Printed circuit board performance testing method, device, equipment and storage medium
Through deep learning technology, combining discrete candidate region division, finite element analysis and least squares support vector regression prediction model, the problem of time-consuming and labor-consuming performance detection of traditional printed circuit boards is solved, and efficient performance detection and virtual pattern design are achieved. Improves design accuracy and development efficiency.
Patent Information
- Application Number
- CN202410647221.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-23
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2044-05-23
AI Technical Summary
Traditional printed circuit board performance detection methods are time-consuming and costly, making it difficult to provide comprehensive performance analysis and optimization suggestions in the early stage of design, resulting in expensive modifications in the later stage of design.
Using deep learning technology, the accuracy of printed circuit board performance detection and the accuracy of virtual pattern design are achieved through discrete candidate region division, finite element analysis, least squares support vector regression prediction model and particle swarm algorithm.
It improves the accuracy of printed circuit board performance detection and the accuracy of virtual pattern design, realizes rapid iteration and improvement of design, and greatly improves development efficiency.
Smart Images

Figure CN118468794B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of deep learning technology, and in particular to a performance detection method, device, equipment and storage medium for a printed circuit board. Background Art
[0002] Printed circuit boards are an integral part of electronic products, and their performance directly affects the reliability and efficiency of the final product. With the rapid development of electronic technology, the performance requirements for printed circuit boards are becoming increasingly higher, including their electrical performance, thermal stability, mechanical strength, etc. Therefore, it is particularly important to develop an efficient and accurate printed circuit board performance testing method to ensure that the printed circuit board design meets strict performance standards and to foresee and solve possible performance problems in the early design stage.
[0003] However, traditional printed circuit board performance testing methods often rely on physical testing and prototyping, which is not only time-consuming and labor-intensive, but also costly. In addition, these methods are difficult to provide comprehensive performance analysis and optimization suggestions in the early stages of design, resulting in expensive design modifications when problems are discovered in the late stages of design. Summary of the invention
[0004] The present application provides a performance detection method, device, equipment and storage medium for a printed circuit board, thereby using deep learning technology to improve the performance detection accuracy and virtual pattern design accuracy of the printed circuit board.
[0005] The first aspect of the present application provides a performance detection method for a printed circuit board, the performance detection method for a printed circuit board comprising:
[0006] Dividing the target printed circuit board into discrete candidate areas to obtain a plurality of discrete candidate areas, and configuring virtual patterns for the plurality of discrete candidate areas to generate an initial virtual pattern design set;
[0007] Performing finite element analysis and performance simulation testing on the target printed circuit board based on the initial virtual pattern design set to obtain a performance test data set;
[0008] Inputting the performance test data set into a preset least squares support vector regression prediction model to perform performance analysis and probability distribution prediction to obtain performance probability distribution;
[0009] The virtual pattern design of the plurality of discrete candidate regions is optimized according to the performance probability distribution to obtain an optimal virtual pattern design set.
[0010] A second aspect of the present application provides a performance detection device for a printed circuit board, the performance detection device for a printed circuit board comprising:
[0011] A configuration module, configured to divide the target printed circuit board into discrete candidate areas to obtain a plurality of discrete candidate areas, and to configure virtual patterns for the plurality of discrete candidate areas to generate an initial virtual pattern design set;
[0012] A testing module, used for performing finite element analysis and performance simulation test on the target printed circuit board based on the initial virtual pattern design set to obtain a performance test data set;
[0013] A prediction module, used for inputting the performance test data set into a preset least squares support vector regression prediction model to perform performance analysis and probability distribution prediction to obtain performance probability distribution;
[0014] An optimization module is used to optimize the virtual pattern design of the plurality of discrete candidate areas according to the performance probability distribution to obtain an optimal virtual pattern design set.
[0015] The third aspect of the present application provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned printed circuit board performance detection method.
[0016] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned printed circuit board performance detection method.
[0017] In the technical solution provided by the present application, by dividing the target printed circuit board into discrete candidate areas and generating an initial virtual pattern design set at the beginning of the design, it is helpful to identify and solve potential design problems. Based on the initial virtual pattern design set, the target printed circuit board is subjected to finite element analysis and performance simulation test to obtain a variety of performance test data sets. The performance test data set is input into the preset least squares support vector regression prediction model to accurately analyze and predict the performance of the printed circuit board, accurately predict the performance of the printed circuit board under various working conditions, and further optimize the virtual pattern design of the discrete candidate areas based on the performance probability distribution to generate the optimal virtual pattern design set. This process can not only improve the performance of the printed circuit board, but also realize the rapid iteration and improvement of the design through optimization technologies such as particle swarm algorithm, greatly improving the development efficiency. Therefore, the present application adopts deep learning technology to improve the performance detection accuracy and virtual pattern design accuracy of the printed circuit board. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 A schematic diagram of an embodiment of a method for detecting the performance of a printed circuit board in an embodiment of the present application;
[0019] Figure 2 This is a schematic diagram of an embodiment of a performance detection device for a printed circuit board in an embodiment of the present application. DETAILED DESCRIPTION
[0020] The embodiments of the present application provide a method, device, equipment and storage medium for performance detection of a printed circuit board, thereby using deep learning technology to improve the performance detection accuracy and virtual pattern design accuracy of the printed circuit board.
[0021] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0022] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the performance detection method of the printed circuit board in the embodiment of the present application includes:
[0023] Step 101, dividing the target printed circuit board into discrete candidate areas to obtain multiple discrete candidate areas, and configuring virtual patterns for the multiple discrete candidate areas to generate an initial virtual pattern design set;
[0024] It is understandable that the execution subject of the present application may be a performance detection device for a printed circuit board, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.
