Automobile wire harness electrical detection method

Abstract

The application relates to the technical field of automobile wire harness detection, and discloses an automobile wire harness electrical property detection method. The method acquires a plurality of electrical property parameter data points of an automobile wire harness, divides a wire harness operation process into a stable time period and a change time period by analyzing the change trend of the electrical property parameter data points, carries out feature fusion processing on the collected electrical property parameter data points in the stable time period, integrates the correlation information between different parameters, generates a fusion feature representation capable of comprehensively reflecting the stable operation state of the wire harness, takes the fusion feature representation generated in the stable time period as a reference basis, calculates the abnormality possibility of the electrical property parameter data points in the change time period, and outputs a risk feature vector containing abnormal information, constructs a defect propagation model based on the risk feature vector, predicts a defect propagation path, generates an adaptive detection threshold according to the prediction result of the defect propagation path and in combination with the current operation state of the wire harness, and realizes accurate detection of the electrical property performance of the automobile wire harness.

Description

Technical Field

[0001] This invention relates to the field of automotive wiring harness testing technology, specifically to a method for testing the electrical properties of automotive wiring harnesses. Background Technology

[0002] As a crucial component of automotive electrical systems, automotive wiring harnesses are responsible for transmitting power and signals, and their electrical performance directly affects the overall operating status and safety of the vehicle. With the continuous improvement of automotive intelligence and electrification, the structure of wiring harnesses is becoming increasingly complex, with a significant increase in the number of wires and connector types, and the corresponding requirements for monitoring electrical parameters are becoming more refined.

[0003] Electrical testing of automotive wiring harnesses primarily employs either offline sampling inspection or online fixed-point monitoring. Offline sampling inspection requires disassembling the wiring harness from the vehicle for testing, which is not only cumbersome and time-consuming but also fails to reflect real-time electrical changes in the wiring harness under actual operating conditions. Furthermore, the limited sampling ratio makes it difficult to cover all potential quality issues, easily leading to missed detections. While online fixed-point monitoring enables real-time detection, it mostly focuses on collecting and judging single data points for a few fixed electrical parameters (such as voltage and current), lacking dynamic analysis of parameter changes.

[0004] During actual vehicle operation, the electrical parameters of the wiring harness exhibit dynamic changes due to factors such as operating conditions (e.g., starting, acceleration, braking), ambient temperature fluctuations, vibration, and shock. These parameters exist in periods of stable operation and periods of rapid change. Existing detection methods fail to effectively distinguish between these two periods. During stable periods, they only assess individual parameter data in isolation, neglecting the correlation between different parameters, thus failing to comprehensively capture the overall electrical characteristics of the wiring harness. During periods of change, the lack of reference based on historical stable characteristics makes it difficult to accurately identify the difference between normal fluctuations and abnormal changes, often leading to misjudgments or delayed assessments.

[0005] When potential defects are detected in wiring harnesses, current technologies cannot further analyze the possible propagation paths of the defects or predict their impact on other electrical components. Repairs can only be carried out after a malfunction has already occurred, increasing vehicle maintenance costs and safety hazards. Furthermore, existing detection methods often use fixed thresholds set manually based on experience, which cannot be dynamically adjusted according to the actual operating status and parameter trends of the wiring harness. This results in poor adaptability to the testing needs of different vehicle models and operating conditions, further reducing the reliability and applicability of the testing. Summary of the Invention

[0006] The purpose of this invention is to provide a method for testing the electrical properties of automotive wiring harnesses, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a method for electrical testing of automotive wiring harnesses, the method comprising:

[0008] Acquire multiple electrical parameter data points of the automotive wiring harness;

[0009] The data points of electrical parameters are divided into stable time periods and changing time periods based on their changing trends.

[0010] Within a stable time period, feature fusion processing is performed on electrical parameter data points to generate fused feature representations;

[0011] Within the time period of change, the probability of anomalies is calculated based on the fused feature representation, and a risk feature vector is output.

[0012] A defect propagation model is constructed based on risk feature vectors to predict defect propagation paths;

[0013] An adaptive detection threshold is generated based on the prediction results to achieve electrical detection.

[0014] Preferably, the acquisition of multiple electrical parameter data points of the automotive wiring harness includes:

[0015] Electrical parameter data points include voltage, current, and resistance values;

[0016] The voltage, current and resistance values ​​of the vehicle's wiring harness are collected in real time by sensors to form time series data points.

[0017] Preferably, the division of stable time periods and changing time periods based on the changing trends of electrical parameter data points includes:

[0018] Calculate the first-order difference value of the electrical parameter data point sequence, and select the moments with positive first-order difference values ​​as growth moments;

[0019] Sort the growth moments according to the first difference value, calculate the second difference value of the sorted sequence, and take the growth moment corresponding to the maximum value of the second difference value as the critical moment.

[0020] The period before the critical moment is a stable period, while the period after the critical moment is a changing period.

[0021] Preferably, the step of performing feature fusion processing on electrical parameter data points within a stable time period to generate a fused feature representation includes:

[0022] Extract multi-dimensional features of electrical parameter data points within a stable time period, including temporal and spatial features;

[0023] A dynamic weighted algorithm is used to fuse multi-dimensional features and calculate the correlation weights of features in different dimensions.

[0024] The preliminary fused features are obtained by weighting the multi-dimensional features according to the relevance weights.

[0025] Redundant information is removed by dimensionality reduction, generating a low-dimensional fusion feature representation.

[0026] Preferably, the step of calculating the probability of anomalies based on fused feature representation and outputting a risk feature vector within the changing time period includes:

[0027] Compare the distance between electrical parameter data points during the change period and the fused feature representations during the stable period;

[0028] Calculate the probability of anomalies for each electrical parameter data point based on the distance value;

[0029] The probability of anomalies in all electrical parameter data points is aggregated to form a risk feature vector.

