

[0063] Extract the time strings stored in a unified standard time format from the maintenance record file. Call the timestamp conversion function to convert all time strings into second-level maintenance timestamp values ​​calculated from 00:00 on January 1, 1970. Allocate dynamic memory space to store the extracted 100 sets of maintenance timestamp values, and call the merge sort algorithm to sort the 100 sets of maintenance timestamp values ​​in strict ascending order from left to right. Initialize the cursor variable, traverse the ascending-sorted maintenance timestamp values, extract the value at the current index position from the value at the previous index position, and perform a difference calculation: 1710000000 - 1709913600 = 86400, to obtain the correlation difference between adjacent maintenance timestamp values. Set a preset time threshold of 259200 seconds (i.e., 3 days), and compare the obtained correlation difference of 86400 seconds with 259200 seconds. Under this operational logic, if the difference 86400 is less than 259200, the memory address pointer of the corresponding vehicle code data is redirected to the same hash table key, completing the aggregation of the same set. If the subsequently calculated difference is 300000, which is greater than 259200, a new hash table key is instantiated in the heap memory to establish an independent set. For all independent or aggregated sets generated in memory, a global auto-incrementing counter is called to write unsigned integer identifiers to each set starting from 1, and the event chain segment number composed of the identifiers is output.

[0064] S113: Call the path segment index sequence and event chain segment number to establish a two-dimensional projection surface for the intersection of the path segment index sequence and event chain segment number, extract all distributed coordinates within the two-dimensional projection surface, assign status values ​​to the distributed coordinates, change the status value of the distributed coordinates with existing fault records to one, change the status value of the distributed coordinates with missing fault records to zero, establish a topological association structure based on the status value arrangement pattern, and obtain a two-dimensional encoding matrix.

[0065] The system reads the one-dimensional path segment index sequence stored in memory as the horizontal X-axis coordinate domain and the event chain segment number as the vertical Y-axis coordinate domain. By calculating the Cartesian product of the two one-dimensional arrays, a two-dimensional continuous space is dynamically allocated in memory to establish a two-dimensional projection surface that intersects the path segment index sequence and the event chain segment number. For the two-dimensional projection surface, a double loop is executed to traverse all distributed coordinates (X, Y) within it, retrieving the vehicle code data area pointed to by the corresponding coordinate. For each coordinate point, a status value assignment operation is performed: the underlying log corresponding to the coordinate is retrieved; if the underlying log contains a non-empty fault identifier, the status value of the distributed coordinate is changed from the default initial state to 1; if the underlying log contains an empty value or only contains a normal status confirmation code, the status value of the distributed coordinate is changed to 0. After completing the traversal, the distributed coordinate array consisting of all 0s and 1s is extracted and serialized into a regular gridded topological association structure according to the absolute increasing order of the X and Y axes. The final output is a two-dimensional encoding matrix with a size of M rows and N columns.

[0066] Please see Figure 3 The specific steps for obtaining the total number of faults in a section are as follows:

[0067] S211: Extract vehicle mileage records, sort the feature values ​​in the vehicle mileage records in ascending order, obtain the preset segment length parameter, perform modular operation and boundary interval division on the sorted vehicle mileage records according to the segment length parameter, assign incremental identifiers according to the division sequence, and establish mileage segment numbers.

[0068] The vehicle's mileage records are accessed via the vehicle terminal interface and stored in the non-volatile memory within the electronic control unit. A dataset containing 500 historical mileage registration nodes is obtained. Feature values ​​representing cumulative mileage are extracted from the dataset, and a heap sort algorithm is used to sort all feature values ​​in ascending order. A preset segment length parameter of 5000 kilometers is set. For the first sorted feature value (e.g., 12345 kilometers), a modulo operation and boundary interval partitioning logic are used: 12345 / 5000 = 2.469 is calculated, and the result is rounded down to 2. This feature value is then assigned to the third boundary interval with a lower boundary of 10000 kilometers and an upper boundary of 15000 kilometers. Following this modulo operation logic, all feature values ​​are iterated and mapped to their corresponding boundary intervals. For all established boundary intervals, integer identifiers (1, 2, 3, etc.) are assigned sequentially in ascending order starting from 0 for each lower boundary value, thus completing the establishment of mileage segment numbers.

