Intelligent testing method and system for wiring harness connection performance

Abstract

The present invention relates to the field of intelligent wire harness evaluation technology, and in particular to a method and system for intelligent testing of wire harness connection performance. The method comprises the following steps: collecting wire harness test data generated by wire harness test equipment and wire harness sensors in real time through edge computing nodes; performing electrical performance detection extraction and mechanical performance detection extraction according to the wire harness test data, and obtaining wire harness electrical performance detection data and wire harness mechanical performance detection data respectively; performing electrical performance depth evaluation according to the wire harness electrical performance detection data to obtain wire harness electrical performance evaluation data, and performing mechanical performance depth evaluation on the wire harness mechanical performance detection data to obtain wire harness mechanical performance evaluation data; performing environmental pressure cross-reference according to the wire harness electrical performance evaluation data and the wire harness mechanical performance evaluation data to obtain wire harness connection performance test evaluation data. The present invention improves the accuracy of wire harness performance evaluation, so that reliable performance data can be obtained in different application fields.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent wire harness evaluation, and in particular to a method and system for intelligent testing of wire harness connection performance. Background Art

[0002] A wiring harness is a component that integrates multiple cables or wires according to electrical or mechanical requirements. It is widely used in the automotive, aviation, industrial automation and other fields. With the increasing complexity of electronic equipment, the connection performance of the wiring harness has become particularly important to the stability, safety and overall performance of the system. However, in actual application environments, the wiring harness is not only affected by electrical properties (such as resistance, current transmission capacity, etc.), but also by mechanical properties (such as tensile strength, bending resistance, etc.) and environmental factors (such as temperature, humidity, vibration, etc.). Therefore, how to accurately and efficiently test and evaluate the connection performance of the wiring harness has become a problem. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a wiring harness connection performance intelligent testing method and system to solve at least one of the above technical problems.

[0004] The present application provides a method for intelligently testing wiring harness connection performance, comprising the following steps:

[0005] Step S1: collecting the wiring harness test data generated by the wiring harness test equipment and the wiring harness sensor in real time through the edge computing node;

[0006] Step S2: performing electrical performance detection extraction and mechanical performance detection extraction according to the wire harness test data, and obtaining wire harness electrical performance detection data and wire harness mechanical performance detection data respectively;

[0007] Step S3: performing an in-depth evaluation of electrical performance according to the wire harness electrical performance test data to obtain wire harness electrical performance evaluation data, and performing an in-depth evaluation of mechanical performance on the wire harness mechanical performance test data to obtain wire harness mechanical performance evaluation data;

[0008] Step S4: Perform environmental pressure cross-reference based on the wire harness electrical performance evaluation data and the wire harness mechanical performance evaluation data to obtain wire harness connection performance test evaluation data.

[0009] In the present invention, edge computing nodes are used to realize real-time collection of harness test data, which reduces the delay of data transmission and ensures the timeliness of test data. Edge computing has preliminary processing capabilities, and can perform noise filtering and data preprocessing while collecting data to improve the accuracy of data. Independent extraction of electrical characteristics (such as resistance, current, voltage) and mechanical characteristics (such as tension and compression strength) of the harness, separate processing of electrical performance and mechanical performance, helps to analyze the performance problems of harnesses with different properties in a targeted manner. Comprehensive coverage of harness performance is achieved, taking into account both electrical functionality and mechanical stability. Electrical performance evaluation uses deep learning algorithms and models to identify potential electrical problems through data mining. Mechanical performance evaluation deeply analyzes the mechanical stability of the harness through intelligent algorithms (such as stress strain analysis or machine learning classification). Cross-analyze electrical and mechanical performance data with environmental stress (such as temperature, humidity, vibration) data. According to the correlation of multi-dimensional data, a comprehensive evaluation report of harness connection performance is generated, which provides test results that are closer to the actual use environment and avoids the limitations of single-dimensional performance evaluation. Able to predict the performance of wiring harnesses under extreme conditions and provide guidance for product design and improvement.

[0010] Preferably, step S1 specifically includes:

[0011] Step S11: collecting initial wiring harness test data generated by wiring harness test equipment and wiring harness sensors in real time through edge computing nodes;

[0012] Step S12: performing interference data separation according to the initial wire harness test data to obtain wire harness test separation data;

[0013] Step S13: filtering and detecting the wire harness test separation data to obtain wire harness test filtering detection data;

[0014] Step S14: performing back-tracking calibration according to the wire harness test filter detection data to obtain wire harness test data.

[0015] In the present invention, the edge computing node performs real-time data collection close to the data source, avoiding the delay caused by long-distance data transmission. With local computing capabilities, it can quickly process large-scale data and reduce dependence on cloud computing resources. Through a specific algorithm, interference signals (such as electromagnetic interference, background noise, etc.) are separated from the initial harness test data, and obvious invalid information is eliminated. The interference source is accurately located and removed by data feature analysis. After separating the interference data, multi-stage filtering operations are performed on the harness test separation data to eliminate high-frequency noise and background fluctuations. Adaptive filtering technology is used to flexibly adjust the filtering parameters for different noise environments. Using the data after filtering detection, retrospective adjustments are made by comparing the reference signal and environmental characteristics to correct data deviations. The calibration operation can solve the error problems caused by test equipment drift and environmental changes.

[0016] Preferably, step S12 is specifically:

[0017] Perform edge collaborative anomaly cleaning based on the initial harness test data to obtain preliminary harness test cleaning data;

[0018] Performing spectrum denoising on preliminary harness test cleanup data to obtain harness test cleanup feature data;

[0019] Performing multi-level feature deconvolution on the wire harness test cleanup feature data to obtain wire harness test deconvolution feature data;

[0020] Performing dimensionality reduction processing on the wire harness test unrolling feature data to obtain wire harness test feature dimensionality reduction data;

[0021] Perform heterogeneous signal analysis on the wire harness test feature dimension reduction data to obtain wire harness test analysis feature data;

[0022] Perform time series debiasing on the wire harness test profiling feature data to obtain wire harness test time series purified data;

[0023] Perform aggregate signal reconstruction on the wiring harness test timing purification data to obtain wiring harness test reconstruction target data;

[0024] The wire harness test reconstruction target data is subjected to verification interference separation to obtain the wire harness test separation data.

[0025] In the present invention, the signal-to-noise ratio of the data is gradually improved through hierarchical optimization, and complex background interference is eliminated. In the initial stage, the edge collaborative anomaly cleaning technology is used to quickly eliminate obvious abnormal signals, and real-time processing is achieved through distributed computing, avoiding the delay and bandwidth bottleneck of traditional centralized data processing. Spectral denoising technology is used to accurately identify and remove high-frequency and random noise, and multi-level feature unrolling technology is used to separate different frequency bands of the signal to remove complex noise interference. Dimensionality reduction processing uses a variety of algorithms (such as principal component analysis PCA and t-SNE) to reduce data redundancy while retaining the key features of the signal. Heterogeneous signal analysis uses heterogeneous graph neural networks to deeply explore the high-order correlations between signal features and further remove redundant interference components. Time series debiasing corrects the deviation in the signal through dynamic time warping, so that the data is more in line with the time characteristics of the harness performance. Aggregate signal reconstruction uses convolutional autoencoders to optimize and reconstruct the signal to ensure the integrity and consistency of the reconstructed data.

[0026] Preferably, the spectrum denoising is specifically:

[0027] Performing waveform decomposition on preliminary harness test cleanup data to obtain harness waveform decomposition data;

[0028] Performing spectral component analysis on the wire beam waveform decomposition data to obtain wire beam spectral component data;

[0029] Performing spectrum peak detection on the spectrum component data of the wire beam to obtain the high-frequency noise characteristic data of the spectrum of the wire beam;

[0030] Performing filter parameter mapping on the high-frequency noise characteristic data of the wire harness spectrum to obtain filter parameter data;

[0031] De-noising the line beam spectrum component data according to the filtering parameter data to obtain the line beam spectrum de-noising data;

[0032] The key frequency features are extracted from the wire harness spectrum denoising data to obtain the wire harness test cleanup feature data.

[0033] In the present invention, the preliminary wire harness test cleanup data is firstly subjected to waveform decomposition, and the data is disassembled into multiple basic waveforms to obtain the wire harness waveform decomposition data. The spectrum component is further decomposed to accurately divide the characteristics of different frequency bands to ensure the accurate identification of noise and target signals. Targeted high-frequency noise detection is provided, which avoids the blind spot of denoising in traditional methods, accurately identifies the peak frequency position of noise, improves the denoising effect, and reduces the effective signal removed by mistake. Targeted filtering is achieved, and the denoising intensity can be automatically adjusted under different noise conditions, reducing the denoising distortion problem caused by fixed parameters in traditional filtering. The dynamically mapped filtering parameters significantly improve the denoising effect and ensure the effective suppression of high-frequency noise. The denoising process is accurate and effective, ensuring that the key feature information in the signal is retained, avoiding signal loss caused by excessive filtering, improving the signal-to-noise ratio of the test data, and making the test results more reliable and accurate. The extracted key frequency features highly represent the performance characteristics of the signal itself, providing more accurate data support for subsequent test evaluation, making the analysis more focused on the core frequency band, and further improving the efficiency and accuracy of data analysis.

[0034] Preferably, the multi-level feature deconvolution is specifically:

[0035] Performing frequency domain conversion on the wire harness test cleanup feature data to obtain the wire harness test frequency domain conversion data;

[0036] Performing frequency domain separation on the wire harness test frequency domain conversion data to obtain the wire harness test frequency domain separation data;

[0037] Perform multi-scale frequency band convolution calculation based on the wire harness test frequency domain separation data to obtain the wire harness test frequency band decoupling data;

[0038] Perform layered unwinding calculation based on the decoupling data of the wire harness test frequency band to obtain the layered unwinding data of the wire harness layer;

[0039] The spectrum is reconstructed based on the layered unwinding data of the wire harness to obtain the unwinding characteristic data of the wire harness test.