[0030] Preferably, the step of constructing a defect propagation model based on risk feature vectors and predicting defect propagation paths includes:

[0031] The risk feature vector is mapped to the three-dimensional spatial coordinates of the automotive wiring harness to generate a preliminary defect concentration distribution.

[0032] The initial defect concentration distribution is dynamically corrected based on historical defect data to obtain the spatiotemporal defect concentration distribution.

[0033] A graph model is constructed based on the automotive wiring harness topology, where nodes represent key locations in the wiring harness and edges represent connections.

[0034] An attention mechanism is used to capture the dependency relationship of defect propagation between nodes, and an initial prediction of defect propagation is output.

[0035] By combining time series prediction models, defect propagation paths and diffusion trends are generated.

[0036] Preferably, generating the adaptive detection threshold based on the prediction result includes:

[0037] Train a normal operating condition generator to simulate normal operating data of automotive wiring harnesses;

[0038] Build a real-time discriminator to compare prediction results with generator data and learn the boundaries between normal and abnormal;

[0039] The results are segmented and judged using a clustering algorithm to generate multi-level adaptive detection thresholds.

[0040] Preferably, the implementation of electrical detection further includes:

[0041] The detection stages are divided according to the adaptive detection threshold, including the first stage, the second stage, and the third stage.

[0042] Different detection strategies are applied to electrical parameter data points at different detection stages.

[0043] Preferably, the step of dividing the detection stage according to the adaptive detection threshold includes:

[0044] The layer with the highest node density in the defect propagation model is set as the first stage;

[0045] The layer with the most child nodes after the first phase is designated as the second phase.

[0046] The second stage is followed by the third stage at the bottom layer.

[0047] Preferably, applying different detection strategies to electrical parameter data points at different detection stages includes:

[0048] The first type of detection strategy is used for the electrical parameter data points in the first stage;

[0049] The second type of detection strategy is used for the electrical parameter data points in the second stage;

[0050] A third type of detection strategy is adopted for the electrical parameter data points in the third stage.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] In the time-segmentation stage, this method can accurately distinguish between stable and changing time periods based on the changing trends of electrical parameter data points, breaking away from the traditional "one-size-fits-all" approach to parameter changes. This time-segmentation processing method allows subsequent detection and analysis to better reflect the actual operating characteristics of the harness. Within stable time periods, feature fusion processing integrates information from multiple electrical parameter data points, moving beyond isolated judgments of single parameters. It fully explores the inherent relationships between different parameters, generating a more comprehensive and accurate fused feature representation that reflects the overall electrical characteristics of the harness under stable operating conditions. This avoids feature omissions caused by single-parameter analysis and provides a more reliable reference benchmark for subsequent anomaly detection.

[0053] Within the changing time period, the probability of anomalies is calculated based on the fused feature representation generated during the stable time period, which can effectively distinguish between normal fluctuations and abnormal changes in parameters. Since the fused feature representation covers the multi-parameter correlation information during the stable operation of the harness, it can be used as a reference to more accurately identify the degree to which parameters deviate from the normal range within the changing time period. The output risk feature vector can clearly present the abnormal information, avoiding the misjudgment or delayed judgment problems caused by the lack of reference benchmarks in traditional detection methods, allowing inspection personnel to grasp the abnormal situation of the harness in a timely manner.

[0054] By constructing a defect propagation model based on risk feature vectors, we can further analyze the propagation patterns of defects in automotive electrical systems and predict defect propagation paths. This process goes beyond simply detecting the presence of defects; it delves into the potential impact of defects on other electrical components, helping inspectors to anticipate fault development trends and take targeted preventative measures to avoid defect propagation leading to more serious faults, thereby reducing costs and safety hazards in automotive maintenance.

[0055] The adaptive detection threshold generated based on the prediction results can match the detection threshold with the actual operating state and defect propagation of the wiring harness. Traditional detection methods use fixed thresholds, which cannot adapt to the detection needs of wiring harnesses under different vehicle models and operating conditions. However, the adaptive detection threshold can be dynamically adjusted according to the current operating characteristics and potential risks of the wiring harness. While ensuring detection accuracy, it improves the adaptability of the detection method, making it applicable to different types of automotive wiring harnesses and different operating conditions, thus expanding the application range of the detection method. Attached Figure Description

[0056] Figure 1 This is a schematic diagram illustrating the working principle of the automotive wiring harness electrical testing method described in this invention.

[0057] Figure 2 A flowchart illustrating the process of acquiring electrical parameter data points for automotive wiring harnesses;

[0058] Figure 3 A flowchart illustrating the workflow for feature fusion processing of electrical parameter data points during a stable time period in automotive wiring harnesses;

[0059] Figure 4 This is a flowchart illustrating the workflow for calculating the probability of abnormal changes in automotive wiring harnesses over a given time period and outputting risk feature vectors. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Please see Figure 1 This invention provides a method for electrical testing of automotive wiring harnesses, the method comprising:

[0062] Electrical parameter data points, such as voltage, current, and resistance values, are acquired during the operation of the automotive wiring harness to form a time-series dataset. Based on the changing trend characteristics of the data points, the entire detection cycle is divided into two characteristic intervals: a stable period and a changing period. Within the stable period, multi-dimensional feature fusion technology is used to extract features and reduce the dimensionality of the electrical parameter data, generating representative fused feature representations. Within the changing period, the deviation of the current data point from the stable state characteristics is calculated to quantify the anomaly risk and construct a risk feature vector. Based on the risk feature vector, a dynamic model of defect propagation is established to predict the diffusion path of defects in the wiring harness network. Finally, the detection threshold is dynamically adjusted based on the prediction results to achieve an adaptive electrical detection process.

