A new energy wiring harness continuity test system

Abstract

The present invention relates to the field of wire harness continuity test technology, and in particular to a new energy wire harness continuity test system, which automatically scans all points of a wire harness continuity test fixture, generates an activation point set, applies signals to the activation points, obtains the electrical connection status between the points, and merges the loop data of the same main points to generate a simplified continuity result set, thereby reducing redundant data and facilitating subsequent processing. The result confirmation module ensures the validity and accuracy of the continuity result set through double verification to prevent good products from being misjudged as defective products. The memory verification module verifies the current continuity result set through memory, uses a neural network to perform intelligent processing on the loop memory, automatically determines whether the wire harness model matches by comparing the continuity result set and the loop memory, and deletes the loop memory that is no longer needed, thereby improving the system's verification capability for different models of wire harnesses. The system effectively improves the efficiency, accuracy and reliability of wire harness continuity testing.

Description

Technical Field

[0001] The invention relates to the technical field of wire harness continuity testing, and in particular to a new energy wire harness continuity testing system. Background Art

[0002] With the rapid development of new energy electric vehicles, the use of new energy wiring harnesses in the whole vehicle has gradually increased, and the continuity test of the wiring harness has become increasingly important in ensuring the stability and safety of the vehicle electrical system. There are some operational limitations and efficiency problems in the existing test system for the conduction of modern new energy wiring harnesses. Especially during the debugging process, the system needs to detect the probes of the test fixture one by one, and manually enter the number to form a loop memory. As the number of wiring harness points increases, the complexity and time cost of the operation also increase. During the numbering process, if the operator misses or enters an error, the loop will not be fully generated, resulting in a lot of time for troubleshooting and re-debugging. In addition, problems with the test fixture are often not discovered in time before testing the wiring harness, which may cause good wiring harnesses to be misjudged as defective products, seriously affecting test efficiency and product quality. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a new energy wiring harness continuity test system, which comprehensively improves the efficiency, accuracy and reliability of new energy wiring harness continuity test through automated scanning verification, generation of loop memory and combined with a neural network automatic cleaning mechanism.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] A new energy wiring harness continuity test system, comprising a point scanning module, a continuity test module, a result confirmation module and a memory verification module connected in sequence;

[0006] The point scanning module is used to insert and scan all points of the wiring harness continuity test fixture to generate activation points;

[0007] The conduction test module is used to apply a test signal to the activated point through a test fixture, traverse and obtain the electrical connection status between each point and other points, generate point loop data, and merge the loop data with the same main point to generate a conduction result set;

[0008] The result confirmation module is used to insert and scan all points after the continuity test module generates the continuity test set and obtain verification points, compare and verify the activation points and the continuity result set through the verification points, and confirm the validity of the continuity result set;

[0009] The memory verification module includes a memory generation unit, a memory retention unit and a memory processing unit;

[0010] The memory generation unit is used to check whether there is a loop memory in the memory retention unit to generate a new loop memory according to the conduction result set and store it in the memory retention unit;

[0011] The memory retention unit is used to temporarily store the loop memory, and is also used to compare and verify the conduction result set received in real time with the temporarily stored loop memory to generate a conduction test verification result;

[0012] The memory processing unit is used to adaptively delete the loop memory temporarily stored in the memory retention unit based on the neural network model.

[0013] Furthermore, inserting and scanning all points of the wiring harness continuity test fixture to generate activation points includes the following steps:

[0014] Scan all points on the harness continuity test fixture in sequence, and mark the points that meet the electrical connection conditions as activated points;

[0015] Record the location information of all activation points and generate an activation point set;

[0016] Transmit the activation point set to the continuity test module.

[0017] Furthermore, when the conduction test module applies a test signal to the activation point through the test fixture, the new energy wiring harness is fixed by the pneumatic switch until all the test signals are applied and the pneumatic switch is released.

[0018] Furthermore, the test signal includes a direct current voltage signal and an alternating current voltage signal.

[0019] Furthermore, the comparison and verification of the activation points and the continuity test set through verification points includes: comparing the position information and the number of points of the verification points, activation points and continuity result set respectively; if the position information and the number of points are consistent, the continuity result set is calibrated to be valid; if the position information and the number of points are inconsistent, the continuity result set is calibrated to be invalid.

[0020] Furthermore, the adaptive deletion of the temporarily stored loop memory in the memory retention unit based on the neural network model includes the following steps:

[0021] Receive a conduction result set calibrated as valid and compare it with the loop memory temporarily stored in the memory retention unit through a pre-trained convolutional neural network model to determine whether the current wiring harness and the wiring harness corresponding to the loop memory are the same model wiring harness;

[0022] If the wiring harness is of the same model, the temporary circuit memory will not be deleted;

[0023] If the wiring harness is of a different model, the temporarily stored circuit memory will be deleted.

