Wiring harness dynamic load and electromagnetic interference coupling test method

Abstract

The invention discloses a wiring harness dynamic load and electromagnetic interference coupling test method, and relates to the technical field of wiring harness testing, and the method comprises the steps: S1, defining a plurality of environment parameter groups, and constructing a test platform; s2, dynamic current pulse and multi-axis vibration excitation are synchronously loaded; s3, applying broadband electromagnetic interference signals and monitoring signal crosstalk amplitude and bit error rate changes; s4, the sensor nodes are dynamically configured according to the topological structure of the wire harness; s5, generating a feature vector; s6, outputting a wire harness performance index; s7, outputting a wire harness fault probability prediction value; s8, obtaining prediction data; s9, triggering a high-speed data recording mode, and storing original sensor data 60 seconds before the fault; and dynamically adjusting the gradient step length and the loading sequence in the environment parameter group. Performance evaluation, fault prediction and service life management can be carried out on the wire harness under complex working conditions, and the reliability and safety of an automobile electrical system are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wire harness testing, and in particular to a method for testing coupling of dynamic load and electromagnetic interference of a wire harness. Background Art

[0002] As a key component of the automotive electrical system, the wiring harness performs the crucial functions of power and signal transmission, and its performance is directly related to the safety and reliability of the entire vehicle. With the rapid development of the new energy vehicle industry, high-voltage wiring harnesses face a more complex operating environment. They must not only withstand physical stresses such as dynamic current pulses and mechanical vibration, but also resist electromagnetic interference from onboard electronic devices and the external environment. Therefore, accurately testing the performance of wiring harnesses under multi-physics coupling conditions has become a critical issue that the industry urgently needs to address.

[0003] Currently, the wiring harness testing technology has the following main limitations: First, the environmental simulation is too simplistic. Traditional testing cannot simulate the complex multi-physics coupling conditions found in actual vehicle operation. For example, current load testing alone fails to account for the impact of vibration and electromagnetic interference on current transmission stability. Vibration testing, on the other hand, lacks the synergy between the electrical and electromagnetic environments, leading to significant deviations between test results and actual operating conditions. This makes it difficult to truly reflect the performance degradation patterns of wiring harnesses under combined stresses.

[0004] Second, data collection and analysis methods are outdated. Existing tests struggle to comprehensively and in real time collect multi-dimensional data such as temperature, strain, and current waveforms at key wiring harness nodes. Furthermore, traditional data processing methods primarily rely on static feature analysis, which is unable to effectively extract timing correlation features under dynamic loads and electromagnetic interference. This results in inaccurate and ineffective wiring harness performance assessments, making it difficult to accurately predict early-stage failures.

[0005] Third, fault prediction and lifespan assessment models are imperfect. Current wiring harness fault prediction lacks in-depth analysis of multi-parameter coupling relationships under complex operating conditions. Furthermore, finite element simulation models lack real-world data, resulting in inefficient parameter correction and difficulty accurately predicting wiring harness degradation trends over long-term operation. This makes it difficult to meet the design and verification requirements for high-reliability, long-life wiring harnesses for new energy vehicles.

[0006] In summary, existing wiring harness testing technology has obvious shortcomings in simulating complex working conditions, multi-dimensional data collection and analysis, and fault prediction. There is an urgent need for a testing method that can realize dynamic load and electromagnetic interference coupling testing, accurate feature extraction, and reliable life assessment to ensure the safe and stable operation of automotive electrical systems. Summary of the Invention

[0007] In view of this, the technical problem to be solved by the present invention is to provide a method for testing the coupling of dynamic load and electromagnetic interference of a wire harness, which can evaluate the performance, predict faults and manage the life of the wire harness under complex working conditions, and improve the reliability and safety of the automotive electrical system.