[0025] Specifically, a three-dimensional model is performed on the target printed circuit board. A three-dimensional model of the printed circuit board is constructed using computer-aided design (CAD) technology, which reflects the physical and electrical characteristics of the circuit board, including the thickness of the board, the distribution of the copper layer, and the preset circuit path. The design space of the model is divided into discrete candidate regions. Each region is defined according to the functional area of the circuit board and the expected application environment. For example, the power supply area may require a higher current carrying capacity, while the signal processing area may require optimized electromagnetic interference (EMI) suppression. The design requirement information of the target printed circuit board is obtained, including specific requirements for functionality, reliability, and durability. The regional granularity of multiple initial candidate regions is optimized according to the circuit board design requirement information, and the size and shape of the region are adjusted to better match the actual use conditions and assembly restrictions of the circuit board. By adjusting each discrete candidate region, its performance and efficiency in actual applications are maximized. Virtual pattern initialization configuration is performed on multiple discrete candidate regions respectively, and one or more virtual pattern design schemes are created for each region. These design schemes are formulated according to the specific needs of the region, which may include optimization of circuit layout, adjustment of conductance path, or consideration of material selection to ensure that each scheme can provide the best performance in its corresponding region. The first virtual pattern design scheme of each discrete candidate area is converted into a set to generate an initial virtual pattern design set. The compatibility between different design schemes and the optimization of overall performance are ensured through the set conversion.
[0026] Step 102: Perform finite element analysis and performance simulation test on the target printed circuit board based on the initial virtual pattern design set to obtain a performance test data set;
[0027] Specifically, obtain the specific material properties of the target printed circuit board, including the chemical and physical properties of the substrate material, conductive layer material and any insulating layer used in the circuit board, such as thermal conductivity, resistivity, elastic modulus, etc. Define the boundary conditions of the circuit board, including various environmental parameters that the circuit board may encounter in actual use, such as temperature range, mechanical load, and electrical load. By defining boundary conditions, simulate the behavior of the circuit board under specific conditions. Construct a finite element model of the three-dimensional printed circuit board model according to the material properties and boundary conditions. Convert the three-dimensional printed circuit board model into a finite element mesh, and the density and accuracy of the mesh directly affect the results of the analysis. In the finite element model, each element is assigned corresponding material properties and corresponding boundary conditions are set. Perform static finite element analysis on the circuit board finite element model based on preset external load parameters, and the model is applied with preset external load parameters, such as pressure, tension or torque, to simulate the stress and strain response of the circuit board under physical load, help identify possible mechanical weaknesses, such as cracks or fracture risk areas, and obtain circuit board stress data. Perform thermal distribution simulation on the circuit board finite element model based on preset temperature condition parameters. Evaluate and predict the thermal stability and thermal stress distribution of the circuit board under the change of operating temperature to ensure that the circuit board can maintain its performance within its operating temperature range. Through simulation, obtain the thermal response data of the circuit board. Perform electrical performance simulation test on the circuit board. Simulate the transmission of electrical signals on the circuit board to evaluate the integrity and noise level of the signal. Electrical performance testing is the key to evaluating the functionality of the circuit board and ensure that the circuit board can meet the predetermined standards in terms of electrical performance. Through testing, obtain the circuit board signal response data, including signal delay, transmission rate and possible interference sources. Generate a performance test data set for the target printed circuit board based on the circuit board stress data, circuit board thermal response data and circuit board signal response data. The data set shows the performance of the circuit board under various conditions, ensuring that the final product can operate stably in various environments and meet all performance expectations.
[0028] Step 103: Input the performance test data set into a preset least squares support vector regression prediction model to perform performance analysis and probability distribution prediction to obtain performance probability distribution;
[0029] It should be noted that curve fitting is performed on the stress data, thermal response data and signal response data of the circuit board to find the most suitable mathematical model for the data so as to accurately describe the behavior and trend of the data. For example, for stress data, a stress curve is obtained by fitting a polynomial or exponential function; thermal response data and signal response data are also fitted by appropriate mathematical functions to obtain thermal response curves and signal response curves, respectively. Feature point identification is performed on the circuit board stress curve, circuit board thermal response curve and circuit board signal response curve, respectively. The key points of the curves are analyzed, such as extreme points, inflection points and other significant change points. After identifying these feature points, the stress feature set, thermal response feature set and signal response feature set are extracted from each curve. Feature normalization and vector encoding conversion are performed on the stress feature set, thermal response feature set and signal response feature set, respectively. Normalization ensures the consistency and comparability of data when comparing and combining different features, while vector encoding conversion converts these features into a format suitable for machine learning model processing. After conversion, stress feature vectors, thermal response feature vectors and signal response feature vectors are obtained. The stress eigenvector, thermal response eigenvector, and signal response eigenvector are fused to obtain a fused eigenvector. Data features from different sources are combined into a single representation to enhance the comprehensiveness and accuracy of model analysis. The fused eigenvector is input into the preset least squares support vector regression prediction model for performance analysis to obtain performance prediction values, which indicate the expected performance of the circuit board under given test conditions. In order to estimate the uncertainty and variability of the performance prediction value, a Monte Carlo algorithm-based method is used to predict the probability distribution of the performance prediction value. The Monte Carlo algorithm introduces randomness to the input data and runs the model multiple times to generate the distribution of the prediction results, obtain the expected value of a performance indicator, as well as its possible fluctuation range and distribution characteristics, and finally obtain the performance probability distribution of the target printed circuit board.