[0024] Furthermore, the training steps of the convolutional neural network model include:

[0025] Collecting historical conduction result sets and loop memories and performing data annotation, wherein the data annotation includes pre-annotating whether the conduction result sets and loop memories are model wiring harnesses;

[0026] Based on the historical conduction result set and the historical loop memory, a conduction result set matrix and a loop memory matrix are respectively constructed and normalized simultaneously;

[0027] The conduction result set matrix and the loop memory matrix are input into the convolutional neural network model, and whether the wiring harness is of the same model is used as the output of the neural network model to calculate the loss value between the classification result score and the actual label threshold;

[0028] Based on the loss value, the convolution kernel weights and bias values ​​in the neural network model are adjusted through the back propagation algorithm.

[0029] Furthermore, the convolution calculation formula of the convolutional neural network model is as follows:

[0030] ;

[0031] in, is the output value of the feature map at position (i, j) in the convolutional neural network model, is the convolution kernel weight matrix; i and j represent the position coordinates of the feature map; p and q represent the position coordinates of the elements in the convolution kernel; For the conduction result set matrix at position The point value at is the loop memory matrix at position The point value at ; k is the radius of the convolution kernel; b is the bias term.

[0032] Furthermore, the memory verification module also includes a memory timing unit, which is used to detect the storage time of the current loop memory in the memory retention unit, and delete the current loop memory in the memory retention unit when the storage time is greater than or equal to the preset cleaning time.

[0033] The beneficial effect of the present invention is that the present invention generates an activation point set by automatically inserting and scanning all points of the wiring harness conduction test fixture, without manual detection and inputting numbers, thereby avoiding the complexity of manual operation and possible omission errors. The automatically generated activation point set ensures that all points with electrical connection conditions are accurately identified, thereby improving the efficiency of test preparation. The conduction test module applies a test signal to the activation point through the test fixture, and traverses to obtain the electrical connection status between each point and other points. The loop data with the same main point is merged to generate a conduction result set. Comprehensive detection of the conduction state is ensured. The data merging process reduces redundant information, improves the simplicity and availability of the conduction result set, and facilitates subsequent analysis and processing. After the conduction test is completed, the result confirmation module inserts and scans all points again to obtain verification points. The verification points are compared and verified with the activation points and the conduction result set. The double verification mechanism effectively prevents test errors caused by problems with the test fixture or the wiring harness itself, and ensures the validity and accuracy of the conduction result set. Abnormal situations can be discovered and corrected immediately, reducing the risk of good products being misjudged as defective products. The memory verification module includes a memory generation unit, a memory retention unit, and a memory processing unit. By checking the loop memory in the memory retention unit, a new loop memory is generated according to the conduction result set, and a neural network is used to adaptively delete the temporarily stored loop memory. A pre-trained convolutional neural network model is used to compare the current conduction result set with the loop memory, and intelligently determine whether the wiring harness model matches and deletes it. The system's test adaptability to different models of wiring harnesses is improved, avoiding test misjudgments caused by mismatches between the model and the model corresponding to the old test memory. The real-time received conduction result set is further compared and verified with the temporarily stored loop memory to generate a conduction test verification result, which improves the efficiency, accuracy and reliability of the conduction test of the new energy wiring harness, and solves the problems of cumbersome operation, high error rate and low efficiency in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a structural schematic diagram of a new energy wiring harness continuity test system in the present invention.

[0035] Figure 2 It is a flow chart of the training steps of the convolutional neural network model in the present invention. DETAILED DESCRIPTION

[0036] See also Figure 1-2 As shown, the present invention relates to a new energy wiring harness continuity test system, comprising a point scanning module, a continuity test module, a result confirmation module and a memory verification module connected in sequence;

[0037] The point scanning module is used to insert and scan all points of the wiring harness continuity test fixture to generate activation points;

[0038] The conduction test module is used to apply a test signal to the activated point through a test fixture, traverse and obtain the electrical connection status between each point and other points, generate point loop data, and merge the loop data with the same main point to generate a conduction result set;

[0039] The result confirmation module is used to insert and scan all points after the continuity test module generates the continuity test set and obtain verification points, compare and verify the activation points and the continuity result set through the verification points, and confirm the validity of the continuity result set;

[0040] The memory verification module includes a memory generation unit, a memory retention unit and a memory processing unit;

[0041] The memory generation unit is used to check whether there is a loop memory in the memory retention unit to generate a new loop memory according to the conduction result set and store it in the memory retention unit;

[0042] The memory retention unit is used to temporarily store the loop memory, and is also used to compare and verify the conduction result set received in real time with the temporarily stored loop memory to generate a conduction test verification result;

[0043] The memory processing unit is used to adaptively delete the loop memory temporarily stored in the memory retention unit based on the neural network model.