[0069] S212: Call the two-dimensional encoding matrix, extract the elements of the abnormal state matrix inside the two-dimensional encoding matrix, delineate the boundary of the numerical interval according to the mileage segment number, perform slicing and truncation on the matrix elements, obtain the segmented sub-matrix feature data, perform non-zero numerical distribution frequency statistical operation on the segmented sub-matrix feature data within a single mileage segment number, and generate the node fault trigger count.

[0070] The system retrieves a ready-to-use two-dimensional encoding matrix based on a memory pointer, scanning the matrix for elements containing the mapping exception state matrix element 1. It reads the mileage segment number obtained in the previous step (e.g., number 3, representing the 10,000 km to 15,000 km range), retrieves all event chain segment numbers occurring within this mileage range, and delineates the numerical interval boundaries in the two-dimensional encoding matrix based on the Y-axis coordinates corresponding to these event chain segment numbers. Using array slicing instructions, it performs row-level slicing on the matrix elements within the delineated boundaries, generating segmented sub-matrix feature data containing only fault records within the current mileage segment. For the segmented sub-matrix feature data generated within a single mileage segment number (e.g., number 3), it performs a non-zero numerical distribution frequency statistical operation column-wise (i.e., by path segment index sequence): it initializes a counter array with the same length as the maximum dimension of the X-axis, accumulates the number of 1s within each column of the sliced ​​matrix, and outputs a one-dimensional integer array composed of the accumulated results of all columns; this array represents the number of node fault triggers.

[0071] S213: Based on the same mileage section number range, lock the internal branch node sequence, call the fault trigger count of the corresponding index position node to construct the interval node fault frequency accumulation array, perform linear summation arithmetic operation on all scalar element values ​​inside the interval node fault frequency accumulation array, map the summation output value to the vertical variable position of the corresponding interval coordinate system, and obtain the total number of section faults.

[0072] Lock all branch node sequences within the same mileage segment number (e.g., number 3) participating in the calculation. Based on the X-axis index position of each node in the branch node sequence, extract the corresponding node fault trigger count from the one-dimensional integer array generated in the previous step. For example, extract the fault trigger counts for nodes 1 to 5 as 2, 0, 1, 3, and 0, respectively. Allocate contiguous space in memory to construct an array accumulating the interval node fault frequency, and write the extracted numerical sequence [2, 0, 1, 3, 0] completely into this array. Perform linear summation arithmetic on all scalar elements in this array: 2 + 0 + 1 + 3 + 0 = 6. Use the calculated output value 6 as the ordinate variable, and perform a Cartesian coordinate mapping operation with mileage segment number 3 as the abscissa variable. Store the final value 6 in the specified memory address to obtain the total number of segment faults for this segment.

[0073] Please see Figure 4 The specific steps for obtaining the node failure probability sequence are as follows:

[0074] S311: Using the coordinates set by the two-dimensional coding matrix, extract the number of node fault triggers and the total number of section faults; collect the node compressive strength and real-time stress values; monitor the insulation layer wear and initial insulation layer thickness; detect the local vibration frequency and reference resonant frequency; using the following formula:

[0075] ;

[0076] The calculation yields the ratio result for a single node.

[0077] in, This represents the ratio result of a single node. This represents the number of times a node failure has been triggered. The total number of faults in the representative section. The normalized value representing real-time stress is obtained by acquiring real-time stress physical data on the nodal surface through strain gauge sensors and inputting it into an extremum normalization function for dimensionless mapping processing. The normalized value representing the compressive strength of a node is obtained by extracting the physical threshold of compressive strength recorded in the preset component file and inputting it into the same extreme value normalization function for dimensionless mapping. The normalized value representing the amount of insulation wear is obtained by detecting the wear thickness of the current wire harness's outer insulation layer and dividing it by the nominal maximum outer diameter, followed by proportional scaling. The normalized value representing the initial thickness of the insulation layer is obtained by reading the initial thickness dimension from the corresponding wire harness factory record file and dividing it by the maximum nominal outer diameter, followed by proportional scaling. The normalized value representing the local vibration frequency is obtained by monitoring the real-time vibration Hertz value at the wire harness connection using a piezoelectric sensor and processing it using the Z-score normalization algorithm. The normalized value representing the reference resonance frequency is obtained by calling the structural intrinsic resonance Hertz value registered in the modal analysis test file and processing it using the Z-score normalization algorithm.