[0030] In this embodiment, the fused feature vector is input into a preset least squares support vector regression prediction model. The model enhances the expressiveness of data features by mapping the fused feature vector to a higher dimensional feature space. In a high-dimensional feature space, data features that may not be easily distinguished in a low-dimensional space can become more obvious, thereby improving the accuracy of the prediction. After mapping to the high-dimensional space, the model generates multiple feature mapping vectors, which represent various possible states and configurations of the data in the new space. The radial basis function kernel in the least squares support vector regression prediction model is used to calculate the similarity between the fused feature vector and each feature mapping vector. The radial basis function kernel uses the formula Calculate, where γ is the width parameter of the kernel, which controls the sensitivity of the function, and Represent the fusion feature vector and the i-th feature mapping vector respectively, and is the square of the Euclidean distance between the two vectors. The similarity between the fused feature vector and each feature vector in the data set is evaluated by the calculation formula. According to the calculated vector similarity, the feature vector with the highest similarity to the fused feature vector is selected from multiple feature mapping vectors as the target feature vector. Finally, the performance prediction value is calculated using the model function of the least squares support vector regression prediction model. The model function is expressed as , where f(x) represents the performance prediction value, N is the total number of feature map vectors, is the Lagrange multiplier, representing the weight of each vector in the model, is the vector similarity calculated by the radial basis function kernel, and b is the bias term of the model. Through this function, the contribution of all relevant feature vectors is weighted and summed to finally obtain a predicted value, which predicts the performance of the printed circuit board under given test conditions and feature configuration.
[0031] In this embodiment, the characteristic variable range corresponding to the performance prediction value is defined based on the Monte Carlo algorithm. The possible range of variation of all key characteristic variables, such as temperature, voltage, frequency and other factors under actual operating conditions is determined. By performing multiple random samplings within the predetermined characteristic variable range, the performance of the target printed circuit board is iteratively predicted based on the characteristic variable range. In each iteration, a value is randomly selected from the possible values of each characteristic variable, and then the performance prediction value of the printed circuit board is calculated using the prediction model based on the randomly selected value. This process is repeated to generate a large number of iterative prediction values, which reflect the prediction results under different conditions, thereby forming a comprehensive data set, which reflects the changes in the prediction results within the entire characteristic variable range. A large number of iterative prediction values are obtained and statistically analyzed for these values. The mean of all iterative prediction values is calculated, and the mean provides a central trend indicator of performance prediction, reflecting the average performance of the printed circuit board under the most likely conditions. The variance of these prediction values is calculated, and the variance is an important indicator for measuring the dispersion of the prediction value, reflecting the range and uncertainty of the performance under different conditions. At the same time, the confidence interval of the prediction value is calculated, and the confidence interval provides a statistical certainty, indicating the range in which the performance prediction value may fall under a specific probability level. Based on the calculated average predicted value, variance, and confidence interval, the performance probability distribution of the target printed circuit board is constructed. The probability distribution can usually be expressed as a probability density function or a cumulative distribution function, which reveals the average level of the expected performance of the printed circuit board and shows the possible fluctuations and uncertainties of its performance.
[0032] Step 104 : Optimizing virtual pattern designs for multiple discrete candidate regions according to the performance probability distribution to obtain an optimal virtual pattern design set.
[0033] Specifically, the global optimization goal and global constraints of each discrete candidate region are defined according to the performance probability distribution. According to the performance probability distribution of the printed circuit board in the prediction model, the goal of achieving the best performance and the constraints to be followed in the design process, such as electrical performance standards, physical size restrictions and environmental adaptability requirements, are set. The virtual pattern parameter range is calibrated for the discrete candidate regions. The design parameters of each region, such as the size, shape, layout, etc. of the pattern, are determined to have their variable ranges, which directly affect the function and performance of the pattern. The parameter range is determined based on the optimization goal and constraints to ensure that all design schemes are explored within an acceptable range. According to the calibrated virtual pattern parameter range, the pattern is adjusted for each discrete candidate region. Each region will generate multiple candidate virtual pattern design schemes according to its specific parameter range. The pattern is adjusted and optimized to adapt to specific performance requirements and manufacturing constraints, and each candidate scheme represents a possible solution. The candidate virtual pattern design schemes of each discrete candidate region are initialized by the preset particle swarm algorithm to generate an initial particle population. The particle swarm optimization algorithm is an optimization technology based on swarm intelligence, which finds the optimal solution through collaboration and information sharing between particles. Each particle represents a potential design solution. All particles in the population jointly explore the design space to find the best solution that meets the global optimization objectives and constraints. By optimizing and solving the initial particle population, the second virtual pattern design solution for each discrete candidate area is obtained. The performance of each particle (design solution) is evaluated, and iterative updates are performed according to the set optimization objectives to gradually approach the optimal solution. After multiple iterations, the final virtual pattern design solution for each area is determined, and these solutions are combined to form the optimal virtual pattern design set for the target printed circuit board.
[0034] In the embodiment of the present application, by dividing the target printed circuit board into discrete candidate areas and generating an initial virtual pattern design set at the beginning of the design, it is helpful to identify and solve potential design problems. Based on the initial virtual pattern design set, the target printed circuit board is subjected to finite element analysis and performance simulation test to obtain a variety of performance test data sets. The performance test data set is input into the preset least squares support vector regression prediction model to accurately analyze and predict the performance of the printed circuit board, accurately predict the performance of the printed circuit board under various working conditions, and further optimize the virtual pattern design of the discrete candidate areas based on the performance probability distribution to generate the optimal virtual pattern design set. This process can not only improve the performance of the printed circuit board, but also realize rapid iteration and improvement of the design through optimization techniques such as particle swarm algorithm, greatly improving development efficiency. Therefore, the present application adopts deep learning technology to improve the performance detection accuracy and virtual pattern design accuracy of the printed circuit board.
[0035] In a specific embodiment, the process of executing step 101 may specifically include the following steps:
[0036] (1) Performing three-dimensional modeling on the target printed circuit board to obtain a three-dimensional printed circuit board model;
[0037] (2) Dividing the design space of the 3D printed circuit board model into discrete candidate regions to obtain multiple initial candidate regions;
[0038] (3) obtaining circuit board design requirement information of the target printed circuit board, and optimizing the region granularity of multiple initial candidate regions according to the circuit board design requirement information to obtain multiple discrete candidate regions;
[0039] (4) Initializing and configuring virtual patterns for the plurality of discrete candidate regions respectively to obtain a first virtual pattern design scheme for each discrete candidate region;
[0040] (5) Performing set transformation on the first virtual pattern design scheme of each discrete candidate area to generate an initial virtual pattern design set.