[0063] Example 1: See Figure 2 In the electrical testing of automotive wiring harnesses, electrical parameter data is acquired through a distributed sensor network. This system deploys dedicated sensor groups at key connection points in the main wiring harness in the engine compartment, the branch wiring harness in the instrument panel, and the chassis wiring harness. Each sensor group contains three measurement modules: a Hall effect current sensor attached to the outer insulation layer of the wire, a non-contact voltage probe connected to the terminal block, and a four-wire resistance measurement module connected to the wiring harness connector. All sensors synchronously acquire data at 100 millisecond intervals and transmit it to the central processing unit via shielded twisted-pair cables. The voltage measurement range covers the 0-48V operating range of the vehicle's electrical system, the current measurement covers the 0-20A load range, and the resistance measurement covers the 0-100Ω wire impedance range. Millisecond-level timestamps and location codes are appended to the acquired data, forming a triplet sequence containing time, space, and parameter dimensions.

[0064] The three independent parameter sequences of voltage, current, and resistance are preprocessed separately, and the first-order difference value of adjacent data points is calculated using the sliding window method. A dynamic threshold mechanism is set: when the absolute value of the difference exceeds three times the sensor's accuracy, the time point is marked as a valid change point. For example, if the voltage sensor accuracy is ±0.1V, the voltage sequence change threshold is set to 0.3V. The valid change points of the three parameter sequences are merged along the time axis to generate a unified candidate sequence of key points. Second-order difference calculation is performed on this sequence, and a local extremum detection algorithm is used to locate points of sudden increase in the rate of change. In the specific implementation, the system scans the second-order difference sequence. When the difference values ​​of three consecutive points satisfy a "low-high-low" distribution and the value of the center point exceeds 200% of the adjacent points, this moment is determined to be a key moment. Using this moment as the dividing point, a 60-second continuous period is defined forward as a stable period, and a 30-second period is defined backward as a changing period. A rolling update strategy is used for the time period boundaries. The difference calculation is re-executed each time a new data packet arrives. When a new key moment is detected and the time offset exceeds 5 seconds, the time period division result is automatically updated.

[0065] In the engine cold start test scenario, the system monitored the battery voltage sequence fluctuating around a reference value of 12.3V. At the instant the ignition signal was triggered, the voltage dropped to 11.8V within 300 milliseconds. This change of 0.5V exceeded the threshold of 0.3V and was marked as a valid change point. Simultaneously, the current jumped from 0.2A to 8.5A, a change of 8.3A exceeding the threshold of 0.15A; the resistance remained constant at 0.05Ω. After merging the voltage and current change points, the system calculated the second-order difference of the merged sequence. The maximum difference peak was detected 310 milliseconds after the ignition signal, and this moment was identified as the critical moment. Based on this, the system divided the time into two periods: a stable period of 60 seconds before ignition (containing 600 sampling points) and a variable period of 30 seconds after ignition (containing 300 sampling points). After the engine entered idle, the system detected a new current drop inflection point after 15 seconds, then updated the critical moment position and redefined the time period.

[0066] Each sensor module has a built-in self-diagnostic function. When a signal drift exceeding 20% ​​of the calibration value is detected, a zero-point calibration procedure is automatically triggered. Data transmission uses CRC checksums, and erroneous data packets trigger an immediate retransmission mechanism. The central processing unit performs logical checks on the received data. For example, if the voltage value at a certain point is zero while the current value is non-zero, it is automatically marked as abnormal data, and redundant sensor data replacement is initiated.

[0067] The output of the time period segmentation is a time period identifier matrix, which includes the start timestamp, end timestamp, and time period type code. This matrix is ​​stored synchronously with the original data stream in a circular buffer for real-time access by subsequent processing modules. The buffer adopts a dual-storage area design: while new data is written to the first storage area, the second storage area provides read-only access for the feature extraction module, realizing a data processing pipeline operation.

[0068] In the braking system wiring harness monitoring case, the system detected abnormal fluctuations in the brake light wiring resistance. The original resistance sequence fluctuated around a baseline value of 2.1Ω. When three consecutive sampling points showed a positive change of more than 0.15Ω, the system marked that period as a potential change segment. The resistance value continued to rise to 3.0Ω within the next 0.5 seconds, with the second-order differential peak appearing at 0.3 seconds, which was identified as a critical moment. Based on this, the system divided the time period before and after the peak and marked the data of that time period as a high-priority detection area. When the driver released the brake pedal, the system detected a drop in resistance value after 0.2 seconds, generating a new critical moment and updating the time period division.

[0069] When no valid change point is detected for 10 consecutive sampling periods, the stable time period range is automatically expanded; when more than three critical moment updates occur within 1 second, the burst mode is activated, shortening the analysis window to 50% of its original value. All time period division records are written to the event log, including timestamps, critical moment coordinates, time period length, and differential feature values ​​on which the division is based, forming a complete time period division trajectory.

[0070] When the charging gun is connected, the system detects that the power battery voltage steadily rises from 400V to 420V. The first-order difference value of the voltage sequence remains below the threshold, and the system maintains a stable state for a certain period. When charging enters the constant voltage stage, the current value linearly decreases from 150A to 50A. The system identifies the critical moment at the inflection point of the current change rate and accurately divides the different electrical characteristic stages of the charging process. Throughout the entire process, the time segmentation response time remains within 50 milliseconds and is updated synchronously with the data acquisition cycle.

[0071] Example 2: See Figure 3 During the stable period of automotive wiring harness electrical testing, the system performs multi-dimensional feature fusion processing. This process extracts twelve feature dimensions from the raw data stream, including six-dimensional features in the time domain and six-dimensional features in the spatial domain. Time-domain feature calculation is based on a sliding time window with a window width of 30 consecutive sampling points. For voltage parameter sequences, the time-domain analysis module calculates the arithmetic mean of the data within the window as the baseline feature, while simultaneously calculating the standard deviation to reflect the degree of fluctuation. The skewness coefficient is used to quantify the asymmetry of the data distribution, and the kurtosis coefficient detects the presence of abnormal peaks. The zero-crossing rate counts the number of times the signal crosses the average value, and the autocorrelation coefficient measures the correlation between consecutive values ​​in the sequence. These features are updated every 5 seconds, using an exponential weighting method for smooth transitions.