[0040] The fine frequency domain separation in the present invention improves the spectral purity of the signal, makes the boundary between noise and useful signals clearer, improves the denoising efficiency and accuracy, and reduces the risk of noise mixing into the target signal. The present invention can extract the multi-scale features of the signal and reduce the feature loss caused by traditional single-scale convolution. It improves the coverage of features in different frequency bands and provides high-quality basic data for subsequent deconvolution. Layered deconvolution effectively removes multiple layers of interfering signals and retains core signal features, significantly improves signal purity, and makes the data more hierarchical and refined. The signal reconstructed through the spectrum has high fidelity, can fully present the core features of the original signal, avoids feature loss during deconvolution and denoising, and enhances the credibility and availability of the reconstructed signal.

[0041] Preferably, the heterogeneous signal analysis is specifically:

[0042] Constructing a graph structure for the wire harness test feature dimensionality reduction data to obtain wire harness test feature graph data;

[0043] A heterogeneous graph is constructed according to the wire harness test characteristic graph data to obtain the wire harness test heterogeneous graph data;

[0044] Extracting image signal features based on the wiring harness test heterogeneous image data to obtain wiring harness test image feature data;

[0045] Perform feature noise stripping according to the feature data of the harness test graph to obtain optimized graph feature data;

[0046] High-order features are reconstructed based on the optimized graph feature data to obtain the wiring harness test analysis feature data.

[0047] The present invention makes the implicit associations between data explicit, improves the accuracy of feature extraction in subsequent analysis, provides a more expressive feature association representation than traditional linear analysis, and provides a more accurate basis for signal analysis. A heterogeneous graph is constructed based on feature graph data to represent different types of signal feature nodes and their diverse relationships. The heterogeneous graph can flexibly define heterogeneous nodes and edges according to different types of features, thereby effectively representing and analyzing the relationships between multiple feature types. The extracted graph feature data can clearly reflect the difference between signal and noise, improve the accuracy of noise separation, realize deep feature extraction of high-dimensional signals, and ensure data integrity and analysis depth. The purity of the signal is significantly improved, high-quality data is provided for feature reconstruction, noise interference on the analysis results is effectively reduced, and the accuracy of overall signal analysis is improved. Providing analysis feature data with high consistency and integrity provides a high-quality foundation for signal evaluation and fault diagnosis, can accurately restore the original appearance of the signal, and avoid the loss of effective signal information during stripping and reconstruction.

[0048] Preferably, step S2 specifically includes:

[0049] Step S21: stratify the test data according to the wiring harness test data to obtain wiring harness test stratification data;

[0050] Step S22: performing classification screening according to the wiring harness test hierarchical data to obtain wiring harness test classification data, wherein the wiring harness test classification data includes wiring harness electrical performance test data and wiring harness mechanical performance test data;

[0051] Step S23: performing electrical performance detection conversion according to the wiring harness electrical performance test data to obtain wiring harness electrical performance detection data;

[0052] Step S24: Perform mechanical property detection conversion according to the wire harness mechanical property test data to obtain the wire harness mechanical property detection data.

[0053] The hierarchical processing in the present invention makes the data structure clearer, facilitates classification and screening, reduces data redundancy, improves the efficiency of data processing, and facilitates targeted analysis and data management. Through accurate classification and screening, the mixing of different performance data is avoided, and the purity of each data group is improved. It is convenient to independently process different types of data, so that electrical and mechanical performance data can be processed and analyzed in a targeted manner. The converted electrical performance test data has high detection accuracy and consistency, providing high-quality input for electrical performance evaluation. Data standardization reduces the deviation in electrical performance testing and makes the analysis results more reliable. The purity and accuracy of mechanical performance test data are enhanced, providing reliable data for subsequent mechanical performance evaluation. The mechanical performance analysis results are made more accurate, which is helpful for the quality control and fault prediction of wiring harnesses in actual application scenarios. The separated data is more in line with the requirements of different tests, which improves the pertinence and test accuracy of the overall test process. It provides favorable support for the independent evaluation and optimization of electrical and mechanical performance, so that the test data can directly provide high-quality input for performance analysis and problem location.

[0054] Preferably, step S3 specifically includes:

[0055] Step S31: using a preset wiring harness electrical evaluation model to perform an in-depth electrical performance evaluation on the wiring harness electrical performance detection data to obtain wiring harness electrical performance evaluation data;

[0056] Step S32: performing stress-strain nonlinear modeling according to the wire harness mechanical property detection data to obtain a wire harness mechanical stress model;

[0057] Step S33: Estimating the mechanical load tolerance according to the wire harness mechanical stress model to obtain wire harness mechanical performance evaluation data;

[0058] The step of constructing the wiring harness electrical evaluation model preset in step S31 includes the following steps:

[0059] Obtain historical wiring harness electrical performance test data and historical electrical performance evaluation label data through a preset wiring harness history database;

[0060] Perform data division on historical wiring harness electrical performance test data to obtain historical wiring harness test data and historical wiring harness verification data;

[0061] Performing primary hidden layer processing on historical harness test data to obtain primary feature data of historical harness;

[0062] Perform multi-head self-attention layer processing on the primary feature data of the historical line bundle to obtain the weighted feature data of the historical line bundle;

[0063] Performing secondary hidden layer processing on the historical line bundle weighted feature data to obtain the historical line bundle secondary feature data;

[0064] The historical wiring harness verification data is used to iteratively train the historical wiring harness secondary feature data and the model is constructed through the historical electrical performance evaluation label data to obtain the wiring harness electrical evaluation model.

[0065] The present invention improves the accuracy of identifying electrical performance anomalies and potential faults, making electrical performance evaluation more comprehensive and reliable. The deep evaluation process can reveal minor anomalies in electrical performance, providing data support for predictive maintenance and performance optimization. The accuracy of mechanical load tolerance evaluation is improved, especially for wiring harnesses under complex stress conditions. Mechanical performance evaluation data can reflect the stability and tolerance of wiring harnesses under various stress conditions, providing an optimization basis for product design and testing. Multi-stage feature extraction and training enable the model to have high generalization ability, which is suitable for a variety of wiring harness test data and reduces the risk of model overfitting. The model can automatically adapt to different electrical test scenarios and data, enhancing the stability and robustness of practical applications. The self-attention mechanism significantly improves the sensitivity of the model to key signals, making electrical performance evaluation more accurate. Accurately identify abnormal signals in electrical performance changes, reduce misjudgment and missed judgment, and improve the reliability of testing and evaluation. The model parameters are continuously optimized through the iterative training process to ensure that the electrical performance evaluation model can accurately adapt to different test data and application scenarios. Dual performance evaluation provides more comprehensive performance analysis results to ensure the accuracy and comprehensiveness of the test.

[0066] Preferably, step S4 is specifically:

[0067] Step S41: obtaining wiring harness environment data;

[0068] Step S42: performing environmental data matching according to the wire harness environmental data, the wire harness electrical performance evaluation data, and the wire harness mechanical performance evaluation data to obtain test environment matching data;

[0069] Step S43: performing cross analysis according to the test environment matching data to obtain parameter cross analysis data;

[0070] Step S44: verifying the environmental adaptability according to the parameter cross-analysis data to obtain the harness environmental adaptability data;

[0071] Step S45: Perform environmental tolerance evaluation based on the wire harness environmental adaptability data to obtain wire harness connection performance test evaluation data.

[0072] In the present invention, by collecting real environmental data (such as temperature, humidity, vibration, pressure, etc.), combined with the harness performance data, it is ensured that the test can accurately reflect the performance of the harness in different actual application environments. Matching analysis ensures the close correlation between electrical and mechanical performance data and specific environmental conditions, and improves the relevance and accuracy of the evaluation. According to the test environment matching data, multiple parameters are cross-analyzed to generate parameter cross-analysis data to identify the interaction between different parameters. Cross-analysis helps to dig out the comprehensive impact of environmental factors on electrical and mechanical performance and provide a deeper understanding of performance. Based on the cross-analysis results, the adaptability of the harness under different environmental conditions is verified, and the harness environmental adaptability data is generated. The adaptability verification can predict the performance of the harness in extreme environments or special application scenarios to ensure its safety and stability. According to the harness environmental adaptability data, the harness tolerance under various environmental pressures (such as high temperature, low temperature, humidity, etc.) is evaluated, and the harness connection performance test evaluation data is generated, which provides a comprehensive evaluation of the harness's ability to resist environmental pressure, so that the reliability of the product in actual application is guaranteed, which helps to identify the performance limit of the harness and provide data support for product optimization and fault prediction.

[0073] Preferably, the present application also provides a wiring harness connection performance intelligent testing system, which is used to execute the wiring harness connection performance intelligent testing method as described above, and the wiring harness connection performance intelligent testing system includes:

[0074] A wiring harness test data acquisition module, which is used to collect wiring harness test data generated by wiring harness test equipment and wiring harness sensors in real time through edge computing nodes;

[0075] A wiring harness performance detection and extraction module is used to perform electrical performance detection and extraction and mechanical performance detection and extraction according to the wiring harness test data, and obtain wiring harness electrical performance detection data and wiring harness mechanical performance detection data respectively;

[0076] A wiring harness performance evaluation module is used to perform an in-depth evaluation of electrical performance based on the wiring harness electrical performance test data to obtain wiring harness electrical performance evaluation data, and to perform an in-depth evaluation of mechanical performance based on the wiring harness mechanical performance test data to obtain wiring harness mechanical performance evaluation data;

[0077] The wiring harness connection performance test evaluation module is used to perform an environmental pressure cross-reference based on the wiring harness electrical performance evaluation data and the wiring harness mechanical performance evaluation data to obtain the wiring harness connection performance test evaluation data.