[0041] Specifically, the target printed circuit board is three-dimensionally modeled, and detailed parameters are input according to the actual circuit board specifications and component layout through computer-aided design (CAD) software to create a three-dimensional model. The three-dimensional model shows the physical structure of the circuit board, including the location, connection points and routing of each electronic component, and can also simulate the interaction of these components in the actual environment. The design space of the three-dimensional printed circuit board model is divided into discrete candidate areas. The algorithm is used to automatically identify the areas with different functions on the circuit board, such as the power management area, the signal processing area, and the interface area, each of which may involve different design requirements and manufacturing technologies. For example, the power management area may require a larger copper foil area to handle high currents, while the signal processing area may require special wiring technology to reduce signal interference. Detailed design requirement information of the target printed circuit board is obtained, including the electrical performance standards, environmental adaptability requirements and physical size restrictions that the circuit board must meet. Based on the design requirement information, the regional granularity of the divided multiple initial candidate areas is optimized. The design parameters of each area are adjusted and optimized to ensure that they can provide the best performance while meeting all specified requirements. The optimization process may involve modifying the size, shape, or layout of the components within the region to accommodate specific functional requirements and manufacturing processes. Virtual pattern initialization configurations are performed for multiple discrete candidate regions. Each region is designed with one or more sets of virtual patterns based on its functional requirements, and these pattern designs take into account multiple aspects such as circuit layout, routing optimization, and thermal management. Each virtual pattern is designed to improve the performance and efficiency of the components in the region while ensuring compatibility with other regions and consistency of the overall circuit board. The first virtual pattern design for each discrete candidate region is collectively transformed to generate an initial virtual pattern design set. Through integration, the unity of the design and the seamless connection between each part are ensured to maximize the overall performance and reliability of the printed circuit board. For example, the motherboard of a high-performance computer needs to manage a large heat load while maintaining high signal integrity. In this case, the signal processing area may be designed with multiple layers of shielding and special signal paths to reduce interference, while the power management area may use larger copper layers and more heat dissipation components to handle heat.
[0042] In a specific embodiment, the process of executing step 102 may specifically include the following steps:
[0043] (1) Obtain the material properties of the target printed circuit board and define the boundary conditions of the target printed circuit board;
[0044] (2) Constructing a finite element model of the three-dimensional printed circuit board model according to material properties and boundary conditions to obtain a finite element model of the circuit board;
[0045] (3) Perform static finite element analysis on the circuit board finite element model based on preset external load parameters to obtain circuit board stress data;
[0046] (4) Performing thermal distribution simulation on the circuit board finite element model based on preset temperature condition parameters to obtain thermal response data of the circuit board;
[0047] (5) Conduct electrical performance simulation test on the circuit board finite element model to obtain circuit board signal response data;
[0048] (6) Generate a performance test data set of the target printed circuit board based on the circuit board stress data, the circuit board thermal response data, and the circuit board signal response data.
[0049] Specifically, the material properties of the target printed circuit board are obtained, including the type of circuit board substrate (such as FR-4), the material of the conductive layer (usually copper), the material of the insulating layer, and the details of any special coatings or treatments. The physical and chemical properties of these materials, such as thermal conductivity, resistivity, elastic modulus, and thermal expansion coefficient, are obtained. The boundary conditions that the circuit board may encounter in actual use are defined, including temperature range, mechanical loads such as bending or vibration, electrical loads, etc. Next, a finite element model is constructed for the three-dimensional printed circuit board model based on the material properties and boundary conditions. The finite element model is created in professional simulation software such as ANSYS or COMSOL, and the software converts the three-dimensional model into a finite element mesh that can be used for detailed analysis. This model reflects the geometry and complex material combination of the circuit board, and each part is modeled according to its material properties. Static finite element analysis is performed on the circuit board finite element model based on preset external load parameters. Based on the preset external load parameters, the model calculates the stress response of the circuit board under these loads. Through analysis, stress data of the circuit board is obtained to help identify possible stress concentration areas or potential structural failure points. Thermal distribution simulation is performed based on preset temperature condition parameters. Consider the operation of the circuit board at different temperatures, simulate its thermal response and the efficiency of thermal conduction. For example, if the circuit board is designed for high-power applications, the simulation may involve operation in an environment of up to 100 degrees Celsius, calculate its temperature distribution and possible hot spots during continuous operation, and evaluate the effectiveness of the heat dissipation solution and necessary improvements. At the same time, the finite element model of the circuit board is tested for electrical performance simulation, including signal integrity and electromagnetic compatibility testing of the circuit on the circuit board. By simulating how the current flows through the conductive path of the circuit board and observing the voltage drop and signal interference, the signal response data of the circuit board is obtained. The collected circuit board stress data, thermal response data, and signal response data are combined to form a comprehensive target printed circuit board performance test data set. Through a comprehensive analysis method, it is ensured that the printed circuit board can meet all technical and performance requirements during the design and production stages.
[0050] In a specific embodiment, the process of executing step 103 may specifically include the following steps:
[0051] (1) performing curve fitting on the circuit board stress data to obtain a circuit board stress curve, performing curve fitting on the circuit board thermal response data to obtain a circuit board thermal response curve, and performing curve fitting on the circuit board signal response data to obtain a circuit board signal response curve;
[0052] (2) Respectively identify characteristic points of the circuit board stress curve, the circuit board thermal response curve, and the circuit board signal response curve to obtain a stress feature set, a thermal response feature set, and a signal response feature set;
[0053] (3) performing feature normalization and vector encoding conversion on the stress feature set, thermal response feature set, and signal response feature set, respectively, to obtain a stress feature vector, a thermal response feature vector, and a signal response feature vector;
[0054] (4) Fusing the stress eigenvector, thermal response eigenvector, and signal response eigenvector to obtain a fused eigenvector;
[0055] (5) Input the fused feature vector into the preset least squares support vector regression prediction model for performance analysis to obtain the performance prediction value;
[0056] (6) Based on the Monte Carlo algorithm, the probability distribution of the performance prediction value is predicted to obtain the performance probability distribution of the target printed circuit board.
