Automobile sensor multi-station test method, device and medium

Abstract

The invention discloses an automobile sensor multi-station test method and device and a medium, and relates to the technical field of automobile manufacturing testing, and the method comprises the steps: training a lightweight gradient boosting tree model through employing a historical test data set in combination with federal learning, obtaining a global optimization test model, and optimizing a test process through employing the global optimization test model, obtaining a parameter configuration adjustment amount and a test sequence priority, and collecting production resume data of the automobile sensor at the same time; the production resume data comprises production equipment data, process parameters, production process relationships and sensor batches; analyzing the production resume data by using a graph convolutional network to obtain production features, and fusing the parameter configuration adjustment amount, the test sequence priority and the production features to generate a self-adaptive test strategy; according to the method, the lightweight gradient boosting tree model is trained through preprocessing of the historical test data and federated learning, the global optimization test model is constructed, and the test efficiency and the resource utilization rate are improved.

Description

Technical Field

[0001] This invention relates to the field of automotive manufacturing testing technology, and in particular to a multi-station testing method, equipment and medium for automotive sensors. Background Technology

[0002] As automobiles become increasingly electronic and intelligent, the testing requirements for automotive sensors are becoming more complex. Multi-station testing, due to its ability to efficiently handle multiple test tasks simultaneously, has gradually become the mainstream approach in the industry. Existing methods for developing test strategies primarily rely on two approaches: statistical process control (SPC) and traditional machine learning. Both optimize various test parameters using past test data, aiming to complete tests faster and produce more consistent results. Additionally, some research has proposed collaborative testing frameworks, which essentially involve sharing data between different test stations to optimize their respective tests. However, overall, this still relies on centralized decision-making or independent decision-making by each station.

[0003] Existing methods also have shortcomings. First, when formulating testing strategies, only the data from the tests themselves are usually used, without fully integrating data from the production process. This leads to a disconnect between test optimization and actual production conditions, making it difficult to flexibly adjust tests based on differences between different batches of products or fluctuations in the production process. Second, when multiple variables influence each other, it is easy for tests to be duplicated or for different tests to conflict with each other, resulting in poor synergy. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multi-station testing method for automotive sensors to solve the problems of low testing efficiency, poor adaptability, and unsatisfactory collaborative effect.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a multi-station testing method for automotive sensors, comprising,

[0008] The test processes and historical test data of automotive sensors from multiple test production lines are collected and preprocessed to obtain a historical test dataset; the historical test data includes test parameters, test timing data and test results.

[0009] A lightweight gradient boosting tree model is trained using historical test datasets and federated learning to obtain a globally optimized test model. The test process is then optimized using this globally optimized test model to obtain parameter configuration adjustment amounts and test order priorities. At the same time, production history data of automotive sensors is collected. The production history data includes production equipment data, process parameters, production process relationships, and sensor batches.

[0010] Graph convolutional networks are used to analyze production history data to obtain production characteristics. Parameter configuration adjustments, test order priorities and production characteristics are then integrated to generate an adaptive testing strategy.

[0011] The test station performs collaborative testing on the automotive sensors according to an adaptive testing strategy to obtain collaborative test data.

[0012] In a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the preprocessing step comprises the following steps:

[0013] The test data is restructured to obtain test data blocks;

[0014] The test data blocks are cleaned and normalized. The mean, maximum, minimum, standard deviation, variance, and range of the normalized test data blocks are calculated to form the test dataset.

[0015] As a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the step of using historical test datasets combined with federated learning to train a lightweight gradient boosting tree model to obtain a globally optimized test model includes the following specific steps.

[0016] Each test production line uses the test dataset to train the lightweight gradient boosting tree model, resulting in a trained lightweight gradient boosting tree model.

[0017] Extract the parameters of the trained lightweight gradient boosting tree model, and use the federated averaging algorithm to aggregate the parameters of the lightweight gradient boosting tree model to obtain the parameters of the globally optimized test model.

[0018] The lightweight gradient boosting tree model is updated using the parameters of the globally optimized test model to obtain the globally optimized test model.

[0019] In a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the step of optimizing the testing process using a global optimization testing model includes the following specific steps.

[0020] Real-time acquisition of test parameters and test time series data; data cleaning and normalization processing of the real-time acquired test parameters and test time series data to obtain real-time normalized data; calculation of the mean, maximum, minimum, standard deviation, variance and range of the real-time normalized data; and concatenation into a real-time feature vector.

[0021] The real-time feature vector is input into the global optimization test model to obtain the test pass probability and the confidence level of the test pass probability;

[0022] Based on the test pass probability and the configuration parameters corresponding to the test process, the first multi-objective optimization function is used to calculate the parameter configuration adjustment amount and test order priority.