[0078] The system locates a specific (X, Y) coordinate position set by the two-dimensional encoding matrix and extracts the node fault trigger count (e.g., extract a value of 5) and the total number of segment faults (e.g., extract a value of 100) from the corresponding memory stack. Simultaneously, it drives an external sensor network to collect corresponding physical quantity data: real-time stress physical data is collected through a miniature metal foil strain gauge sensor attached to the wire harness surface, obtaining a value of 120 MPa; the physical threshold of the compressive strength of this type of wire harness, set to 200 MPa, is retrieved from the preset component file. Based on this, the system calls an extreme value normalization function to process the stress data, setting the minimum value to 0 MPa and the maximum value to 300 MPa. The calculation 120 / 300 = 0.4 yields the normalized value of the real-time stress, and the calculation 200 / 300 = 0.667 yields the normalized value of the node compressive strength. The actual wear thickness of the outer insulation layer of the current wiring harness was measured to be 0.5 mm using a laser thickness gauge. The initial insulation thickness was read from the factory record file as 2.0 mm, and the maximum nominal outer diameter was extracted as 5.0 mm. The normalized value of the insulation wear was calculated as 0.5 / 5.0 = 0.1. The normalized value of the initial insulation thickness was then calculated as 2.0 / 5.0 = 0.4. A piezoelectric accelerometer installed near the node was used to monitor the real-time vibration frequency at 55 Hz. The natural resonant frequency registered in the modal analysis test file was 50 Hz. The Z-score normalization algorithm (mean set to 45, standard deviation set to 10) was used to calculate (55-45) / 10 = 1.0, obtaining the normalized value of the local vibration frequency. The normalized value of the reference resonant frequency was then calculated as (50-45) / 10 = 0.5.

[0079] Table 1. Detailed list of sensor data and normalization results:

[0080]

[0081] Table 1 lists the raw physical quantities acquired by various sensors and their values ​​after dimensionless or standardized processing.

[0082] Extract all the parameters after the above processing, and solve for the result by arithmetic substitution according to the given formula:

[0083] ;

[0084] The calculated ratio for a single node is 2.25. This result indicates that the current node, under conditions of high-frequency vibration and a certain degree of insulation wear, has a significantly higher risk of failure evolution than the normal operating baseline. The advantage of this formula lies in its effective enhancement of the weight of early physical degradation characteristics in the overall failure assessment by introducing the relative deviation between stress and strength and combining it with the square root amplification effect of the wear ratio.

[0085] S312: For the single node ratio result, obtain the harness node number and mileage section number, lock the cross positioning point according to the horizontal axis associated with the harness node number and the vertical axis associated with the mileage section number, write the single node ratio result into the storage area mapped by the cross positioning point, and establish a state mapping feature set.

[0086] Extract the calculated output value of 2.25, and simultaneously retrieve the harness node number (e.g., 12) and corresponding mileage segment number (e.g., 3) of the current processing target from the associated context memory dictionary. Using harness node number 12 as the horizontal X-axis coordinate index and mileage segment number 3 as the vertical Y-axis coordinate index, execute coordinate addressing instructions in the pre-initialized relational database table structure to accurately locate the intersection point. Obtain write permissions to the memory block corresponding to the intersection point, and write the single node ratio result 2.25 directly into the storage area mapped to the intersection point in double-precision floating-point format. Repeat the above physical write instructions until all intersection points are covered, and establish a complete state mapping feature set composed of double-precision floating-point numbers in the database.

[0087] S313: Call the mileage segment number to extract the increasing sorting pattern, perform linear extraction and recombination on the single node ratio results recorded in the state mapping feature set according to the increasing sorting pattern, merge discrete numerical terms, and construct a node failure probability sequence;

[0088] The system reads the ordinate dimension data from the state mapping feature set, retrieves the mileage segment numbers, and extracts their increasing sorting pattern from 1 to the maximum segment number (e.g., increasing from 1 to 20) according to an integer auto-incrementing rule. It then locks a specific harness node number (e.g., number 12) with a fixed abscissa dimension. Following the aforementioned increasing sorting pattern, it recursively reads the single-node ratio results recorded when the mileage segment numbers are 1 to 20 from the database's continuous addresses using a cursor. For example, it reads a series of discrete double-precision floating-point values ​​such as 0.5, 0.8, and 2.25. These 20 discrete value items are concatenated in the memory buffer according to their chronological order and stored in a one-dimensional floating-point array, completing the linear extraction and recombination operation. Finally, it disables write access to this one-dimensional floating-point array and outputs the memory address pointer, thus constructing the specific node fault probability sequence.