[0078] The beneficial effects of the present invention are that the local data processing of the edge computing node reduces the transmission delay, ensures that the collected data is closer to the actual use environment, and improves the accuracy of the data. Through a special detection and extraction method, the electrical performance and mechanical performance data are separated so that each performance indicator can be deeply evaluated without interference. The in-depth evaluation of electrical performance uses an advanced electrical evaluation model to conduct an in-depth analysis of the electrical performance data to identify potential performance degradation and abnormal conditions. The mechanical performance evaluation simulates the mechanical response of the wiring harness under various load environments through stress-strain nonlinear modeling, thereby obtaining the mechanical tolerance and stability data of the wiring harness. The electrical performance evaluation model performs multi-level feature extraction and training based on historical wiring harness electrical performance data, and uses multi-stage optimization structures such as primary hidden layers, multi-head self-attention mechanisms, and secondary hidden layers to enable the model to adapt to a variety of electrical environments. Through iterative training and label data verification, the model has a strong generalization ability and can identify and adapt to different types of wiring harness electrical characteristics and abnormal conditions. The present invention provides an environmental simulation that is closer to reality, ensuring that the evaluation results reflect the performance of the wiring harness in a real environment and improving the accuracy and representativeness of the evaluation. Cross-reference analysis helps identify the comprehensive impact of environmental factors on wiring harness performance, reveals the complex relationship between environmental pressure and performance indicators, and provides a scientific basis for the application of wiring harnesses in various environments. Through environmental adaptability verification and tolerance evaluation, the performance limits and weak points of wiring harnesses under extreme conditions can be discovered in advance, which helps to improve the safety of wiring harnesses in extreme application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0080] Figure 1 A flowchart showing a method for intelligently testing wiring harness connection performance according to an embodiment is shown;

[0081] Figure 2 A flowchart showing a method for collecting data from a wiring harness test according to an embodiment of the present invention is provided;

[0082] Figure 3A flowchart showing a method for detecting and extracting wire harness performance according to an embodiment of the present invention is provided;

[0083] Figure 4 A flowchart showing a method for evaluating wire harness performance according to an embodiment of the present invention is provided;

[0084] Figure 5 A flowchart of the steps of a method for testing and evaluating the connection performance of a wiring harness according to an embodiment is shown. DETAILED DESCRIPTION

[0085] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0086] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0087] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0088] During the test, the edge computing node is connected to the wiring harness test equipment and sensors to collect electrical and mechanical test data in real time. When testing an automotive wiring harness, the real-time collected data includes electrical parameters (such as voltage and current) and mechanical parameters (such as tensile force and deformation). Electrical parameters: voltage = 12V, 13V, 12.5V, 14V (collected once per second), mechanical parameters: tensile force = 50N, 52N, 51.5N, 54N (collected once per second), preliminary wiring harness test data (output data stream of edge computing node): [(12V, 50N), (13V, 52N), (12.5V, 51.5N), (14V, 54N)].

[0089] The collected preliminary harness test data was processed in layers, and the data was classified into electrical performance test data and mechanical performance test data, which were extracted for in-depth analysis. The voltage and current data were statistically analyzed and filtered to extract the core parameters of electrical performance. The tensile force and deformation data were modeled to extract the mechanical response of the harness under different stresses. The electrical performance test data extracted: average voltage = 12.88V (by data averaging), the mechanical performance test data extracted: average tensile force = 51.63N (by data averaging). The calculated harness electrical performance test data: 12.88V, harness mechanical performance test data: 51.63N.

[0090] After obtaining the electrical performance test data and mechanical performance test data, they are deeply evaluated. Using the preset electrical evaluation model, the electrical stability and abnormal changes of the harness are calculated by inputting the test data (average voltage = 12.88V). Assume that the model derives a harness electrical performance stability score = 92 points (out of 100 points), which is obtained by the model automatically learning and verifying the historical harness data. Perform nonlinear stress modeling on the mechanical test data and estimate the harness's tolerance under dynamic loads. For example, based on the average tensile force (51.63N), the model derives a harness mechanical tolerance score = 85 points (out of 100 points). Electrical performance evaluation data: stability score = 92 points, mechanical performance evaluation data: tolerance score = 85 points. Harness electrical performance evaluation data: 92, harness mechanical performance evaluation data: 85.

[0091] After evaluating the electrical and mechanical performance, environmental data is added for cross-reference analysis to confirm the adaptability of the wiring harness in the actual environment. The test environment is a high-temperature cabin in a car, and the environmental data includes temperature = 80°C and humidity = 70%. Parameter matching is performed based on the wiring harness electrical performance evaluation data (92 points), mechanical performance evaluation data (85 points) and environmental data (80°C, 70% humidity). Through cross-analysis, the model calculates the adaptability score of the wiring harness in this environment = 80 points, indicating that the wiring harness can still maintain stable performance under the current high temperature and high humidity conditions. Based on the cross-analysis data, the tolerance of the wiring harness at higher temperatures and higher humidity is further estimated, and the tolerance limit conditions (temperature 90°C, humidity 75%) are obtained. The system simulates the tolerance of the wiring harness under extreme conditions to help evaluate the extreme use environment of the wiring harness.

[0092] See also Figures 1 to 5 , the present application provides a wiring harness connection performance intelligent testing method, comprising the following steps:

[0093] Step S1: collecting the wiring harness test data generated by the wiring harness test equipment and the wiring harness sensor in real time through the edge computing node;

[0094] Specifically, first, an edge computing node is set to connect to the wire harness test equipment and sensors for real-time collection of electrical and mechanical performance data of the wire harness. The edge computing node is installed in a physical location close to the wire harness to reduce data transmission delays and ensure the real-time nature of the collection. During the test of the wire harness, the edge node receives real-time electrical parameters (such as voltage, current) and mechanical parameters (such as tensile force, deformation, etc.) through the interface with the sensor. The edge computing node performs preliminary data preprocessing, including data format conversion and noise filtering. For example, the collected raw electrical signal is first filtered to eliminate high-frequency interference, and then the data is formatted, such as converting the collected analog signal into a standard digital format and marking a timestamp. After collecting the data, the edge node will package it in a specified data format and send it to the storage module to form a preliminary wire harness test data stream that is updated in real time.

[0095] Step S2: performing electrical performance detection extraction and mechanical performance detection extraction according to the wire harness test data, and obtaining wire harness electrical performance detection data and wire harness mechanical performance detection data respectively;

[0096] Specifically, based on the collected preliminary test data, the data is layered and classified. First, the edge computing node layers the data according to its attributes and separates the electrical parameter and mechanical parameter data. After stratification, the electrical and mechanical performance data are extracted separately. For electrical data, the key characteristic values ​​of voltage and current are extracted, and their statistical characteristics such as mean and variance are calculated, or the signal amplitude and frequency are sampled to obtain basic data that can reflect the electrical characteristics. For mechanical data, stress and strain analysis is performed, and the tensile force and deformation data are smoothed based on the time series to better analyze the trend of the data. The final electrical and mechanical data groupings will be normalized according to different detection standards to ensure the accuracy of subsequent analysis.

[0097] Step S3: performing an in-depth evaluation of electrical performance according to the wire harness electrical performance test data to obtain wire harness electrical performance evaluation data, and performing an in-depth evaluation of mechanical performance on the wire harness mechanical performance test data to obtain wire harness mechanical performance evaluation data;

[0098] Specifically, after obtaining the electrical and mechanical performance test data, they are deeply analyzed to obtain detailed performance evaluation results. The electrical performance in-depth evaluation uses the electrical evaluation model to first parse the input data and map the electrical performance data to multiple hidden layers, with each layer extracting the features of the signal step by step. The intermediate hidden layer adjusts the model structure by updating the weight and bias parameters to reflect the various characteristics of the electrical performance. Weights are introduced into the extracted features to highlight the key features in the signal and suppress the noise features, and the comprehensive score of the electrical performance of the harness is output. For the in-depth evaluation of mechanical performance, stress-strain analysis is first performed to construct a mechanical model, and multiple sets of tensile force and deformation data are used to calculate the mechanical tolerance of the harness, and then nonlinear modeling is combined to predict its mechanical stability under various loads. After the evaluation, the mechanical performance score and its tolerance limit under various load conditions are output.

[0099] In order to conduct an in-depth evaluation of the electrical performance of the wiring harness, a time series prediction model based on an LSTM neural network can be used. Electrical performance data collected from wiring harness test equipment and sensors. The input data includes: current (unit: A), voltage (unit: V), resistance (unit: Ω), temperature (unit: °C), and other electrical parameters (such as power, fluctuation frequency). Electrical data is a time-based sequence data. The model inputs multidimensional feature values ​​within a time window each time. The time window can be set to 10 seconds, 1 minute, etc. The window size is 60 (for example, 60 data are collected within 1 minute), then the shape of the input tensor is (batch_size, time_steps=60, features=4), where features=4 represents the four features of current, voltage, resistance, and temperature. The LSTM model includes two processing layers, such as the LSTM layer and the hidden layer. The main task of the LSTM layer is to extract key time-dependent features from the input time series. Through its inherent memory unit and gating mechanism, the LSTM layer can remember long-term time dependencies and selectively forget unimportant information. It is particularly suitable for processing sequence data with long-term dependencies (such as historical data of current signals or changing trends of harness test data). The specific structure includes weight calculation through the forget gate and selection of memory data through the input gate to perform candidate memory calculation (including activation calculation or other candidate calculation models), and send it to the output gate for output by controlling the output of the time step. The main task of the hidden layer is to further transform and nonlinearly map the time series features output by the LSTM layer. The hidden layer can enhance the expression ability of the model by applying different activation functions, so that the features are more suitable for output prediction. In most cases, the hidden layer is a fully connected layer in the present invention, which integrates and transforms the time series features transmitted from the LSTM layer to generate a higher-level feature representation. The hidden layer consists of one or more fully connected layers, including intermediate structures such as weight matrices, biases, and activation functions. The output of the model is the predicted probability of failure (ranging between 0 and 1), or a range related to the label such as damage rate or damage condition. By setting a threshold (for example, 0.5), the output value is converted into a judgment of failure or non-fault. Since fault detection is a binary classification problem (fault / non-fault), binary cross entropy is selected as the loss function for model training.

[0100] Step S4: Perform environmental pressure cross-reference based on the wire harness electrical performance evaluation data and the wire harness mechanical performance evaluation data to obtain wire harness connection performance test evaluation data.