[0023] In a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the step of using a graph convolutional network to analyze production history data and obtain production characteristics includes the following specific steps.

[0024] By using production equipment, process parameters, and sensor batches as nodes and production process relationships as edges, a production knowledge graph is formed.

[0025] Graph convolutional networks are used to learn node embedding representations from the production knowledge graph to obtain node embedding representations.

[0026] Cluster analysis is used to extract production features from node embedding representations.

[0027] As a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the step of fusing parameter configuration adjustment, test sequence priority, and production characteristics to generate an adaptive testing strategy includes establishing a second multi-objective optimization function, using the parameter configuration adjustment and test sequence priority as the initial solution of the second multi-objective optimization function, using the production characteristics as the constraint conditions of the second multi-objective optimization function, and using the second multi-objective optimization function to solve for the optimal test parameter configuration and test station scheduling sequence to obtain the adaptive testing strategy.

[0028] In a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the testing stations perform collaborative testing on the automotive sensors according to an adaptive testing strategy. The specific steps are as follows.

[0029] Convert adaptive testing strategies into test parameter configuration instructions and test sequence instructions;

[0030] The testing station adjusts the parameter configuration according to the test parameter configuration instructions, and the mechanical conveying device adjusts the test sequence priority according to the test sequence instructions.

[0031] As a preferred embodiment of the multi-station testing method for automotive sensors described in this invention, the specific steps for calculating the parameter configuration adjustment amount and test order priority using a first multi-objective optimization function are as follows:

[0032] Establish a first multi-objective optimization function, take the configuration parameters corresponding to the test process as the initial solution of the first multi-objective optimization function, and take the test pass probability and the confidence level of the test pass probability as the input of the first multi-objective optimization function;

[0033] The status data of the test station is collected as the constraint condition of the first multi-objective optimization function, and Pareto optimal solution set is generated iteratively through selection operation, crossover operation and mutation operation;

[0034] Extract parameter configuration adjustments and test order priorities from the Pareto optimal solution set.

[0035] In a second aspect, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the multi-station testing method for automotive sensors as described in the first aspect of the present invention.

[0036] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the multi-station testing method for automotive sensors as described in the first aspect of the present invention.

[0037] The beneficial effects of this invention are as follows: By preprocessing historical test data and training a lightweight gradient boosting tree model through federated learning, a globally optimized test model is constructed, realizing dynamic and nonlinear optimization of test parameters and order, thereby improving test efficiency and resource utilization; by analyzing production history data through graph convolutional networks, complex correlation features between equipment, processes, and batches are mined, making the test strategy closely integrated with actual production conditions, enhancing adaptability to production fluctuations, reducing test blind spots and redundancy, and improving overall collaborative efficiency. Attached Figure Description

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a flowchart of a multi-station testing method for automotive sensors.

[0040] Figure 2 This is a schematic diagram of data preprocessing.

[0041] Figure 3 A schematic diagram for training a globally optimized test model for federated learning.

[0042] Figure 4 This is a schematic diagram of production feature extraction. Detailed Implementation

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0044] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0045] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0046] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a multi-station testing method for automotive sensors, including the following steps:

[0047] S1: Collect and preprocess the test processes and historical test data of automotive sensors from multiple test production lines to obtain a historical test dataset; the historical test data includes test parameters, test timing data and test results.

[0048] S1.1: Collect test process and historical test data from the data acquisition interface of all test stations on the test production line. Historical test data includes voltage and current of electrical characteristic test stations, temperature and humidity of environmental simulation test stations, timestamps of test command issuance and response, and test result judgment signals.

[0049] It should also be noted that the voltage and current of the electrical characteristic test station and the temperature and humidity of the environmental simulation test station are the test parameters, the timestamps of the test command issuance and response are the test timing data, and the test result judgment signal is the test result.

[0050] S1.2: Reorganize the historical test data according to the structure of test production line - test station - automotive sensor batch - acquisition time to generate test data blocks including data header and data body;

[0051] The header of the test data block records the test production line number, test station number, automotive sensor batch number, and acquisition time of historical test data. The data body of the test data block records the test parameters, test timing data, and test results in the order of acquisition time.

[0052] S1.3: Perform data cleaning on the test data block, including removing duplicate data, filling blank data with linear interpolation, and removing test parameters that exceed the range of the automotive sensors;

[0053] The cleaned test data blocks are normalized using min-max normalization, which uses the range of the automotive sensor as the maximum and minimum values.

