Model selection method and device of electronic component, electronic equipment and storage medium

Abstract

The invention relates to the technical field of integrated circuits, and provides an electronic component type selection method and device, electronic equipment and a storage medium. A component parameter library is obtained and multi-dimensional normalization processing is performed on the component parameter library to obtain a standardized parameter tensor, so that principal component analysis is performed on the standardized parameter tensor to obtain a dimensionality reduction feature space, and spectral clustering analysis is performed on the dimensionality reduction feature space to obtain a plurality of functional equivalence classes. And performing reliability degradation modeling on the function equivalence class to obtain an MTBF prediction matrix, performing multi-target optimization on the MTBF prediction matrix to obtain a Pareto leading-edge solution set, and finally performing fuzzy decision on the Pareto leading-edge solution set to obtain a model selection scheme of the electronic component. According to the method, a comprehensive component type selection process is formed through multi-dimensional normalization processing, dimension reduction feature space analysis, spectral clustering analysis, reliability modeling, multi-objective optimization and fuzzy decision, so that components meeting requirements are accurately and efficiently selected.

Description

Technical Field

[0001] The present application relates to the field of integrated circuit technology, and in particular, to a method and device for selecting electronic components, an electronic device, and a storage medium. Background Art

[0002] With the continuous development of electronic product technology and the increasing variety and specifications of components, reasonable component selection has become particularly important. If the selection is improper, it will not only affect the function realization of the product, but may also lead to system instability, increased failure rate, and even potential safety hazards. Therefore, accurate and efficient component selection methods play an extremely important role in electronic engineering.

[0003] Existing electronic component selection methods usually rely on single technical parameters and static performance evaluations, and do not fully consider the long-term reliability and adaptability of components. For example, many methods only focus on the current performance indicators of components, such as power consumption, speed, size, etc., but ignore the reliability performance of components during long-term use, such as failure rate, lifespan, etc. Therefore, components may not be able to work stably in actual applications for a long time, increasing the cost of system maintenance and fault repair. Summary of the Invention

[0004] In view of this, the present application provides a method and device for selecting electronic components, an electronic device, and a storage medium to solve the problem of inaccurate selection of electronic components.

[0005] The first aspect of the present application provides a method for selecting electronic components, the method including:

[0006] Obtaining a component parameter library and performing multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor;

[0007] Performing principal component analysis on the standardized parameter tensor to obtain a dimensionality-reduced feature space;

[0008] Performing spectral clustering analysis on the dimensionality-reduced feature space to obtain function equivalence classes;

[0009] Performing reliability degradation modeling on the function equivalence classes to obtain an MTBF prediction matrix;

[0010] Performing multi-objective optimization on the MTBF prediction matrix to obtain a Pareto front solution set;

[0011] Performing fuzzy decision-making on the Pareto front solution set to obtain a selection scheme for electronic components.

[0012] In an alternative embodiment, the performing multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor includes:

[0013] Extract the parameter set of the target component from the component parameter library according to the preset test conditions to obtain the three-dimensional data set of each target component under the test conditions;

[0014] Identify the functional relationship characteristics of various parameter subsets in the three-dimensional data set to obtain the corresponding normalization methods for various parameter subsets;

[0015] Perform multi-dimensional normalization processing on the parameter subsets according to the normalization methods to obtain a standardized parameter value set;

[0016] Construct a data structure based on the target component, the test conditions, and the standardized parameter value set to generate a standardized parameter tensor corresponding to each target component.

[0017] In an alternative embodiment, the performing principal component analysis on the standardized parameter tensor to obtain a dimensionality-reduced feature space includes:

[0018] Perform an unfolding process on the standardized parameter tensor to obtain a two-dimensional matrix;

[0019] Calculate the covariance matrix of the two-dimensional matrix to obtain a covariance matrix;

[0020] Perform eigenvalue decomposition on the covariance matrix to obtain a first set of eigenvalue and eigenvector combinations;

[0021] Calculate the contribution rate of each eigenvalue in the first set of eigenvalue and eigenvector combinations according to the eigenvalues, and screen the eigenvectors in the first set of eigenvalue and eigenvector combinations according to the contribution rate and a preset contribution threshold to obtain a set of principal component vectors whose contribution rate is greater than or equal to the contribution threshold;

[0022] Perform a projection process on the standardized parameter tensor according to the set of principal component vectors to obtain the dimensionality-reduced feature space.

[0023] In an alternative embodiment, the performing spectral clustering analysis on the dimensionality-reduced feature space to obtain functional equivalence classes includes:

[0024] Calculate the similarity between the data in the dimensionality-reduced feature space to obtain a similarity matrix between the target components;

[0025] Perform a Gaussian kernel function process on the similarity matrix to obtain a weighted similarity matrix;

[0026] Construct a Laplacian matrix according to the weighted similarity matrix, and perform eigenvalue decomposition on the Laplacian matrix to obtain a second set of eigenvalue and eigenvector combinations;

[0027] Perform clustering processing on the target component according to the second eigenvalue and the set of eigenvector combinations to obtain the functional equivalence class.

[0028] In an alternative embodiment, the reliability degradation modeling of the functional equivalence class to obtain the MTBF prediction matrix includes:

[0029] Extract the working environment parameters and stress factor data for each target component in the functional equivalence class, and construct a working environment model for each target component according to the working environment parameters and the stress factor data;

[0030] Perform temperature stress analysis and Arrhenius formula calculation according to the working environment model to obtain the life coefficient of each target component in the functional equivalence class under a preset specific environment;

[0031] Perform reliability degradation modeling on the target components in each functional equivalence class according to the life coefficient to obtain the failure rate set of each target component under different working environments;

[0032] Construct a failure distribution model according to the failure rate set and the preset failure mode of the corresponding target component;

[0033] Perform mean time between failures prediction on each target component according to the failure distribution model to obtain the MTBF prediction matrix of each functional equivalence class.