[0044] In some embodiments, first, the hardware architecture of the point scanning module is based on a multi-channel high-precision scanning interface. The module performs a comprehensive scan of all points on the continuity test fixture by automatically inserting scanning technology. In terms of hardware implementation, a signal acquisition device and a digital processing controller are used to detect the voltage and current status of each point in real time. The built-in signal amplifier and filter of the system are used to ensure the reliable capture of weak signals and remove noise interference, thereby ensuring the accuracy of point status detection. The controller marks the points with electrical connection conditions through a built-in algorithm to generate an activation point set. The processing speed of the point scanning depends on the multi-threaded processing capability of the controller, which can achieve efficient parallel scanning and ensure that each point in a large-scale wiring harness system can be quickly and accurately identified. The generated activation point set will be transmitted to the continuity test module through the bus. The whole process avoids the steps of manual input and numbering, greatly improves the operating efficiency, and reduces the occurrence of human errors. The continuity test module is responsible for applying the test signal and collecting the electrical connection status between each point. The module uses a high-precision signal generator and a digital switching circuit. The signal generator can generate various types of test signals, including DC voltage signals and AC voltage signals. To ensure accurate signal transmission, the test system uses a programmable logic controller (PLC) and a high-speed switch matrix to achieve automatic switching of test signals. The signal access of each point is controlled by a high-speed relay or solid-state relay to ensure fast response and high reliability of signal application. At the same time, the system is equipped with precision current sensors and voltage acquisition modules to dynamically monitor the electrical parameters of each point in the test fixture to generate point loop data. The system uses FPGA (field programmable gate array) for data processing, merges loop data with the same main point, and generates a simplified conduction result set. FPGA has high throughput and parallel computing capabilities, can quickly process multi-point data, ensure comprehensive detection of the conduction state, and update the conduction result set in real time. The result confirmation module is used to verify the validity of the conduction result. The result confirmation module also uses multi-channel scanning technology. After the conduction test is completed, all points are scanned again to generate a verification point set. The verification point is compared point by point with the activation point and the conduction result set through the hardware acceleration processing unit. The system uses a high-precision comparison algorithm in the verification module to compare the location information, signal strength and other parameters of the point to ensure data consistency. During the comparison process, the system will automatically detect abnormal points and alert the user through the built-in alarm system. The processing core of the verification module uses a dual-core ARM processor and integrates a high-precision comparison algorithm. It can complete the data comparison of large-scale points within milliseconds and ensure that the test results of each point can be fed back in a timely and accurate manner. The memory verification module is the intelligent core component of the system, which mainly includes a memory generation unit, a memory retention unit and a memory processing unit.The memory generation unit automatically generates loop memory based on the test result set, and the built-in high-speed storage unit can realize large-capacity loop data storage. The memory retention unit is responsible for temporarily storing loop memory and updating the conduction test data in real time. The unit is connected to the main processor of the system through the PCIe interface to ensure high-speed data transmission and processing. The hardware architecture adopts a dual redundant storage system to ensure data reliability and real-time performance. During the storage process, the system will dynamically detect the validity of the loop memory and judge the accuracy of the loop by comparing the historical loop memory with the current conduction result set. The memory processing unit is based on the hardware accelerator of the neural network and uses GPU acceleration to perform inference calculations of the convolutional neural network. The neural network extracts features and recognizes patterns from the loop memory and the conduction result set to determine whether the current wiring harness model matches the loop memory. During the data comparison process, the system can automatically identify inconsistent wiring harness models and delete the no longer used loop memory in a timely manner according to the judgment results. This process achieves efficient matching of wiring harness data through the multi-layer feature extraction structure in the convolutional neural network model to avoid misjudgment caused by different wiring harness models. This intelligent processing method enables the system to dynamically adapt to different wiring harness models and ensure the effectiveness of verification.

[0045] Furthermore, inserting and scanning all points of the wiring harness continuity test fixture to generate activation points includes the following steps:

[0046] Scan all points on the harness continuity test fixture in sequence, and mark the points that meet the electrical connection conditions as activated points;

[0047] Record the location information of all activation points and generate an activation point set;

[0048] Transmit the activation point set to the continuity test module.