[0054] Calculate the mean, maximum, minimum, standard deviation, variance, and range of the test parameters and test time series data in the normalized test data block, and concatenate them into test parameter feature vectors and test time series data feature vectors respectively. Then, concatenate the test parameter feature vectors and test time series data feature vectors into a statistical feature vector.

[0055] The statistical feature vectors are combined with the corresponding test results and data headers to form a historical test dataset.

[0056] S2: Use historical test datasets combined with federated learning to train a lightweight gradient boosting tree model to obtain a globally optimized test model. Use the globally optimized test model to optimize the test process, obtain parameter configuration adjustment amounts and test order priorities, and at the same time collect production history data of automotive sensors; production history data includes production equipment data, process parameters, production process relationships and sensor batches.

[0057] S2.1: All test production lines use normalized test data blocks to train the lightweight gradient boosting tree model: Set the learning rate, number of decision trees, maximum depth of decision trees, and minimum data size of leaf nodes for the lightweight gradient boosting tree model; initialize the predicted values ​​of the lightweight gradient boosting tree model; use feature vectors and test results as test samples and input them into the lightweight gradient boosting tree model; perform multiple iterations on the lightweight gradient boosting tree model until the required number of decision trees is reached to complete the training.

[0058] Furthermore, for each iteration, the log loss function is used to calculate the log loss between the current predicted value and the test result. The partial derivative of the log loss with respect to the current predicted value is calculated as the error gradient. The histogram algorithm is used to divide each dimension of the feature vector into buckets, and the boundary of each bucket is used as a candidate split point. The split gain of each candidate split point is calculated based on the error gradient. The candidate split point with the largest split gain is selected as the optimal split point. After each split, the optimal split point is repeatedly selected until the maximum depth of the decision tree or the minimum data size of the leaf nodes is reached, resulting in a new decision tree. The average error gradient of all test samples in the final optimal split point is used as the predicted value of the new decision tree. The product of the new decision tree's predicted value and the learning rate is added to the current predicted value to obtain the predicted value of the new lightweight gradient boosting tree model.

[0059] S2.2: Upload the parameters of all trained lightweight gradient boosting tree models to the central federated learning server. The central federated learning server uses the federated averaging algorithm to aggregate the parameters of the gradient boosting tree models to obtain the parameters of the globally optimized test model.

[0060] The central federated learning server distributes the parameters of the globally optimized test model to all test production lines. All test production lines replace the parameters of the gradient boosting tree model with the parameters of the globally optimized test model to obtain the globally optimized test model.

[0061] S2.3: When the automotive sensors enter the testing production line, the test parameters and test timing data of the automotive sensors are collected in real time. The data cleaning and normalization processes in steps S1.2 and S1.3 are performed on the real-time collected test parameters and test timing data to obtain real-time normalized data. The mean, maximum value, minimum value, standard deviation, variance and range of the real-time normalized data are calculated respectively and concatenated into a real-time feature vector.

[0062] S2.4: Input the real-time feature vector into the global optimization test model. Each decision tree in the global optimization test model performs forward propagation inference to obtain the predicted value of each decision tree. Use the learning rate as the weight of the predicted value, and perform a weighted sum of the predicted values ​​of all decision trees to obtain the predicted value of the global optimization test model. Use the predicted value of the global optimization test model as the input of the sigmoid function to calculate the probability of passing the test.

[0063] Calculate the standard deviation of the predicted values ​​of all decision trees, use the standard deviation of the predicted values ​​of all decision trees as input to the sigmoid function, and use the sigmoid function to calculate the confidence level of the probability of passing the test.

[0064] S2.5: Collect the configuration parameters corresponding to the current test process, including the upper and lower limits of voltage and current thresholds for electrical characteristic test stations, the upper and lower limits of temperature and humidity ranges for environmental simulation test stations, and the execution order priority of each test station;

[0065] The voltage and current threshold limits of the electrical characteristic test station and the temperature and humidity range limits of the environmental simulation test station are normalized in step S1.3. The execution order priority of each test station is one-hot encoded to obtain the preprocessed configuration parameters. The test pass probability, the confidence level of the test pass probability and the preprocessed configuration parameters are concatenated into an input vector.

[0066] S2.6: With the optimization objectives of maximizing test efficiency, maximizing test accuracy, and minimizing test resource energy consumption, the first multi-objective optimization function is constructed, with the expression as follows:

[0067] ;

[0068] in, For the first multi-objective optimization function value, For the input vector, For testing efficiency, it is defined as the number of automotive sensor tests completed per unit time. The accuracy rate is defined as the ratio of the number of correctly predicted vehicle sensors to the total number of vehicle sensors tested. To test resource energy consumption, it is defined as the total energy consumption of equipment at each test station;

[0069] The expression for calculating test efficiency is as follows:

[0070] ;

[0071] in, To test the number of sensors, This is the test time under the current configuration parameters. To test the confidence level of the probability of passing, The time reduction factor is determined by fitting the mathematical relationship between the test time and the confidence level of the test pass probability using the least squares method.