[0089] Please see Figure 5 The specific steps for obtaining the path weight parameters are as follows:

[0090] S411: Call the harness node number, perform cutting and isolation on its internal character sequence, extract the independent characters of the first segment to form the main loop number, cut off the remaining characters of the second segment to form the branch line number, obtain the preset topology rules, perform key-value alignment on the node mapping main loop number and branch line number according to the topology rules, and establish a node topology identifier set;

[0091] Retrieve the harness node number string sequence (e.g., "PwrMain-Branch04") from the underlying hardware topology configuration file. Execute a string splitting and isolation instruction using a preset hyphen "-" as the delimiter, extracting the independent characters "PwrMain" before the delimiter and storing them in the main loop number variable; extract the remaining characters "Branch04" after the delimiter and store them in the branch line number variable. Read the preset topology rule file stored in JSON format from disk and parse its hierarchical dependencies. According to the parsed topology rules, initialize a hash map table in memory, setting the extracted main loop number "PwrMain" as the primary key and the branch line number "Branch04" as the subkey. Perform key-value alignment with the original node absolute index values ​​to construct a nested dictionary data structure, completing the establishment of the node topology identifier set.

[0092] S412: Based on the node failure probability sequence, obtain the preset proportional coefficient, sort the values ​​within the node failure probability sequence in descending order, call the preset risk limit value to truncate and filter the sorted values, lock the corresponding nodes above the limit, perform cross-pairing operation on the nodes, and obtain the target node combination.

[0093] The algorithm reads the ratio results stored in the node failure probability sequence and retrieves the preset ratio coefficient of 1.0 from the configuration list. It then uses a quicksort algorithm to sort the values ​​within the sequence in descending order, resulting in a sorted value such as [2.25, 1.8, 1.5, 0.9, 0.4]. A preset risk threshold of 1.0 is invoked, and a traversal filtering logic is executed: the sorted values ​​are compared from left to right with 1.0. If a value less than 1.0 is encountered (e.g., 0.9), a truncation instruction is immediately triggered, discarding all data ending at 0.9. The values ​​above the threshold (2.25, 1.8, 1.5) are mapped back to their absolute node indices, locking these three nodes. A double loop iterates through these three nodes, combining them in pairs to form data pairs such as (node ​​A, node B), (node ​​A, node C), and (node ​​B, node C). Cross-pairing operations are performed, and reflexive and symmetric duplicates are removed. The final output is an array of target node combinations consisting entirely of unique data pairs.

[0094] S413: For the target node combination, call the node topology identifier set to extract the main loop number and branch line number corresponding to the paired node. Perform equality comparison on the main loop number and branch line number within the paired node. When the main loop number and branch line number are the same, read the first constant. When the main loop number is the same and the branch line number is different, read the second constant. When the main loop number is different, read the third constant. Assign the extracted constant to the corresponding link to generate path weight parameters.

[0095] The first constant is obtained by extracting the reference conduction coefficient between the two ends inside a single physical connection based on the preset physical attribute file of the wire harness, inputting the extracted reference conduction coefficient into the normalization function to perform mapping processing, and directly setting the dimensionless value output by the mapping as the first constant.

[0096] The second constant is obtained by calling a preset circuit interference feature set, retrieving the electromagnetic coupling attenuation factor between different parallel branches under the same power supply network, performing absolute value extraction and scaling calculation on the electromagnetic coupling attenuation factor, and setting the calculated output value as the second constant.

[0097] The third constant is obtained by acquiring the pre-stored full vehicle wiring harness insulation test log file, extracting the leakage current transfer rate noise parameter between independent power supply circuits under physical isolation, and setting the leakage current transfer rate noise parameter as the third constant after smoothing filtering and standardization.

[0098] For a specific data pair in the target node combination array (e.g., containing node A and node B), the node topology identifier set generated in the previous stage is invoked to extract the main loop number "PwrMain" and branch line number "Branch04" for node A, and the main loop number "PwrMain" and branch line number "Branch05" for node B. The string equality comparison function is initiated: the first comparison is performed, and the main loop number of node A equals the main loop number of node B; the second comparison is performed, and the branch line number of node A does not equal the branch line number of node B. For branches that meet the condition of "same main loop number and different branch line number," an instruction to read the second constant is issued according to the logical branch control flow.