[0034] In an alternative embodiment, the multi-objective optimization of the MTBF prediction matrix to obtain the Pareto front solution set includes:

[0035] Normalize the reference values in the MTBF prediction matrix to obtain a set of target values;

[0036] Perform multi-objective optimization modeling on the set of target values to construct an objective function;

[0037] Optimize the objective function through a preset non-dominated sorting genetic algorithm to obtain a number of non-dominated solutions;

[0038] Generate the Pareto front solution set according to the non-dominated solutions.

[0039] In an alternative embodiment, the fuzzy decision-making on the Pareto front solution set to obtain the selection scheme of electronic components includes:

[0040] Construct a membership function according to the non-dominated solutions in the Pareto front solution set to obtain the membership values of each non-dominated solution under different preset objectives;

[0041] Perform a weighted process on the membership degree values to obtain weighted membership degree values;

[0042] According to the weighted membership degree values, calculate the comprehensive satisfaction degree for each non-dominated solution to obtain the comprehensive satisfaction degree index of each non-dominated solution;

[0043] Rank the non-dominated solutions in the Pareto front solution set according to the comprehensive satisfaction degree index to obtain the optimal non-dominated solution corresponding to the maximum comprehensive satisfaction degree index;

[0044] Obtain the selection scheme of the electronic component according to the target component parameters in the optimal non-dominated solution.

[0045] The second aspect of the present application provides a selection device for electronic components, and the device includes:

[0046] A parameter tensor module, configured to obtain a component parameter library and perform multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor;

[0047] A feature dimension reduction module, configured to perform principal component analysis on the standardized parameter tensor to obtain a dimension-reduced feature space;

[0048] A function equivalence module, configured to perform spectral clustering analysis on the dimension-reduced feature space to obtain function equivalence classes;

[0049] A prediction matrix module, configured to perform reliability degradation modeling on the function equivalence classes to obtain an MTBF prediction matrix;

[0050] A front solution set module, configured to perform multi-objective optimization on the MTBF prediction matrix to obtain a Pareto front solution set;

[0051] A selection scheme module, configured to perform fuzzy decision-making on the Pareto front solution set to obtain a selection scheme of the electronic component.

[0052] The third aspect of the present application provides an electronic device, and the electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above-mentioned selection method for electronic components are implemented.

[0053] The fourth aspect of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned selection method for electronic components are implemented.

[0054] In summary, the present application at least includes the following beneficial technical effects:

[0055] 1. By building a reliability degradation model and an MTBF prediction matrix, the lifespan of components can be estimated, which helps to comprehensively consider their long-term stability during component selection, especially applicable to applications with high reliability requirements.

[0056] 2. Adopting a multi-objective optimization method based on the Pareto front solution set and non-dominated sorting genetic algorithm can find the best balance among multiple objectives, ensuring that the selected component solution achieves optimized results in multiple dimensions such as performance, cost, and reliability, and avoiding the limitations that may be brought by single-objective optimization.

[0057] 3. By comprehensively ranking the satisfaction of different solutions through a fuzzy decision-making method, the final selected component solution better meets the user's demand preferences and provides more flexible decision-making support.

[0058] 4. Considering multiple factors such as the performance, cost, lifespan, and reliability of components, and providing the optimal solution through an optimization algorithm, it can significantly improve the decision-making efficiency and quality during component selection and reduce the interference of subjective human factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0060] Figure 1 is a flowchart of a method for selecting electronic components provided by an embodiment of the present application;

[0061] Figure 2 is a functional module diagram of a device for selecting electronic components provided by an embodiment of the present application;

[0062] Figure 3 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, rather than all embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0064] Such as Figure 1As shown in the figure, it is a flowchart of the method for selecting electronic components provided by the embodiments of the present application. The method for selecting electronic components provided by the embodiments of the present application includes the following steps.

[0065] Step S1: Obtain a component parameter library and perform multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor.

[0066] Among them, the component parameter library includes but is not limited to various types of components and their corresponding various attributes, such as electrical parameters, mechanical parameters, and environmental parameters, etc. Component parameters are the basic descriptions of component performance, constituting the basic data of components, usually stored in the Datasheet of electronic components, or can be obtained through other relevant documents or databases (such as data provided by suppliers, experimental data, etc.).

[0067] Extract the parameter set of the target component from the component parameter library according to the preset test conditions. Among them, the preset test conditions include but are not limited to external conditions such as temperature, humidity, frequency, voltage, etc. that will affect the performance of the component.

[0068] Assume that the test conditions include temperature (such as 25°C, 50°C, 75°C) and voltage (such as 5V, 12V, 24V). The performance data of each component will be different under these different conditions. The parameter data under each test condition constitutes a dimension. By testing under different conditions or according to existing test data, extract the performance parameters (i.e., the parameter set) of each target component under different test conditions to obtain a three-dimensional data set. Among them, the three-dimensional data set includes component numbers, parameter types, and test conditions.

[0069] It should be understood that the three-dimensional data set represents the parameter values under the test conditions. Since different parameters may have different effects on the test results, especially for those parameters with non-linear relationships, such as temperature coefficients, resistance changes, etc., it is necessary to perform feature recognition on these parameters. Function relationship feature recognition determines how each type of parameter should be normalized by analyzing the change trend of the same type of parameters of the same component under different test conditions. For some non-linear parameters, piecewise normalization may be required, and they are processed separately for different temperature or frequency segments, while for parameters with linear relationships, conventional normalization methods can be used.

[0070] For parameters with linear relationships, use the standard normalization method for processing. The formula is as follows:

[0071]

[0072] Among them, is the normalized parameter value, x jkis the parameter value of the original j-th component under the k-th test condition, max(X k ) and min(X k ) are the maximum and minimum values of all parameter values under the k-th test condition respectively.