[0072] The expression for calculating the energy consumption of the test resources is as follows:

[0073] ;

[0074] in, The baseline power for the current test station is determined by taking the average of the historical power data for that station. For the first Test time under the current configuration parameters For the first The energy sensitivity coefficient of the current configuration parameter will be the first one. The actual energy consumption of the current configuration parameters is used as the dependent variable. Performing linear regression on the independent variable yields the regression coefficients of the first variable. The energy sensitivity coefficient of the current configuration parameters. For the first The current configuration parameters, For the first Initial configuration parameters;

[0075] The minimum-maximum normalization method is used to normalize the test efficiency and test accuracy respectively, and the negative normalization method is used to normalize the test resource energy consumption. The test accuracy is the ratio of the number of correctly predicted car sensors to the number of car sensors tested.

[0076] The preprocessed configuration parameters are used as the initial solution of the first multi-objective optimization function. The test pass probability and the confidence level of the test pass probability are used as the input of the first multi-objective optimization function. The real-time collected status data of the test station are used as the constraint conditions of the first multi-objective optimization function. The first multi-objective optimization function is solved using a non-dominated sorting genetic algorithm with an elite strategy.

[0077] Furthermore, the non-dominated sorting genetic algorithm with an elitist strategy iteratively generates the Pareto optimal solution set of the first multi-objective optimization function through initialization population selection, crossover, and mutation operations.

[0078] When initializing the population, each individual in the population represents a candidate solution, which includes the adjustment amount of the candidate parameter configuration and the priority of the candidate test order;

[0079] When performing selection operations, the non-dominated sorting genetic algorithm of the elite strategy uses the tournament selection method to select candidate solutions in the population to obtain parent individuals;

[0080] When performing crossover operations in the non-dominated sorting genetic algorithm of the elite strategy, the crossover probability is set, and simulated binary crossover is used to perform crossover operations on the parent individuals to obtain the offspring individuals.

[0081] When performing mutation operations in the non-dominated sorting genetic algorithm of the elite strategy, the mutation probability is set, and Gaussian mutation is used to randomly perturb the parameter configuration adjustment amount and test order priority of the offspring individuals to obtain new offspring individuals. The parent individuals, offspring individuals and new offspring individuals are then merged into a new population.

[0082] During the iteration process, a non-dominated sorting genetic algorithm with an elitist strategy is simulated and run in a computer environment to calibrate the number of iterations of the non-dominated sorting genetic algorithm with an elitist strategy. The number of iterations of the non-dominated sorting genetic algorithm with an elitist strategy is set, and non-dominated sorting and crowding calculation are performed on each generation of the population. Starting from the first layer of non-dominated sorting, individuals in the population are extracted as elite individuals in descending order of crowding and retained to the next generation of the population. The iteration terminates after the number of iterations of the non-dominated sorting genetic algorithm with an elitist strategy is reached. All individuals in the new population after the iteration terminates constitute the Pareto optimal solution set of the first multi-objective optimization function. Tournament selection, simulated binary crossover, Gaussian mutation non-dominated sorting, and crowding calculation are all existing technologies and will not be described in detail.

[0083] S2.7: Extract parameter configuration adjustments and test order priorities from the Pareto optimal solution set of the first multi-objective optimization function: Collect historical test pass probabilities and corresponding historical test results. Among all historical test results that passed, the proportion of historical test pass probabilities that were true positives is used as the true positive rate; among all historical test results that failed, the proportion of historical test pass probabilities that were true negatives is used as the false positive rate. Plot the true positive rate on the ordinate and the false positive rate on the abscissa, and calculate the true positive rate and false positive rate under different historical test pass probabilities. Plot an ROC curve, and determine the distance from the top left corner. The historical test pass probability corresponding to the most recent point on the ROC curve is used as the interception threshold. For example, if the point (0.0, 1.0) on the ROC curve is closest to the top left corner (0.0, 1.0), then the historical test pass probability of 0.4 corresponding to the point (0.0, 1.0) on the ROC curve is selected as the interception threshold. The historical test pass probability that maximizes the difference between the true positive rate and the false positive rate is used as the pass threshold. For example, when the historical test pass probability is 0.5, the difference between the true positive rate and the false positive rate is maximized, so the historical test pass probability of 0.5 is used as the pass threshold.