[0008] To solve the above technical problems, the technical solution of the present invention is as follows: A method for testing the coupling of dynamic load and electromagnetic interference of a wire harness, comprising the following steps: S1. Define a number of environmental parameter groups, which include dynamic current pulse gradient, multi-axis vibration frequency gradient, electromagnetic interference frequency band gradient, and corrosion acceleration factor concentration gradient; construct a test platform, which includes a dynamic current generator, a multi-axis vibration table, an electromagnetic interference source, a data acquisition system, and a three-axis orthogonal electromagnetic shielding cabin. The shielding effectiveness of the cabin is ≥80dB, and anechoic materials are arranged inside the cabin with a coverage rate of more than 90%; S2. Apply a dynamic current pulse load to the wire harness based on the dynamic current pulse gradient, apply a multi-axis vibration excitation to the wire harness based on the multi-axis vibration frequency gradient, and synchronously load the dynamic current pulse and the multi-axis vibration excitation through a preset phase difference. The phase difference range is 0° to 180°, simulating the composite environment of transient high current and mechanical stress during the actual operation of the vehicle; S3. Apply a broadband electromagnetic interference signal to the wire harness on the test platform. The wire harness includes a high-voltage wire harness and a low-voltage signal wire harness laid adjacent to it; monitor the signal crosstalk amplitude and the change of bit error rate between the high-voltage wire harness and the low-voltage signal wire harness in real time through a bit error rate monitoring device; S4. Real-time collect the temperature, strain, and high-frequency current waveform data of several specified sensor nodes in the wire harness through a distributed fiber optic grating sensor network. The sensor nodes are dynamically configured according to the wire harness topology; S5. Extract the features of the dynamic data collected at several sensor nodes to generate feature vectors, which include the temperature gradient change rate, the dynamic current harmonic distortion rate, and the vibration spectrum energy distribution; S6. Input the feature vectors into the wire harness parameter performance analysis model. The wire harness parameter performance analysis model is deployed on an edge computing device, and the time-series and spatial features are extracted through a double-layer bidirectional gated recurrent unit to obtain the target high-dimensional feature vectors, and the parameter performance regression analysis is carried out through a classification regression layer to output the wire harness performance indicators; S7. Input the feature vectors into the wire harness fault prediction network. The wire harness fault prediction network is deployed on an edge computing device, extract the hidden features through a bidirectional long short-term memory layer, and output the wire harness fault probability prediction value after fusion through a fully connected layer; S8. Modify the electro-thermal-mechanical coupling parameters of the wire harness based on the digital twin simulation model, combine the measured data of S3 and the measured data of S4 to predict the remaining service life and fault threshold of the wire harness, and obtain the prediction data; S9. When the predicted value of the wiring harness fault probability exceeds the threshold or the fault threshold predicted and output by S8 is breached, trigger the high-speed data recording mode, and store the original sensor data 60 seconds before the fault; feedback the prediction result of S8 to S1, and dynamically adjust the gradient step size and loading sequence in the environmental parameter group.

[0009] Preferably, in S2, the loading range of the dynamic current pulse load is 200A to 500A, the pulse duration is 1ms to 10ms, the frequency is 1Hz to 100Hz, and the time deviation between the rising edge of the dynamic current pulse and the vibration peak moment is ≤1ms; The multi-axis vibration excitation includes the vibration frequency gradients in the X-axis, Y-axis, and Z-axis directions. The vibration frequency ranges of the X-axis, Y-axis, and Z-axis are all 10Hz to 500Hz, and the acceleration amplitude is 0.5g to 20g.

[0010] Preferably, in S3, the injection frequency band of the broadband electromagnetic interference signal is 1MHz to 6GHz, the interference intensity is 10V / m to 100V / m, and the superposition methods include radiation interference and conduction interference; The radiation interference includes forming an electromagnetic field around the wiring harness through an electromagnetic radiator, and the conduction interference includes directly connecting to the wiring harness through a coupler for conduction. The power ratio of the radiation interference to the conduction interference is 3:1 to 1:3.

[0011] Preferably, in S4, the arrangement density of the distributed fiber Bragg grating sensor network is to set one sensor node every 10cm to 20cm, the spatial resolution is ±2mm, the temperature measurement accuracy is ±0.5°C, the strain measurement accuracy is ±1με, the sampling frequency is 10Hz to 10kHz, the fiber Bragg grating sensor is shielded by a metal braided layer, the shielding effectiveness is ≥60dB, and the data transmission adopts the wavelength division multiplexing method; The sensor node and the bit error rate monitoring device in S3 are synchronized through the PTP precise time protocol, and the clock deviation is ≤50μs.

[0012] Preferably, in S5, the calculation formula for the temperature gradient change rate is: dT / dt=(T(t + Δt)-T(t)) / Δt, where T(t) is the temperature at the current moment, T(t + Δt) is the temperature at the next moment, and Δt is the sampling interval. The sampling interval is 0.1 second to 10 seconds. The temperature gradient change rate is used to evaluate the thermal stability and fault risk of the wiring harness, and the threshold range is 0.1°C / s to 5°C / s; When the temperature gradient change rate dT / dt > 5°C / s, cut off the dynamic current load, and at the same time stop the multi-axis vibration excitation and turn off the electromagnetic interference source.

[0013] Preferably, in S5, feature extraction includes performing fast Fourier transform analysis on the dynamic current waveform to extract the total harmonic distortion rate and the proportion of high-frequency component energy, performing time series analysis on the temperature data to calculate the temperature gradient change rate and the standard deviation of temperature fluctuation, and performing wavelet transform analysis on the vibration data to extract the vibration spectrum energy distribution.