[0049] In some embodiments, the harness continuity test fixture realizes automatic detection of all points through a high-precision insertion scanning mechanism. The hardware part is mainly composed of a multi-channel scanning interface, a high-precision signal processor and a control unit. The specific working process is as follows: First, the scanning interface directly connects to each test point on the fixture to detect the electrical state of each point one by one. The design of the scanning interface adopts multi-channel parallel sampling technology, which can process the signal state of multiple points at the same time. The electrical state of each point is judged by collecting its voltage and current changes. When the electrical parameters of the point meet the preset connection conditions, it is marked as an activated point. The signal processor in the hardware system plays a vital role. It is responsible for receiving real-time signals from the multi-channel scanning interface and ensuring the stability and accuracy of the data through built-in filtering and signal amplification modules. The filtering module is used to remove high-frequency noise that may occur during the scanning process, while the signal amplification module enhances weak electrical signals to ensure that the signal can be accurately captured and identified by the subsequent processing unit. The signal processor is equipped with a high-speed analog-to-digital converter (ADC) to convert analog signals into digital signals for subsequent data processing and recording. In the data processing stage, the system performs marking and recording of activated points through an integrated control unit. The control unit uses FPGA as the core hardware. The parallel computing capability of FPGA ensures real-time processing during large-scale point scanning. Specifically, the control unit uses a preset algorithm to judge the electrical connection conditions of each point based on the electrical status data transmitted by the signal processor. Points that meet the conditions will be marked as activated points, and their position information will be automatically recorded. The position information here includes the physical location of the point (specific coordinates on the fixture) and whether it is electrically conductive. The storage unit inside the FPGA is used to temporarily store the information of these activated points and generate a unified set of activated points.

[0050] Furthermore, when the conduction test module applies a test signal to the activation point through the test fixture, the new energy wiring harness is fixed by the pneumatic switch until all the test signals are applied and the pneumatic switch is released.

[0051] Specifically, when the conduction test module applies a test signal to the activation point, the system first uses a pneumatic switch to fix the wiring harness. The pneumatic switch consists of a pneumatic control device and a mechanical fixing device. The pneumatic switch monitors the air pressure state in real time through a built-in pressure sensor to ensure the stability and precise control of the fixing force. The hardware architecture of the system includes a pneumatic control module, which uses a high-precision solenoid valve and a gas pressure regulator to control the opening and closing state of the pneumatic switch. The operating principle of the pneumatic switch is that when the system starts the conduction test, the pneumatic control module starts the air pressure through the solenoid valve to push the mechanical fixing device to clamp the wiring harness to ensure that the wiring harness remains stationary during the test. The fixing process of the pneumatic switch is highly automated and accurate. Through the feedback loop in the pneumatic control module, the system can dynamically adjust the air pressure according to the pressure data collected by the sensor in real time to ensure that the wiring harness is subjected to a moderate fixing force and is not damaged by excessive clamping. At the same time, the design of the pneumatic switch takes into account the fixing requirements of different types of wiring harnesses, and can adjust the clamping force and angle according to the specifications and shapes of the wiring harness to ensure the stable fixation of different types of wiring harnesses. During the continuity test, the continuity test module applies the test signal to the activation point through the test fixture. The process of applying the signal is managed by the control unit, which is responsible for applying the preset test signal to each point one by one according to the position information of the activation point. In terms of hardware implementation, the test fixture is connected to the signal generator, and the signal path is switched through a multi-way switch matrix to ensure that the signal can accurately reach each point. The test signal types include DC voltage, AC voltage and other pulse signals in specific frequency bands to meet the test requirements of different wiring harnesses. The entire signal application process controls the signal amplitude, frequency and time through a precision control algorithm to ensure the accuracy of the test. During the test signal application process, the pneumatic switch continues to keep the wiring harness fixed to ensure that the wiring harness is not subject to external physical interference such as vibration or displacement during the test. Since the pneumatic switch provides a stable mechanical fixation, the wiring harness maintains its initial position unchanged during the entire test process, ensuring that the connection state during the signal application process will not change due to the influence of external physical factors. When the continuity test module completes the application of the test signal to all activation points, the system controls the release operation of the pneumatic switch. At this point, the pneumatic control module releases the air pressure through the solenoid valve, and the mechanical fixture releases the harness. The release process of the pneumatic switch is also precisely timed to ensure that the harness is released only after all test signals have been applied, thus avoiding any errors caused by premature release of the harness during the test.

[0052] Furthermore, the test signal includes a direct current voltage signal and an alternating current voltage signal.