[0084] When the test pass probability in the input vector is not lower than the pass threshold, the candidate parameter configuration adjustment amount and candidate test order priority with the highest test efficiency are extracted from the Pareto optimal solution set of the first multi-objective optimization function and used as the parameter configuration adjustment amount and test order priority.

[0085] When the probability of passing the test in the input vector is not higher than the interception threshold, the candidate parameter configuration adjustment amount and candidate test order priority with the highest test accuracy are extracted from the Pareto optimal solution set of the first multi-objective optimization function, and used as the parameter configuration adjustment amount and test order priority.

[0086] When the test pass probability in the input vector is between the interception threshold and the pass threshold, the candidate parameter configuration adjustment amount and candidate test order priority with the highest normalized test resource energy consumption are extracted from the Pareto optimal solution set of the first multi-objective optimization function, and used as the parameter configuration adjustment amount and test order priority.

[0087] The parameter configuration adjustments are added to the preprocessed configuration parameters, and the execution order of the test stations is sorted according to the test order priority to obtain the optimized test process.

[0088] S2.8: Collect production history data of automotive sensors from the data interface of Manufacturing Execution Management, including production equipment data, process parameters, production process relationships and sensor batches.

[0089] It should also be noted that step S2 uses federated learning to aggregate multi-production line data to train a lightweight gradient boosting tree model, generating a globally optimized test model. This achieves the integration and utilization of site-wide test knowledge while ensuring data privacy. Based on the globally optimized test model, the testing process is optimized, improving testing efficiency and accuracy while reducing testing resource energy consumption.

[0090] S3: Use graph convolutional networks to analyze production history data, obtain production characteristics, and integrate parameter configuration adjustments, test order priorities, and production characteristics to generate an adaptive testing strategy.

[0091] S3.1: Production equipment data is used as equipment nodes. The attributes of equipment nodes include equipment number, equipment type and equipment status. Process parameters are used as parameter nodes. The attributes of parameter nodes include parameter type, parameter value and parameter unit. Sensor batches are used as batch nodes. The attributes of batch nodes include batch number, production time and material information. Production process relationships are used as edges. The attributes of edges include relationship type, relationship strength and timestamp, forming a production knowledge graph.

[0092] S3.2: Use graph convolutional networks to learn node embedding representations for the production knowledge graph: For each node, perform one-hot encoding on equipment type, equipment status, parameter type, parameter unit, material information and relationship type, and use min-max normalization to normalize parameter values, production time, relationship strength and timestamp.

[0093] The corresponding normalized and one-hot encoded attributes are combined into the node vector of the corresponding node. All node vectors are concatenated into a node feature matrix with the number of nodes as the number of rows and the number of dimensions of the node vector as the number of columns. Historical production history data is collected, and the number of times each relationship type between each pair of nodes appears in the historical production history data is counted. Maximum value normalization is used for mapping to obtain the corresponding edge weights. All edge weights are concatenated into an adjacency matrix with the number of rows and columns equal to the number of nodes.

[0094] The encoder of the graph convolutional network is initialized by inputting the node feature matrix and the adjacency matrix into the encoder of the graph convolutional network. The first graph convolutional layer is used to perform convolution operation on the node feature matrix and the adjacency matrix to obtain the first layer node feature representation. The second graph convolutional layer is used to perform convolution operation on the first layer node feature representation and the adjacency matrix to obtain the second layer node feature representation. The third graph convolutional layer is used to perform convolution operation on the second layer node feature representation and the adjacency matrix to obtain the initial node embedding matrix.

[0095] The initial node embedding matrix is ​​input into the decoder. A fully connected layer is used to perform nonlinear transformation and mapping operations on the initial node embedding matrix, and outputs the reconstructed node feature matrix and the predicted adjacency matrix respectively. The mean square error of the reconstructed node feature matrix and the node feature matrix is ​​calculated. The cross entropy loss function is used to calculate the cross entropy loss of the predicted adjacency matrix and the adjacency matrix. The average of the mean square error and the cross entropy loss is calculated as the total loss.

[0096] The chain rule is used to calculate the partial derivatives of the total loss with respect to the parameters of the encoder and decoder, which are used as the gradient of the total loss. The gradient descent method is then used to iteratively update the parameters of the encoder and decoder until the average total loss of a fixed training period is one order of magnitude lower than the average total loss of the previous fixed training period. At this point, the training is complete, and the updated graph convolutional network is obtained.