[0057] Specifically, curve fitting is performed on the circuit board stress data, thermal response data, and signal response data to help understand the basic trends of the data and provide information on the performance of the circuit board under various working conditions. Taking stress data as an example, these data are usually derived from tests under different levels of mechanical pressure or temperature conditions on the circuit board, and the data points include the response of stress to mechanical load or temperature changes. These points are used to generate a stress curve, and polynomial or exponential functions are usually selected as fitting models to better describe the relationship between stress and load. Similarly, thermal response data, which describes the physical behavior of the circuit board at different ambient temperatures, is also fitted into a curve to evaluate and predict the thermal stability and thermal efficiency of the circuit board in actual use. Signal response data, about the performance of the circuit board during high-frequency signal transmission, such as signal attenuation and noise level, is analyzed through curve fitting to determine the behavior of signal quality as it changes with frequency. After the curve fitting is completed, feature point identification is performed on these curves to extract key information from the curves, such as the maximum value, minimum value, inflection point, etc. of the curve, which are key indicators for evaluating the performance of the circuit board. A stress feature set, a thermal response feature set, and a signal response feature set are formed for the stress curve, thermal response curve, and signal response curve, respectively. These feature sets are normalized and vectorized to standardize the feature data to the same scale, eliminating the impact of different dimensions and making the subsequent machine learning model processing more efficient and accurate. The normalized feature points are encoded as feature vectors. The feature vectors are fused to form a comprehensive fused feature vector. The fusion process considers the correlation and importance between different features, and uses algorithms such as principal component analysis (PCA) or other multivariate analysis methods to ensure that the final feature vector can fully reflect the overall performance of the circuit board. The fused feature vector is input into the preset least squares support vector regression prediction model, and its data fitting ability is used to predict the performance of the circuit board based on the input feature vector. Least squares support vector regression predicts the performance value by establishing a method to find the best plane in a high-dimensional space, which is as close as possible to all the converted data points. In order to comprehensively evaluate the possible performance of the circuit board in future practical applications, the performance prediction value is predicted by probability distribution based on the Monte Carlo algorithm. By repeatedly randomly sampling the input of the prediction model and recording the prediction results each time, a probability distribution of predicted performance is generated. This distribution helps the system understand the potential performance of the circuit board under various uncertain future conditions and make more information-based design decisions based on this.
[0058] In a specific embodiment, the execution step inputs the fused feature vector into a preset least squares support vector regression prediction model for performance analysis, and the process of obtaining the performance prediction value may specifically include the following steps:
[0059] (1) Input the fused feature vector into the preset least squares support vector regression prediction model;
[0060] (2) Mapping the fused feature vector to a high-dimensional feature space through a least squares support vector regression prediction model, and generating multiple feature mapping vectors in the high-dimensional feature space;
[0061] (3) The radial basis function kernel in the least squares support vector regression prediction model is used to calculate the similarity between the fused feature vector and each feature mapping vector, and the vector similarity between the fused feature vector and each feature mapping vector is obtained, where the radial basis function kernel is:
[0062] ;
[0063] in, represents the vector similarity, represents the degree parameter of the radial basis function kernel, represents the fused feature vector, represents the i-th feature map vector, express and The square of the Euclidean distance between them;
[0064] (4) Select the vector with the highest similarity from multiple feature mapping vectors as the target feature vector based on vector similarity;
[0065] (5) The performance prediction value of the target feature vector is calculated by the model function in the least squares support vector regression prediction model to obtain the corresponding performance prediction value, where the model function is:
[0066] ;
[0067] in, represents the performance prediction value, N represents the total number of feature mapping vectors, represents the Lagrange multiplier, represents the vector similarity, Represents the bias term.
[0068] Specifically, the fused feature vector is input into the preset least squares support vector regression prediction model. The fused feature vector integrates multiple performance indicators of the circuit board, such as mechanical stress, thermal response and electrical signal characteristics. The fused feature vector is mapped to a high-dimensional feature space through the kernel function of the least squares support vector regression prediction model. The kernel function can transform data from the original feature space to a higher-dimensional space, which is usually mathematically easier to handle. Using the radial basis function kernel , γ is a pre-selected parameter that controls the width of the function, affecting the sensitivity and possibility of overfitting of the model. In high-dimensional space, each fused feature vector is mapped into multiple feature mapping vectors, which now represent the high-dimensional expression of the original data. The model calculates the similarity between the fused feature vector and each high-dimensional feature mapping vector. The radial basis function kernel is used to measure the proximity of two vectors in high-dimensional space. The calculation of similarity considers the square of the Euclidean distance between the vectors, and then transforms them through an exponential function to obtain a similarity value between 0 and 1. This similarity value reflects the relative position relationship between the two vectors in the feature space. By comparing the similarity between all feature mapping vectors and the fused feature vector, the model selects the feature mapping vector with the highest similarity as the target feature vector. The higher the similarity of the vector, the more accurately the corresponding performance prediction result can reflect the actual situation represented by the fused feature vector. The performance prediction value of the target feature vector is calculated through the model function in the least squares support vector regression prediction model, and the model function is used to calculate the performance prediction value of the target feature vector. , where f(x) represents the predicted performance value, N is the total number of feature map vectors, is the Lagrange multiplier associated with each vector (learned during model training), and b is the bias term. This function predicts performance by weighting the similarity of all feature vectors, where the weights are the Lagrange multipliers, reflecting the importance of each vector in the prediction model. For example, suppose the test data of a circuit board under different temperature and voltage conditions is processed to form a fused feature vector. After the vector is processed by the least squares support vector regression prediction model, it may be found that it has the highest similarity with a feature mapping vector under 150 degrees Celsius and 3.3 volts. Based on the highest similarity, the model will predict the performance of the circuit board under such extreme conditions, such as predicting the maximum withstand voltage or heat dissipation efficiency of the circuit board.