[0053] It should be noted that during the conduction test, the test fixture is connected to the signal generator through a multi-channel signal switching matrix. The signal generator has the ability to output various types of signals, especially the generation of DC voltage signals and AC voltage signals. The signal generator of the system adopts high-precision digital control technology, which can generate accurate voltage signals according to preset test conditions. For DC voltage signals, the system provides stable DC output through a precision voltage source module. The voltage value can be adjusted by the control unit within the set range. The common test range ranges from a few millivolts to hundreds of volts. In the hardware design, the output of the DC voltage source passes through a multi-stage filtering circuit to ensure the purity and stability of the signal and avoid the influence of external electromagnetic interference on the signal. The generation of AC voltage signals is completed by a programmable signal generator. The signal generator can output AC voltage signals of different frequencies and amplitudes, and supports AC signal output ranging from low frequency (such as 50Hz) to high frequency (such as MHz level) to test the electrical characteristics of the wiring harness in different frequency bands. The signal generator is equipped with a frequency synthesis module and a waveform modulation module to ensure the precise control of the output AC signal in frequency, phase and amplitude to meet the test requirements of the complex electrical characteristics of the new energy wiring harness. Furthermore, it is possible to support signal output of multiple waveform types, such as sine wave, square wave, triangle wave, etc., so as to more comprehensively detect the conduction performance of the wiring harness under different waveform conditions. During the conduction test, the system dynamically switches the test signal through a multi-way switch matrix to ensure that each activation point can receive the test of DC voltage signal and AC voltage signal in turn. The DC voltage signal is used to test the electrical performance of the wiring harness under static connection conditions, mainly to detect the conduction state, impedance value and other parameters of the wiring harness. The AC voltage signal is used for dynamic detection. The system can determine whether the wiring harness has abnormal conduction or failure during the frequency change process by detecting the electrical response of the point at different frequencies. This process can comprehensively evaluate the electrical connection quality of the wiring harness and ensure the stability of the wiring harness in the actual working environment.

[0054] Furthermore, the comparison and verification of the activation points and the continuity test set through verification points includes: comparing the position information and the number of points of the verification points, activation points and continuity result set respectively; if the position information and the number of points are consistent, the continuity result set is calibrated to be valid; if the position information and the number of points are inconsistent, the continuity result set is calibrated to be invalid.

[0055] It should be noted that after the continuity test is completed, the system verifies the activation points and the continuity result set through the verification points. First, the generation process of the verification points is the same as that of the activation points. The system uses the point scanning module to insert and scan all the points of the wiring harness continuity test fixture again, and records all the points that meet the electrical connection conditions. This process is implemented by a high-precision multi-channel scanning interface, and the hardware configuration ensures the real-time scanning and the accuracy of data acquisition. The system organizes the scanned point data into a verification point set, which includes the location information and connection status of each point. First, after generating the verification point set, the system verifies it one by one through the location information and point quantity comparison algorithm. Specifically, the system first uses an algorithm based on hash mapping to match the verification points, activation points, and points in the continuity result set. Through hash mapping, the system can quickly generate a unique hash value for each point, so that the location information of each point is mapped to a unique numerical representation. Secondly, for the comparison process, the algorithm designs a double-layer verification mechanism. The first level of verification is based on the location information of the verification points, activation points and conduction result sets. Through the hash value generated by the hash algorithm, the system can quickly match the points with completely consistent location information in the two sets. For inconsistent points, the algorithm will further call the secondary comparison mechanism to compare the specific physical coordinates and electrical connection status to determine whether there is a slight deviation caused by floating noise or other accidental factors. If there is such a slight deviation, the algorithm will use an algorithm based on Euclidean distance to correct it to ensure that the slight offset between points does not affect the overall comparison result. The second level of verification is based on the verification algorithm of the number of points. The system uses a simple counting algorithm to ensure that the number of points in the verification points, activation points and conduction result sets is consistent. This process mainly uses a statistical algorithm to calculate the number of points in each set. If the number of the three is consistent, it will further enter the comparison of detailed location information and status; if the number is inconsistent, the system will directly determine that the conduction result set is invalid, and adjust the test strategy through an adaptive algorithm to automatically rescan the verification points to reduce human intervention and operation time. Furthermore, in order to ensure data consistency and comparison efficiency, a parallel processing algorithm can be used. By splitting the data set into multiple small blocks, the algorithm can compare the verification points and activation points of each small block in parallel. Specifically, the comparison of position information and quantity is performed in multiple threads at the same time, which significantly improves the processing speed of large-scale harness testing. The coordination of parallel processing is managed by the master control algorithm, which distributes data to each processing core through a scheduling strategy and summarizes the processing results. During the multi-threaded processing process, the algorithm also introduces a locking mechanism to ensure that the processing data of each thread will not encounter race conditions and ensure the accuracy of the results.

