An electrical performance testing platform and method

Abstract

This application relates to the field of data processing technology and discloses an electrical performance testing platform and method. The platform includes: a modeling module for digitally modeling a target circuit to obtain a digital circuit model, and performing finite element analysis on the digital circuit model based on preset operating condition parameters to obtain time-domain analog signal data and thermal effect data; a transformation module for performing short-time Fourier transform on the time-domain analog signal data to obtain a two-dimensional circuit spectrum, and performing distribution mapping on the thermal effect data to obtain a thermal effect distribution map; an extraction module for extracting response features from the two-dimensional circuit spectrum using discrete Wolterra series to obtain a response feature set, and extracting features from the thermal effect distribution map to obtain a thermal effect feature set; and a classification module for classifying fault types using a hidden Markov model to obtain potential fault types and state transition probabilities. This application improves the accuracy and efficiency of electrical performance testing of circuits.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an electrical performance testing platform and method. Background Art

[0002] In the design and manufacturing process of electronic equipment, reliability testing of electrical performance is a key step to ensure product quality and stability. With the rapid development of electronic technology, the complexity of circuits continues to increase, and traditional electrical performance testing methods are gradually unable to meet the requirements of modern electronic products for high efficiency and high accuracy. Especially when dealing with complex integrated circuits and high-frequency circuits, how to accurately simulate and predict various faults that circuits may encounter in actual applications has become a technical problem that needs to be solved urgently.

[0003] Currently, electrical performance testing usually relies on physical testing or simplified model analysis, which often ignores the effects of material properties, environmental factors, and long-term operation. For example, it is difficult for traditional testing methods to capture the performance changes of circuits under high temperature, high humidity or other extreme working conditions in real time. Minor changes under these conditions may cause circuit failure or performance degradation. In addition, physical testing is costly and time-consuming, which is not conducive to rapid iteration and optimization of design, so the accuracy and efficiency of existing technologies are low. Summary of the invention

[0004] The present application provides an electrical performance testing platform and method for improving the accuracy and efficiency of electrical performance testing of circuits.

[0005] In a first aspect, the present application provides an electrical performance testing method, the electrical performance testing method comprising:

[0006] Digitally modeling the target circuit to obtain a digital circuit model, and performing finite element analysis on the digital circuit model based on preset working condition parameters to obtain time domain analog signal data and thermal effect data;

[0007] Performing short-time Fourier transform on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram, and performing distribution mapping on the thermal effect data to obtain a thermal effect distribution diagram;

[0008] Extracting response features from the two-dimensional circuit spectrum diagram using discrete Volterra series to obtain a response feature set, and extracting features from the thermal effect distribution diagram to obtain a thermal effect feature set;

[0009] The response feature set and the thermal effect feature set are input into a preset hidden Markov model to classify the fault types, so as to obtain the potential fault types and state transition probabilities of the target circuit.

[0010] In a second aspect, the present application provides an electrical performance testing platform and method, the electrical performance testing platform comprising:

[0011] A modeling module, used to digitally model the target circuit to obtain a digital circuit model, and perform finite element analysis on the digital circuit model based on preset working condition parameters to obtain time domain analog signal data and thermal effect data;

[0012] A transformation module, used for performing short-time Fourier transformation on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram, and performing distribution mapping on the thermal effect data to obtain a thermal effect distribution diagram;

[0013] An extraction module, configured to extract response features from the two-dimensional circuit spectrum diagram by using a discrete Volterra series to obtain a response feature set, and to extract features from the thermal effect distribution diagram to obtain a thermal effect feature set;

[0014] The classification module is used to input the response feature set and the thermal effect feature set into a preset hidden Markov model to classify the fault types, so as to obtain the potential fault types and state transition probabilities of the target circuit.

[0015] In the technical solution provided by this application, by introducing digital modeling and finite element analysis, complex circuit responses can be simulated in a virtual environment, rather than relying solely on physical testing. In addition, by using short-time Fourier transform and discrete Volterra series analysis, combined with hidden Markov models, it is not only possible to more accurately predict the performance of circuits under different working conditions, but also to effectively identify and classify potential fault types, thereby significantly improving the efficiency and accuracy of testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0017] Figure 1 This is a schematic diagram of an embodiment of the electrical performance testing method in the embodiment of the present application;

[0018] Figure 2 This is a schematic diagram of an embodiment of an electrical performance testing platform in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The embodiments of the present application provide an electrical performance test platform and method. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present 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 interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. 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.