[0072] Spatial feature extraction considers the physical layout of the wiring harness network. The difference between adjacent measurement points is used to calculate the reading differences of sensors at different locations at the same time, reflecting the line impedance distribution. The regional mean ratio is selected as the voltage ratio between the main wiring harness and branch wiring harnesses to identify potential voltage drop anomalies. Topological distance weighted sums are used to calculate the attenuation of readings from distant sensors based on the node distances in the wiring harness connection diagram. Spatial autocorrelation coefficients analyze the parameter synchronization of different branches within the same electrical circuit. The spatial feature update cycle is synchronized with the temporal feature update cycle to ensure spatiotemporal feature alignment.

[0073] Dynamic weight allocation employs feature correlation analysis. A 12×12 feature correlation matrix is ​​constructed, and the Pearson correlation coefficient between each pair of features is calculated. Eigenvalue decomposition is performed on the correlation matrix, and the eigenvectors corresponding to the first three principal components are selected as the weight benchmarks. The initial weights for time-domain features are set to 0.6, and the initial weights for spatial-domain features are set to 0.4. The weights of each dimension are dynamically adjusted based on the principal component analysis results. During the adjustment process, for feature groups with a correlation exceeding 0.8, the weight allocation of secondary features within the group is reduced; for features with strong independence, their weight coefficients are appropriately increased. Weight updates adopt a gradual adjustment strategy, with each adjustment not exceeding 20% ​​of the previous period's value.

[0074] The weighted fusion process employs a layered architecture: the first layer performs internal weighted summation on the six-dimensional features in the time domain to generate a comprehensive time-domain index. The second layer performs internal weighted summation on the six-dimensional features in the spatial domain to generate a comprehensive spatial index. The third layer merges the two comprehensive indices according to their final weights to form a preliminary fused feature. An outlier filtering mechanism is implemented during the fusion process; when a feature value in a certain dimension deviates from the historical mean by more than three standard deviations, the weight of that dimension is temporarily reduced to 50% of the baseline value.

[0075] The initial 12-dimensional fusion features are input into the dimensionality reduction module, with a perplexity of 15 and a learning rate of 200. During algorithm execution, the early exaggeration factor is dynamically adjusted, initially set to 4 and linearly decreased to 1 during iteration. The dimensionality reduction result outputs 3D spatial coordinates, preserving the topological relationships of the original features. Density clustering is performed on the dimensionality-reduced 3D feature points, with a neighborhood radius of 0.3 and a minimum sample size of 5. The clustering results generate several feature clusters, and the center point coordinates and boundary range of each cluster are calculated. Finally, the three feature clusters with the highest density are selected as the fusion feature representation for that time period, and their center point coordinates, number of points within the cluster, and distribution radius are stored.

[0076] In the case study of new energy vehicle charging condition monitoring, the system detected stable characteristics during the connection phase between the charging pile and the vehicle. Time-domain analysis showed that the mean voltage remained at 398.5V, with a standard deviation of 0.8V, skewness of 0.15, kurtosis of 2.3, zero-crossing rate of 0.2 times / second, and autocorrelation coefficient of 0.92. Spatial characteristics showed that the voltage ratio between the main harness and the charging port branch was 1.02, the maximum difference between adjacent sensors was 0.5V, the topological weighted sum was 1.15, and the spatial autocorrelation coefficient was 0.88. Dynamic weight calculations yielded a dominant weight of 0.65 in the time domain and 0.35 in the spatial domain. The dimensionality-reduced 3D features formed two main clusters with center point coordinates of (0.12, -0.05, 0.08) and (-0.03, 0.11, -0.02), and cluster radii of 0.15 and 0.12, respectively. The system recorded this state as the baseline feature representation of the charging connection phase.

[0077] Each time a new fused feature representation is generated, its Euclidean distance to the historical benchmark features is calculated. If the distance value exceeds the warning threshold for three consecutive cycles, a feature relearning process is initiated. This process expands the sliding window size to 60 sampling points, adds a secondary feature extraction stage, and recalculates the weights of each dimension. If the relearned features still deviate from the historical benchmark, a system calibration prompt is generated, suggesting checking the sensor's operating status or the physical connection of the wiring harness.

[0078] In a vehicle vibration test scenario, the system detected feature changes caused by a loose wiring harness connector. The initial fused feature center point in a stable state was (0.05, 0.07, -0.03), which gradually drifted to (0.18, -0.12, 0.05) after vibration began. The system detected that the feature drift exceeded a threshold and automatically triggered a vibration compensation mode. This mode enhances the weighting of spatial features, particularly increasing the weight of the difference feature between adjacent sensors from 0.1 to 0.25, thus more sensitively capturing changes in physical connections. The compensated fused feature center point was corrected to (0.08, -0.05, 0.02), more accurately reflecting the actual electrical characteristics.

[0079] The baseline feature representation is stored in a high-speed cache for easy real-time comparison; historical feature sequences are stored on a solid-state drive and indexed by timestamps; long-term feature models are archived to a cloud database, supporting feature sharing across devices. Each time a feature is updated, the system generates a feature change log, recording the feature parameters, weight distribution, and dimensionality reduction projection information before and after the adjustment, forming a complete feature evolution trajectory.

[0080] When the ambient temperature dropped to -20℃, the system detected a significant change in the wiring harness resistance characteristics. The weight of the resistance-related dimension in the original fused features was 0.15; the system automatically adjusted it to 0.25 to enhance the focus on temperature-sensitive features. The dimensionality-reduced feature space exhibited a new cluster distribution, which the system labeled as a low-temperature mode feature representation. After the vehicle ran for 30 minutes, the wiring harness temperature rose again, and the features gradually reverted to the standard clusters. The system then correspondingly reduced the weight of the resistance features, reverting to the normal-temperature detection mode. Throughout the entire process, the feature fusion system maintained stable operation without any misjudgments or missed detections.