[0056] Furthermore, the adaptive deletion of the temporarily stored loop memory in the memory retention unit based on the neural network model includes the following steps:

[0057] Receive a conduction result set calibrated as valid and compare it with the loop memory temporarily stored in the memory retention unit through a pre-trained convolutional neural network model to determine whether the current wiring harness and the wiring harness corresponding to the loop memory are the same model wiring harness;

[0058] If the wiring harness is of the same model, the temporary circuit memory will not be deleted;

[0059] If the wiring harness is of a different model, the temporarily stored circuit memory will be deleted.

[0060] It should be noted that, first, after the conduction test is completed, the system inputs the conduction result set marked as valid into the neural network model for processing. The conduction result set contains the electrical connection status of each point and its corresponding loop information. The system uses a pre-trained convolutional neural network model to compare the conduction result set with the temporary loop memory in the memory retention unit to determine whether the current wiring harness matches the wiring harness model stored in the loop memory. In the specific implementation, the convolutional neural network (CNN) model first preprocesses the conduction result set. The preprocessing step includes a normalization operation to standardize the input electrical connection status data to meet the input requirements of the neural network. After normalization, the conduction result set is converted into a two-dimensional matrix form, and each matrix element represents the conduction status of the corresponding point. Next, the system inputs this matrix into the convolution layer of the CNN for feature extraction. The convolution layer performs a sliding window operation on the input data through a series of filters (convolution kernels) to extract the potential spatial relationship and feature patterns in the loop memory and the conduction result set. The result of the convolution operation generates a set of feature maps, which represent the characteristic patterns of the conduction result set at different scales and levels. In order to further extract more discriminative features, the CNN model also contains multiple activation layers and pooling layers. The activation layer usually uses the ReLU function (rectified linear unit) to perform nonlinear mapping on the output of the convolution operation, thereby introducing the nonlinear expression ability of the model. The pooling layer is used for downsampling, reducing the size of the feature map, and retaining the most significant feature information to improve computational efficiency. After multiple layers of convolution, activation, and pooling operations, the system will obtain a vector representation that has been processed with high-dimensional features, called an embedding. This embedding contains the core features of the conduction result set. In the same neural network model, the system will perform the same processing on the loop memory in the memory retention unit to extract its feature embedding representation. At this time, the system will calculate the similarity between the embedding representation of the conduction result set and the embedding representation of the loop memory. When the system determines that the current wiring harness does not match the loop memory, it means that the staff has started to test the next model of wiring harness. The memory processing unit will trigger the deletion algorithm to clear the loop memory that does not match the current conduction result set to ensure the continuity and efficiency of the test. When the memory generation unit generates a new memory loop and stores it in the memory retention unit, the staff can check the memory loop of the memory retention unit and determine whether it meets the loop standard of the wiring harness.

[0061] Furthermore, the training steps of the convolutional neural network model include:

[0062] Collecting historical conduction result sets and loop memories and performing data annotation, wherein the data annotation includes pre-annotating whether the conduction result sets and loop memories are model wiring harnesses;

[0063] Based on the historical conduction result set and the historical loop memory, a conduction result set matrix and a loop memory matrix are respectively constructed and normalized simultaneously;

[0064] The conduction result set matrix and the loop memory matrix are input into the convolutional neural network model, and whether the wiring harness is of the same model is used as the output of the neural network model to calculate the loss value between the classification result score and the actual label threshold;

[0065] Based on the loss value, the convolution kernel weights and bias values ​​in the neural network model are adjusted through the back propagation algorithm.