[0101] Specifically, after obtaining the electrical performance evaluation data and mechanical performance evaluation data, environmental parameters are added for cross-analysis. First, the environmental data (such as temperature, humidity, and vibration) of the wiring harness are collected. These environmental data are matched with the performance data to form a data group associated with the environment and performance. Next, the system performs a multi-dimensional cross-analysis of electrical, mechanical, and environmental parameters. During the analysis process, the system tests the impact of environmental changes on electrical and mechanical properties one by one. For example, in a high temperature environment, the humidity conditions are gradually increased and the changes in the electrical performance score and the mechanical tolerance score are monitored to form a characteristic curve. Finally, based on the results of the cross-analysis, the adaptability score and tolerance limit conditions of the wiring harness in a specific environment are calculated, and the comprehensive performance evaluation data is output to demonstrate the reliability and durability of the wiring harness in an actual environment.

[0102] Specifically, the electrical and mechanical performance evaluation data are combined to evaluate the tolerance under environmental pressure. When the working environment temperature of the wiring harness exceeds the set range, the system will automatically perform additional in-depth analysis of the electrical and mechanical performance data and provide a comprehensive performance evaluation based on the current environmental conditions. Using multimodal analysis technology, the electrical and mechanical data are mapped to a unified evaluation space for data fusion. Combined with deep learning networks (such as the fusion model CNN+RNN architecture), comprehensive evaluation of different data modalities is achieved.

[0103] Preferably, step S1 specifically includes:

[0104] Step S11: collecting initial wiring harness test data generated by wiring harness test equipment and wiring harness sensors in real time through edge computing nodes;

[0105] Specifically, first, an interface connection with the harness test equipment and sensors is set up in the edge computing node to receive real-time electrical and mechanical test data generated by the sensor. The specific data includes electrical and mechanical characteristics such as voltage, current, tensile force, and deformation. The edge computing node converts the analog signal collected by the sensor into a digital signal and attaches a timestamp to each data record to maintain the integrity and consistency of the time series in subsequent processing. In order to reduce the load on the edge computing node, the initial data will use data compression technology to reduce the amount of transmitted data during the collection process, such as sampling at a fixed frequency (sampling frequency of 10 times per second) to reduce redundant information while ensuring that key data is not missed.

[0106] Step S12: performing interference data separation according to the initial wire harness test data to obtain wire harness test separation data;

[0107] Specifically, the initially collected data contains a large amount of external interference information, such as electromagnetic interference, background noise, etc. The edge computing node detects interference by analyzing the statistical characteristics of the data, such as by calculating the peak value and variance of the electrical signal. If the voltage and current fluctuations are found to be beyond a reasonable range, the outliers are identified. For the outliers identified in the data, the node uses a feature separation method to perform interference stripping. First, the continuity and difference of the data points are detected, and the abnormal peaks in the signal are separated from the background noise. After the separation is completed, the data stripped of interference is stored in an independent data group, retaining the valid signal. After the interference data is stripped, the remaining data is the wire harness test separation data, which removes most of the external interference factors.

[0108] Step S13: filtering and detecting the wire harness test separation data to obtain wire harness test filtering detection data;

[0109] Specifically, after interference separation, the edge computing node further filters and detects the data, mainly processing the residual noise in the electrical and mechanical signals. The filtering process separates the low-frequency and high-frequency components and removes high-frequency noise from the signal to improve the clarity of the data. The specific implementation includes first converting the data into the frequency domain and decomposing it into multiple frequency components. Then, by setting the filtering threshold, the low-frequency part is retained and the high-frequency part is suppressed. For example, for voltage signals, its components are retained within the range below the set high-frequency noise threshold. The filtered data will be transmitted back to the processing module of the edge node to form the wiring harness test filtering detection data to ensure that the data only contains valid signal components and avoid noise interference with the analysis.

[0110] Specifically, the Kalman filter is used to fuse the current data with the predicted data to reduce data fluctuations. There is noise in the observed data and state transitions of the system. Through the iterative calculation of the Kalman filter, the optimal estimate of the state can be obtained each time, thereby reducing the error caused by noise. At each time step, the Kalman filter is mainly divided into two steps: prediction and update. In the prediction stage, the filter generates the predicted state and prediction error covariance of the current time step based on the dynamic model of the system, and obtains the predicted data based on the historical state. Next, in the update stage, the filter fuses the predicted data with the current observed data and calculates a weight coefficient (i.e., the Kalman gain). The Kalman gain is automatically adjusted according to the size of the prediction error and observation noise to determine the weight of the current observed data and the predicted data. Finally, by weighted fusion of the predicted data and the observed data, the filter outputs an updated state estimate. The key to the Kalman filter is to balance the weight between the current observed data and the predicted data. For observed data with large noise, the Kalman filter will reduce the weight of the observed data and rely more on the predicted data of the model, thereby reducing the fluctuation caused by noise; when the observed data is relatively stable, the filter will increase its reliance on the observed data, making the estimated state closer to the actual data.

[0111] Step S14: performing back-tracking calibration according to the wire harness test filter detection data to obtain wire harness test data.

[0112] Specifically, there is a certain time deviation and baseline offset in the filtered data. To this end, the edge computing node performs back-tracking calibration to calibrate the signal by comparing it with the previous benchmark data. The first step of back-tracking calibration is to establish a benchmark signal model, and use the statistically stable part of the historical data as the benchmark model. Every time new data is collected, the difference between the new data and the benchmark signal is compared. If it is found that the characteristic deviations such as voltage and tensile force in the electrical and mechanical data are large, the benchmark model is used to correct the data. The last step of the calibration is to adjust the deviated signal through signal back-tracking processing to keep it consistent with the benchmark signal to reduce the deviation in the collected data, and the calibrated data is stored as harness test data.

[0113] Specifically, back-testing calibration corrects current data through historical data or benchmark data to improve data consistency. Use back-testing analysis technology to calibrate current data through historical trends and adjust deviations to a reasonable range. Calibrate current data based on historical means. Calculate the historical mean of similar data and fine-tune current data to ensure that the data conforms to historical distribution, such as calculating a sliding average within a certain time window to correct current data. Furthermore, a linear regression or support vector regression (SVR) model is used to use historical data to back-test and correct current values, where the weight parameters or mathematical parameters of linear regression or support vector regression are trained using historical data.

[0114] Preferably, step S12 is specifically:

[0115] Perform edge collaborative anomaly cleaning based on the initial harness test data to obtain preliminary harness test cleaning data;

[0116] Specifically, the edge computing node detects abnormal values ​​in the collected data by real-time collaboration with the test sensors. First, the fluctuation amplitude of electrical data (such as voltage and current) and mechanical data (such as tensile force and deformation) is calculated to identify abnormal large fluctuations in the data. The sliding window method is used to analyze the stability of the data and set the abnormal threshold. For example, when the voltage fluctuates by more than 0.5V within a certain sliding window, it is marked as abnormal. The system dynamically removes these abnormal data according to the set threshold. The cleaned data is retained in the preliminary harness test cleanup data, removing most of the interference fluctuations and discrete abnormal data to ensure the basic stability of the data.

[0117] Performing spectrum denoising on preliminary harness test cleanup data to obtain harness test cleanup feature data;

[0118] Specifically, a spectrum analysis is performed on the preliminary harness test cleanup data, and the time domain data is converted to the frequency domain in order to identify noise components of different frequencies. For electrical data and mechanical data, their spectrum characteristics are analyzed, and high-frequency noise components in the spectrum are found. A spectrum denoising method is used to set a specific noise frequency range to suppress or remove noise frequency bands, such as noise signals above 10Hz. Through inverse transformation, the denoised data is converted back to a time domain signal to generate harness test cleanup feature data. Spectral denoising ensures that only key frequency bands that affect performance are retained in the data, suppressing the impact of background noise.

[0119] Performing multi-level feature deconvolution on the wire harness test cleanup feature data to obtain wire harness test deconvolution feature data;

[0120] Specifically, the wire harness test cleanup feature data is subjected to multi-level feature unwrapping processing to separate the subtle features of different frequency bands and feature layers in the signal. The first layer of multi-level unwrapping targets low-frequency features to extract the overall trend of the signal; the middle layer extracts the intermediate frequency features of the signal to capture the main change patterns of voltage and current in the electrical signal; the high level is used to separate high-frequency features to further eliminate residual tiny noise. The full-band features of the signal are generated after the unwrapping results of each level are superimposed, ensuring that the signal removes unnecessary subtle fluctuations while retaining the main features.

[0121] Performing dimensionality reduction processing on the wire harness test unrolling feature data to obtain wire harness test feature dimensionality reduction data;

[0122] Specifically, the deconvoluted feature data is processed for dimensionality reduction to reduce the data dimension and remove redundant information. The dimensionality reduction process selects the main components of the signal and maps the main features to a lower dimensional space. During dimensionality reduction, more than 80% of the data variance is retained to ensure that the main features of the signal are not lost, forming simplified feature data, greatly reducing the amount of data and redundancy, and providing a clear feature structure for subsequent signal analysis.

[0123] Perform heterogeneous signal analysis on the wire harness test feature dimension reduction data to obtain wire harness test analysis feature data;

[0124] Specifically, the graph structure is constructed for the dimension-reduced data, and the features of different frequency bands are used as nodes. The correlation between nodes represents the correlation between different features. Based on the heterogeneous signal analysis of the graph structure, multiple characteristic patterns in the data are identified, and the heterogeneity of the nodes is weighted to more accurately remove redundant interference signals. Heterogeneous analysis generates wire harness test analysis feature data, providing clear separation for key characteristic patterns in the signal.

[0125] Perform time series debiasing on the wire harness test profiling feature data to obtain wire harness test time series purified data;

[0126] Specifically, the analyzed data is debiased in time series to eliminate the long-term trend offset in the data. The time series trend of the data is analyzed to find the overall offset. The drift deviation in the time series is corrected using a time alignment algorithm. For example, when the mean deviation of the signal offset exceeds the set threshold, the data is traced back to the reference time axis. The debiased data generates time series cleansing data to ensure the consistency of the data in the time dimension.