[0020] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 , an embodiment of the electrical performance testing method in the embodiment of the present application includes:

[0021] Step S101, digitally modeling the target circuit to obtain a digital circuit model, and performing finite element analysis on the digital circuit model based on preset working condition parameters to obtain time domain analog signal data and thermal effect data;

[0022] Specifically, the circuit design file of the target circuit is obtained. The design file usually contains detailed structural information of the circuit, such as the type and quantity of components and the connection relationship between them. The circuit design file is parsed by special software to extract the structural data of the circuit, including the physical position and connection method of each component. The circuit structure data is input into the circuit modeling software for three-dimensional modeling processing. The software generates a three-dimensional representation of the circuit based on the input structural data using graphical tools to form an initial three-dimensional circuit model. The model visually shows the physical layout of the circuit. In order to transform the three-dimensional model into a digital circuit model that can be used for performance analysis, the model is assigned material properties. This includes setting properties such as resistivity and dielectric constant for various materials in the circuit, which determine the behavior of the circuit when current passes through and its response to electromagnetic fields. The boundary conditions and initial electrical conditions of the digital circuit model are set, such as voltage, current, and ambient temperature, which simulate the environment that the circuit may encounter in actual work. Based on the set working condition parameters, the finite element analysis method is used to perform circuit signal analysis and thermal simulation analysis on the digital circuit model. Circuit signal analysis simulates the propagation process of current and voltage in the circuit to obtain time domain analog signal data, which can show the electrical performance of the circuit under specific working conditions. At the same time, thermal simulation analysis simulates the thermal effect caused by the passage of current in the circuit after a long period of operation, thereby obtaining thermal effect data, which helps to evaluate the stability and safety of the circuit under high load or harsh environment.

[0023] Step S102, performing short-time Fourier transform on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram, and performing distribution mapping on the thermal effect data to obtain a thermal effect distribution diagram;

[0024] Specifically, the time-domain analog signal data is converted into digital signal data, and the digital signal processor ensures that the analog signal is converted into digital form through appropriate sampling, quantization and encoding. The digital signal data is subjected to short-time Fourier transform to obtain time-frequency domain signal data. The short-time Fourier transform obtains the frequency information that changes with time by sliding a window on the signal and obtains the time-frequency domain signal data. The conversion reveals the frequency component changes of the signal at different time points by dividing the signal into overlapping segments and applying Fourier transform to each segment, which is helpful for analyzing the dynamic response of the circuit. The time-frequency domain signal data is converted into a frequency domain signal spectrum to obtain the initial circuit spectrum. Due to the boundary effect of the short-time Fourier transform and the spectrum leakage that may be caused by the window selection, the initial circuit spectrum needs to be processed by window function to optimize the spectrum representation. The purpose of the window function processing is to reduce the discontinuity effect of the signal edge, and the resolution and leakage characteristics of the spectrum can be improved by an appropriate window function. In order to enhance the quality and readability of the spectrum, the optimized circuit spectrum is filtered and the main frequency component is analyzed. Filtering can remove noise and non-main frequency components and highlight the main frequency response of the circuit. The final result is a two-dimensional circuit spectrum diagram, which helps the system identify possible design problems or performance bottlenecks, thereby guiding the optimization of circuit design and fault diagnosis.

[0025] Extract key information from the thermal effect data obtained from the finite element analysis, including heat flux calculation for multiple target points selected in the digital circuit model. Heat flux, that is, the amount of heat passing through a unit area per unit time, is an important indicator for evaluating the thermal stability of a circuit in an actual working environment. According to the heat flux data of each target point, the thermal behavior of the circuit under different conditions, including heat accumulation and heat dissipation, is analyzed. The three-dimensional coordinate information of multiple target points is obtained. An initial temperature distribution map is constructed based on the coordinates and the corresponding heat flux data. This chart shows the heat flux of each target point. The initial temperature distribution map is smoothed using the Kriging interpolation function. Kriging interpolation is an efficient spatial interpolation method that estimates the values ​​of unsampled points based on the model and considers the statistical characteristics of spatial data, such as variogram and structural characteristics. In this way, the blank areas of the temperature distribution are effectively filled, making the temperature distribution more uniform and continuous. The smoothed thermal effect distribution map provides comprehensive visual information to reveal the thermal distribution characteristics of each part of the circuit, and can also help the system identify potential thermal problem areas, and then carry out targeted thermal management and optimization measures.