[0069] In a specific embodiment, the execution step performs probability distribution prediction on the performance prediction value based on the Monte Carlo algorithm to obtain the performance probability distribution of the target printed circuit board, which may specifically include the following steps:
[0070] (1) Define the characteristic variable range corresponding to the performance prediction value based on the Monte Carlo algorithm;
[0071] (2) Iteratively predict the performance of the target printed circuit board based on the range of characteristic variables to obtain multiple iterative prediction values;
[0072] (3) Calculating the mean of multiple iterative prediction values to obtain an average prediction value, calculating the variance of multiple iterative prediction values to obtain a variance, and calculating the confidence interval of multiple iterative prediction values to obtain a confidence interval;
[0073] (4) Construct the performance probability distribution of the target printed circuit board based on the mean prediction value, variance, and confidence interval.
[0074] Specifically, the characteristic variable range corresponding to the performance prediction value is defined based on the Monte Carlo algorithm. Determine the various working conditions that the printed circuit board may be subjected to, such as temperature, humidity, voltage, etc. The range of variation of these conditions is crucial for performance testing. For example, if a circuit board is used in an environment with an operating temperature ranging from -40°C to 125°C, then each temperature point within this range may become a value of a characteristic variable. Similarly, if the voltage may vary between 3V and 12V, the voltage within this range must also be taken into consideration. Perform iterative performance prediction on the target printed circuit board based on the characteristic variable range. The value of each characteristic variable is randomly selected by the Monte Carlo algorithm, such as randomly selecting a temperature and a voltage in each iteration. Assume that 1000 iterations are performed, and the temperature and voltage combination selected in each iteration is different, and then use the performance prediction model to calculate the expected performance in each case. For example, the first iteration may use -10°C and 5V, and the predicted performance value is 95% reliable; the second iteration may use 120°C and 11V, and the predicted performance value is 85% reliable. Perform statistical analysis on these 1000 iteration prediction values. Calculating the average prediction of these values gives the average performance of the board under all the considered operating conditions. For example, if the average performance reliability of all iterations is 90%, this represents the average expected reliability under all simulated operating conditions. At the same time, calculating the variance of these predictions helps understand the dispersion of the predictions. The larger the variance, the greater the performance variation under different operating conditions. For example, a variance of 25 means that the performance predictions fluctuate widely around the mean, indicating that the performance varies significantly under different conditions. Calculate the confidence intervals for these iterative predictions. The calculation of confidence intervals usually involves statistical methods, such as using t distribution or normal distribution, depending on the characteristics of the data distribution. Construct the performance probability distribution of the target printed circuit board based on the average prediction, variance, and confidence interval. The probability distribution plot shows the possible range of variation of the board performance and its probability under all simulated operating conditions. For example, use this data to draw a probability density function plot, where the x-axis represents the performance reliability and the y-axis represents the probability of the performance occurring. The view helps evaluate the performance of the board under the most unfavorable conditions, so as to make necessary design adjustments to enhance the reliability and robustness of the product.
[0075] In a specific embodiment, the process of executing step 104 may specifically include the following steps:
[0076] (1) Define the global optimization objectives and global constraints for multiple discrete candidate regions based on the performance probability distribution;
[0077] (2) calibrating the virtual pattern parameter ranges of multiple discrete candidate regions respectively to obtain the virtual pattern parameter range of each discrete candidate region;
[0078] (3) adjusting the virtual patterns of the plurality of discrete candidate regions according to the virtual pattern parameter range to obtain a candidate virtual pattern design scheme for each discrete candidate region;
[0079] (4) Initializing the population of candidate virtual pattern design schemes for each discrete candidate area using a preset particle swarm algorithm to obtain an initial particle population;
[0080] (5) Optimizing and solving the initial particle population according to the global optimization goal and global constraints to obtain a second virtual pattern design solution for each discrete candidate area;
[0081] (6) Generate an optimal virtual pattern design set of the target printed circuit board according to the second virtual pattern design scheme of each discrete candidate area.
[0082] Specifically, global optimization objectives and global constraints for multiple discrete candidate regions are defined based on the performance probability distribution. The performance probability distribution provides performance data of the circuit board under different working conditions, including the variability and reliability of key parameters such as temperature tolerance, electrical performance, and mechanical strength. For example, if the performance test shows that the performance of the circuit board deteriorates under high temperature conditions, the optimization goal may include improving the temperature resistance under these conditions. At the same time, global constraints may include material cost, manufacturing complexity, and environmental standards. The virtual pattern parameter range is calibrated for each discrete candidate region. The design parameters of each region, such as copper track width, aperture size, or layer spacing, are set within a specific range that takes into account material properties, manufacturing capabilities, and expected functional requirements. For example, for a high-speed signal transmission area, narrower copper tracks and smaller apertures may be required to reduce signal loss and crosstalk. According to the virtual pattern parameter range, virtual pattern adjustments are performed on multiple discrete candidate regions to generate candidate virtual pattern design solutions for each region. These design solutions maximize performance indicators by adjusting pattern parameters, such as minimizing resistance or maximizing heat dissipation efficiency. For example, software tools are used to simulate the thermal distribution under different pattern configurations to determine the optimal heat dissipation pattern layout. The particle swarm optimization algorithm is used to initialize and optimize the candidate virtual pattern design solutions for each discrete candidate area. Particle swarm optimization is an algorithm based on swarm intelligence. Each "particle" represents a potential design solution. The particle updates its position and velocity by tracking and imitating the most successful solution in history. For example, if the design solution represented by a particle produces an abnormally high heat dissipation efficiency, other particles will move closer to this design direction and gradually improve the design through an iterative process. Finally, the second virtual pattern design solutions for each discrete candidate area obtained by the particle swarm algorithm are integrated together to form the optimal virtual pattern design set for the target printed circuit board. This design set reflects the best balance of all key performance indicators, ensuring that the circuit board meets or exceeds the performance goals while meeting all design constraints.