[0127] Perform aggregate signal reconstruction on the wiring harness test timing purification data to obtain wiring harness test reconstruction target data;

[0128] Specifically, the multi-dimensional signals of the time series purification data are aggregated and reconstructed to form a signal structure. Aggregate reconstruction integrates different signal features into a complete time series through multi-layer signal aggregation. During the reconstruction process, the signal is smoothed to ensure the coherence and consistency of the reconstructed signal. The reconstructed signal forms the target data for the harness test reconstruction, which contains complete signal information and removes interference factors.

[0129] The wire harness test reconstruction target data is subjected to verification interference separation to obtain the wire harness test separation data.

[0130] Specifically, the verification method is used to perform the final interference separation on the reconstructed target data to ensure that there are no residual interference factors in the signal. The verification process compares the reconstructed target data with the reference data to detect the difference. When interference features are detected, these residual noises are stripped away through the verification filter to retain the pure features of the signal. The wire harness test separation data is generated to ensure that the signal data is highly pure and provide accurate input for performance testing and evaluation.

[0131] Preferably, the spectrum denoising is specifically:

[0132] Performing waveform decomposition on preliminary harness test cleanup data to obtain harness waveform decomposition data;

[0133] Specifically, the preliminarily cleaned harness test data is decomposed into basic waveforms to facilitate subsequent spectrum analysis. Waveform decomposition can be performed by transforming the signal into multiple wavelet coefficients, each of which corresponds to a specific frequency band in the signal, through methods such as wavelet transform. For each sampling time point, its wavelet coefficient is calculated to decompose the time signal into basic waveforms of different frequency bands.

[0134] Performing spectral component analysis on the wire beam waveform decomposition data to obtain wire beam spectral component data;

[0135] Specifically, the decomposed waveform data is subjected to spectral component analysis, and each basic waveform is converted into the frequency domain to extract the spectral components of each frequency band. The specific process includes performing a fast Fourier transform on each waveform to convert it from the time domain to the frequency domain. After the spectral components are generated, the amplitude of each frequency component can be clearly observed, providing the frequency distribution information of the signal. Beam spectral component data: contains the amplitude information of each frequency, such as [frequency 1, amplitude 1], [frequency 2, amplitude 2], ... [frequency N, amplitude N], [frequency N+1, amplitude N+1], showing the energy distribution of different frequency components.

[0136] Performing spectrum peak detection on the spectrum component data of the wire beam to obtain the high-frequency noise characteristic data of the spectrum of the wire beam;

[0137] Specifically, peak detection is performed on the spectrum component data to identify possible high-frequency noise components. By calculating the amplitude of each frequency point, the frequency band with a significantly higher amplitude than other frequencies is found. Threshold detection is set. For example, when the amplitude of a certain frequency is greater than the set reference value (such as twice the mean), it is marked as a noise frequency band.

[0138] Performing filter parameter mapping on the high-frequency noise characteristic data of the wire harness spectrum to obtain filter parameter data;

[0139] Specifically, suitable filter parameters are generated based on the high-frequency noise characteristic data so as to effectively suppress the noise frequency band in subsequent processing. The filter parameter mapping includes setting the filter cutoff frequency and bandwidth for the noise frequency band. By setting the threshold and bandwidth of the noise frequency band, the parameters of the low-pass or band-stop filter are generated accordingly. These parameters are stored as a filter configuration.

[0140] De-noising the line beam spectrum component data according to the filtering parameter data to obtain the line beam spectrum de-noising data;

[0141] Specifically, the spectral component data is denoised using the filter parameter data, specifically including using a low-pass filter or a band-stop filter to remove the high-frequency noise band. Through filter design, the high-frequency components are suppressed or removed, and the low-frequency signals are retained to reduce the impact of noise. For example, the frequency band below 10Hz is low-pass filtered to remove the high-frequency noise components. The spectrum data obtained after processing is smoothed in the high-frequency part to retain the main signal while removing the high-frequency noise.

[0142] The key frequency features are extracted from the wire harness spectrum denoising data to obtain the wire harness test cleanup feature data.

[0143] Specifically, in the denoised spectrum data, key frequency features are extracted, such as the main frequency of the electrical signal and the fundamental frequency of the mechanical signal. These characteristic frequencies can help identify the main performance characteristics of the wiring harness. By statistically analyzing the frequency distribution information, the low-frequency band with the highest energy in the spectrum data is extracted to form the feature set of the main signal. The extracted key frequency features are further used for performance evaluation to ensure that the denoised data only retains the important frequencies related to performance evaluation.

[0144] Preferably, the multi-level feature deconvolution is specifically:

[0145] Performing frequency domain conversion on the wire harness test cleanup feature data to obtain the wire harness test frequency domain conversion data;

[0146] Specifically, first, the wire harness test cleanup feature data is converted to the frequency domain so that the spectral characteristics of the signal can be analyzed in the frequency domain. Frequency domain conversion can convert the time domain signal into a frequency domain representation through methods such as Fourier transform. In this step, the sampling point of each time domain signal is converted into amplitude and phase information on the spectrum so that the energy distribution of each frequency band in the signal can be observed.

[0147] Performing frequency domain separation on the wire harness test frequency domain conversion data to obtain the wire harness test frequency domain separation data;

[0148] Specifically, through frequency domain separation, each frequency band in the frequency domain conversion data is separated independently, and the signal is divided into low frequency, medium frequency and high frequency bands according to the frequency. It helps to identify and isolate the characteristics of different frequencies in the signal. For example, data with a frequency below 1Hz is classified as low frequency, data between 1Hz and 5Hz is classified as medium frequency, and data above 5Hz is classified as high frequency.

[0149] Perform multi-scale frequency band convolution calculation based on the wire harness test frequency domain separation data to obtain the wire harness test frequency band decoupling data;

[0150] Specifically, multi-scale convolution calculations are performed on the separated low-frequency, medium-frequency and high-frequency band data to further extract the detailed features of each frequency band. Multi-scale convolution calculations process the signal features of different frequency bands through convolution kernels of different sizes. For example, a larger convolution kernel is used to extract the global features of low frequencies, and a smaller convolution kernel is used to extract the local features of high frequencies. Each convolution operation generates features of a specific frequency band, which facilitates more fine separation of the hierarchy of the signal during layered deconvolution.

[0151] Specifically, multi-scale convolution is performed on the separated frequency band data to further extract the characteristics of each frequency band and decouple the complex signal components. Convolution kernels of different scales (such as 3, 5, 7, etc.) are used to act on each frequency band data to enhance the detection of different frequency features. The convolution layer in the convolutional neural network (CNN) is used to simulate the frequency band convolution, or the frequency band data is processed by sliding window convolution. For example, a convolution with a window size of 3 is applied to low-frequency data, a convolution with a window size of 5 is applied to medium-frequency data, and a convolution with a window size of 7 is applied to high-frequency data.

[0152] Perform layered unwinding calculation based on the decoupling data of the wire harness test frequency band to obtain the layered unwinding data of the wire harness layer;

[0153] Specifically, layered unconvolution is performed based on the frequency band decoupling data, and the features at different levels in the signal are gradually separated through multi-layer unconvolution operations. The low level of layered unconvolution captures the global features of the signal, and the high level captures finer details. For example, the first layer of unconvolution extracts the overall fluctuation trend of the signal; the second layer of unconvolution separates the local feature changes in the mid-frequency band; and the third layer of unconvolution is used to identify high-frequency noise or subtle fluctuations. The results of each layer of unconvolution form a hierarchical structure, so that the main components and secondary components of the signal are separated hierarchically.

[0154] Specifically, the decoupled frequency band data is layered and unconvolved to further remove the superimposed spectral components so that each layer of data represents a specific frequency feature. The data is further layered through wavelet decomposition to extract different details and approximate layers of the signal. Use wavelet transform (such as discrete wavelet transform DWT) to decompose the signal at multiple levels and extract signal layers of different resolutions. Use DWT to decompose the intermediate frequency data into three layers to generate three detail layers and one approximate layer. The specific implementation method is as follows: import pywt# Use wavelet decomposition for layered unconvolution coeffs = pywt.wavedec(mid_freq_convolved, 'db4', level=3)# coeffs contains three detail layers and one approximate layer.

[0155] Specifically, more importantly, the decoupling data of the wire harness test frequency band is layered and unwrapped to obtain the wire harness frequency band layered data, wherein the wire harness frequency band layered data includes the wire harness frequency band high-level data, the wire harness frequency band middle-level data and the wire harness frequency band low-level data; Gaussian convolution calculation is performed on the wire harness frequency band high-level data to obtain the first wire harness frequency band feature data; self-attention convolution calculation is performed on the wire harness frequency band middle-level data to obtain the second wire harness frequency band feature data; clustering processing and small convolution calculation are performed on the wire harness frequency band low-level data to obtain the third wire harness frequency band feature data; the spectrum of the first wire harness frequency band feature data, the second wire harness frequency band feature data and the third wire harness frequency band feature data is reconstructed to obtain the wire harness layer layered unwrapped data. The high-level data mainly contains high-frequency signal components and noise. Gaussian convolution is a smoothing filtering method that is suitable for suppressing high-frequency noise while retaining the global trend, so it is particularly effective when processing high-level frequency band data. The middle-level data contains lower intermediate frequency signals and is the key characteristic frequency band in the signal. Self-attention convolution allows the model to find important feature areas in the signal and give them higher weights, so it is suitable for strengthening the recognition and extraction of important information in mid-level frequency band data. Low-level data represents low-frequency components and is the global feature of the signal. Clustering processing combined with small convolutions to process low-level data can perform deeper pattern recognition of global changes in the signal. Clustering processing can classify the main patterns in the frequency band, while small convolutions are used to accurately extract the features of each category.

[0156] The spectrum is reconstructed based on the layered unwinding data of the wire harness to obtain the unwinding characteristic data of the wire harness test.

[0157] Specifically, after the layered unwrapping is completed, spectrum reconstruction is performed to integrate the layered data into a complete signal to retain the core features and remove unnecessary noise. During the reconstruction process, the components of each layer are combined according to the weights so that the global features of the low frequency are retained and the high frequency noise is weakened. Through the inverse transformation, the frequency domain data is restored to the time domain to form the wire harness test unwrapping feature data. The reconstructed signal presents clear main features and removes unnecessary subtle noise.