[0073] For parameters with non-linear relationships, a piecewise normalization method is used for processing, and the formula is as follows:

[0074]

[0075] Among them, is the normalized parameter value, x jk is the parameter value of the original j-th component under the k-th test condition, max(X k ) and min(X k ) are the maximum and minimum values of all parameter values under the k-th test condition respectively, is the standard deviation, and is the mean value.

[0076] By normalization processing, the value of each parameter is converted into a dimensionless numerical value, and the range is usually restricted between 0 and 1 to ensure the comparability of parameters under different test conditions. Among them, for each piecewise region in the piecewise normalization processing, separate normalization is carried out for each piece to ensure the consistency of non-linear parameters within each region. Finally, the parameters after normalization processing are summarized to form a set of standardized parameter values.

[0077] According to each target component number, parameter type, and test condition, a three-dimensional data structure (i.e., a standardized parameter tensor) is constructed. Among them, the standardized parameter tensor contains all the standardized parameter values of each target component under different test conditions. The data structure of the standardized parameter tensor is that the first dimension is the component number, the second dimension is the parameter type, and the third dimension is the test condition.

[0078] Step S2: Perform principal component analysis on the standardized parameter tensor to obtain a reduced-dimensional feature space.

[0079] After obtaining the standardized parameter tensor, to meet the input data structure required for principal component analysis, the standardized parameter tensor needs to be converted from a three-dimensional structure to a two-dimensional matrix. Assume that the dimension of the standardized parameter tensor is n×p×t, where n is the number of components, p is the number of parameter types, and t is the number of test conditions. This three-dimensional tensor is expanded into a two-dimensional matrix M n×(p×t) . Each row of the two-dimensional matrix is a vector of the standardized parameter values of a component under multiple test conditions, and each column of the two-dimensional matrix corresponds to the standardized value of a certain parameter (such as voltage, current, etc.) of each component under different conditions.

[0080] It should be understood that the covariance matrix is the core in principal component analysis. The covariance matrix reveals the correlation between different dimensions of the data. Through the covariance matrix, the variation relationships of each feature (i.e., parameter) under different components and test conditions can be understood, providing a data basis for principal component analysis. The covariance matrix is calculated using a two-dimensional matrix through the following formula:

[0081]

[0082] where is the covariance matrix, with a size of (p×t)×(p×t). is the transpose of matrix M. M is a two-dimensional matrix with a size of n×(p×t), and each row represents the standardized parameters of a component. n is the number of components. The covariance matrix describes the correlation of component parameters under different test conditions. The larger the covariance, the stronger the relationship between the two parameters, and vice versa. In principal component analysis, the goal is to extract the features that best represent the data variation from the covariance matrix through eigenvalue decomposition.

[0083] By performing eigenvalue decomposition on the covariance matrix, eigenvalues and eigenvectors are extracted from the covariance matrix. The eigenvalues represent the amount of variance explained by each principal component, while the eigenvectors represent the directions of each principal component. Finally, the eigenvalues and eigenvectors are associated, and the associated data is aggregated to form the first set of eigenvalue and eigenvector combinations. The eigenvalue decomposition formula is as follows:

[0084] Cv i =λ i v i

[0085] where C is the covariance matrix. v i is the i-th eigenvector, representing the direction of the principal component. λ i is the i-th eigenvalue, representing the variance explained by the i-th principal component, with a scalar size. Each eigenvector corresponds to a principal component, and its eigenvalue represents the importance of the principal component. A larger eigenvalue indicates that the principal component can explain more variance of the data, so it is the most important principal component.

[0086] Furthermore, by calculating the contribution rate of each eigenvalue in the first set of eigenvalue and eigenvector combinations, the number of principal components to be retained is determined. Among them, the contribution rate represents the proportion of each principal component in the data variance. By setting a preset contribution threshold, those principal components that can explain most of the variance can be selected, so as to achieve the purpose of dimensionality reduction. The formula for calculating the contribution rate is as follows:

[0087]

[0088] where λ′ iis the contribution rate of the i-th eigenvalue, λ i is the i-th eigenvalue, is the sum of all eigenvalues. When the cumulative contribution rate reaches a preset contribution threshold (e.g., 85%), stop retaining more principal components, which can effectively reduce the data dimension while retaining as much information as possible.

[0089] Finally, project the standardized parameter tensor according to the selected set of principal component vectors to obtain a reduced-dimensional feature space, thereby mapping the original multi-dimensional data to a low-dimensional space (i.e., the reduced-dimensional feature space), simplifying the complexity of the data while retaining the main features of the data.

[0090] Step S3: Perform spectral clustering analysis on the reduced-dimensional feature space to obtain functional equivalence classes.

[0091] It should be understood that spectral clustering analysis constructs a graph representing the similarity between data, and uses the Laplacian matrix of this graph for clustering, thereby finding groups of components with similarity or the same function in the data. In the embodiments of the present application, spectral clustering analysis is based on the data in the reduced-dimensional feature space, and finally obtains functional equivalence classes by clustering the target components.

[0092] Specifically, the first step of spectral clustering is to calculate the similarity between components. Usually, Euclidean distance, Manhattan distance or other appropriate distance metrics are used to measure the similarity degree of two components in the feature space. A matrix representing the similarity between target components is obtained through similarity calculation. The calculation formula of the similarity matrix is as follows:

[0093] S ij = exp(-γ||x i -x j || 2 )

[0094] where s ij is the similarity between the i-th component and the j-th component. x j and x i respectively represent the vector representations of the i-th and j-th components in the reduced-dimensional feature space. ||x i -x j || is the Euclidean distance between the i-th and j-th components. γ is a parameter used to control the similarity bandwidth (i.e., how to scale the distance). A larger γ will make the values of the similarity matrix approach zero, and a smaller γ will make the similarity values larger.