[0066] It should be noted that, first of all, the training step starts with data collection and labeling. The system extracts the continuity result set and the corresponding loop memory from the historical records. These historical data represent the results of multiple wiring harness continuity tests, and each result set contains a large number of point continuity states and their corresponding loop information. In order for the neural network to learn the characteristic relationship of the wiring harness, the system needs to label the collected data. Specifically, the labeling process records the matching between each continuity result set and the loop memory. If the continuity result set and the loop memory correspond to the same model of wiring harness, the data is marked as a "positive example"; if they are not the same model, they are marked as a "negative example". This labeling provides basic training labels for the subsequent supervised learning process, so that the model can learn how to distinguish different models of wiring harnesses. Next, the system enters the data preprocessing and matrix construction stage. The continuity result set and loop memory will be converted into two-dimensional matrix forms, called the continuity result set matrix and the loop memory matrix. In these matrices, the rows and columns of the matrix correspond to the point connection information of the wiring harness, and each element represents the continuity state of the corresponding point. In order to improve the convergence speed and stability of the model, the system normalizes these matrices. The purpose of normalization is to adjust the values ​​of the conduction state and loop information to the same range (usually between 0 and 1) to prevent large values ​​from dominating the learning process of the model. Normalization can use the minimum-maximum normalization method. After preprocessing and normalization, the system inputs the conduction result set matrix and the loop memory matrix into the convolutional neural network (CNN) model. The architecture design of the CNN model includes multiple convolutional layers, pooling layers, and fully connected layers. In the convolutional layer, the convolution kernel (filter) is used to extract features from the input matrix. During the extraction process, the convolution kernel slides on the matrix with a fixed-size window and performs a weighted summation operation on the local area. The output of the convolutional layer is a feature map, which contains the association information of the conduction result set and the loop memory in the spatial dimension. For example, the convolution operation can identify the local pattern of point connections in the wiring harness, which helps to distinguish the conduction characteristics of different models of wiring harnesses. The role of the pooling layer is to downsample the feature map output by the convolution layer, reduce the size of the feature map, and retain the most important feature information. The maximum pooling operation is usually used to select the maximum value in the local area to further reduce the computational complexity and prevent overfitting. After the feature extraction is completed, the system enters the classification stage of the model. After multiple layers of convolution and pooling operations, the feature map will be flattened into a one-dimensional vector and input into the fully connected layer to finally generate the classification result. The output classification result is a binary label, indicating whether the conduction result set and the loop memory belong to the same model harness. The system calculates the loss value between the classification result and the actual label, and the cross entropy loss function can be used to measure the accuracy of the classification.Based on the loss value, the system will update the convolution kernel weights and bias values ​​of the model through the backpropagation algorithm. The core of backpropagation is to calculate the gradient of the loss value relative to each parameter, and use the gradient descent algorithm to minimize the loss value. Specifically, the system will calculate the contribution of each layer of parameters to the final loss through the chain rule, and update the convolution kernel weights and bias values ​​layer by layer. After multiple rounds of training, the model can classify the conduction result set and loop memory with a high degree of accuracy to determine whether they belong to the same model of wiring harness. With the optimization of the model weights, the prediction accuracy of the system continues to improve, providing intelligent support for subsequent wiring harness conduction tests.

[0067] Furthermore, the convolution calculation formula of the convolutional neural network model is as follows:

[0068] ;

[0069] in, is the output value of the feature map at position (i, j) in the convolutional neural network model, is the convolution kernel weight matrix; i and j represent the position coordinates of the feature map; p and q represent the position coordinates of the elements in the convolution kernel; For the conduction result set matrix at position The point value at is the loop memory matrix at position The point value at ; k is the radius of the convolution kernel; b is the bias term.

[0070] Specifically, the convolution calculation extracts local features in the data by applying the convolution kernel to local areas of the conduction result set matrix and the loop memory matrix. The convolution operation is essentially a local weighted summation that can capture features with spatial correlation in the input matrix. For example, in the comparison process of the conduction test and the loop memory, the convolution operation can identify the connection pattern between the harness points and the feature differences between different harness models. The weight matrix in the convolution kernel is updated by the back propagation algorithm. The system calculates the gradient of the weight to the loss based on the loss value of the model, and adjusts the weight of each convolution kernel according to the gradient descent method. After multiple training iterations, the convolution kernel can automatically learn the local feature patterns in the data, thereby improving the accuracy of the model matching the harness model. Through this convolution operation, the system can extract useful features from the conduction result set and compare them with the historical data in the loop memory, helping the model to capture the feature patterns of harness connections at different spatial scales, and finally determine whether the current harness matches the historical loop memory.

[0071] Furthermore, the memory verification module also includes a memory timing unit, which is used to detect the storage time of the current loop memory in the memory retention unit, and delete the current loop memory in the memory retention unit when the storage time is greater than or equal to the preset cleaning time.

[0072] Specifically, the memory timing unit is tightly integrated with the system's continuity test process. Whenever a new continuity test result is generated, the system calls the memory verification module to compare the current continuity result set with the historical loop memory in the memory retention unit. During the comparison process, the memory timing unit detects this access and immediately refreshes the storage time of the corresponding loop memory, clears its timer, and restarts the storage time. Each loop memory is assigned a timestamp when it is stored, indicating the start time of its storage. When a new continuity result set is generated and needs to be compared, the system detects this comparison behavior and triggers the memory timing unit to update the timestamp. At this time, the timing unit will use the current system time as the new timestamp, replace the old timestamp, and mark the latest storage time starting point of the loop memory. After the timestamp is updated, the timer of the timing unit will be reset and the storage time will be calculated again. This operation ensures that the loop memory will not be mistakenly deleted after being used due to the previous storage time being too long. Each time the loop memory is accessed, the storage time will be cleared and re-timed, extending its duration in the storage unit.