[0081] Example 3: See Figure 4 During the changing time period, the system performs anomaly probability calculation and defect propagation model construction. The anomaly detection module first calculates the distance metric between the current data point and the fused feature representation of the stable time period. An improved Mahalanobis distance calculation method is adopted, considering the distribution shape and anisotropy of feature clusters. For detection points in the three-dimensional feature space... With the center of the feature cluster Distance calculation introduces the covariance matrix of the cluster. :

[0082]

[0083] in: This represents the three-dimensional coordinates of the current detection point. The coordinates of the center point of the feature cluster are: This is the covariance matrix of all sample points for the feature cluster within a stable time period. This distance metric reflects the degree of deviation of the detected point from the overall distribution of the feature cluster. The system sets a dynamic distance threshold, initially three times the feature cluster radius, which is dynamically adjusted based on historical detection results. When the distance value exceeds the threshold, the anomaly probability calculation process is triggered.

[0084] The anomaly probability transformation employs a two-parameter sigmoid function, mapping distance values ​​to the [0,1] interval. The center point of the transformation function is set at 1.2 times the distance threshold, and the slope parameter is adaptively adjusted according to the density of the feature clusters. A gentler transformation curve is used for sparsely distributed feature clusters, while a steeper curve is used for dense feature clusters. The anomaly probability calculation result for each data point is stored together with the timestamp and location code to form an anomaly event record.

[0085] The first level of aggregation is performed in the time dimension, with a sliding window width of 15 sampling points, calculating the maximum, mean, and standard deviation of the anomaly probability within the window. The second level of aggregation is performed in the spatial dimension, based on the harness topology, weighting the anomaly features of adjacent detection points, with the weights proportional to the electrical coupling strength between nodes. The third level of aggregation extracts trend features, calculating the first and second differences of anomaly indicators for three consecutive time windows to capture the dynamic evolution of risk. The final generated risk feature vector contains 15 dimensions, with each dimension's values ​​standardized to eliminate the influence of dimensions.

[0086] The automotive wiring harness system is abstracted as a directed graph structure. Nodes represent key locations such as electrical connection points, branch switches, and connectors, while edges represent wire connections. Node features include three parts: basic attributes (wire diameter, material, rated current), real-time status (current risk feature vector), and historical data (past fault records). Edge features include connection length, bending angle, and insulation material type. The graph structure is represented by an adjacency matrix, where matrix element values ​​reflect the electrical connection strength between nodes.

[0087] Each graph attention layer contains eight independent attention heads, each calculating the propagation weights between nodes. A spatial attenuation factor is introduced in the attention weight calculation to account for the impact of path length on defect propagation.

[0088] For nodes and Attention coefficient The calculation considers three factors: the strength of the foundation connection. Risk feature similarity and spatial decay factor ,in Indicates the physical distance between nodes. This represents the attenuation coefficient. The model aggregates neighborhood information step by step through a three-layer graph attention network, ultimately resulting in each node obtaining a risk representation that includes the global context.

[0089] The input sequence contains risk feature vectors from the most recent 10 time steps, and temporal patterns are extracted using causal convolutional layers. The network consists of four dilated convolutional blocks with dilation coefficients of 1, 2, 4, and 8, covering the entire input sequence with a receptive field. Each convolutional block is followed by a gated activation unit to control the intensity of information transmission. The output layer predicts the defect concentration distribution for the next 5 time steps, including the expected defect value and propagation direction probability for each node.

[0090] In a case study of high-voltage wiring harness monitoring for new energy vehicles, the system detected a sudden increase in the probability of anomalies at nodes near the charging interface. Distance calculations showed that the distance between this point and the stable feature cluster reached 4.2 times the cluster radius, triggering a high-risk warning. The risk feature vector showed an average anomaly probability of 0.82, a spatial clustering degree of 0.75, and a trend slope of 0.15 in this area. Graph attention network analysis indicated that the anomaly was most likely to propagate along the high-voltage bus towards the battery management system, predicting that the defect concentration at the battery-side nodes would reach the warning value after three time steps. Based on this, the system generated a targeted detection plan, increasing the sampling frequency of the relevant lines to 200Hz.

[0091] After each detection period, the actual observations are compared with the predicted values, and the loss function is calculated. When the prediction error exceeds a threshold for three consecutive periods, model parameter fine-tuning is triggered. The fine-tuning process uses mini-batch gradient descent with a learning rate set to one-tenth of the initial value to prevent overfitting. Model version management uses incremental storage, retaining parameter snapshots of the most recent five versions to support fast rollback.

[0092] When the ignition coil operates, it generates a strong electromagnetic pulse, causing abnormal fluctuations in the readings of multiple sensors. The risk feature vector accurately identifies the spatiotemporal characteristics of the electromagnetic interference, distinguishing it from abnormal patterns caused by actual defects. The graph attention network analyzes the discontinuities in the propagation path to determine that the anomaly is due to external interference rather than an internal defect in the wiring harness, thus avoiding false alarms. The system automatically activates an anti-interference mode, temporarily adjusting the feature weight allocation to reduce sensitivity to transient changes.

[0093] The wire harness network is projected onto a 3D space according to its actual layout. The color depth of the nodes represents the defect concentration, and the arrow direction indicates the predicted propagation path. The visualization interface supports dragging the timeline to show the dynamic process of defect propagation. Engineers can interactively view detailed risk characteristics and prediction basis for any node to assist in diagnostic decision-making.

[0094] The system ran continuously for 300 hours during durability testing, processing over 2 million detection data points. Model parameters converged stably, and prediction accuracy remained stable. When new anomaly patterns were detected, the system adjusted parameters within 10 cycles through an online learning mechanism to adapt to the new operating conditions. Throughout the process, resource consumption remained at a reasonable level, with CPU utilization not exceeding 25% and memory usage consistently below 4GB.