[0095] In spectral clustering, the similarity matrix is often processed using the Gaussian kernel function. The purpose is to adjust the similarity between different components, increasing the weights between components with greater similarity so that they are more likely to be clustered into the same class. The similarity matrix processed by the Gaussian kernel function is weighted and reflects the true similarity relationship between different components in the data. The Gaussian kernel function formula is as follows:

[0096]

[0097] where W ij represents the weighted similarity between the i-th component and the j-th component. x j and x i respectively represent the vector representations of the i-th and j-th components in the reduced-dimensional feature space. ||x i -x j || represents the Euclidean distance between the i-th and j-th components. σ is the bandwidth parameter of the kernel function, which controls the width of the similarity function. Through the Gaussian kernel function, the weights between components with higher similarity can be effectively amplified, and the weights between distant components can be reduced, which helps to improve the clustering effect and ensure that components with similar characteristics are clustered into the same class.

[0098] In spectral clustering, the Laplacian matrix is constructed based on the weighted similarity matrix and represents the connection relationships in the graph. Decomposing the eigenvalues of the Laplacian matrix helps to reveal the internal structure in the data. Through eigenvalue decomposition, the eigenvectors with the largest variances can be extracted, and these eigenvectors determine the classification of data points in space. The Laplacian matrix and its eigenvalue decomposition formula are as follows:

[0099]

[0100] where L is the Laplacian matrix. D is the degree matrix, which is a diagonal matrix, where represents the degree of the i-th component (i.e., the sum of the similarities between this component and other components). S is the similarity matrix. v i is the i-th eigenvector, representing the distribution of the data in a certain principal component direction. λ i is the i-th eigenvalue, representing the variance size corresponding to the eigenvector. The Laplacian matrix and its eigenvalue decomposition help to extract the main directions of the data, thereby obtaining the second eigenvalue and eigenvector combination set composed of eigenvalues and corresponding eigenvectors.

[0101] Referring to the process of calculating the contribution rate and screening the principal component vectors in step S2, vector screening is performed on the second eigenvalue and eigenvector combination set to obtain a number of the most representative eigenvectors. And these eigenvectors are composed into a secondary feature space. Then, according to a preset clustering algorithm (such as K-means clustering, hierarchical clustering, etc.), clustering analysis is performed on the eigenvectors in the secondary feature space, and all target components are grouped according to similar functions through the clustering analysis to obtain function equivalence classes. Each cluster in the function equivalence class corresponds to a group of target components with similar functions.

[0102] Step S4: Perform reliability degradation modeling on the function equivalence classes to obtain an MTBF prediction matrix.

[0103] It should be understood that in the process of selecting electronic components, reliability is a very crucial factor, especially in ensuring product performance and long-term stability. Mean Time Between Failures (MTBF) is an important indicator for measuring the reliability of electronic components, which is used to predict the average time between failures of components in a specific working environment. Reliability degradation modeling is used to provide reliability predictions (i.e., MTBF prediction matrix) for the target components in each function equivalence class to help select the most reliable components.

[0104] Specifically, the performance and lifespan of target components are affected by various environmental factors, mainly including factors such as temperature, humidity, and voltage. These environmental factors can cause components to age or be damaged. Therefore, for each target component in the function equivalence class, parameter data related to its working environment (i.e., working environment parameters) is extracted, such as working temperature, humidity, load voltage, frequency, etc. In addition, stress is also a main factor that accelerates the aging process of components and affects their lifespan. Therefore, data related to stress (i.e., stress factor data) is also obtained for each target component, such as voltage stress, temperature stress, etc. According to the extracted environmental parameters and stress factor data, a working environment model for each component is constructed, and the working environment model reflects the performance changes and reliability changes of the component under different working environments.

[0105] Temperature is one of the key factors affecting the lifespan of components. Especially in a high-temperature environment, the loss rate of components will increase significantly. Therefore, based on the working environment model, it is necessary to perform temperature stress analysis on the components and calculate their life coefficients through the Arrhenius formula. The Arrhenius formula is as follows:

[0106]

[0107] where, MTBF Tis the mean time between failures (MTBF) at temperature T. MTBF0 is the initial mean time between failures at the reference temperature T0. E a is the activation energy, a constant representing the aging of materials or components at different temperatures. k is the Boltzmann constant, a constant used to describe the relationship between thermal energy and temperature. T0 is the reference temperature. T is the target operating temperature. The Arrhenius formula helps calculate the life coefficient under different temperature conditions by considering the acceleration of temperature on the component life, and then determines the reliability of the component, and can estimate the life performance (i.e., life coefficient) of the component in a preset specific working environment.

[0108] In addition to the influence of temperature on the life of the target component, the failure rate of the target component will change during long-term operation. Further, the performance degradation of the target component under different working conditions is modeled according to a preset failure rate model, and finally a reliability index is generated. Among them, the failure rate describes the probability of the target component failing per unit time, and usually increases with the increase of the usage time. The failure rate model includes the early failure stage, the stable operation stage, and the aging failure stage. Common failure rate models are as follows:

[0109] Weibull distribution model: Used to describe the statistical distribution of component life and widely applied in reliability engineering.

[0110] Exponential distribution model: Used to describe the failure probability in the stable stage and often used in reliability assessment.

[0111] Normal distribution model: Used to model the failure time or characteristics, especially applicable when the failure time is symmetrically distributed.

[0112] Through the failure rate model, the failure rate of the target component at different time periods can be modeled, and the failure probability (i.e., failure rate set) of the component can be predicted based on comprehensive consideration of environmental factors, stress factors, and life coefficient.

[0113] After obtaining the failure rate of each target component, the next step is to construct a failure distribution model. The failure distribution model describes the failure probability distribution of components under different failure modes. The failure modes may include electrical failure, mechanical failure, temperature failure, etc. Based on the failure rate data and preset failure modes, a failure distribution model is constructed, and statistical models such as Weibull distribution and normal distribution are usually used to describe the probability and life of failure. Among them, the setting of the failure mode needs to measure the structure and working environment of the component. The failure modes of each target component are usually different and may involve problems such as short circuit, overheating, and wear of electronic components.