[0158] Specifically, the spectrum is reconstructed on the basis of multi-level deconvolution to generate clear and denoised deconvolution feature data. The layered data are recombined and inversely transformed back to the frequency domain to obtain the denoised spectrum data. Through wavelet reconstruction or frequency domain merging, the layered data are superimposed and inversely transformed to the time domain. Wavelet reconstruction is performed on low-frequency, medium-frequency and high-frequency data respectively, and the layered data are merged into a complete spectrum.

[0159] Preferably, the heterogeneous signal analysis is specifically:

[0160] Constructing a graph structure for the wire harness test feature dimensionality reduction data to obtain wire harness test feature graph data;

[0161] Specifically, first, a graph structure is constructed based on the dimensionality reduction data of the harness test features to represent the relationship between different features in the signal. Each feature data point (such as different frequency band characteristics of voltage and tensile force) is used as a node in the graph, and the edges between the nodes represent the degree of association between the features. For each node, the connection weights between the nodes are calculated based on distance metrics or correlation. For example, similar frequency components will be assigned higher connection weights, while nodes with larger frequency differences will have lower connection weights. After the graph structure is formed, it can well represent the signal feature relationship in the dimensionality reduction data and provide a basis for the construction of heterogeneous graphs.

[0162] Specifically, the reduced dimension data of the wiring harness test features is constructed into a graph structure to capture the correlation between the data. Each signal feature after dimensionality reduction is regarded as a node in the graph, and the similarity or correlation between different features is defined as the edge of the graph. The edge weights between nodes are calculated by methods such as cosine similarity or Euclidean distance. The higher the similarity, the higher the edge weight between nodes. If the reduced dimension feature data contains the main components of current and voltage, nodes A and B in the graph are constructed respectively. The similarity between A and B is calculated by cosine similarity to determine the edge weight.

[0163] A heterogeneous graph is constructed according to the wire harness test characteristic graph data to obtain the wire harness test heterogeneous graph data;

[0164] Specifically, based on the feature graph data, a heterogeneous graph is constructed so that different types of signal features (such as electrical and mechanical features, or different frequency bands) are represented in the form of heterogeneous nodes. Heterogeneous graphs can express complex relationships between heterogeneous feature nodes, such as the mutual influence of electrical and mechanical features. In the graph, the node category of each feature is identified by the node type, and the association relationship between heterogeneous nodes is reflected by cross-category connections. For example, the node with an electrical frequency of 0.5 Hz is associated with the node with a mechanical frequency of 1 Hz through a low-weight connection, while the electrical high-frequency noise node has a high-weight connection. The construction of heterogeneous graphs enables data with different characteristics to be represented in layers and described in terms of association, making it easier to isolate different types of signal features during processing.

[0165] Specifically, different types of signal features (such as current, voltage, and temperature) are integrated into a heterogeneous graph to capture the diverse associations between features. In the constructed graph structure, different types of nodes are added to represent different signal sources or features. For example, current features and voltage features belong to different types of nodes. Edges are created between different types of nodes to capture the heterogeneous relationships between different signal features. The weight of the edge can be calculated based on the similarity or correlation across types of signals. For example, the features of current and voltage are represented as different types of nodes A and B in the graph. If the current and voltage have a high correlation in certain frequency components, such as change frequency, change acceleration, regional correlation, or time change synergy, a weighted edge is added between A and B.

[0166] Extracting image signal features based on the wiring harness test heterogeneous image data to obtain wiring harness test image feature data;

[0167] Specifically, based on the heterogeneous graph data, the signal features of each node are extracted from the graph. The influence of each node's neighbor nodes is calculated to identify important signal features. In the process of graph signal feature extraction, similar features of each node are aggregated. For example, the electrical frequency bands 0.5Hz and 1Hz have a high correlation, so their features will be aggregated to enhance representativeness. Similar aggregation is also adopted for the stretching and deformation nodes in the mechanical frequency band. The extracted graph signal feature set contains the main feature information of each node in the heterogeneous graph, retains the core features of the signal and removes irrelevant node noise.

[0168] Perform feature noise stripping according to the feature data of the harness test graph to obtain optimized graph feature data;

[0169] Specifically, feature noise stripping is performed in the extracted graph feature data. Nodes with noise features in the heterogeneous graph (usually high-frequency nodes or relatively independent nodes) are identified and stripped. The feature noise stripping process removes noise nodes with low similarity to the main feature nodes based on the metric values ​​in the graph. For example, if the high-frequency noise node (10Hz) in the electrical node has a low similarity to the main signal node (0.5Hz), it will be stripped from the feature data. After stripping the noise nodes, the retained feature set will focus on the main features of the signal, eliminating high-frequency or independent noise interference.

[0170] Specifically, a noise reduction algorithm (such as low-pass filtering) is used or noise features are removed based on a threshold, or principal component analysis (PCA) is used to remove low variance components and retain the main features.

[0171] High-order features are reconstructed based on the optimized graph feature data to obtain the wiring harness test analysis feature data.

[0172] Specifically, high-order feature reconstruction is performed using optimized graph feature data. Multi-level signal features are combined into a complete signal structure by weighted merging of core node features in heterogeneous graphs. In high-order reconstruction, signals are weighted and combined based on the aggregation characteristics of nodes. For example, low-frequency components in electrical features are given higher weights, and high-frequency components in mechanical features are given lower weights to achieve high-fidelity reconstruction of signals. The reconstructed signal retains the key features of electrical and mechanical features, removes noise and minor components, and forms wiring harness test analysis feature data.

[0173] Specifically, the optimized graph features are reconstructed at a high level to generate harness test profiling feature data. The optimized features are reconstructed using a reconstruction algorithm (such as an autoencoder) to obtain a higher-level feature representation. The autoencoder is applied to encode and decode the graph features to generate high-order features with stronger representation capabilities. The autoencoder is used to encode and decode the optimized 4-dimensional features to generate new high-order features as profiling feature data. Furthermore, the autoencoder includes two encoding layers and one decoding layer, the two encoding layers are a mask encoding layer and an interpretation encoding layer, and the decoding layer is a multi-head self-attention decoding layer. The role of the mask encoding layer is to mask certain parts of the input data so that the model can learn to effectively extract features even when some inputs are missing. Specifically, the mask encoding layer randomly sets or masks a part of the input data, which can help the model enhance its robustness to noise and missing data. This process is similar to data enhancement, but at the feature level, the model is forced to recognize the core pattern of the input data, rather than relying too much on each specific value, using random masks or masks with prior knowledge, such as masking specific parts according to data characteristics (such as known interference frequency bands). The role of the interpretation coding layer is to interpret and process the output data of the mask coding layer through further feature extraction, so that the model can pay more attention to the local details and complex relationships of the features after obtaining the main pattern. The interpretation coding layer consists of multiple layers of neural networks or convolutional layers, which further compress the input features to obtain a more refined representation. The core of the interpretation coding layer is to introduce an explanatory mechanism. The interpretation coding layer adaptively assigns different weights to different parts of the input features according to the output characteristics of the mask coding layer, thereby giving higher weights to important features and lower weights to unimportant or noise features. This layer can also assign different weights to different features based on the importance of the features, so that the model is more inclined to learn important features during the interpretation of the features. The core task of the multi-head self-attention decoding layer is to decode the encoded low-dimensional features into high-dimensional representations and achieve global fusion of information in the process. Through the self-attention mechanism, the decoding layer can flexibly capture the dependencies between different features. The main steps of the self-attention decoding layer are as follows. In the multi-head self-attention mechanism, the input features are split into multiple heads, and each head independently calculates different attention weights. The advantage of multi-head attention is that it can capture the association between features from different angles. The self-attention mechanism encodes each input feature into three parts: query, key, and value, calculates the attention weight of each feature to other features, and generates a weighted combined output. The multi-head attention mechanism allows the model to capture long-distance dependencies between features, enabling the decoding layer to fuse the interactions between features during the reconstruction process, rather than just local time points or spatial points.After independent feature decoding, the output of each head is concatenated or weighted to form the final decoded output. Through the feature fusion of multiple heads, the model can pay attention to global and local information during the reconstruction process and flexibly adapt to the association between complex features.

[0174] Preferably, step S2 specifically includes:

[0175] Step S21: stratify the test data according to the wiring harness test data to obtain wiring harness test stratification data;

[0176] Specifically, first, the data is divided into a hierarchical structure according to the source or characteristics of the harness test data. Test data usually contains electrical performance data (such as voltage, current) and mechanical performance data (such as tensile force, displacement), which are different in nature and test indicators, so hierarchical processing helps further analysis. Use attribute labels or field tags to stratify data. For example, the collected data will be divided according to the "electrical" or "mechanical" label. During the stratification process, check whether the characteristics of each type of data meet the typical range of electrical or mechanical properties. The stratified data stream is stored in a structured format, and each layer only contains characteristic data of a certain nature, providing a clear hierarchical structure for subsequent classification and screening.

[0177] Step S22: performing classification screening according to the wiring harness test hierarchical data to obtain wiring harness test classification data, wherein the wiring harness test classification data includes wiring harness electrical performance test data and wiring harness mechanical performance test data;

[0178] Specifically, on the basis of the hierarchical data, further classification and screening are carried out to extract performance data that meets specific test requirements. First, the characteristic value of each data point is checked in the electrical and mechanical layers to determine whether it meets the set electrical or mechanical performance standards. The data is screened by logical conditions. For example, the range of voltage and current data in the electrical layer should be within the set value; the values ​​of tensile force and displacement in the mechanical layer should also meet the design requirements. The data that meets the conditions is retained in the classified data set. After the classification and screening is completed, two main performance data sets are generated: wiring harness electrical performance test data and wiring harness mechanical performance test data, ensuring that the classified data can be directly used for specialized testing of electrical or mechanical performance.