[0095] Example 4: Adaptive detection threshold generation employs a generative adversarial network (GAN) framework, comprising a data generation module and a real-time discrimination module. The data generation module uses a Long Short-Term Memory (LSTM) autoencoder structure. The encoder contains three LSTM layers, each with 128 neurons, and the input window width is set to 30 time steps. The decoder uses two LSTM layers, learning the data distribution under normal operating conditions through reconstruction loss constraints. The training dataset contains 200 hours of fault-free operating condition records, covering 12 standard operating modes such as vehicle start-up, driving, and charging. The real-time discrimination module uses a one-dimensional convolutional neural network architecture, containing four convolutional blocks: the first block has 32 5×1 convolutional kernels, the second block has 64 3×1 convolutional kernels, the third block has 128 3×1 convolutional kernels, and the fourth block has 256 1×1 convolutional kernels. Each convolutional layer is followed by max pooling and batch normalization operations, and the final output layer uses the Sigmoid activation function to generate anomaly scores.

[0096] In the initial stage, the generator is trained using only clean, normal data to minimize the mean squared error of reconstruction. The second stage injects 5% synthetic anomalous data into the training data, including impulse interference, gradual drift, and step abrupt changes. The third stage increases the anomalous proportion to 15% and adds real-world fault case data. Discriminator training and generator training alternate, with the discrimination accuracy evaluated after each training round. The next stage begins when the validation set accuracy exceeds 95% for three consecutive rounds.

[0097] After the input window data passes through the fourth convolutional block, a 256-dimensional feature vector is extracted as the clustering input. An improved K-means++ algorithm is used to initialize the cluster centers, and three clusters are set to correspond to three levels of detection thresholds. A feature weighting mechanism is introduced into the clustering process, assigning higher weights to feature dimensions with strong discriminative power. After each clustering iteration, the discriminator output mean of samples within a cluster is calculated, and the cluster center positions are adjusted to maximize the differences between clusters. The final generated three-level threshold features are shown in Table 1.

[0098] Table 1: Adaptive detection threshold feature table.

[0099] Threshold level Feature density range Discriminator output range Typical triggering scenarios Level 1 threshold High-density areas (0.8-1.0) 0.05-0.15 Sensor noise / transient interference Secondary threshold Medium density region (0.5-0.8) 0.15-0.35 poor connector contact / local insulation damage Level 3 threshold Low-density areas (0.0-0.5) 0.35-0.95 Wiring harness short circuit / open circuit / severe aging

[0100] In the case of high-voltage wiring harnesses in new energy vehicles, the system identifies the main node of the battery management system as a high-risk defect area and divides it into the first-stage detection area, which includes one main node and three direct child nodes. The second stage selects two main paths from the main node to the motor controller, covering 12 intermediate nodes. The third stage includes the remaining 46 terminal nodes distributed throughout the vehicle body. The detection parameters for each stage are configured as follows: The first stage uses three redundant sensors for synchronous acquisition at a sampling rate of 200Hz, and data verification includes extreme value checks, slope consistency, and spatial correlation verification; the second stage uses dual sensors (main and backup) at a sampling rate of 100Hz, and implements trend prediction and propagation pattern matching; the third stage uses a single sensor at a sampling rate of 20Hz, and performs threshold monitoring and event-triggered acquisition.

[0101] During brake system wiring harness monitoring, the system detected an anomaly in the right rear wheel speed sensor wiring. The discriminator output value fluctuated between 0.28 and 0.32, triggering a level-two threshold alarm. The system automatically upgraded the relevant wiring node to the second-stage detection mode, increasing the sampling rate from 20Hz to 100Hz. The trend analysis module detected that the resistance value was continuously increasing at a rate of 0.5Ω / second. When the discriminator output exceeded 0.35, the system upgraded the node to the first-stage detection mode and initiated three-sensor redundancy verification. Diagnosis confirmed that the increased contact resistance was caused by connector pin oxidation. After replacing the component, the system returned to the basic monitoring mode within 15 seconds.

[0102] The threshold update mechanism includes two triggering conditions: periodic updates are executed every 24 hours, retraining the discriminator and updating cluster centers based on the data collected that day; event-triggered updates are initiated when a new abnormal pattern is detected, collecting 50 consecutive abnormal samples before performing incremental training. A protection mechanism is implemented during the update process: when the new threshold deviates from the historical benchmark by more than 20%, the old threshold version is retained and runs in parallel, and the main version is switched after 48 hours of verification and confirmation.

[0103] The system demonstrated adaptability during fast charging tests of electric vehicles. When the charging power increased from 50kW to 150kW, the discriminator detected a new current ripple pattern. Initially misjudged as abnormal (output value 0.25), the system initiated a threshold update after collecting 30 minutes of data. New cluster analysis identified this pattern as a normal characteristic of high-power charging, classifying it within the first-level threshold range (output value 0.12). After the update, the system correctly distinguished between a genuine insulation fault (output value 0.52) and the high-power charging ripple, avoiding false alarms and system shutdowns.

[0104] The operation log records the complete trajectory of operational state transitions. During testing with hybrid vehicles, the system recorded 27 stage switching events per day: 19 from level 1 to level 2 detection, 6 from level 2 to level 3, and 2 skipping levels to level 1. Switching cause statistics show that 65% were triggered by momentary interference, 30% by progressive fault development, and 5% by sudden changes in environmental conditions. Log data is stored as a triplet sequence of timestamp-event type-trigger parameter, supporting subsequent pattern analysis and system optimization.

[0105] This implementation method demonstrated stability in humid and hot environments. When the ambient humidity reached 95%, the system detected intermittent abnormal signals at the door wiring harness nodes. The discriminator output value fluctuated between 0.18 and 0.25, consistently remaining within the first-level threshold range. The system maintained the third-stage detection mode, recording events without triggering alarms. Three days later, when the output value exceeded 0.35, the system upgraded to the second-stage detection, confirming that partial discharge was caused by moisture in the wire insulation. Throughout the process, the threshold system effectively distinguished between environmental interference and actual faults, preventing premature triggering of maintenance alarms.