[0114] Finally, using the failure distribution model, the MTBF of each target component is predicted. MTBF is a key indicator in reliability analysis, representing the mean time between failures of the target component under specified operating conditions. The MTBF of each target component will be predicted based on the failure distribution model, and an MTBF prediction matrix will be generated. The MTBF prediction formula is usually calculated based on the Weibull distribution, and the formula is as follows:

[0115]

[0116] Among them, MTBF is the mean time between failures of the component. η is the scale parameter, representing the life scale of the component. Γ is the gamma function, used to calculate the statistic under a specific failure distribution. β is the shape parameter, reflecting the change of the failure mode. λ is the failure rate, representing the failure occurrence rate per unit time. Through the failure mode and distribution model of the target component, the mean time between failures (MTBF) of the target component under a specific working environment is calculated. An MTBF prediction matrix of the target components in each functional equivalence class can be obtained to help evaluate the reliability of the components.

[0117] Step S5: Perform multi-objective optimization on the MTBF prediction matrix to obtain the Pareto front solution set.

[0118] In the process of selecting electronic components, the optimization process involves not only a single objective, but multiple objectives need to be considered comprehensively, such as MTBF, cost, size, etc. Among them, multi-objective optimization aims to find the best compromise point among various objectives, forming the Pareto front solution set, which provides the best choice among different design requirements.

[0119] Specifically, in order to enable all objectives to be compared on a unified scale, it is necessary to normalize each objective value in the MTBF prediction matrix. The multiple objectives (such as cost, size, MTBF, etc.) included in the MTBF prediction matrix have different dimensions and numerical ranges. Therefore, the normalization step is necessary to eliminate the dimension difference and enable different objective values to be compared. The normalization process of the reference values in the MTBF prediction matrix is similar to the standard normalization process of the linear relationship parameters in step S1. For details, please refer to the standard normalization process of the linear relationship parameters in step S1, which will not be elaborated here. By mapping the objective values to between 0 and 1, the dimension difference of different objective values is eliminated, enabling each objective value to be optimized on the same scale.

[0120] In multi-objective optimization, each objective value in the objective value set (such as MTBF, cost, size) needs to be considered during the optimization process, and there are usually conflicting relationships among these objectives. For example, increasing MTBF may increase cost or size. Therefore, the objective function needs to comprehensively consider these conflicting objectives through a certain modeling method. Based on the normalized objective value set, multiple objective functions are constructed. Common objective functions include: a cost function for reflecting the cost of the target component, a size function for reflecting the size or volume of the target component, and a reliability function for reflecting the mean time between failures of the target component. The expression form of each objective function usually depends on specific optimization requirements (such as minimizing cost and size, maximizing MTBF). The multi-objective optimization form can be expressed as the following formula:

[0121] f2 = Cost; f3 = Size

[0122] Among them, f1 is the first objective function (usually cost or other minimization objectives), such as cost in this case. f2 is the second objective function (usually reliability-related objectives), such as MTBF in this case. f3 is the third objective function (usually volume, size-related objectives), such as size in this case.

[0123] In multi-objective optimization, one of the most commonly used optimization algorithms is the non-dominated sorting genetic algorithm (NSGA-II). The non-dominated sorting genetic algorithm can find non-dominated solutions among multiple objective functions through multiple generations of evolution of the objective functions, thereby obtaining the Pareto front solution set. Specifically, initialize a population containing multiple solutions. Each solution represents a design scheme of a target component, and each solution contains the values of the objective functions (such as MTBF, cost, and size). Use the objective functions to evaluate the fitness of each solution in the population. Fitness is usually evaluated by calculating the values of the objective functions and the non-dominated relationships among the solutions. By selecting solutions with better fitness for crossover and mutation operations to generate a new population, and iterating multiple times, the NSGA-II algorithm will generate a set of non-dominated solutions, and thus generate the Pareto front solution set by analyzing the relationships among the non-dominated solutions. In the Pareto front, no non-dominated solution can be better than other non-dominated solutions without sacrificing other objectives. Therefore, the Pareto front solution set contains the best compromise solutions among all objectives.

[0124] Step S6: Perform fuzzy decision-making on the Pareto front solution set to obtain the selection scheme of the electronic component.

[0125] After multi-objective optimization and the generation of the Pareto front solution set, fuzzy decision-making is used to rank and select non-dominated solutions in order to determine the final selection scheme of electronic components according to specific design requirements. Among them, fuzzy decision-making can find the best compromise among multiple optimization objectives and help designers make reasonable decisions, especially when there are conflicts among multiple objectives. By performing fuzzy decision-making on the non-dominated solutions in the Pareto front solution set, the weights of each objective and the design priorities can be comprehensively considered, and finally a selection scheme of electronic components that meets the requirements can be obtained.

[0126] In fuzzy decision-making, it is first necessary to construct membership functions for each objective (such as cost, MTBF, size, etc.). The membership function is used to measure the "fitness" or "goodness" of each solution under a specific objective. It converts the objective value into a dimensionless numerical value, indicating the degree of satisfaction of the solution under this objective. The membership value of each objective reflects the proximity of the objective value to the design expectation value. The range of membership values is usually from 0 to 1, where 0 means completely not meeting the objective and 1 means completely meeting the objective. The formula for constructing the membership function is as follows:

[0127]

[0128] where, μ i (x j ) represents the membership value of the i-th objective on the j-th solution. x j represents the actual value of the j-th solution on the i-th objective. m i represents the expected value of the i-th objective, usually the design objective or reference value. σ i represents the standard deviation or bandwidth of the i-th objective, which controls the width of the membership function. The membership function formula calculates the membership value through the Gaussian function. The higher the membership value, the closer the non-dominated solution is to the design expectation value under this objective. By constructing the membership function based on the non-dominated solutions in the Pareto front solution set, a satisfaction score (i.e., membership value) can be assigned to each objective.