[0073] The above implementation modes are merely descriptions of the preferred implementation modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary engineering and technical personnel in the field shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A new energy wiring harness continuity test system, characterized in that: It includes a point scanning module, a conduction test module, a result confirmation module and a memory verification module which are connected in sequence; The point scanning module is used to insert and scan all points of the wiring harness continuity test fixture to generate activation points; The conduction test module is used to apply a test signal to the activated point through a test fixture, traverse and obtain the electrical connection status between each point and other points, generate point loop data, and merge the loop data with the same main point to generate a conduction result set; The result confirmation module is used to insert and scan all points after the continuity test module generates the continuity test set and obtain verification points, compare and verify the activation points and the continuity result set through the verification points, and confirm the validity of the continuity result set; The memory verification module includes a memory generation unit, a memory retention unit and a memory processing unit; The memory generation unit is used to check whether there is a loop memory in the memory retention unit to generate a new loop memory according to the conduction result set and store it in the memory retention unit; The memory retention unit is used to temporarily store the loop memory, and is also used to compare and verify the conduction result set received in real time with the temporarily stored loop memory to generate a conduction test verification result; The memory processing unit is used to adaptively delete the loop memory temporarily stored in the memory retention unit based on the neural network model.

2. A new energy harness continuity test system according to claim 1, characterized in that: Inserting and scanning all points of the wiring harness continuity test fixture to generate activation points includes the following steps: Scan all points on the harness continuity test fixture in sequence, and mark the points that meet the electrical connection conditions as activated points; Record the location information of all activation points and generate an activation point set; Transmit the activation point set to the continuity test module.

3. A new energy wiring harness continuity test system according to claim 1, characterized in that: When the conduction test module applies a test signal to the activation point through the test fixture, the new energy wiring harness is fixed by the pneumatic switch until all the test signals are applied and the pneumatic switch is released.

4. A new energy harness continuity test system according to claim 1, characterized in that: The test signal includes a DC voltage signal and an AC voltage signal.

5. A new energy wiring harness continuity test system according to claim 1, characterized in that: The comparison and verification of the activation points and the continuity test set through verification points includes: comparing the position information and the number of points of the verification points, activation points and continuity result set respectively; if the position information and the number of points are consistent, the continuity result set is calibrated to be valid; if the position information and the number of points are inconsistent, the continuity result set is calibrated to be invalid.

6. A new energy harness continuity test system according to claim 1, characterized in that: The adaptive deletion of the temporarily stored loop memory in the memory retention unit based on the neural network model comprises the following steps: Receive a conduction result set calibrated as valid and compare it with the loop memory temporarily stored in the memory retention unit through a pre-trained convolutional neural network model to determine whether the current wiring harness and the wiring harness corresponding to the loop memory are the same model wiring harness; If the wiring harness is of the same model, the temporary circuit memory will not be deleted; If the wiring harness is of a different model, the temporarily stored circuit memory will be deleted.

7. A new energy harness continuity test system according to claim 6, characterized in that: The training steps of the convolutional neural network model include: Collecting historical conduction result sets and loop memories and performing data annotation, wherein the data annotation includes pre-annotating whether the conduction result sets and loop memories are wire harnesses of the same model; Based on the historical conduction result set and the historical loop memory, a conduction result set matrix and a loop memory matrix are respectively constructed and normalized simultaneously; The conduction result set matrix and the loop memory matrix are input into the convolutional neural network model, and whether the wiring harness is of the same model is used as the output of the neural network model to calculate the loss value between the classification result output by the neural network model and the actual label; wherein the classification result is a binary label, indicating whether the conduction result set and the loop memory belong to the same model wiring harness; Based on the loss value, the convolution kernel weights and bias values ​​in the neural network model are adjusted through the back propagation algorithm.

8. A new energy harness continuity test system according to claim 7, characterized in that: The convolution calculation formula of the convolutional neural network model is as follows: ； in, is the output value of the feature map at position (i, j) in the convolutional neural network model, is the convolution kernel weight matrix; i and j represent the position coordinates of the feature map; p and q represent the position coordinates of the elements in the convolution kernel; For the conduction result set matrix at position The point value at is the loop memory matrix at position The point value at ; k is the radius of the convolution kernel; b is the bias term.

9. A new energy harness continuity test system according to claim 1, characterized in that: The memory verification module also includes a memory timing unit, which is used to detect the storage time of the current loop memory in the memory retention unit, and delete the current loop memory in the memory retention unit when the storage time is greater than or equal to the preset cleaning time.

Images