[0014] Preferably, in S6, the double-layer bidirectional gated recurrent unit includes a first-layer bidirectional recurrent neural network and a second-layer gated recurrent unit. The hidden layer dimension of the first-layer bidirectional recurrent neural network is 64, and the hidden layer dimension of the second-layer gated recurrent unit is 32. The time series features are weighted and fused through an attention mechanism, and after dimensionality reduction, a target high-dimensional feature vector with a dimension of 128 is generated.

[0015] Preferably, in S7, the bidirectional long short-term memory layer includes an attention mechanism module. The hidden layer dimensions of the forward LSTM and the backward LSTM are both 64. The attention mechanism module calculates the dot product of the query vector and the time series features to generate attention weights, and the weighted features are input into the fully connected layer. The fully connected layer includes 3 hidden layers, and the number of neurons is 128, 64, and 32 respectively.

[0016] Preferably, in S8, the digital twin simulation model is constructed based on finite element analysis, including an electro-thermal coupling sub-model, a thermal-mechanical coupling sub-model, and an aging prediction sub-model. The model parameters are corrected through the Bayesian optimization algorithm to predict the remaining service life of the wire harness at the time nodes of 1000 hours, 5000 hours, and 10000 hours.

[0017] Preferably, in S8, the determination basis of the fault threshold is: when the wire harness resistance change rate exceeds 50% or the signal error rate exceeds 10 -4 %, a fault warning signal is triggered; the calculation formula for the resistance change rate is: (ΔR) / R0 = ((Rt - R0)) / R0 × 100%, where R0 is the initial reference resistance, Rt is the real-time working resistance, and ΔR is the absolute change in resistance; The digital twin simulation model corrects the parameters through the Bayesian optimization algorithm, and takes the root mean square error of the measured and simulated data < 5% as the convergence condition, and rechecks the model accuracy every 24 hours. When the continuous 3 - time check error > 5%, the aging prediction sub-model is reconstructed; The resistance change rate is calculated by comparing the current resistance value with the initial resistance value, and the signal error rate is determined by measuring the ratio of the number of error bits to the total number of transmitted bits per unit time by an error detector; The aging prediction sub-model introduces the hydrogen sulfide concentration as a corrosion acceleration factor, and the concentration gradient is 10 ppm to 1000 ppm, and the concentration gradient is dynamically adjusted according to the type of wire harness sheath material: 10 ppm to 500 ppm for rubber materials, and 200 ppm to 1000 ppm for silicone materials.

[0018] After adopting the above technical solutions, the beneficial effects of the present invention are as follows: First, the present invention can significantly improve the authenticity of composite environment simulation. Among them, through the 0°-180° phase difference control, precise synchronous loading of dynamic current pulses and multi-axis vibration is achieved (S2), fundamentally solving the inherent defect of the separation of mechanical and electrical stress time sequences in traditional tests. This design significantly improves the reproduction accuracy of extreme working conditions such as rapid acceleration and braking by simulating the composite environment of transient high current and mechanical stress during actual vehicle operation. The phase difference adjustable mechanism can precisely match the electromechanical coupling characteristics of different vehicle models, and the strict alignment of the rising edge of the dynamic current pulse and the vibration peak effectively suppresses the contact resistance fluctuation and terminal fretting wear caused by mechanical vibration, providing a real physical environment basis for the study of the performance degradation of wire harnesses under composite stresses.

[0019] Second, the present invention can break through the synergy of multi-physical field monitoring. Among them, the collaborative architecture of a three-axis orthogonal electromagnetic shielding cabin (shielding effectiveness ≥ 80 dB) and a distributed fiber Bragg grating sensor network (S1+S4) constructs a multi-physical field synchronous monitoring system in an electromagnetic pure environment. The high coverage rate of the absorbing material in the shielding cabin combined with the metal braided layer shielding design of the fiber optic sensor can still ensure the acquisition accuracy of temperature, strain, and current waveform data under strong electromagnetic interference conditions. Through the PTP precise time protocol, clock synchronization between sensor nodes and the bit error rate monitoring device is achieved, solving the pain point of data spatio-temporal misalignment in traditional step-by-step tests, and providing a highly consistent data basis for the crosstalk analysis of high-voltage wire harnesses and low-voltage signal wire harnesses and the study of electro-thermal-mechanical coupling effects.