[0129] In multi-objective decision-making, designers will assign different weights to different objectives to reflect the importance of each objective. For some design objectives (such as MTBF), a higher weight may be given, while other objectives (such as size) can be assigned a lower weight. The weighting process will help adjust the influence of each objective on the final decision according to the objective priorities. The formula for calculating the weighted membership value is as follows:

[0130]

[0131] where, represents the membership value of the weighted i-th objective on the j-th solution. μ i (xj ) represents the original membership value of the i-th objective on the j-th solution. W i is the weight factor of the i-th objective, indicating the importance of this objective in the decision-making, and satisfying Σw i = 1. By weighting the membership values, the important objectives have a greater impact on the final decision-making, ensuring that the decision-making process meets the design requirements and priorities.

[0132] After calculating the weighted membership values, these weighted membership values are integrated to obtain the comprehensive satisfaction index of each non-dominated solution. Among them, the comprehensive satisfaction index summarizes the satisfaction degrees of multiple objectives, representing the comprehensive performance of each solution on all objectives. The calculation of the comprehensive satisfaction is a key step in the fuzzy decision-making, which helps to evaluate the overall advantages and disadvantages of each solution. The calculation formula of the comprehensive satisfaction is as follows:

[0133]

[0134] where, S j is the comprehensive satisfaction value of the j-th solution, representing the comprehensive performance of this solution on all objectives. is the weighted membership value, representing the satisfaction degree of the j-th solution under the i-th objective. W i is the weight factor of the i-th objective. n is the number of objectives. Through the calculation formula of the comprehensive satisfaction, the performances of each non-dominated solution on various objectives are weighted and accumulated to obtain a comprehensive satisfaction value. A solution with a high comprehensive satisfaction represents that this non-dominated solution performs well on multiple objectives.

[0135] Finally, according to the calculated comprehensive satisfaction index, the non-dominated solutions in the Pareto front solution set are sorted. The purpose of the sorting is to select the non-dominated solution that best meets the design requirements (i.e., the non-dominated solution with the highest comprehensive satisfaction after sorting) according to the comprehensive satisfaction index of each non-dominated solution, so as to determine the selection scheme of the electronic components. Usually, the higher the comprehensive satisfaction value of a solution, the more excellent it is on multiple objectives and the more in line with the design requirements, so it is preferentially selected.

[0136] This application is applied to the field of integrated circuit technology. By obtaining a component parameter library and performing multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor, principal component analysis is then performed on the standardized parameter tensor to obtain a dimensionality-reduced feature space, and spectral clustering analysis is performed on the dimensionality-reduced feature space to obtain multiple functionally equivalent classes. Furthermore, reliability degradation modeling is performed on the functionally equivalent classes to obtain an MTBF prediction matrix. At the same time, multi-objective optimization is performed on the MTBF prediction matrix to obtain a Pareto front solution set. Finally, fuzzy decision-making is performed on the Pareto front solution set to obtain a selection scheme for electronic components. Through multi-dimensional normalization processing, dimensionality-reduced feature space analysis, spectral clustering analysis, reliability modeling, multi-objective optimization, and fuzzy decision-making, this application constitutes a comprehensive component selection process, thereby accurately and efficiently selecting components that meet the requirements.

[0137] As Figure 2 shown, it is a functional module diagram of a device for selecting electronic components provided by an embodiment of this application.

[0138] In some embodiments, the device 2 for selecting electronic components may include multiple functional modules composed of computer program segments. The computer programs of each program segment in the device 2 for selecting electronic components can be stored in the memory of the server and executed by at least one processor to execute (see details in Figure 1 description) the functions of the method for selecting electronic components.

[0139] In this embodiment, according to the functions it performs, the device 2 for selecting electronic components can be divided into multiple functional modules. The functional modules may include: a parameter tensor module 21, a feature dimensionality reduction module 22, a functional equivalence module 23, a prediction matrix module 24, a front solution set module 25, and a selection scheme module 26. What this invention calls a module refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, and is stored in the memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0140] The parameter tensor module 21 is used to obtain a component parameter library and perform multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor.

[0141] In an alternative embodiment, the parameter tensor module 21 is specifically used to include:

[0142] Extract the parameter set of the target component from the component parameter library according to the preset test conditions to obtain a three-dimensional data set of each target component under the test conditions;

[0143] Identify the functional relationship characteristics of various parameter subsets in the three-dimensional dataset to obtain the normalization methods corresponding to the various parameter subsets;

[0144] Perform multi-dimensional normalization processing on the parameter subset according to the normalization method to obtain a standardized parameter value set;

[0145] Construct a data structure based on the target component, the test conditions, and the standardized parameter value set to generate a standardized parameter tensor corresponding to each target component.

[0146] The feature dimensionality reduction module 22 is used to perform principal component analysis on the standardized parameter tensor to obtain a dimensionality-reduced feature space.

[0147] In an alternative embodiment, the feature dimensionality reduction module 22 is specifically configured to include:

[0148] Perform an unfolding process on the standardized parameter tensor to obtain a two-dimensional matrix;

[0149] Calculate the covariance matrix for the two-dimensional matrix to obtain a covariance matrix;

[0150] Perform eigenvalue decomposition on the covariance matrix to obtain a first set of eigenvalue and eigenvector combinations;

[0151] Calculate the contribution rate of each eigenvalue in the first set of eigenvalue and eigenvector combinations, and screen the eigenvectors in the first set of eigenvalue and eigenvector combinations according to the contribution rate and a preset contribution threshold to obtain a set of principal component vectors whose contribution rate is greater than or equal to the contribution threshold;

[0152] Perform a projection process on the standardized parameter tensor according to the set of principal component vectors to obtain the dimensionality-reduced feature space.

[0153] The functional equivalence module 23 is used to perform spectral clustering analysis on the dimensionality-reduced feature space to obtain functional equivalence classes.