[0106] The first phase of the inspection task occupies 40% of the processor resources, the second phase allocates 30%, the third phase allocates 20%, and the remaining 10% is used for system maintenance. When multiple nodes trigger upgrades simultaneously, priority ranking based on defect concentration is implemented. In the battery pack harness monitoring case, the system handled 4 first-level alarm nodes and 7 second-level alarm nodes simultaneously. Processor resources were dynamically adjusted to: 15% for each first-level node and 5% for each second-level node, ensuring the inspection needs of critical areas were met.

[0107] Example 5: The first-stage detection strategy implements a multi-level verification mechanism for core defect areas. In the monitoring of the main node of the new energy vehicle battery management system, the system deploys three sets of redundant current sensors, each containing three different detection units based on Hall effect sensors, Rogowski coils, and shunts. The data acquisition cycle is set to 5 milliseconds, and the three independent channels simultaneously record the charging and discharging current waveforms. The verification process is divided into three levels: the first level performs instantaneous value comparison, taking the median of the three sensor readings as the benchmark value, and setting an allowable deviation range of ±5%; the second level analyzes short-term trends, performing correlation tests on the slopes of 10 consecutive sampling points, requiring a Pearson correlation coefficient ≥0.95; the third level verifies the law of conservation of energy, comparing the differences in the integral values ​​of the input and output currents, with an allowable error threshold set at 0.3%. When a sensor's data exceeds the deviation range three times consecutively, the system automatically marks it as a failed channel, switches to the backup link, and issues a calibration prompt. In a simulated test of a loose battery terminal, the system detected a 12% difference in the readings of the three sets of sensors within 200 milliseconds, and simultaneously, the current integral deviation exceeded 0.8%, triggering a three-level linkage alarm. The second-stage detection strategy establishes a dynamic analysis model for the wiring harness propagation path. In the case of monitoring the wiring from the body control module to the door actuator, the system installs dual sensors to collect voltage signals from the input and output terminals of the door control unit. The monitoring window is set to a 100-millisecond cycle, and three key propagation features are extracted in each cycle: signal delay time is calculated using a cross-correlation function to capture the time lag between the controller command and the actuator response; voltage attenuation rate is measured as the peak-to-peak ratio of the output to the input; and high-frequency noise ratio is analyzed using Fourier transform to determine the proportion of spectral components above 20kHz. The feature update cycle is 50 milliseconds, and the system maintains a feature queue of the most recent 20 cycles for trend modeling. An autoregressive moving average model is established to predict the feature value change in the next cycle, with a confidence interval set at the predicted value ± 2 standard deviations. When insulation wear occurs in the door hinge wiring harness, the system detects that the voltage attenuation rate exceeds the upper limit of the prediction interval for five consecutive cycles, while the high-frequency noise ratio suddenly increases from 3% to 12%. The system immediately initiates the path scanning mode, increases the sampling frequency to 5kHz, and locks the wear point at the location of the third conductor at the door seam bend. The third-stage detection strategy employs an intelligent event triggering mechanism. The in-vehicle entertainment system wiring harness monitoring case covers 46 terminal nodes, each configured with a single sensor scanning voltage parameters at a 5Hz base frequency. The system establishes two levels of triggering conditions: a normal trigger threshold set at ±15% of the rated voltage, and an instantaneous jump threshold set at a change rate exceeding 50% between adjacent sampling points. When a high-power device is connected to the cigarette lighter socket, the system detects a sudden drop in 12V voltage to 9.6V, with an instantaneous change rate of 56%. This event activates a 10ms high-speed sampling mode, continuously recording for 200 cycles.The feature extraction module only calculates the mean and standard deviation. The mean data is input into a pre-trained decision tree classifier. The model contains five decision branches: if the mean is below 9V and the standard deviation exceeds 0.5V, it is classified as a short circuit risk; if the mean fluctuates between 9-10V, it is classified as an overload condition. In this incident, the system detected a stable mean of 9.8V and a standard deviation of 0.2V, which was determined to be a normal load change. After the device was removed, the system resumed basic monitoring mode within 15 seconds, and no false alarms were triggered throughout the entire process.

[0108] During the monitoring of the electric power steering system, when the basic scan detects a ±0.5V periodic fluctuation in the torque sensor signal, the system activates the event trigger mode. Preliminary analysis shows the standard difference has increased to 0.38V, and the classifier outputs boundary confidence values. The system automatically requests a secondary review, transmitting 10 seconds of data to the regional processing unit. The second-stage analysis identifies an 85% correlation between the fluctuation frequency and vehicle speed, and the signal delay conforms to a normal propagation curve, but an abnormal increase of 1.8dB in the noise component is detected. This result triggers a strategy linkage request to the first-level detection, initiating simultaneous acquisition of data from three sensors in the core area. Redundancy verification confirms that the wiring harness grounding impedance has increased from 0.1Ω to 0.8Ω, pinpointing the fault location as oxidation of the steering column connector. The system automatically generates a diagnostic report and marks the fault location coordinates, permanently upgrading the relevant nodes to the first-level monitoring strategy and allocating dedicated communication bandwidth and computing resources.

[0109] During a 30Hz sinusoidal sweep frequency test conducted in the electromagnetic compatibility laboratory, an increased frequency of voltage fluctuations in the reversing radar circuit was detected in the third stage. The basic scan mode recorded 8 ±10% fluctuations within 20 seconds, indicating insufficient confidence from the conventional classifier output. The system requested a second-level analysis to confirm that the fluctuation pattern was synchronized with the vibration frequency, but the propagation characteristics conformed to normal path parameters. At this critical moment, no abnormal triggering occurred in the first-level core area, and the resource scheduling center automatically allocated idle computing units for in-depth verification. The third-level collaborative analysis confirmed that the interference source was vibration causing poor sensor contact, recommending tightening the connectors rather than replacing the wiring harness. The entire decision-making process took 800 milliseconds, saving unnecessary maintenance operations while maintaining continuous system monitoring capabilities.