[0099] Table 2. Comparison of Constant Value Assignment and Acquisition Sources:

[0100]

[0101] Table 2 shows the criteria for determining node pairs at different topological levels and the specific constant values ​​assigned to them.

[0102] When reading the first constant, the system directly accesses the wiring harness physical attribute archive database, extracts the measured baseline continuity coefficient between specified physical connection endpoints as 0.08 ohms, inputs it into the normalization function: calculate 1.0 - 0.08 / 1.6 = 0.95, and maps the output dimensionless value 0.95 as the first constant. When reading the second constant, the system accesses the circuit interference feature set, retrieves the electromagnetic coupling attenuation factor between parallel branches as -40 dB, extracts the absolute value as 40, performs a scaling calculation of 40 / 100 = 0.4, then subtracts this value from 1 to calculate 1.0 - 0.4 = 0.60, which is set as the second constant. When reading the third constant, the system parses the full vehicle wiring harness insulation test log file, extracts the leakage current transfer rate noise floor parameter under physical isolation as 0.3 μA, and after moving average smoothing filtering, the value stabilizes at 0.3 μA. Standardization is then performed to calculate 0.3 / 2.0 = 0.15, and 0.15 is set as the third constant. The selected constant (such as 0.60 in the current case) is assigned to the corresponding link variable representing the connection strength of the node, thus generating the path weight parameter matrix.

[0103] Please see Figure 6 The specific steps for obtaining the results of fault data mining and analysis are as follows:

[0104] S511: Call the node failure probability sequence, extract the sequence element values, set the network hidden layer connection weights based on the path weight parameters, perform matrix inner product operation on the sequence element values ​​and connection weights to extract linear transformation feature quantities, collect preset scalars as bias term parameters, perform arithmetic addition operation on the bias term parameters and linear transformation feature quantities to obtain intermediate network tensors, compare the tensor values ​​with zero values, remove negative values, and obtain the node product values;

[0105] A one-dimensional input vector (e.g., containing the values ​​[2.25, 1.8, 1.5]) is constructed from all sequence elements in the node failure probability sequence. The connection matrix constructed from the aforementioned path weight parameters is then used as the hidden layer connection weight matrix. In the microprocessor's vector operation unit, the inner product operation between the one-dimensional input vector and the connection weight matrix is ​​performed: for example, the calculation process for the first node is 2.25*0.95+1.8*0.60+1.5*0.15=2.1375+1.08+0.225=3.4425, and the result 3.4425 is extracted as a linear transformation feature. A preset scalar constant -1.2 is acquired via the read-only memory bus as a bias term parameter. An arithmetic addition operation is performed between the bias term parameter -1.2 and the linear transformation feature 3.4425: 3.4425+(-1.2)=2.2425, obtaining the intermediate network tensor. The Rectified Linear Activation Function (ReLU) is invoked to compare the tensor value 2.2425 with zero. Since 2.2425 is greater than 0, the logic for eliminating negative values ​​is not triggered, and the original value of 2.2425 is retained, resulting in the nodal product value. If the sum is negative, it is forcibly overwritten as 0.

[0106] S512: For the target node combination, extract the internal node mapping coordinates, lock the feature temporary storage interval according to the mapping coordinates, call the node product value to extract the scalar term, perform aggregation extraction on all node scalar terms within the coverage of the same target node combination, load the aggregation term into the feature temporary storage interval and perform linear accumulation and summation arithmetic calculation to generate the clustering degree parameter.

[0107] Read the memory mapping address corresponding to the target node combination and extract the mapping coordinates of all nodes within the combination. Based on the boundary extreme values ​​of the mapping coordinates, dynamically allocate and lock an independently addressable feature temporary storage interval in memory. For the node product value array output in the previous step, extract all independent scalar items through memory offsets. Retrieve all node scalar items currently within the coverage of the same target node combination (e.g., containing three values: 2.2425, 1.8500, and 0.5000), perform aggregation extraction operations on these values, and write them into the feature temporary storage interval. Within the temporary storage interval, call the accumulator hardware instruction to perform linear summation arithmetic calculation on these three floating-point numbers: 2.2425 + 1.8500 + 0.5000 = 4.5925. Directly assign the output result 4.5925 to the pre-declared aggregate feature variable to generate the clustering degree parameter.