[0179] Step S23: performing electrical performance detection conversion according to the wiring harness electrical performance test data to obtain wiring harness electrical performance detection data;

[0180] Specifically, the classified electrical performance test data is subjected to detection conversion to extract key signal features. The detection conversion process includes signal correction, normalization, and feature extraction to ensure that the electrical data is accurate during processing. First, signal correction is performed to adjust the deviations in the voltage and current data to the reference value. Next, normalization is performed to scale different test values ​​to a uniform magnitude for unified processing. Feature extraction mainly targets the key characteristics of electrical data, such as average voltage, current fluctuation range, and spectral characteristics. The extracted features constitute the electrical performance detection data.

[0181] Step S24: Perform mechanical property detection conversion according to the wire harness mechanical property test data to obtain the wire harness mechanical property detection data.

[0182] Specifically, the wire harness mechanical property test data is tested and converted to obtain key characteristic values ​​related to the mechanical properties. Mechanical detection conversion usually involves stress-strain analysis, data smoothing and feature extraction. First, the tensile force and displacement data are processed through the stress-strain model to obtain the mechanical response of the wire harness under different stress conditions. Secondly, a smoothing algorithm is applied to eliminate the noise in the data to obtain stable mechanical properties. Finally, feature extraction is performed to obtain key parameters of mechanical properties, such as average tensile force, maximum strain value and load range. Mechanical property test data can help determine the mechanical stability and tolerance of the wire harness.

[0183] Preferably, step S3 is specifically:

[0184] Step S31: using a preset wiring harness electrical evaluation model to perform an in-depth electrical performance evaluation on the wiring harness electrical performance detection data to obtain wiring harness electrical performance evaluation data;

[0185] Specifically, based on the harness electrical performance test data, the performance evaluation is performed using the constructed harness electrical evaluation model. The evaluation model consists of multiple neural network layers, which can identify complex features in the data and output electrical performance scores. First, the electrical performance test data is input into the primary hidden layer of the model, which extracts basic features from the data and converts the original voltage and current data into primary features. Then, the multi-head self-attention layer is used to enhance the representation of important features in the data. The self-attention layer calculates the correlation between the features and assigns higher weights to features with strong correlation, so that the model can identify key characteristics in the electrical signal. In the secondary hidden layer, the weighted features are further refined to obtain a comprehensive electrical performance score. The output electrical performance evaluation data includes the stability and reliability scores of the electrical system, providing support for subsequent evaluations.

[0186] Step S32: performing stress-strain nonlinear modeling according to the wire harness mechanical property detection data to obtain a wire harness mechanical stress model;

[0187] Specifically, a nonlinear stress-strain model is established through the mechanical properties test data of the wire harness to reflect the mechanical response of the wire harness under different load conditions. The model is based on nonlinear equations and can accurately depict the complex relationship between stress and strain. First, based on the detected tensile force and displacement data, the stress and strain values ​​at each load level are calculated. The data is then substituted into the nonlinear equation for fitting to obtain the model parameters of the curve. During the fitting process, the stress-strain curve is iterated and optimized multiple times until the model parameters are stable. The completed stress model can reflect the strain changes of the wire harness under stress conditions and form a basic model of mechanical properties.

[0188] Step S33: Estimating the mechanical load tolerance according to the wire harness mechanical stress model to obtain wire harness mechanical performance evaluation data;

[0189] Specifically, the mechanical stress model is used to evaluate the tolerance of the wiring harness under various load conditions through multiple simulation calculations. This process calculates the ultimate strain and deformation of the wiring harness under different stress conditions by substituting different load values ​​into the stress model. According to the tolerance standard, a threshold is set (for example, the maximum allowable deformation is 0.5mm), and the strain value output by the model is tested at different load levels to determine whether it exceeds the threshold. The evaluation results include the maximum load that the wiring harness can withstand within a safe range, that is, the maximum load tolerance of the mechanical system without exceeding 0.5mm deformation.

[0190] The step of constructing the wiring harness electrical evaluation model preset in step S31 includes the following steps:

[0191] Obtain historical wiring harness electrical performance test data and historical electrical performance evaluation label data through a preset wiring harness history database;

[0192] Specifically, historical electrical performance test data and corresponding evaluation labels are extracted from the harness history database. The data includes past voltage and current information, and the label data is the electrical performance score (such as qualified or unqualified) annotated by experts. After the data collection is completed, each historical record is labeled to form a set of paired training data for model training and evaluation.

[0193] Perform data division on historical wiring harness electrical performance test data to obtain historical wiring harness test data and historical wiring harness verification data;

[0194] Specifically, the historical data is divided into a training set and a validation set to ensure that the model training process has good generalization. The data is usually divided in an 8:2 ratio, with 80% of the data used for training and 20% for validation. After the division, the data is divided into historical harness test data (training set) and historical harness validation data, which are used for model training and accuracy evaluation respectively.

[0195] Performing primary hidden layer processing on historical harness test data to obtain primary feature data of historical harness;

[0196] Specifically, the historical harness test data is input into the primary hidden layer (preliminary convolution layer calculation, pooling calculation, activation calculation and full connection layer calculation) for basic feature extraction. The primary hidden layer calculates the basic statistical features of the voltage and current of each record (such as mean, variance, etc.) to preliminarily characterize the electrical performance. After processing by the primary hidden layer, a feature data set with preliminary features is generated for further feature optimization.

[0197] Perform multi-head self-attention layer processing on the primary feature data of the historical line bundle to obtain the weighted feature data of the historical line bundle;

[0198] Specifically, the primary feature data is input into the multi-head self-attention layer, and the correlation between different features is calculated and weighted. The self-attention layer can capture the interaction between each feature to ensure that the model focuses on the most important electrical performance features. The weight of each feature combination reflects its contribution to the final performance evaluation. The output weighted feature data contains the key features after self-attention processing.

[0199] Performing secondary hidden layer processing on the historical line bundle weighted feature data to obtain the historical line bundle secondary feature data;

[0200] Specifically, the weighted feature data is further processed through the secondary hidden layer (further convolution layer calculation, pooling calculation, activation calculation and full connection layer calculation) to extract refined feature representation. The secondary hidden layer converts the electrical features into high-level features suitable for model evaluation through nonlinear transformation. The output data set of the secondary hidden layer contains the final extracted electrical performance features, which provides support for the model output score.

[0201] The historical wiring harness verification data is used to iteratively train the historical wiring harness secondary feature data and the model is constructed through the historical electrical performance evaluation label data to obtain the wiring harness electrical evaluation model.

[0202] Specifically, the secondary feature data is iteratively trained using the complete square error loss function using historical harness verification data. The secondary feature data of the training set is paired with the evaluation labels and input into the model. By calculating the error and back-propagating it, the model parameters are gradually adjusted to fit the historical electrical performance evaluation. After training, the model is able to generate a comprehensive evaluation of the electrical performance based on the input electrical feature scores.

[0203] Preferably, step S4 is specifically:

[0204] Step S41: obtaining wiring harness environment data;

[0205] Specifically, data related to the harness test environment is collected, including environmental parameters such as temperature, humidity, and vibration intensity that may affect the performance of the harness. This data can be collected in real time through sensors deployed at the test site. The collection frequency of environmental data should be consistent with the collection frequency of harness performance test data to ensure data alignment. After collection, the environmental data is cleaned to remove outliers or noise interference to obtain accurate environmental data as the basis for analysis.

[0206] Step S42: performing environmental data matching according to the wire harness environmental data, the wire harness electrical performance evaluation data, and the wire harness mechanical performance evaluation data to obtain test environment matching data;

[0207] Specifically, the environmental data is matched with the electrical performance evaluation data and the mechanical performance evaluation data to analyze the performance of the wiring harness under specific environmental conditions. The matching process is based on data feature comparison to ensure that data measured under the same conditions can be associated. By setting a conditional matching algorithm, such as matching data within the range of temperature, humidity and vibration intensity, the association of electrical and mechanical performance with environmental data is established. After matching, the test environment matching data is generated to describe the performance of the wiring harness under specific conditions.

[0208] Step S43: performing cross analysis according to the test environment matching data to obtain parameter cross analysis data;

[0209] Specifically, according to the test environment matching data, cross-analysis is performed to explore the combined impact of environmental factors on electrical and mechanical performance. Cross-analysis identifies possible impact patterns by comparing environmental parameters such as temperature, humidity, and vibration intensity with electrical and mechanical performance data in a multivariate manner. Through methods such as regression analysis or multivariate analysis of variance, the trend of changes in environmental changes (such as temperature rise or increased vibration) on electrical and mechanical performance is calculated to quantify the impact of environmental factors. The results of the cross-analysis can reveal the impact of each environmental parameter on the performance of the wiring harness, for example, the impact of temperature on electrical performance is greater than that on mechanical performance.

[0210] Step S44: verifying the environmental adaptability according to the parameter cross-analysis data to obtain the harness environmental adaptability data;

[0211] Specifically, the stability of the wiring harness under specific environmental conditions is evaluated through environmental adaptability verification. The verification process is based on parameter cross-analysis data, and the adaptability of the wiring harness under environmental changes is judged by comparing the electrical and mechanical properties with the set adaptability standards. During the adaptability verification process, the performance fluctuations of the wiring harness when the temperature, humidity, and vibration intensity change are checked to confirm whether the fluctuation range is within the allowable threshold. For example, when the temperature variation range is ±5°C, the electrical performance score fluctuation is ≤3 points, which is considered to be qualified for adaptability. After verification, the wiring harness environmental adaptability data is generated, indicating the performance adaptability of the wiring harness in a specific environment.

[0212] Step S45: Perform environmental tolerance evaluation based on the wire harness environmental adaptability data to obtain wire harness connection performance test evaluation data.

[0213] Specifically, an environmental tolerance assessment is conducted on the basis of environmental adaptability verification to analyze the tolerance limit of the wiring harness under extreme environmental conditions. The evaluation process compares the environmental adaptability data of the wiring harness with the environmental tolerance standard to identify the performance under extreme temperature, humidity and vibration conditions. By setting extreme conditions (such as an upper temperature limit of 90°C, an upper humidity limit of 80%, and an upper vibration limit of 10g), the electrical performance and mechanical load tolerance are tested. If the electrical score and mechanical load remain within the qualified range under extreme conditions, the wiring harness is considered to have passed the environmental tolerance assessment, and the wiring harness connection performance test evaluation data is generated, detailing the tolerance and performance of the wiring harness under various environmental conditions.