[0026] Step S103, extracting response features from the two-dimensional circuit spectrum diagram using discrete Volterra series to obtain a response feature set, and extracting features from the thermal effect distribution diagram to obtain a thermal effect feature set;

[0027] Specifically, feature extraction is performed on the two-dimensional circuit spectrum diagram to identify and record key frequency response features that can represent the circuit behavior. These features may include peak frequency, energy density, and spectrum shape. The extracted multiple circuit response features are converted into feature vectors to obtain initial feature vectors. The initial feature vectors are deeply extracted through discrete Volterra series. Discrete Volterra series is a mathematical tool that can capture the nonlinear relationship between input and output, and can extract more complex and in-depth nonlinear response features from the initial feature vector. Feature dimension reduction is performed on multiple nonlinear response features. Feature dimension reduction is an important part of optimizing data processing. It reduces the computational burden by reducing the dimension of data while retaining the most influential features, which helps to improve the efficiency and accuracy of subsequent analysis. After dimensionality reduction, the core features are clustered and similar features are grouped into the same group, which helps to understand the behavior pattern of the circuit at a higher level. The multiple feature groups obtained are set mapped to combine the features of each group to form a complete response feature set.

[0028] The thermal effect distribution map is analyzed by the spatiotemporal graph convolutional network, and the thermal interaction relationship between each region is extracted to obtain the thermal interaction relationship matrix. This matrix reflects the interaction between different circuit components in heat transfer and distribution, and provides a data basis for understanding how each component affects each other. The spatiotemporal graph convolutional network is used to analyze the dynamic changes of the thermal effect distribution map in the time dimension to obtain the time-varying feature matrix. This matrix shows the thermal behavior of the circuit at different time points and helps identify those areas that show potential thermal risks over time. The thermal interaction relationship matrix and the time-varying feature matrix are normalized and projected to the feature subspace respectively. Normalization is to eliminate the influence of different dimensions and make the data comparable on the same scale, while the feature subspace projection is to extract the most representative and informative features, which helps the accuracy and efficiency of subsequent analysis. After processing, the target interaction relationship matrix and the target time-varying matrix are obtained. The spatial attention mechanism analysis and feature fusion are performed on the target interaction relationship matrix and the target time-varying matrix respectively. The spatial attention mechanism allows the model to focus on the most important regional features, thereby improving the accuracy of the analysis. Feature fusion merges the information of the two matrices to generate a spatiotemporal fusion feature matrix that comprehensively reflects the temporal and spatial dynamics. Multiple thermal effect features are calculated based on the spatiotemporal fusion feature matrix and set transformed to form a thermal effect feature set containing all key thermal features.

[0029] Step S104: input the response feature set and the thermal effect feature set into a preset hidden Markov model to classify the fault types, and obtain the potential fault types and state transition probabilities of the target circuit.

[0030] Specifically, the response feature set and the thermal effect feature set are converted into feature vectors, and the original features are converted into a numerical vector format suitable for mathematical processing to obtain a response feature vector and a thermal effect feature vector. The response feature vector and the thermal effect feature vector are vector spliced ​​to obtain a spliced ​​feature vector, and the electrical response information and thermal effect information of the circuit are integrated so that the hidden Markov model can simultaneously consider these two aspects of information for more comprehensive analysis and judgment. The spliced ​​feature vector is input into a preset hidden Markov model. The model includes a state transition probability matrix and an emission probability matrix, wherein the state transition probability matrix describes the probability of the system transferring from one state to another, and the emission probability matrix describes the probability of observing a certain observation value under a given model state. The state transition probability of the spliced ​​feature vector is calculated by a forward algorithm. The forward algorithm is used to calculate the probability of the model reaching a certain state under a given observation sequence. The calculation results help to understand the behavior mode and fault development trend of the circuit under different test conditions. By applying the Viterbi algorithm, the most likely state sequence is solved. The Viterbi algorithm is a dynamic programming algorithm used to find the path that is most likely to generate an observation data set, that is, a state sequence. This algorithm can not only predict the current state of the circuit, but also predict the state transitions that the circuit may encounter in the future. By analyzing the state sequence, the potential fault types of the target circuit can be identified.

[0031] In the embodiments of the present application, by introducing digital modeling and finite element analysis, complex circuit responses can be simulated in a virtual environment, rather than relying solely on physical testing. In addition, by using short-time Fourier transform and discrete Volterra series analysis, combined with hidden Markov models, not only can the performance of the circuit under different working conditions be more accurately predicted, but also potential fault types can be effectively identified and classified, thereby significantly improving the efficiency and accuracy of the test.