[0110] Data from cold-region testing sites showed that low temperatures caused a periodic increase in the false alarm rate in the third stage. The system automatically established a temperature compensation model: when the ambient temperature was below -10℃, the normal trigger threshold for all terminal nodes was relaxed to ±18%, the instantaneous jump threshold was increased to 65%, and the classifier feature weights were tilted towards the mean parameter. Simultaneously, the self-check cycle was shortened to 2 hours, and sensor drift compensation was enhanced. When the vehicle entered a warm workshop, the system detected that the connector temperature had risen above 15℃ and automatically restored the standard parameter configuration. In the annual periodic data statistics, this dynamic adjustment mechanism kept the false alarm rate in winter within a fluctuation range of 3%, comparable to that in summer.

[0111] A hybrid vehicle experienced intermittent anomalies during heavy rain, with multiple nodes triggering alarms alternately in the third stage. The system prioritized resource allocation to address braking system nodes that directly jeopardized driving safety, while reducing motor cooling system monitoring to a minimum. An anomaly pattern propagation map was established, identifying the wiper circuit as the initial source of interference. A two-level isolation strategy was implemented: first, the power supply to the wiper control module was disconnected to eliminate interference; second, a second-level in-depth inspection was performed on critical system nodes to confirm the integrity of the wiring harness. While ensuring the vehicle's basic functions, the system identified the root cause of signal crosstalk caused by rainwater seeping into the combination switch, guiding targeted component replacement rather than comprehensive wiring harness repair.

[0112] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0113] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for testing the electrical properties of automotive wiring harnesses, characterized in that, Including the following steps: Acquire multiple electrical parameter data points of the automotive wiring harness; The data points of electrical parameters are divided into stable time periods and changing time periods based on their changing trends. Within a stable time period, feature fusion processing is performed on electrical parameter data points to generate fused feature representations; Within the time period of change, the probability of anomalies is calculated based on the fused feature representation, and a risk feature vector is output. A defect propagation model is constructed based on risk feature vectors to predict defect propagation paths; An adaptive detection threshold is generated based on the prediction results to achieve electrical detection. Within the changing time period, the probability of anomalies is calculated based on the fused feature representation, and a risk feature vector is output, including: Compare the distance between electrical parameter data points during the change period and the fused feature representations during the stable period; Calculate the probability of anomalies for each electrical parameter data point based on the distance value; Aggregate the anomaly probabilities of all electrical parameter data points to form a risk feature vector; The step of constructing a defect propagation model based on risk feature vectors and predicting defect propagation paths includes: The risk feature vector is mapped to the three-dimensional spatial coordinates of the automotive wiring harness to generate a preliminary defect concentration distribution. The initial defect concentration distribution is dynamically corrected based on historical defect data to obtain the spatiotemporal defect concentration distribution. A graph model is constructed based on the automotive wiring harness topology, where nodes represent key locations in the wiring harness and edges represent connections. An attention mechanism is used to capture the dependency relationship of defect propagation between nodes, and an initial prediction of defect propagation is output. By combining time series prediction models, defect propagation paths and diffusion trends are generated; The process of generating an adaptive detection threshold based on the prediction results includes: Train a normal operating condition generator to simulate normal operating data of automotive wiring harnesses; Build a real-time discriminator to compare prediction results with generator data and learn the boundaries between normal and abnormal; The results are segmented and judged using a clustering algorithm to generate multi-level adaptive detection thresholds.

2. The method for testing the electrical properties of automotive wiring harnesses according to claim 1, characterized in that, The acquisition of multiple electrical parameter data points of the automotive wiring harness includes: Electrical parameter data points include voltage, current, and resistance values; The voltage, current and resistance values ​​of the vehicle's wiring harness are collected in real time by sensors to form time series data points.

3. The method for testing the electrical properties of automotive wiring harnesses according to claim 2, characterized in that, The division of stable and changing time periods based on the changing trends of electrical parameter data points includes: Calculate the first-order difference value of the electrical parameter data point sequence, and select the moments with positive first-order difference values ​​as growth moments; Sort the growth moments according to the first difference value, calculate the second difference value of the sorted sequence, and take the growth moment corresponding to the maximum value of the second difference value as the critical moment. The period before the critical moment is a stable period, while the period after the critical moment is a changing period.

4. The method for electrical testing of automotive wiring harnesses according to claim 3, characterized in that, The step of performing feature fusion processing on electrical parameter data points within a stable time period to generate a fused feature representation includes: Extract multi-dimensional features of electrical parameter data points within a stable time period, including temporal and spatial features; A dynamic weighted algorithm is used to fuse multi-dimensional features and calculate the correlation weights of features in different dimensions. The preliminary fused features are obtained by weighting the multi-dimensional features according to the relevance weights. Redundant information is removed by dimensionality reduction, generating a low-dimensional fusion feature representation.

5. The method for testing the electrical properties of automotive wiring harnesses according to claim 1, characterized in that, The implementation of electrical detection further includes: The detection stages are divided according to the adaptive detection threshold, including the first stage, the second stage, and the third stage. Different detection strategies are applied to electrical parameter data points at different detection stages.

6. The method for testing the electrical properties of automotive wiring harnesses according to claim 5, characterized in that, The step of dividing the detection stage according to the adaptive detection threshold includes: The layer with the highest node density in the defect propagation model is set as the first stage; The layer with the most child nodes after the first phase is designated as the second phase. The second stage is followed by the third stage at the bottom layer.

7. The method for electrical testing of automotive wiring harnesses according to claim 6, characterized in that, The application of different detection strategies to electrical parameter data points at different detection stages includes: The first type of detection strategy is used for the electrical parameter data points in the first stage; The second type of detection strategy is used for the electrical parameter data points in the second stage; A third type of detection strategy is adopted for the electrical parameter data points in the third stage.

Images