[0083] The above describes the performance detection method of the printed circuit board in the embodiment of the present application. The following describes the performance detection device of the printed circuit board in the embodiment of the present application. Figure 2 In one embodiment of the present application, a performance detection device for a printed circuit board includes:
[0084] The configuration module 201 is used to divide the target printed circuit board into discrete candidate areas to obtain multiple discrete candidate areas, and to configure virtual patterns for the multiple discrete candidate areas to generate an initial virtual pattern design set;
[0085] A testing module 202 is used to perform finite element analysis and performance simulation test on a target printed circuit board based on an initial virtual pattern design set to obtain a performance test data set;
[0086] Prediction module 203, used to input the performance test data set into a preset least squares support vector regression prediction model to perform performance analysis and probability distribution prediction to obtain performance probability distribution;
[0087] The optimization module 204 is used to optimize the virtual pattern design of multiple discrete candidate areas according to the performance probability distribution to obtain an optimal virtual pattern design set.
[0088] Through the collaborative cooperation of the above-mentioned components, by dividing the target printed circuit board into discrete candidate areas and generating an initial virtual pattern design set at the beginning of the design, it is helpful to identify and solve potential design problems. Based on the initial virtual pattern design set, the target printed circuit board is subjected to finite element analysis and performance simulation testing to obtain a variety of performance test data sets. The performance test data set is input into the preset least squares support vector regression prediction model to accurately analyze and predict the performance of the printed circuit board, accurately predict the performance of the printed circuit board under various working conditions, and further optimize the virtual pattern design of the discrete candidate areas based on the performance probability distribution to generate the optimal virtual pattern design set. This process can not only improve the performance of the printed circuit board, but also realize rapid iteration and improvement of the design through optimization technologies such as particle swarm algorithm, greatly improving development efficiency. Therefore, this application uses deep learning technology to improve the performance detection accuracy and virtual pattern design accuracy of the printed circuit board.
[0089] The present application also provides a computer device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the steps of the performance detection method for the printed circuit board in the above-mentioned embodiments.
[0090] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the performance detection method for a printed circuit board.
[0091] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0092] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.
[0093] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for testing the performance of a printed circuit board, characterized in that: The performance detection method of the printed circuit board comprises: Dividing the target printed circuit board into discrete candidate areas to obtain a plurality of discrete candidate areas, and configuring virtual patterns for the plurality of discrete candidate areas to generate an initial virtual pattern design set; Performing finite element analysis and performance simulation testing on the target printed circuit board based on the initial virtual pattern design set to obtain a performance test data set; Input the performance test data set into a preset least squares support vector regression prediction model for performance analysis and probability distribution prediction to obtain a performance probability distribution; specifically including: defining a characteristic variable range corresponding to a performance prediction value based on a Monte Carlo algorithm; performing iterative performance prediction on the target printed circuit board based on the characteristic variable range to obtain a plurality of iterative prediction values; performing mean calculation on the plurality of iterative prediction values to obtain an average prediction value, performing variance calculation on the plurality of iterative prediction values to obtain a variance, and performing confidence interval calculation on the plurality of iterative prediction values to obtain a confidence interval; constructing a performance probability distribution of the target printed circuit board according to the average prediction value, the variance and the confidence interval; The virtual pattern design of the multiple discrete candidate areas is optimized according to the performance probability distribution to obtain an optimal virtual pattern design set; specifically comprising: defining a global optimization target and a global constraint condition for the multiple discrete candidate areas according to the performance probability distribution; calibrating the virtual pattern parameter range of the multiple discrete candidate areas respectively to obtain a virtual pattern parameter range for each discrete candidate area; adjusting the virtual patterns of the multiple discrete candidate areas according to the virtual pattern parameter range to obtain a candidate virtual pattern design scheme for each discrete candidate area; initializing a population of candidate virtual pattern design schemes for each discrete candidate area by a preset particle swarm algorithm to obtain an initial particle population; optimizing and solving the initial particle population according to the global optimization target and the global constraint condition to obtain a second virtual pattern design scheme for each discrete candidate area; and generating an optimal virtual pattern design set for the target printed circuit board according to the second virtual pattern design scheme for each discrete candidate area.
2. The performance testing method of a printed circuit board according to claim 1, characterized in that: The step of dividing the target printed circuit board into discrete candidate areas to obtain a plurality of discrete candidate areas, and performing virtual pattern configuration on the plurality of discrete candidate areas to generate an initial virtual pattern design set includes: Performing three-dimensional modeling on the target printed circuit board to obtain a three-dimensional printed circuit board model; Dividing the design space of the three-dimensional printed circuit board model into discrete candidate regions to obtain a plurality of initial candidate regions; Acquire circuit board design requirement information of the target printed circuit board, and optimize the region granularity of the multiple initial candidate regions according to the circuit board design requirement information to obtain multiple discrete candidate regions; Initializing and configuring virtual patterns for the plurality of discrete candidate regions respectively to obtain a first virtual pattern design scheme for each discrete candidate region; The first virtual pattern design scheme of each discrete candidate area is converted into a set to generate an initial virtual pattern design set.