[0108] S513: Obtain the node traversal search directory, extract the clustering parameters corresponding to all nodes based on the search directory, sort the parameter values ​​in descending order according to the size rule, obtain the position number, assign the feature dimension mapping index to the node according to the position number, align and concatenate the feature dimension mapping index with the clustering parameter, construct the data tuple array, and establish the fault data mining analysis results.

[0109] The node traversal directory list is read from the root directory. Based on the index pointers provided by this directory list, the set of clustering parameters corresponding to all nodes is extracted from memory. For all parameter values ​​in the set (e.g., including 4.5925, 3.1200, 5.8800), the merge sort algorithm is called to perform a strict descending sort according to the numerical size, resulting in the sorted sequence [5.8800, 4.5925, 3.1200]. The sorted sequence is traversed, and a loop counter is used to assign position numbers starting from 1 (i.e., first, second, third place). Based on the physical topological coordinates of the original node (e.g., X1Y1), feature dimension mapping indices are assigned to the nodes according to the position numbers. A structure array space is allocated in memory, with the feature dimension mapping index as a prefix and the clustering parameter as the data body. Alignment and concatenation are performed using memory copy instructions. For example, by splicing together the data structure [1, X1Y1, 5.8800], and cyclically generating a data tuple array of all nodes, the tuple array is encapsulated and output to establish a complete fault data mining and analysis result file, which can be directly called for subsequent visualization or maintenance decisions.

[0110] A system for mining and analyzing vehicle wiring harness fault data, the system being used to implement the aforementioned method for mining and analyzing vehicle wiring harness fault data, the system comprising:

[0111] The two-dimensional coding processing module extracts vehicle code data and wiring harness node numbers from the vehicle fault diagnosis box and maintenance records, parses the wiring harness node numbers to construct a path segment index sequence, clusters the time difference of the code data, generates event chain segment numbers, and combines the path segment index sequence and event chain segment numbers to construct a two-dimensional coding matrix.

[0112] The section fault summary module divides mileage records to generate mileage section numbers, calculates the number of node fault triggers under a single mileage section number in the two-dimensional coding matrix, accumulates the number of node fault triggers, and generates the total number of section faults.

[0113] The fault probability analysis module calls the coordinate positions set by the two-dimensional coding matrix, compares the number of node fault triggers with the total number of faults in the section, generates single-node ratio results, arranges the single-node ratio results, and constructs a node fault probability sequence.

[0114] The path weight allocation module parses the harness node number, extracts the main loop number and branch line number, extracts the target node combination based on the node failure probability sequence, performs logic comparison, if the main loop number and branch line number are the same, configures the first constant, if only the main loop number is the same, configures the second constant, otherwise configures the third constant, and generates path weight parameters.

[0115] The clustering analysis module integrates the node failure probability sequence and path weight parameters, inputs them into the multilayer perceptron model to perform calculations, generates node product values, merges the node product values ​​into the clustering parameters, and arranges the clustering parameters to construct the failure data mining analysis results.

[0116] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for mining and analyzing vehicle wiring harness fault data, characterized in that, Includes the following steps: S1: Extract vehicle code data and wiring harness node numbers from vehicle fault diagnosis box and maintenance records, parse wiring harness node numbers to construct path segment index sequence, cluster code data time difference to generate event chain segment number, and combine path segment index sequence and event chain segment number to construct two-dimensional coding matrix; S2: Divide the mileage records to generate mileage segment numbers, calculate the number of node fault triggers under a single mileage segment number in the two-dimensional coding matrix, accumulate the number of node fault triggers, and generate the total number of segment faults; S3: Call the coordinate position set by the two-dimensional encoding matrix, compare the number of node fault triggers with the total number of faults in the section, generate a single node ratio result, arrange the single node ratio results, and construct a node fault probability sequence; S4: Parse the harness node number, extract the main loop number and branch line number, extract the target node combination according to the node failure probability sequence, perform logic comparison, if the main loop number and branch line number are the same, configure the first constant, if only the main loop number is the same, configure the second constant, otherwise configure the third constant, and generate path weight parameters; S5: Integrate the node fault probability sequence and path weight parameters, input them into the multilayer perceptron model to perform calculations, generate node product values, merge the node product values ​​into the clustering parameters, and arrange the clustering parameters to construct the fault data mining and analysis results. The specific steps for obtaining the fault data mining and analysis results are as follows: S511: Call the node failure probability sequence, extract the sequence element values, set the network hidden layer connection weights based on the path weight parameters, perform matrix inner product operation on the sequence element values ​​and connection weights to extract linear transformation feature quantities, collect preset scalars as bias term parameters, perform arithmetic addition operation on the bias term parameters and linear transformation feature quantities to obtain intermediate network tensors, compare the tensor values ​​with zero values, remove negative values, and obtain the node product values; S512: For the target node combination, extract the internal node mapping coordinates, lock the feature temporary storage interval according to the mapping coordinates, call the node product value to extract the scalar term, perform aggregation extraction on all node scalar terms within the coverage of the same target node combination, load the aggregation term into the feature temporary storage interval and perform linear accumulation and summation arithmetic calculation to generate the clustering degree parameter. S513: Obtain the node traversal search directory, extract the clustering parameter corresponding to all nodes based on the search directory, sort the parameter values ​​in descending order according to the size rule, obtain the position number, assign the feature dimension mapping index to the node according to the position number, align and concatenate the feature dimension mapping index with the clustering parameter, construct the data tuple array, and establish the fault data mining analysis results.