[0020] Third, the reliability of fault prediction is essentially enhanced. By adopting the dual-channel parallel processing architecture of the GRU performance analysis model and the LSTM fault prediction model (S6+S7), the deep analysis ability of the time series model for dynamic feature vectors is fully utilized. The GRU network extracts the spatio-temporal evolution law of the wire harness performance index through a double-layer bidirectional structure, while the LSTM network focuses on the identification of the long-term dependence of fault features. The complementary fusion of the two models significantly improves the detection ability of hidden faults such as microcracks in the insulation layer and fatigue fractures of conductors. Combining the dual decision-making mechanism of the fault probability prediction value and the physical threshold (such as the resistance change rate) (S9), a multi-level protection system from early warning to fuse protection is constructed, greatly reducing the misjudgment risk caused by the failure of a single sensor.

[0021] IV. Comprehensively optimize the test efficiency and cost structure. The dynamic adjustment mechanism of the corrosion acceleration factor concentration gradient (S1) and the feedback closed-loop of the digital twin prediction results (S9) form an adaptive optimization system for environmental parameters. This design automatically matches the hydrogen sulfide concentration gradient according to the type of wire harness sheath material (rubber / silicone), ensuring the authenticity of the material degradation mechanism while accelerating aging; the electro-thermal-mechanical coupling parameters are corrected in real time through the digital twin simulation model, and the prediction results are fed back to the test parameter definition link, realizing the iteration of the test load sequence. This closed-loop architecture significantly compresses the equivalent aging test cycle and reduces the energy consumption cost of repeated tests, providing an efficient verification platform for wire harness design. Description of the Drawings

[0022] The present invention will be further described below in conjunction with the drawings and embodiments.

[0023] Figure 1 It is a schematic structural diagram of the wire harness dynamic load and electromagnetic interference coupling test method according to an embodiment of the present invention. Detailed Embodiments

[0024] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0025] The wire harness dynamic load and electromagnetic interference coupling test method of the present invention includes the following steps: S1. Define several groups of environmental parameters, including dynamic current pulse gradient, multi-axis vibration frequency gradient, electromagnetic interference frequency band gradient, and corrosion acceleration factor concentration gradient; construct a test platform, which includes a dynamic current generator, a multi-axis vibration table, an electromagnetic interference source, a data acquisition system, and a three-axis orthogonal electromagnetic shielding chamber. The shielding efficiency of the chamber is ≥80 dB, and anechoic materials are arranged inside the chamber with a coverage rate of more than 90%; S2. Apply a dynamic current pulse load to the wire harness based on the dynamic current pulse gradient, apply a multi-axis vibration excitation to the wire harness based on the multi-axis vibration frequency gradient, and synchronously load the dynamic current pulse and the multi-axis vibration excitation through a preset phase difference. The phase difference range is 0° to 180°, simulating the transient high-current and mechanical stress composite environment during actual vehicle operation; S_{ S4. Real-time collect the temperature, strain and high-frequency current waveform data of several specified sensor nodes in the wire harness through a distributed fiber Bragg grating sensor network, and the sensor nodes are dynamically configured according to the wire harness topology structure; S5. Extract the features of the dynamic data collected at several sensor nodes to generate feature vectors, where the feature vectors include the temperature gradient change rate, the dynamic current harmonic distortion rate, and the vibration spectrum energy distribution; S6. Input the feature vectors into the wire harness parameter performance analysis model. The wire harness parameter performance analysis model is deployed on an edge computing device, performs time-series and spatial feature extraction through a double-layer bidirectional gated recurrent unit to obtain target high-dimensional feature vectors, and performs parameter performance regression analysis through a classification regression layer to output wire harness performance indicators; S7. Input the feature vectors into the wire harness fault prediction network. The wire harness fault prediction network is deployed on an edge computing device, extracts hidden features through a bidirectional long short-term memory layer, and outputs the wire harness fault probability prediction value after fusion through a fully connected layer; S8. Modify the electro-thermal-mechanical coupling parameters of the wire harness based on the digital twin simulation model, and combine the measured data in S3 and the measured data in S4 to predict the remaining service life and fault threshold of the wire harness to obtain prediction data; S9. When the wire harness fault probability prediction value exceeds the threshold or the fault threshold predicted and output in S8 is breached, trigger the high-speed data recording mode to store the original sensor data 60 seconds before the fault; feedback the prediction result in S8 to S1 to dynamically adjust the gradient step size and loading sequence in the environmental parameter group.

[0026] Specifically, in S1, the dynamic current pulse gradient is used to simulate the transient load changes during vehicle operation (such as motor start / stop, energy recovery). The gradient setting is classified according to the wire harness type: the low-voltage wire harness adopts a step change from 50A to 200A (step size 10A), and the high-voltage wire harness adopts a step change from 200A to 500A (step size 50A). Progressive degradation is excited through incremental loading to capture the non-linear characteristics of the terminal contact resistance.