[0097] The node feature matrix and adjacency matrix are input into the encoder of the updated graph convolutional network. The updated first graph convolutional layer is used to convolve the node feature matrix and adjacency matrix to obtain the updated first-layer node feature representation. The updated second graph convolutional layer is used to convolve the updated first-layer node feature representation and adjacency matrix to obtain the updated second-layer node feature representation. The updated third graph convolutional layer is used to convolve the updated second-layer node feature representation and adjacency matrix to generate the node embedding representation of all nodes.

[0098] S3.3: Extract production features from node embedding representations using cluster analysis: Calculate the sum of squared errors for different numbers of clusters, the expression is:

[0099] ;

[0100] in, For the sum of squared errors, For cluster indexing, For node embedding representation, For the first A set of clusters, For the first The cluster centers of each cluster. For node embedding representation to the first The square of the Euclidean distance between the cluster centers of the clusters;

[0101] Plot the K-SSE curve with the number of clusters on the horizontal axis and the sum of squared errors on the vertical axis. Smooth the K-SSE curve using the moving average method, calculate the curvature of each point on the smoothed K-SSE curve, and take the number of clusters corresponding to the point with the maximum curvature as the optimal number of clusters for node embedding representation.

[0102] Using node embedding representation and optimal cluster size as inputs to the K-means algorithm, the cluster centers are iteratively calculated and sample points are assigned to obtain the device cluster feature vector, parameter cluster feature vector, and batch cluster feature vector. The device cluster feature vector, parameter cluster feature vector, and batch cluster feature vector are concatenated to obtain the production feature vector.

[0103] Furthermore, the optimal number of node embedding representations for clustering are randomly selected from the node embedding representations of the device nodes as the initial cluster centers. The Euclidean distance from the node embedding representation of each device node to all current cluster centers is calculated. The node embedding representation of each device node is assigned to the cluster center with the smallest Euclidean distance to form a device cluster. The mean of the node embedding representations of all device nodes in each device cluster is calculated and used as the cluster center of the device cluster.

[0104] The cluster centers of the device cluster are compared with the initial cluster centers. If the cluster centers of the device cluster are different from the initial cluster centers, the initial cluster centers are replaced with the new cluster centers, and new cluster centers are recalculated. Node embedding representations are assigned to each device node, and new cluster centers of the device cluster are calculated until the new cluster centers of the device cluster are the same as the initial cluster centers. All the new cluster centers of the device cluster are then concatenated to form the device cluster feature vector. The generation methods of the parameter cluster feature vector and the batch cluster feature vector are the same as those of the device cluster feature vector, and will not be elaborated further.

[0105] S3.4: With the optimization objectives of maximizing test stability, maximizing quality consistency, and maximizing production resource utilization, a second multi-objective optimization function is established, expressed as:

[0106] ;

[0107] in, For the second multi-objective optimization function value, To test stability, To ensure consistent quality, To improve the utilization rate of production resources;

[0108] The expression for calculating test stability is:

[0109] ;

[0110] in, This is a set of test results under the optimized test process configuration parameters. for The variance;

[0111] The expression for calculating quality consistency is:

[0112] ;

[0113] in, This represents the total number of automotive sensor batches. For automotive sensor batch indexing, For the first The average performance of a batch of automotive sensors under the optimized test procedure configuration parameters. This represents the overall average performance across all batches.

[0114] The expression for calculating the utilization rate of production resources is:

[0115] ;

[0116] in, As a weight for equipment utilization rate, To improve equipment utilization, For parameter utilization weights, For parameter utilization, and The grid search method is used to search historical production history data for the weighted combination that maximizes production resource utilization. For example, the grid search shows that when... , When the production resource utilization rate is at its maximum, 0.6 and 0.4 are selected as the weights for equipment utilization rate and parameter utilization rate, respectively.

[0117] S3.5: Use the parameter configuration adjustment amount and test order priority from step S2.7 as the initial solution of the second multi-objective optimization function, generate the feature vector as the constraint condition of the second multi-objective optimization function, and use the multi-objective particle swarm optimization algorithm to solve the second multi-objective optimization function.

[0118] Furthermore, the multi-objective particle swarm optimization algorithm is simulated and run in a computer environment to calibrate the particle swarm size and the number of iterations of the multi-objective particle swarm optimization algorithm. The particle swarm size and the number of iterations of the multi-objective particle swarm optimization algorithm are set, and the initial position and initial velocity of each particle are randomly generated. The particle position represents the value of the parameter configuration adjustment amount, and the particle velocity represents the change range of the parameter configuration adjustment amount.