[0154] In an alternative embodiment, the functional equivalence module 23 is specifically configured to include:

[0155] Calculate the similarity between the data in the dimensionality-reduced feature space to obtain a similarity matrix between the target components;

[0156] Perform a Gaussian kernel function process on the similarity matrix to obtain a weighted similarity matrix;

[0157] Construct a Laplacian matrix based on the weighted similarity matrix, and perform eigenvalue decomposition on the Laplacian matrix to obtain a second set of eigenvalue and eigenvector combinations;

[0158] Perform clustering processing on the target component according to the second eigenvalue and the set of eigenvector combinations to obtain the functional equivalence class.

[0159] The prediction matrix module 24 is used to perform reliability degradation modeling on the functional equivalence class to obtain the MTBF prediction matrix.

[0160] In an optional embodiment, the prediction matrix module 24 specifically includes:

[0161] Extract the working environment parameters and stress factor data for each target component in the functional equivalence class, and construct a working environment model for each target component according to the working environment parameters and the stress factor data;

[0162] Perform temperature stress analysis and Arrhenius formula calculation according to the working environment model to obtain the life coefficient of each target component in the functional equivalence class under a preset specific environment;

[0163] Perform reliability degradation modeling on the target components in each functional equivalence class according to the life coefficient to obtain the failure rate set of each target component under different working environments;

[0164] Construct a failure distribution model according to the failure rate set and the preset failure mode of the corresponding target component;

[0165] Perform mean time between failures prediction on each target component according to the failure distribution model to obtain the MTBF prediction matrix of each functional equivalence class.

[0166] The Pareto front set module 25 is used to perform multi-objective optimization on the MTBF prediction matrix to obtain the Pareto front set.

[0167] In an optional embodiment, the Pareto front set module 25 specifically includes:

[0168] Normalize the reference values in the MTBF prediction matrix to obtain a set of target values;

[0169] Perform multi-objective optimization modeling on the set of target values to construct an objective function;

[0170] Optimize the objective function through a preset non-dominated sorting genetic algorithm to obtain a number of non-dominated solutions;

[0171] Generate the Pareto front set according to the non-dominated solutions.

[0172] The selection solution module 26 is configured to perform fuzzy decision-making on the Pareto front solution set to obtain a selection solution for electronic components.

[0173] In an alternative embodiment, the selection solution module 26 is specifically configured to include:

[0174] Construct a membership function based on the non-dominated solutions in the Pareto front solution set to obtain the membership values of each non-dominated solution under different preset objectives;

[0175] Perform weighted processing on the membership values to obtain weighted membership values;

[0176] Calculate the comprehensive satisfaction degree of each non-dominated solution based on the weighted membership values to obtain the comprehensive satisfaction index of each non-dominated solution;

[0177] Sort the non-dominated solutions in the Pareto front solution set according to the comprehensive satisfaction index to obtain the optimal non-dominated solution corresponding to the maximum comprehensive satisfaction index;

[0178] Obtain a selection solution for electronic components based on the target component parameters in the optimal non-dominated solution.

[0179] It should be understood that the various change methods and specific embodiments in the methods provided in the above embodiments are equally applicable to the selection device for electronic components in this embodiment. Through the foregoing detailed description of the selection method for electronic components, those skilled in the art can clearly know the implementation method of the selection device for electronic components in this embodiment. For the sake of simplicity of the specification, it will not be described in detail here.

[0180] As Figure 3 shown, it is a schematic structural diagram of an electronic device provided by an embodiment of the present application.

[0181] In a preferred embodiment of the present invention, the electronic device 3 may include, but is not limited to: a memory 31, at least one processor 32, and at least one communication bus 33.

[0182] Those skilled in the art should understand that Figure 3 the structure of the electronic device 3 shown does not constitute a limitation on the embodiments of the present invention. The electronic device 3 may further include more or fewer other hardware or software than shown, or different component arrangements.

[0183] In some embodiments, the electronic device 3 is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits, programmable gate arrays, digital processors, and embedded devices, etc.

[0184] It should be noted that the electronic device 3 is only an example. Other existing or future electronic products that can be adapted to this application should also be included within the protection scope of this application and are hereby incorporated by reference.

[0185] In some embodiments, a computer program is stored in the memory 31. When the computer program is executed by the at least one processor 32, all or part of the steps in the method for selecting electronic components as described above are implemented. The memory 31 includes a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically-erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM), or other optical disc memories, magnetic disk memories, tape memories, or any other computer-readable medium that can be used to carry or store data. Further, the computer-readable storage medium mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.

[0186] In some embodiments, the at least one processor 32 is the control core (Control Unit) of the electronic device 3, connecting various components of the entire electronic device 3 through various interfaces and circuits. By running or executing the programs or modules stored in the memory 31, and by calling the data stored in the memory 31, various functions of the electronic device 3 are executed and data is processed. For example, when the at least one processor 32 executes the computer program stored in the memory 31, all or part of the steps in the method for selecting electronic components in the embodiments of the present application are implemented; or all or part of the functions of the device for selecting electronic components are implemented. The at least one processor 32 can be composed of integrated circuits. For example, it can be composed of a single packaged integrated circuit, or can be composed of multiple integrated circuits with the same or different functions packaged together, including a combination of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips.

[0187] In some embodiments, the at least one communication bus 33 is arranged to implement connection communication between the memory 31, the at least one processor 32, etc. Although not shown, the electronic device 3 may further include a power source (such as a battery) for powering each component. Preferably, the power source can be logically connected to the at least one processor 32 through a power management device, so as to implement functions such as management of charging, discharging, and power consumption management through the power management device. The power source may further include any components such as one or more DC or AC power sources, a recharge device, a power failure detection circuit, a power converter or inverter, a power status indicator, etc. The electronic device 3 may further include various sensors, a Bluetooth module, a Wi-Fi module, etc., which will not be elaborated herein.

[0188] The integrated units implemented in the form of software function modules as described above can be stored in a computer-readable storage medium. The above software function modules are stored in a storage medium and include several instructions for causing an electronic device (which may be a personal computer, an electronic device, or a network device, etc.) or a processor to execute parts of the methods described in various embodiments of the present application.