2. The method for mining and analyzing vehicle wiring harness fault data according to claim 1, characterized in that, The two-dimensional coding matrix includes a topological mapping vector, span evolution coordinates, and spatial location identifiers. The total number of segment faults includes an overall loss benchmark, cumulative fatigue frequency, and global degradation index. The node fault probability sequence includes local risk propensity, life cycle decay curve, and spatial vulnerability characteristics. The path weight parameters include structural coupling coefficient, hierarchical correlation factor, and spatial linkage influence. The fault data mining and analysis results include a core high-risk map, state evolution benchmark, and health assessment baseline.

3. The method for mining and analyzing vehicle wiring harness fault data according to claim 2, characterized in that, The specific steps for obtaining the two-dimensional encoding matrix are as follows: S111: Extract vehicle code data and wiring harness node numbers from vehicle fault diagnosis box and maintenance records. Perform character segmentation on wiring harness node numbers to extract branch path segment numbers. Calculate the hierarchical position parameters attached to the branch path segment numbers. Sort the branch path segment numbers according to the hierarchical position parameters, establish hierarchical mapping relationship, and generate path segment index sequence. S112: Extract maintenance timestamp values ​​based on the vehicle reporting data. Sort the extracted multiple sets of maintenance timestamp values ​​in ascending order according to their value size. Calculate the correlation difference between adjacent maintenance timestamp values. Call a preset time threshold and compare the difference with the time threshold. When the difference is lower than the time threshold, aggregate the corresponding vehicle reporting data into the same set. When the difference is higher than the time threshold, establish an independent set. Assign a corresponding identifier to each set to obtain the event chain segment number. S113: Call the path segment index sequence and event chain segment number to construct a two-dimensional projection surface for the intersection of the path segment index sequence and event chain segment number, extract all distributed coordinates within the two-dimensional projection surface, assign status values ​​to the distributed coordinates, change the status value of the distributed coordinates with existing fault records to one, change the status value of the distributed coordinates with missing fault records to zero, establish a topological association structure based on the arrangement pattern of status values, and obtain a two-dimensional encoding matrix.

4. The method for mining and analyzing vehicle wiring harness fault data according to claim 3, characterized in that, The specific steps for obtaining the total number of faults in the section are as follows: S211: Extract vehicle mileage records, sort the feature values ​​in the vehicle mileage records in ascending order, obtain the preset segment length parameter, perform modular operation and boundary interval division on the sorted vehicle mileage records according to the segment length parameter, assign incremental identifiers according to the division sequence, and establish mileage segment numbers. S212: Call the two-dimensional encoding matrix, extract the elements of the abnormal state mapping matrix inside the two-dimensional encoding matrix, delineate the boundary of the numerical interval according to the mileage segment number, perform slicing and truncation on the matrix elements, obtain segmented sub-matrix feature data, perform non-zero numerical distribution frequency statistical operation on the segmented sub-matrix feature data within a single mileage segment number, and generate the node fault trigger count. S213: Based on the same mileage segment number range, lock the internal branch node sequence of the interval, call the fault trigger count of the corresponding index position node to construct the interval node fault frequency accumulation array, perform linear summation arithmetic operation on all scalar element values ​​inside the interval node fault frequency accumulation array, map the summation output value to the vertical variable position of the corresponding interval coordinate system, and obtain the total number of segment faults.