[0027] The multi-axis vibration frequency gradient refers to the three-axis independent frequency sequence (10Hz to 500Hz for the X / Y / Z axes) that reproduces the road spectrum vibration environment. The core feature is that it is synchronously loaded with the current pulse at a phase difference of 0° to 180°, accurately simulating the electromechanical coupling effect of the engine vibration peak and the motor current rising edge during hard acceleration. The acceleration amplitude gradient is set to 0.5g to 20g to match different suspension stiffnesses.

[0028] The electromagnetic interference frequency band gradient refers to the frequency segmentation sequence covering typical interference sources of in-vehicle electronics: the 1 MHz to 100 MHz frequency band simulates the switching noise of power devices (such as inverter dV / dt), and the 1 GHz to 6 GHz frequency band simulates high-frequency communication interference (such as radar / 5G). The power ratio of radiation interference (antenna emission) to conduction interference (coupler injection) is dynamically adjustable.

[0029] The corrosion acceleration factor concentration gradient refers to the hydrogen sulfide concentration sequence for different sheath materials: rubber-like materials use a low concentration gradient of 10 ppm to 500 ppm, and silicone-like materials use a high concentration gradient of 200 ppm to 1000 ppm. The electrochemical corrosion process is accelerated by increasing the concentration, and the aging test cycle is shortened.

[0030] In S2, the phase difference refers to the timing offset between the dynamic current pulse and the multi-axis vibration excitation, which is quantified in angular measure (in the range of 0° to 180°). Among them: 0° phase difference: the rising edge of the current pulse coincides exactly with the vibration peak moment, simulating the rapid acceleration condition (the peak current of the motor is synchronized with the engine vibration); 90° phase difference: the half-peak point of the current pulse corresponds to the vibration peak, simulating the shift shock condition; 180° phase difference: the end moment of the current pulse aligns with the vibration peak, simulating the braking energy recovery condition (the current decay is coupled with the suspension vibration). It can reproduce the electromechanical stress coupling effect in actual vehicle operation (such as the terminal fretting wear caused by vibration is most significant at the current peak), and solves the physical distortion problem caused by asynchronous loading of current and vibration in traditional tests.

[0031] The loading range of the dynamic current pulse load is 200 A to 500 A, the pulse duration is 1 ms to 10 ms, the frequency is 1 Hz to 100 Hz, and the time deviation between the rising edge of the dynamic current pulse and the vibration peak moment ≤ 1 ms; the multi-axis vibration excitation includes the vibration frequency gradient in the X-axis, Y-axis, and Z-axis directions. The vibration frequency ranges of the X-axis, Y-axis, and Z-axis are all 10 Hz to 500 Hz, and the acceleration amplitude is 0.5 g to 20 g.

[0032] Among them, the dynamic current pulse refers to a single transient load event (such as a current waveform with an amplitude of 300 A and a duration of 5 ms), which is used to simulate specific vehicle conditions (such as motor start and stop); the dynamic current pulse gradient is a parameterized pulse sequence.

[0033] In S3, the injection frequency band of the broadband electromagnetic interference signal is 1 MHz to 6 GHz, the interference intensity is 10 V / m to 100 V / m, and the superposition methods include radiation interference and conduction interference; the radiation interference includes forming an electromagnetic field around the wire harness through an electromagnetic radiator, and the conduction interference includes directly connecting to the wire harness through a coupler for conduction. The power ratio of radiation interference to conduction interference is 3:1 to 1:3.

[0034] The electromagnetic radiator utilizes a broadband log-periodic antenna, driven by a power amplifier and positioned parallel to the axis, one meter above the wiring harness. By adjusting the transmit power (10V / m to 100V / m), a controllable electromagnetic field is generated within the shielded cabin. Field strength calibration is performed at 16 points on a 1.5m x 1.5m vertical plane in accordance with the ISO11452-2 standard, ensuring field intensity fluctuations within the target wiring harness area are ≤±3dB. An X / Y / Z tri-directional antenna array simulates a multi-angle radiation environment. Combined with the bulkhead's absorbing material (coverage >90%), it suppresses reflected standing waves, inducing common-mode currents in the wiring harness conductors, equivalent to the radiated interference effect of a real vehicle inverter.

[0035] In S4, the distributed fiber Bragg grating sensor network is arranged with a sensor node every 10cm to 20cm, with a spatial resolution of ±2mm, a temperature measurement accuracy of ±0.5℃, a strain measurement accuracy of ±1με, and a sampling frequency of 10Hz to 10kHz. The fiber Bragg grating sensors are shielded by a metal braid with a shielding effectiveness of ≥60dB, and data transmission uses wavelength division multiplexing. The sensor nodes and the bit error rate monitoring equipment in S3 are synchronized using the PTP precise time protocol, with a clock deviation of ≤50μs.