[0214] Preferably, the present application also provides a wiring harness connection performance intelligent testing system, which is used to execute the wiring harness connection performance intelligent testing method as described above, and the wiring harness connection performance intelligent testing system includes:

[0215] A wiring harness test data acquisition module, which is used to collect wiring harness test data generated by wiring harness test equipment and wiring harness sensors in real time through edge computing nodes;

[0216] A wiring harness performance detection and extraction module is used to perform electrical performance detection and extraction and mechanical performance detection and extraction according to the wiring harness test data, and obtain wiring harness electrical performance detection data and wiring harness mechanical performance detection data respectively;

[0217] A wiring harness performance evaluation module is used to perform an in-depth evaluation of electrical performance based on the wiring harness electrical performance test data to obtain wiring harness electrical performance evaluation data, and to perform an in-depth evaluation of mechanical performance based on the wiring harness mechanical performance test data to obtain wiring harness mechanical performance evaluation data;

[0218] The wiring harness connection performance test evaluation module is used to perform an environmental pressure cross-reference based on the wiring harness electrical performance evaluation data and the wiring harness mechanical performance evaluation data to obtain the wiring harness connection performance test evaluation data.

[0219] Therefore, from any point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present invention is limited by the attached application documents rather than the above description, and it is intended that all changes falling within the meaning and scope of equivalent elements of the application documents are included in the present invention.

[0220] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for intelligent testing of wiring harness connection performance, characterized in that: The following steps are involved: Step S1: collecting the wiring harness test data generated by the wiring harness test equipment and the wiring harness sensor in real time through the edge computing node; Step S2: performing electrical performance detection extraction and mechanical performance detection extraction according to the wire harness test data, and obtaining wire harness electrical performance detection data and wire harness mechanical performance detection data respectively; Step S3: performing an in-depth evaluation of electrical performance according to the wire harness electrical performance test data to obtain wire harness electrical performance evaluation data, and performing an in-depth evaluation of mechanical performance on the wire harness mechanical performance test data to obtain wire harness mechanical performance evaluation data; Step S4: cross-reference the environmental pressure according to the wire harness electrical performance evaluation data and the wire harness mechanical performance evaluation data to obtain the wire harness connection performance test evaluation data; Step S1 is specifically as follows: Step S11: collecting initial wiring harness test data generated by wiring harness test equipment and wiring harness sensors in real time through edge computing nodes; Step S12: performing interference data separation according to the initial wire harness test data to obtain wire harness test separation data; Step S13: filtering and detecting the wire harness test separation data to obtain wire harness test filtering detection data; Step S14: performing backtracking calibration according to the wire harness test filter detection data to obtain wire harness test data; Step S12 is specifically as follows: Perform edge collaborative anomaly cleaning based on the initial harness test data to obtain preliminary harness test cleaning data; Performing spectrum denoising on preliminary harness test cleanup data to obtain harness test cleanup feature data; Performing multi-level feature deconvolution on the wire harness test cleanup feature data to obtain wire harness test deconvolution feature data; Performing dimensionality reduction processing on the wire harness test unrolling feature data to obtain wire harness test feature dimensionality reduction data; Perform heterogeneous signal analysis on the wire harness test feature dimension reduction data to obtain wire harness test analysis feature data; Perform time series debiasing on the wire harness test profiling feature data to obtain wire harness test time series purified data; Perform aggregate signal reconstruction on the wiring harness test timing purification data to obtain wiring harness test reconstruction target data; Perform verification interference separation on the wiring harness test reconstruction target data to obtain wiring harness test separation data; The multi-level feature deconvolution is specifically as follows: According to the decoupling data of the wiring harness test frequency band, layered unconvolution is performed to obtain the wiring harness frequency band layered data, wherein the wiring harness frequency band layered data includes the wiring harness frequency band high-level data, the wiring harness frequency band middle-level data and the wiring harness frequency band low-level data; Gaussian convolution calculation is performed on the wiring harness frequency band high-level data to obtain the first wiring harness frequency band feature data; self-attention convolution calculation is performed on the wiring harness frequency band middle-level data to obtain the second wiring harness frequency band feature data; clustering processing and small convolution calculation are performed on the wiring harness frequency band low-level data to obtain the third wiring harness frequency band feature data; the spectrum of the first wiring harness frequency band feature data, the second wiring harness frequency band feature data and the third wiring harness frequency band feature data is reconstructed to obtain the wiring harness layer layered unconvolution data.

2. The method according to claim 1, characterized in that The specific spectrum denoising is as follows: Performing waveform decomposition on preliminary harness test cleanup data to obtain harness waveform decomposition data; Performing spectral component analysis on the wire beam waveform decomposition data to obtain wire beam spectral component data; Performing spectrum peak detection on the spectrum component data of the wire beam to obtain the high-frequency noise characteristic data of the spectrum of the wire beam; Performing filter parameter mapping on the high-frequency noise characteristic data of the wire harness spectrum to obtain filter parameter data; De-noising the line beam spectrum component data according to the filtering parameter data to obtain the line beam spectrum de-noising data; The key frequency features are extracted from the wire harness spectrum denoising data to obtain the wire harness test cleanup feature data.

3. The method according to claim 1, characterized in that The multi-level feature deconvolution is specifically as follows: Performing frequency domain conversion on the wire harness test cleanup feature data to obtain the wire harness test frequency domain conversion data; Performing frequency domain separation on the wire harness test frequency domain conversion data to obtain the wire harness test frequency domain separation data; Perform multi-scale frequency band convolution calculation based on the wire harness test frequency domain separation data to obtain the wire harness test frequency band decoupling data; Perform layered unwinding calculation based on the decoupling data of the wire harness test frequency band to obtain the layered unwinding data of the wire harness layer; The spectrum is reconstructed based on the layered unwinding data of the wire harness to obtain the unwinding characteristic data of the wire harness test.

4. The method according to claim 1, characterized in that The heterogeneous signal analysis is as follows: Constructing a graph structure for the wire harness test feature dimensionality reduction data to obtain wire harness test feature graph data; A heterogeneous graph is constructed according to the wire harness test characteristic graph data to obtain the wire harness test heterogeneous graph data; Extracting image signal features based on the wire harness test heterogeneous image data to obtain wire harness test image feature data; Perform feature noise stripping according to the feature data of the harness test graph to obtain optimized graph feature data; High-order features are reconstructed based on the optimized graph feature data to obtain the wiring harness test analysis feature data.

5. The method according to claim 1, characterized in that Step S2 is specifically as follows: Step S21: stratify the test data according to the wiring harness test data to obtain wiring harness test stratification data; Step S22: performing classification screening according to the wiring harness test hierarchical data to obtain wiring harness test classification data, wherein the wiring harness test classification data includes wiring harness electrical performance test data and wiring harness mechanical performance test data; Step S23: performing electrical performance detection conversion according to the wiring harness electrical performance test data to obtain wiring harness electrical performance detection data; Step S24: Perform mechanical property detection conversion according to the wire harness mechanical property test data to obtain the wire harness mechanical property detection data.

6. The method according to claim 1, characterized in that Step S3 is specifically as follows: Step S31: using a preset wiring harness electrical evaluation model to perform an in-depth electrical performance evaluation on the wiring harness electrical performance detection data to obtain wiring harness electrical performance evaluation data; Step S32: performing stress-strain nonlinear modeling according to the wire harness mechanical property detection data to obtain a wire harness mechanical stress model; Step S33: Estimating the mechanical load tolerance according to the wire harness mechanical stress model to obtain wire harness mechanical performance evaluation data; The step of constructing the wiring harness electrical evaluation model preset in step S31 includes the following steps: Obtain historical wiring harness electrical performance test data and historical electrical performance evaluation label data through a preset wiring harness history database; Perform data division on historical wiring harness electrical performance test data to obtain historical wiring harness test data and historical wiring harness verification data; Performing primary hidden layer processing on historical harness test data to obtain primary feature data of historical harness; Perform multi-head self-attention layer processing on the primary feature data of the historical line bundle to obtain the weighted feature data of the historical line bundle; Performing secondary hidden layer processing on the historical line bundle weighted feature data to obtain the historical line bundle secondary feature data; The historical wiring harness verification data is used to iteratively train the historical wiring harness secondary feature data and the model is constructed through the historical electrical performance evaluation label data to obtain the wiring harness electrical evaluation model.

7. The method according to claim 1, characterized in that Step S4 is specifically as follows: Step S41: obtaining wiring harness environment data; Step S42: performing environmental data matching according to the wire harness environmental data, the wire harness electrical performance evaluation data, and the wire harness mechanical performance evaluation data to obtain test environment matching data; Step S43: performing cross analysis according to the test environment matching data to obtain parameter cross analysis data; Step S44: verifying the environmental adaptability according to the parameter cross-analysis data to obtain the harness environmental adaptability data; Step S45: Perform environmental tolerance evaluation based on the wire harness environmental adaptability data to obtain wire harness connection performance test evaluation data.

8. A wiring harness connection performance intelligent testing system, characterized in that: Used to perform the wiring harness connection performance intelligent testing method according to claim 1, the wiring harness connection performance intelligent testing system comprises: A wiring harness test data acquisition module, which is used to collect wiring harness test data generated by wiring harness test equipment and wiring harness sensors in real time through edge computing nodes; A wiring harness performance detection and extraction module is used to perform electrical performance detection and extraction and mechanical performance detection and extraction according to the wiring harness test data, and obtain wiring harness electrical performance detection data and wiring harness mechanical performance detection data respectively; A wiring harness performance evaluation module is used to perform an in-depth evaluation of electrical performance based on the wiring harness electrical performance test data to obtain wiring harness electrical performance evaluation data, and to perform an in-depth evaluation of mechanical performance based on the wiring harness mechanical performance test data to obtain wiring harness mechanical performance evaluation data; The wiring harness connection performance test evaluation module is used to perform an environmental pressure cross-reference based on the wiring harness electrical performance evaluation data and the wiring harness mechanical performance evaluation data to obtain the wiring harness connection performance test evaluation data.

Images