[0189] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0190] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units. They can be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0191] The above are all preferred embodiments of the present application. The protection scope of the present application is not limited thereby. Therefore, any equivalent changes made according to the structure, shape, and principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for selecting electronic components, characterized in that: The method comprises: Acquire a component parameter library and perform multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor; Performing principal component analysis on the standardized parameter tensor to obtain a reduced-dimensional feature space; Performing spectral clustering analysis on the dimension-reduced feature space to obtain functional equivalent classes; Performing reliability degradation modeling on the functional equivalence class to obtain an MTBF prediction matrix; Performing multi-objective optimization on the MTBF prediction matrix to obtain a Pareto frontier solution set; Fuzzy decision making is performed on the Pareto frontier solution set to obtain a selection scheme for electronic components.

2. The method for selecting electronic components according to claim 1, characterized in that: The performing multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor includes: Extracting a parameter set of a target component from the component parameter library according to a preset test condition to obtain a three-dimensional data set of each target component under the test condition; Performing functional relationship feature recognition on various parameter subsets in the three-dimensional data set to obtain normalization methods corresponding to various parameter subsets; Performing multi-dimensional normalization processing on the parameter subset according to the normalization method to obtain a standardized parameter value set; A data structure is constructed according to the target components, the test conditions, and the standardized parameter value set to generate a standardized parameter tensor corresponding to each target component.

3. The method for selecting electronic components according to claim 1, characterized in that: The performing principal component analysis on the standardized parameter tensor to obtain a dimension-reduced feature space comprises: Performing an expansion process on the standardized parameter tensor to obtain a two-dimensional matrix; Performing covariance matrix calculation on the two-dimensional matrix to obtain a covariance matrix; Performing eigenvalue decomposition on the covariance matrix to obtain a combination set of first eigenvalues ​​and eigenvectors; Calculating the contribution rate of each eigenvalue according to the eigenvalues ​​in the first eigenvalue and eigenvector combination set, and screening the eigenvectors in the first eigenvalue and eigenvector combination set according to the contribution rate and a preset contribution threshold, so as to obtain a principal component vector set whose contribution rate is greater than or equal to the contribution threshold; The standardized parameter tensor is projected according to the principal component vector set to obtain the reduced-dimensional feature space.

4. The method for selecting electronic components according to claim 2, characterized in that: The performing spectral clustering analysis on the dimension-reduced feature space to obtain functional equivalent classes comprises: Performing similarity calculation on the data in the dimension-reduced feature space to obtain a similarity matrix between the target components; Performing Gaussian kernel function processing on the similarity matrix to obtain a weighted similarity matrix; Constructing a Laplace matrix according to the weighted similarity matrix, and performing eigenvalue decomposition on the Laplace matrix to obtain a combination set of second eigenvalues ​​and eigenvectors; The target components are clustered according to the second eigenvalue and eigenvector combination set to obtain the functional equivalence class.

5. The method for selecting electronic components according to claim 2, characterized in that: The reliability degradation modeling of the functional equivalence class to obtain the MTBF prediction matrix includes: Extracting working environment parameters and stress factor data for each target component in the functional equivalence class, and constructing a working environment model for each target component according to the working environment parameters and the stress factor data; Performing temperature stress analysis and Arrhenius formula calculation according to the working environment model to obtain the life coefficient of each target component in the functional equivalence class under a preset specific environment; Performing reliability degradation modeling on target components in each functional equivalence class according to the life coefficient to obtain a failure rate set of each target component under different working environments; Constructing a failure distribution model according to the failure rate set and the corresponding preset failure modes of target components; The mean time between failures is predicted for each target component according to the failure distribution model to obtain an MTBF prediction matrix for each functional equivalence class.

6. The method for selecting electronic components according to claim 1, characterized in that: The multi-objective optimization of the MTBF prediction matrix to obtain the Pareto frontier solution set includes: Normalizing the reference values ​​in the MTBF prediction matrix to obtain a target value set; Performing multi-objective optimization modeling on the target value set to construct an objective function; Optimizing the objective function by using a preset non-dominated sorting genetic algorithm to obtain a plurality of non-dominated solutions; The Pareto front solution set is generated according to the non-dominated solutions.

7. The method for selecting electronic components according to claim 6, characterized in that: The fuzzy decision making of the Pareto frontier solution set to obtain the selection scheme of electronic components includes: Constructing a membership function according to the non-dominated solutions in the Pareto frontier solution set to obtain the membership value of each non-dominated solution under different preset objectives; Performing weighted processing on the membership value to obtain a weighted membership value; Calculating the comprehensive satisfaction of each non-dominated solution according to the weighted membership value to obtain a comprehensive satisfaction index of each non-dominated solution; sorting the non-dominated solutions in the Pareto frontier solution set according to the comprehensive satisfaction index to obtain the optimal non-dominated solution corresponding to the maximum comprehensive satisfaction index; According to the target component parameters in the optimal non-dominated solution, a selection scheme for electronic components is obtained.

8. A device for selecting electronic components, characterized in that: The device comprises: A parameter tensor module, used for acquiring a component parameter library and performing multi-dimensional normalization processing on the component parameter library to obtain a standardized parameter tensor; A feature dimension reduction module, used for performing principal component analysis on the standardized parameter tensor to obtain a dimension-reduced feature space; A functional equivalence module, used for performing spectral clustering analysis on the dimension-reduced feature space to obtain functional equivalence classes; A prediction matrix module, used for performing reliability degradation modeling on the functional equivalence class to obtain an MTBF prediction matrix; A frontier solution set module, used for performing multi-objective optimization on the MTBF prediction matrix to obtain a Pareto frontier solution set; The selection scheme module is used to make fuzzy decisions on the Pareto frontier solution set to obtain a selection scheme for electronic components.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method for selecting electronic components according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for selecting an electronic component according to any one of claims 1 to 7 are implemented.

Images