[0036] Wiring harness topology refers to the three-dimensional spatial layout and electrical connections of a wiring harness system, including the trunk cable routing, branch node distribution, connector locations, and parallel sections of high and low voltage wiring harnesses. Topological characteristics include: 1. Physical path topology, which refers to the routing and curvature radius of the wiring harness through the sheet metal holes / clip attachment points; 2. Electrical connection topology, which refers to the terminal plug-in relationship and grounding point distribution hierarchy; and 3. Electromagnetic coupling topology, which refers to the length and spacing of adjacent high and low voltage wiring harnesses.

[0037] Distributed fiber optic sensors dynamically configure nodes based on this structure: they are densely deployed (spacing 10 to 20 cm) at key locations such as branch intersections, high curvature areas, and connector near fields, and sparsely deployed in straight sections (spacing > 30 cm), to achieve precise monitoring of stress concentration areas and electromagnetic sensitive areas.

[0038] In S5, the temperature gradient change rate is calculated as follows: dT / dt=(T(t+Δt)-T(t)) / Δt, where T(t) is the temperature at the current moment, T(t+Δt) is the temperature at the next moment, Δt is the sampling interval, and the sampling interval is 0.1 seconds to 10 seconds. The temperature gradient change rate is used to evaluate the thermal stability and failure risk of the wiring harness, and the threshold range is 0.1℃ / s to 5℃ / s. When the temperature gradient change rate dT / dt is greater than 5℃ / s, the dynamic current load is cut off, the multi-axis vibration excitation is stopped, and the electromagnetic interference source is turned off.

[0039] In S5, feature extraction includes performing fast Fourier transform analysis on the dynamic current waveform to extract the total harmonic distortion rate and the proportion of high-frequency component energy, performing time series analysis on the temperature data to calculate the temperature gradient change rate and the standard deviation of temperature fluctuations, and performing wavelet transform analysis on the vibration data to extract the vibration spectrum energy distribution.

[0040] In S6, the double-layer bidirectional gated recurrent unit includes a first-layer bidirectional recurrent neural network and a second-layer gated recurrent unit. The hidden layer dimension of the first-layer bidirectional recurrent neural network is 64, and the hidden layer dimension of the second-layer gated recurrent unit is 32. The time series features are weighted and fused through the attention mechanism, and a target high-dimensional feature vector with a dimension of 128 is generated after dimensionality reduction.

[0041] The wire harness parameter performance analysis model is an evaluation system based on a deep learning architecture, deployed on edge computing devices to achieve real-time processing. Its core uses a double-layer bidirectional gated recurrent unit (GRU): the first-layer bidirectional network extracts the time series evolution features of sensor data (such as the temperature fluctuation trend), and the second-layer gated unit captures the spatial correlation features (such as the correlation of multi-node strain distribution). The spatio-temporal information is fused through the attention mechanism to generate a 128-dimensional high-dimensional feature vector. Finally, it is mapped to quantization performance indicators by the classification regression layer, including key parameters such as the dynamic resistance change rate, the insulation impedance attenuation rate, and the thermal imbalance coefficient, comprehensively reflecting the electrical, mechanical, and thermal management states of the wire harness.

[0042] In S7, the wire harness fault prediction network is a diagnostic system based on a deep time series model, deployed on edge computing devices to achieve millisecond-level response. Its core uses a bidirectional long short-term memory layer (BiLSTM): the forward and backward LSTMs respectively extract the historical and future dependence relationships of the feature vectors, and the key time series features (such as sudden temperature changes and sudden increases in current harmonics) are weighted and fused through the attention mechanism to generate a hidden state vector. The fully connected layer further fuses the multi-dimensional hidden features, and finally outputs a fault probability prediction value in the range of 0 to 1, covering typical failure modes such as terminal oxidation, insulation layer cracking, and conductor fatigue, realizing early risk warning.

[0043] The bidirectional long short-term memory layer contains an attention mechanism module. The hidden layer dimensions of the forward LSTM and the backward LSTM are both 64. The attention mechanism module calculates the dot product of the query vector and the time series features to generate attention weights, and the weighted features are input into the fully connected layer. The fully connected layer contains 3 hidden layers, and the number of neurons is 128, 64, and 32 respectively.

[0044] In S8, the digital twin simulation model is constructed based on finite element analysis, including an electro-thermal coupling sub-model, a thermal-mechanical coupling sub-model, and an aging prediction sub-model. The model parameters are corrected through the Bayesian optimization algorithm to predict the remaining service life of the wire harness at time nodes of 1000 hours, 5000 hours, and 10000 hours.