[0119] The test stability, quality consistency, and production resource utilization rate of each particle are calculated using the expression for production resource utilization rate, and then concatenated into a fitness vector. The fitness vector of each particle is compared with the best fitness vector of the historical particles. The fitness vector in which all elements are not lower than the corresponding elements in the best fitness vector of the historical particles, and at least one element in the fitness vector is higher than the corresponding element in the best fitness vector of the historical particles, is called the fitness vector dominating the best fitness vector of the historical particles. If the fitness vector dominates the best fitness vector of the historical particles, the corresponding particle position is used to replace the particle position corresponding to the best fitness vector of the historical particles, forming a global optimal position set.

[0120] Based on the global optimal position set, the velocity update formula and position update formula of the multi-objective particle swarm optimization algorithm are used to calculate the new particle velocity and new particle position of each particle respectively.

[0121] Repeat the steps of calculating the fitness vector to form a global optimal position set, calculating the velocity and position of new particles, until the number of iterations of the multi-objective particle swarm optimization algorithm is reached. The fitness vectors of all particles are sorted in a non-dominated manner, and the current particle position and current particle velocity of all non-dominated particles are used as the Pareto optimal solution set of the second multi-objective optimization function.

[0122] The decision matrix is ​​formed by combining all fitness vectors in the Pareto optimal solution set of the second multi-objective optimization function. The rows of the decision matrix correspond to the solutions in the Pareto optimal solution set of the second multi-objective optimization function. Vector normalization is used to normalize the decision matrix. Based on historical production history data, the entropy weight method is used to calculate the weights of production equipment, process parameters, and sensor batches. The weights of production equipment, process parameters, and sensor batches are used to weight each column of the normalized decision matrix to obtain the weighted decision matrix.

[0123] The maximum values ​​in each column of the weighted decision matrix are combined to form the ideal solution, and the minimum values ​​in each column are combined to form the negative ideal solution. The Euclidean distance from each Pareto optimal solution in the Pareto optimal solution set of the second multi-objective optimization function to the ideal solution and to the negative ideal solution are calculated. The sum of the Euclidean distances of the ideal solution and the negative ideal solution is taken as the Euclidean distance span, and the ratio of the Euclidean distance of the negative ideal solution to the Euclidean distance span is taken as the relative proximity. The Pareto optimal solutions with the largest relative proximity are superimposed on the configuration parameters of the optimized test process to obtain the fusion parameter configuration. The execution order of the test stations is sorted according to the test order priority to obtain the station execution sequence, forming an adaptive test strategy.

[0124] It should also be noted that: Step S3 constructs a production knowledge graph and applies a graph convolutional network for node embedding representation learning, thereby mining production equipment data, process parameters, production process relationships, and complex nonlinear correlations between sensor batches, and extracting high-value production features.

[0125] By establishing a second multi-objective optimization function and using features as constraints, and integrating them with parameter configuration adjustment and test order priority, an adaptive testing strategy closely coupled with actual production conditions is generated.

[0126] The adaptive testing strategy enhances the adaptability of the testing scheme to production fluctuations, improves testing stability, quality consistency and production resource utilization, and achieves closed-loop optimization from testing to production.

[0127] S4: The test station performs collaborative testing on the automotive sensors according to the adaptive testing strategy to obtain collaborative test data.

[0128] S4.1: Use inverse minimum-maximum normalization to inverse normalize the fusion parameter configuration to obtain the test parameter configuration instruction, and use a priority encoder to encode the station execution sequence into a test sequence instruction;

[0129] The test parameter configuration command is sent to the test station controller. The test station controller adjusts the voltage output range and current sampling frequency of the electrical characteristic test station to the voltage output range and current sampling frequency in the test parameter configuration command. The temperature control curve and humidity control parameters of the environmental simulation test station are set according to the test parameter configuration command.

[0130] The test sequence command is sent to the mechanical conveyor controller, which then rearranges the execution sequence of the test station according to the test sequence command.

[0131] The test station performs collaborative testing according to the adjusted parameter configuration and the execution sequence of the test station, and collects collaborative test data in real time, including collaborative test parameters, collaborative test timing data and collaborative test results.

[0132] This embodiment also provides a computer device applicable to a multi-station testing method for automotive sensors, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multi-station testing method for automotive sensors as proposed in the above embodiment.