3. The performance testing method of a printed circuit board according to claim 2, characterized in that: The step of performing finite element analysis and performance simulation test on the target printed circuit board based on the initial virtual pattern design set to obtain a performance test data set includes: Acquiring material properties of the target printed circuit board and defining boundary conditions of the target printed circuit board; Constructing a finite element model of the three-dimensional printed circuit board model according to the material properties and the boundary conditions to obtain a finite element model of the circuit board; Performing static finite element analysis on the circuit board finite element model based on preset external load parameters to obtain circuit board stress data; Performing thermal distribution simulation on the circuit board finite element model based on preset temperature condition parameters to obtain thermal response data of the circuit board; Conducting electrical performance simulation test on the circuit board finite element model to obtain circuit board signal response data; A performance test data set of the target printed circuit board is generated according to the circuit board stress data, the circuit board thermal response data and the circuit board signal response data.
4. The performance testing method of a printed circuit board according to claim 3, characterized in that: The step of inputting the performance test data set into a preset least squares support vector regression prediction model to perform performance analysis and probability distribution prediction to obtain performance probability distribution includes: Performing curve fitting on the circuit board stress data to obtain a circuit board stress curve, performing curve fitting on the circuit board thermal response data to obtain a circuit board thermal response curve, and performing curve fitting on the circuit board signal response data to obtain a circuit board signal response curve; Respectively identifying characteristic points of the circuit board stress curve, the circuit board thermal response curve, and the circuit board signal response curve to obtain a stress feature set, a thermal response feature set, and a signal response feature set; Performing feature normalization and vector encoding conversion on the stress feature set, the thermal response feature set, and the signal response feature set, respectively, to obtain a stress feature vector, a thermal response feature vector, and a signal response feature vector; Performing feature vector fusion on the stress feature vector, the thermal response feature vector and the signal response feature vector to obtain a fused feature vector; Inputting the fused feature vector into a preset least squares support vector regression prediction model for performance analysis to obtain a performance prediction value; The performance prediction value is predicted by probability distribution based on the Monte Carlo algorithm to obtain the performance probability distribution of the target printed circuit board.
5. The performance testing method of a printed circuit board according to claim 4, characterized in that: The step of inputting the fused feature vector into a preset least squares support vector regression prediction model for performance analysis to obtain a performance prediction value includes: Inputting the fused feature vector into a preset least squares support vector regression prediction model; Mapping the fused feature vector to a high-dimensional feature space through the least squares support vector regression prediction model, and generating a plurality of feature mapping vectors in the high-dimensional feature space; The similarity between the fused feature vector and each feature mapping vector is calculated by using the radial basis function kernel in the least squares support vector regression prediction model to obtain the vector similarity between the fused feature vector and each feature mapping vector, wherein the radial basis function kernel is: ; in, represents the vector similarity, represents the degree parameter of the radial basis function kernel, represents the fused feature vector, represents the i-th feature map vector, express and The square of the Euclidean distance between them; According to the vector similarity, select the one with the highest similarity from the multiple feature mapping vectors as the target feature vector; The performance prediction value of the target feature vector is calculated by the model function in the least squares support vector regression prediction model to obtain the corresponding performance prediction value, wherein the model function is: ; in, represents the performance prediction value, N represents the total number of feature mapping vectors, represents the Lagrange multiplier, represents the vector similarity, Represents the bias term.
6. A performance detection device for a printed circuit board, characterized in that: The performance detection device of the printed circuit board comprises: A configuration module, configured to divide the target printed circuit board into discrete candidate areas to obtain a plurality of discrete candidate areas, and to configure virtual patterns for the plurality of discrete candidate areas to generate an initial virtual pattern design set; A testing module, used for performing finite element analysis and performance simulation test on the target printed circuit board based on the initial virtual pattern design set to obtain a performance test data set; A prediction module is used to input the performance test data set into a preset least squares support vector regression prediction model to perform performance analysis and probability distribution prediction to obtain a performance probability distribution; specifically comprising: defining a characteristic variable range corresponding to a performance prediction value based on a Monte Carlo algorithm; performing iterative performance prediction on the target printed circuit board based on the characteristic variable range to obtain a plurality of iterative prediction values; performing mean calculation on the plurality of iterative prediction values to obtain an average prediction value, performing variance calculation on the plurality of iterative prediction values to obtain a variance, and performing confidence interval calculation on the plurality of iterative prediction values to obtain a confidence interval; and constructing a performance probability distribution of the target printed circuit board according to the average prediction value, the variance and the confidence interval; An optimization module is used to optimize the virtual pattern design of the multiple discrete candidate areas according to the performance probability distribution to obtain an optimal virtual pattern design set; specifically comprising: defining the global optimization target and global constraint conditions of the multiple discrete candidate areas according to the performance probability distribution; calibrating the virtual pattern parameter range of the multiple discrete candidate areas respectively to obtain the virtual pattern parameter range of each discrete candidate area; adjusting the virtual pattern of the multiple discrete candidate areas according to the virtual pattern parameter range to obtain a candidate virtual pattern design scheme for each discrete candidate area; initializing the population of candidate virtual pattern design schemes for each discrete candidate area by a preset particle swarm algorithm to obtain an initial particle population; optimizing and solving the initial particle population according to the global optimization target and the global constraint conditions to obtain a second virtual pattern design scheme for each discrete candidate area; generating the optimal virtual pattern design set of the target printed circuit board according to the second virtual pattern design scheme of each discrete candidate area.
7. A computer device, characterized in that: The computer device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory to enable the computer device to execute the performance detection method for a printed circuit board according to any one of claims 1 to 5.
8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the performance detection method of the printed circuit board as described in any one of claims 1 to 5 is implemented.
Citation Information
Patent Citations
PCB reliability prediction and method and computer equipment
CN108197385A
Flexible PCB structure design method based on wet chemical etching joint simulation
CN112966467A
Cited By
Circuit board performance detection system and method based on artificial intelligence
CN122432544A