[0045] In S8, the basis for determining the fault threshold is as follows: when the wire harness resistance change rate exceeds 50% or the signal error rate exceeds 10 -4 %, a fault warning signal is triggered; the calculation formula for the resistance change rate is: (ΔR) / R0 = ((Rt - R0)) / R0 × 100%, where R0 is the initial reference resistance, Rt is the real-time working resistance, and ΔR is the absolute change in resistance; The digital twin simulation model corrects parameters through the Bayesian optimization algorithm, and converges when the root mean square error of the measured and simulated data < 5%. The model accuracy is re-verified every 24 hours. When the verification error > 5% for three consecutive times, the aging prediction sub-model is reconstructed; The resistance change rate is calculated by comparing the current resistance value with the initial resistance value, and the signal error rate is determined by measuring the ratio of the number of error bits to the total number of transmitted bits per unit time with an error code meter; The aging prediction sub-model introduces the hydrogen sulfide concentration as a corrosion acceleration factor, with a concentration gradient of 10 ppm to 1000 ppm, and dynamically adjusts the concentration gradient according to the wire harness sheath material type: 10 ppm to 500 ppm for rubber materials, and 200 ppm to 1000 ppm for silicone materials.

[0046] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for testing the coupling of dynamic load and electromagnetic interference of a wire harness, characterized in that, It includes the following steps: S1. Define several groups of environmental parameters, which include dynamic current pulse gradient, multi-axis vibration frequency gradient, electromagnetic interference frequency band gradient, and corrosion acceleration factor concentration gradient; construct a test platform, which includes a dynamic current generator, a multi-axis vibration table, an electromagnetic interference source, a data acquisition system, and a three-axis orthogonal electromagnetic shielding cabin. The shielding effectiveness of the cabin is ≥80 dB, and anechoic materials are set inside the cabin with a coverage rate of more than 90%; S2. Apply a dynamic current pulse load to the wire harness based on the dynamic current pulse gradient, apply a multi-axis vibration excitation to the wire harness based on the multi-axis vibration frequency gradient, and synchronously load the dynamic current pulse and the multi-axis vibration excitation through a preset phase difference. The phase difference range is 0° to 180°, simulating the transient high-current and mechanical stress composite environment during the actual operation of the vehicle; S3. Apply a broadband electromagnetic interference signal to the wire harness on the test platform. The wire harness includes a high-voltage wire harness and a low-voltage signal wire harness laid adjacent to it; monitor the signal crosstalk amplitude and the change of bit error rate between the high-voltage wire harness and the low-voltage signal wire harness in real time through a bit error rate monitoring device; S4. Real-time collect the temperature, strain, and high-frequency current waveform data of several specified sensor nodes in the wire harness through a distributed fiber optic grating sensor network. The sensor nodes are dynamically configured according to the wire harness topology; S5. Extract features from the dynamic data collected at several sensor nodes to generate feature vectors, which include the temperature gradient change rate, the dynamic current harmonic distortion rate, and the vibration spectrum energy distribution; S6. Input the feature vectors into the wire harness parameter performance analysis model. The wire harness parameter performance analysis model is deployed on an edge computing device, and temporal-spatial feature extraction is performed through a double-layer bidirectional gated recurrent unit to obtain target high-dimensional feature vectors, and parameter performance regression analysis is performed through a classification regression layer to output the wire harness performance indicators; S7. Input the feature vectors into the wire harness fault prediction network. The wire harness fault prediction network is deployed on an edge computing device, extract hidden features through a bidirectional long short-term memory layer, and output the wire harness fault probability prediction value after fusion through a fully connected layer; S8. Modify the electro-thermal-mechanical coupling parameters of the wire harness based on the digital twin simulation model, and predict the remaining service life and fault threshold of the wire harness in combination with the measured data of S3 and the measured data of S4 to obtain prediction data; S9. When the wire harness fault probability prediction value exceeds the threshold or the fault threshold predicted and output in S8 is breached, trigger the high-speed data recording mode and store the original sensor data 60 seconds before the fault; feedback the prediction result of S8 to S1 to dynamically adjust the gradient step size and loading sequence in the environmental parameter group.

2. The method for testing the coupling of dynamic load and electromagnetic interference of the wire harness according to claim 1, wherein In S2, the loading range of the dynamic current pulse load is 200 A to 500 A, the pulse duration is 1 ms to 10 ms, the frequency is 1 Hz to 100 Hz, and the time deviation between the rising edge of the dynamic current pulse and the vibration peak moment is ≤1 ms; The multi-axis vibration excitation includes the vibration frequency gradient in the X-axis, Y-axis, and Z-axis directions. The vibration frequency ranges of the X-axis, Y-axis, and Z-axis are all 10 Hz to 500 Hz, and the acceleration amplitude is 0.5 g to 20 g.

3. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 1, characterized in that, In S3, the injection frequency band of the broadband electromagnetic interference signal is 1 MHz to 6 GHz, the interference intensity is 10 V / m to 100 V / m, and the superposition methods include radiation interference and conduction interference; The radiation interference includes forming an electromagnetic field around the wire harness through an electromagnetic radiator, and the conduction interference includes conducting directly through a coupler connected to the wire harness. The power ratio of radiation interference to conduction interference is 3:1 to 1:

3.

4. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 1, characterized in that, In S4, the arrangement density of the distributed fiber Bragg grating sensor network is to set one sensor node every 10 cm to 20 cm, the spatial resolution is ±2 mm, the temperature measurement accuracy is ±0.5 °C, the strain measurement accuracy is ±1 με, the sampling frequency is 10 Hz to 10 kHz, the fiber Bragg grating sensor is shielded by a metal braided layer, the shielding effectiveness ≥ 60 dB, and the data transmission adopts the wavelength division multiplexing method; The sensor node and the bit error rate monitoring device in S3 are synchronized through the PTP precise time protocol, and the clock deviation ≤ 50 μs.

5. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 1, wherein In S5, the calculation formula for the temperature gradient change rate is: dT / dt = (T(t + Δt) - T(t)) / Δt, where T(t) is the temperature at the current moment, T(t + Δt) is the temperature at the next moment, Δt is the sampling interval, the sampling interval is 0.1 second to 10 seconds, and the temperature gradient change rate is used to evaluate the thermal stability and fault risk of the wire harness, and the threshold range is 0.1 °C / s to 5 °C / s; When the temperature gradient change rate dT / dt > 5 °C / s, the dynamic current load is cut off, and at the same time, the multi-axis vibration excitation is stopped and the electromagnetic interference source is turned off.

6. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 5, wherein In S5, feature extraction includes performing fast Fourier transform analysis on the dynamic current waveform to extract the total harmonic distortion rate and the proportion of high-frequency component energy, performing time series analysis on the temperature data to calculate the temperature gradient change rate and the standard deviation of temperature fluctuation, and performing wavelet transform analysis on the vibration data to extract the vibration spectrum energy distribution.

7. The method for testing the coupling of the dynamic load and electromagnetic interference of the wire harness according to claim 1, wherein, In S6, the double-layer bidirectional gated recurrent unit includes the first-layer bidirectional recurrent neural network and the second-layer gated recurrent unit. The hidden layer dimension of the first-layer bidirectional recurrent neural network is 64, and the hidden layer dimension of the second-layer gated recurrent unit is 32. The time series features are weighted and fused through the attention mechanism, and a target high-dimensional feature vector with a dimension of 128 is generated after dimensionality reduction.

8. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 1, wherein In S7, the bidirectional long short-term memory layer contains an attention mechanism module. The hidden layer dimensions of the forward LSTM and the backward LSTM are both 64. The attention mechanism module calculates the dot product of the query vector and the time series features to generate attention weights, and the weighted features are input into the fully connected layer. The fully connected layer contains 3 hidden layers, and the number of neurons is 128, 64, and 32 respectively.

9. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 1, wherein In S8, the digital twin simulation model is constructed based on finite element analysis, including an electro-thermal coupling sub-model, a thermal-mechanical coupling sub-model, and an aging prediction sub-model. The model parameters are corrected through the Bayesian optimization algorithm, and the remaining service life of the wire harness at the time nodes of 1000 hours, 5000 hours, and 10000 hours is predicted.

10. The method for testing the coupling of dynamic load and electromagnetic interference of a wire harness according to claim 9, characterized in that In S8, the determination basis of the fault threshold is as follows: when the change rate of the harness resistance exceeds 50% or the signal error rate exceeds 10 -4 , a fault warning signal is triggered; The calculation formula for the resistance change rate is: (ΔR) / R0 = ((Rt - R0)) / R0 × 100%, where R0 is the initial reference resistance, Rt is the real-time working resistance, and ΔR is the absolute resistance change; The digital twin simulation model corrects parameters through the Bayesian optimization algorithm, with the root mean square error of the measured and simulated data < 5% as the convergence condition, and the model accuracy is re-verified every 24 hours. When the verification error > 5% for three consecutive times, the aging prediction sub-model is reconstructed; The resistance change rate is calculated by comparing the current resistance value with the initial resistance value, and the signal bit error rate is determined by measuring the ratio of the number of bit errors to the total number of transmitted bits per unit time with a bit error tester; The aging prediction sub-model introduces the hydrogen sulfide concentration as the corrosion acceleration factor, with the concentration gradient ranging from 10 ppm to 1000 ppm, and the concentration gradient is dynamically adjusted according to the type of wire harness sheath material: 10 ppm to 500 ppm for rubber materials, and 200 ppm to 1000 ppm for silicone materials.

Images