[0133] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0134] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements a multi-station testing method for automotive sensors as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0135] In summary, this invention constructs a globally optimized test model by preprocessing historical test data and training a lightweight gradient boosting tree model using federated learning. This achieves dynamic and nonlinear optimization of test parameters and order, improving test efficiency and resource utilization. Furthermore, by analyzing production history data through graph convolutional networks, it uncovers complex correlations between equipment, processes, and batches, closely integrating the test strategy with actual production conditions. This enhances adaptability to production fluctuations, reduces test blind spots and redundancy, and improves overall collaborative efficiency.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-station testing method for automotive sensors, characterized in that: include, The test processes and historical test data of automotive sensors from multiple test production lines are collected and preprocessed to obtain a historical test dataset; the historical test data includes test parameters, test timing data and test results. A lightweight gradient boosting tree model is trained using historical test datasets and federated learning to obtain a globally optimized test model. The test process is then optimized using this globally optimized test model to obtain parameter configuration adjustment amounts and test order priorities. At the same time, production history data of automotive sensors is collected. The production history data includes production equipment data, process parameters, production process relationships, and sensor batches. Graph convolutional networks are used to analyze production history data to obtain production characteristics. Parameter configuration adjustments, test order priorities and production characteristics are then integrated to generate an adaptive testing strategy. The test station performs collaborative testing on the automotive sensors according to an adaptive testing strategy to obtain collaborative test data.

2. The multi-station testing method for automotive sensors as described in claim 1, characterized in that: The preprocessing process involves the following steps. The historical test data is restructured to obtain test data blocks; The test data blocks are cleaned and normalized. The mean, maximum, minimum, standard deviation, variance, and range of the normalized test data blocks are calculated to form the test dataset.

3. The multi-station testing method for automotive sensors as described in claim 1, characterized in that: The process involves using historical test datasets combined with federated learning to train a lightweight gradient boosting tree model, resulting in a globally optimized test model. The specific steps are as follows: Each test production line uses the test dataset to train the lightweight gradient boosting tree model, resulting in a trained lightweight gradient boosting tree model. Extract the parameters of the trained lightweight gradient boosting tree model, and use the federated averaging algorithm to aggregate the parameters of the lightweight gradient boosting tree model to obtain the parameters of the globally optimized test model. The lightweight gradient boosting tree model is updated using the parameters of the globally optimized test model to obtain the globally optimized test model.

4. The multi-station testing method for automotive sensors as described in claim 1, characterized in that: The optimization of the test process using a global optimization test model involves the following specific steps. Real-time acquisition of test parameters and test time series data; data cleaning and normalization processing of the real-time acquired test parameters and test time series data to obtain real-time normalized data; calculation of the mean, maximum, minimum, standard deviation, variance and range of the real-time normalized data; and concatenation into a real-time feature vector. The real-time feature vector is input into the global optimization test model to obtain the test pass probability and the confidence level of the test pass probability; Based on the test pass probability and the configuration parameters corresponding to the test process, the first multi-objective optimization function is used to calculate the parameter configuration adjustment amount and test order priority.

5. The multi-station testing method for automotive sensors as described in claim 1, characterized in that: The process of using graph convolutional networks to analyze production history data and obtain production characteristics involves the following specific steps. By using production equipment, process parameters, and sensor batches as nodes and production process relationships as edges, a production knowledge graph is formed. Graph convolutional networks are used to learn node embedding representations from the production knowledge graph to obtain node embedding representations. Cluster analysis is used to extract production features from node embedding representations.

6. The multi-station testing method for automotive sensors as described in claim 5, characterized in that: The process of integrating parameter configuration adjustment, test sequence priority, and production characteristics to generate an adaptive testing strategy includes establishing a second multi-objective optimization function, using parameter configuration adjustment and test sequence priority as the initial solution of the second multi-objective optimization function, using production characteristics as the constraint conditions of the second multi-objective optimization function, and using the second multi-objective optimization function to solve for the optimal test parameter configuration and test station scheduling sequence to obtain the adaptive testing strategy.

7. The multi-station testing method for automotive sensors as described in claim 1, characterized in that: The test station performs collaborative testing on the automotive sensors according to an adaptive testing strategy. The specific steps are as follows. Convert adaptive testing strategies into test parameter configuration instructions and test sequence instructions; The testing station adjusts the parameter configuration according to the test parameter configuration instructions, and the mechanical conveying device adjusts the test sequence priority according to the test sequence instructions.

8. The multi-station testing method for automotive sensors as described in claim 4, characterized in that: The specific steps for calculating the parameter configuration adjustment amount and test order priority using the first multi-objective optimization function are as follows: Establish a first multi-objective optimization function, take the configuration parameters corresponding to the test process as the initial solution of the first multi-objective optimization function, and take the test pass probability and the confidence level of the test pass probability as the input of the first multi-objective optimization function; The status data of the test station is collected as the constraint condition of the first multi-objective optimization function, and Pareto optimal solution set is generated iteratively through selection operation, crossover operation and mutation operation; Extract parameter configuration adjustments and test order priorities from the Pareto optimal solution set.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the multi-station testing method for automotive sensors as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the multi-station testing method for automotive sensors as described in any one of claims 1 to 8.

Images