5. The method for mining and analyzing vehicle wiring harness fault data according to claim 4, characterized in that, The specific steps for obtaining the node failure probability sequence are as follows: S311: Call the coordinate position set by the two-dimensional coding matrix, extract the number of node fault triggers and the total number of section faults, collect the node compressive strength and real-time stress value, monitor the insulation layer wear and initial insulation layer thickness, detect the local vibration frequency and reference resonance frequency, and calculate the single node ratio result. S312: For the single node ratio result, obtain the harness node number and mileage segment number, lock the cross positioning point according to the horizontal axis associated with the harness node number and the vertical axis associated with the mileage segment number, write the single node ratio result into the storage area mapped by the cross positioning point, and establish a state mapping feature set. S313: Call the mileage segment number to extract the increasing sorting pattern, and perform linear extraction and recombination on the single node ratio result recorded in the state mapping feature set according to the increasing sorting pattern, merge discrete numerical terms, and construct a node failure probability sequence.

6. The method for mining and analyzing vehicle wiring harness fault data according to claim 5, characterized in that, The specific formula for obtaining the ratio result of a single node is as follows: ； in, This represents the ratio result of a single node. This represents the number of times a node failure has been triggered. The total number of faults in the representative section. The normalized value representing the real-time stress. The normalized value representing the compressive strength of the node. Normalized value representing the amount of insulation wear. The normalized value representing the initial thickness of the insulation layer. The normalized value representing the local vibration frequency. This represents the normalized value of the reference resonant frequency.

7. The method for mining and analyzing vehicle wiring harness fault data according to claim 5, characterized in that, The specific steps for obtaining the path weight parameters are as follows: S411: Call the harness node number, perform cutting and isolation on its internal character sequence, extract the independent characters of the first segment to form the main loop number, cut off the remaining characters of the second segment to form the branch line number, obtain the preset topology rule, perform key-value alignment on the node mapping of the main loop number and the branch line number according to the topology rule, and establish a node topology identifier set; S412: Based on the node failure probability sequence, obtain a preset proportional coefficient, sort the values ​​within the node failure probability sequence in descending order, call a preset risk threshold value to truncate and filter the sorted values, lock the corresponding nodes above the threshold, perform cross-pairing operation on the nodes, and obtain the target node combination. S413: For the target node combination, call the node topology identifier set to extract the main loop number and branch line number corresponding to the paired node, perform equality judgment comparison on the main loop number and branch line number within the paired node respectively, read the first constant when the main loop number and branch line number are the same, read the second constant when the main loop number is the same and the branch line number is different, read the third constant when the main loop number is different, assign the extracted constant to the corresponding link, and generate path weight parameters.

8. A system for mining and analyzing vehicle wiring harness fault data, characterized in that, The system is used to implement the vehicle wiring harness fault data mining and analysis method according to any one of claims 1-7, the system comprising: The two-dimensional coding processing module extracts vehicle code data and wiring harness node numbers from the vehicle fault diagnosis box and maintenance records, parses the wiring harness node numbers to construct a path segment index sequence, clusters the time difference of the code data, generates event chain segment numbers, and combines the path segment index sequence and event chain segment numbers to construct a two-dimensional coding matrix. The section fault summary module divides mileage records to generate mileage section numbers, calculates the number of node fault triggers under a single mileage section number in the two-dimensional coding matrix, accumulates the number of node fault triggers, and generates the total number of section faults. The fault probability analysis module calls the coordinate position set by the two-dimensional coding matrix, compares the number of node fault triggers with the total number of faults in the section, generates a single node ratio result, arranges the single node ratio results, and constructs a node fault probability sequence. The path weight allocation module parses the harness node number, extracts the main loop number and branch line number, extracts the target node combination and performs logic comparison according to the node failure probability sequence, if the main loop number and branch line number are the same, configures the first constant, if only the main loop number is the same, configures the second constant, otherwise configures the third constant, and generates path weight parameters. The clustering analysis module integrates the node failure probability sequence and path weight parameters, inputs them into the multilayer perceptron model to perform calculations, generates node product values, merges the node product values ​​into the clustering parameters, and arranges the clustering parameters to construct the failure data mining analysis results.

Images