[0032] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0033] (1) Obtaining the circuit design file of the target circuit, and parsing the circuit components and connection relationships of the circuit design file to obtain circuit structure data;

[0034] (2) Inputting the circuit structure data into the circuit modeling software for 3D modeling processing to obtain an initial 3D circuit model;

[0035] (3) Assigning material resistivity and dielectric constant properties to the initial three-dimensional circuit model to obtain a digital circuit model;

[0036] (4) Set the boundary conditions and initial electrical conditions of the digital circuit model, configure the simulation environment, and obtain the working condition parameters;

[0037] (5) Performing circuit signal analysis on the digital circuit model based on the working condition parameters to obtain time domain analog signal data, and performing thermal simulation analysis on the digital circuit model to obtain thermal effect data.

[0038] Specifically, obtain the circuit design files of the target circuit. These files are usually stored in the electronic design automation (EDA) software format, such as CAD or Altium Designer format, and contain complete design information of the circuit, including component layout, connection method and component specifications. Through parsing tools or software plug-ins, extract detailed information of circuit components and their connection relationships from the design files. The obtained circuit structure data includes the location and type of each component in the circuit and the electrical connection relationship between them. Input the structure data into the circuit modeling software, such as ANSYS or SolidWorks Electrical, which can use the input data to generate a three-dimensional geometric model of the circuit. The initial three-dimensional circuit model is a visual representation that shows the physical structure of the circuit but does not yet include electrical properties. Assign material properties to the initial three-dimensional circuit model, including properties such as resistivity and dielectric constant of each electronic component. These properties determine how current flows through the circuit and how the circuit behaves in the electromagnetic field. For example, in a circuit model containing copper wires and silicon semiconductors, the resistivity of copper and the dielectric constant of silicon will directly affect the accuracy of the simulation results. Set boundary conditions and initial electrical conditions. Boundary conditions may include the operating temperature range of the circuit, voltage supply limitations, and physical interface conditions, while initial electrical conditions involve the voltage and current settings of the circuit before simulation begins. These conditions, together with the configuration of the simulation environment, constitute the working parameters of the circuit simulation. Based on the set working condition parameters, the circuit simulation software is used to perform circuit signal analysis and thermal simulation analysis on the digital circuit model. Circuit signal analysis focuses on the behavior of the circuit under specific electrical loads and conditions, including signal integrity, noise level, and signal delay. At the same time, thermal simulation analysis focuses on the thermal stability of the circuit during long-term operation or under specific electrical loads. The thermal effect data generated by the analysis can help the system evaluate the thermal problems that the circuit may encounter in actual operation. For example, for a high-power amplifier circuit, circuit signal analysis can reveal nonlinear distortion and signal attenuation problems during signal amplification, while thermal simulation analysis can help discover overheating problems caused by long-term operation. These hot spots may affect the reliability and life of components.

[0039] In a specific embodiment, the process of performing a short-time Fourier transform on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram may specifically include the following steps:

[0040] (1) Convert the time domain analog signal data into digital signal to obtain digital signal data;

[0041] (2) Perform short-time Fourier transform on the digital signal data to obtain time-frequency domain signal data, where the function of short-time Fourier transform is:

[0042] (3) Convert the time-frequency domain signal data into a frequency domain signal spectrum to obtain an initial circuit spectrum diagram;

[0043] (4) Performing window function processing on the initial circuit spectrum to obtain an optimized circuit spectrum;

[0044] (5) The optimized circuit spectrum diagram is filtered and the main frequency component is analyzed to obtain a two-dimensional circuit spectrum diagram.

[0045] Specifically, the analog signal is sampled by an analog-to-digital converter, and the amplitude of the signal is measured at a fixed time interval (sampling frequency). The sampling frequency should be at least twice the highest frequency component in the signal to avoid aliasing effects. The continuous amplitude values ​​obtained by sampling are quantized, that is, converted into discrete values. This process may introduce quantization noise, but appropriate quantization bit depth can reduce this effect. The discrete values ​​are encoded into digital form to form digital signal data. Short-time Fourier transform is applied to analyze the changes of the signal in time and frequency. The short-time Fourier transform realizes the time-frequency analysis of the signal by sliding a fixed-length window on the signal and calculating the Fourier transform at each window position. The formula for short-time Fourier transform is: is the input time domain signal, that is, the original analog or digitized electrical signal, where Represents time, the function represents the signal at time The instantaneous amplitude of Represents the window function, which is used to select a local segment of the signal for analysis in each step of Fourier transform. is the center position of the window function. Slide across the signal, which means at different points in time Perform local analysis on the signal. The choice of window function (such as Hanning window, Hamming window, Gaussian window, etc.) will affect the results of spectrum analysis, especially in terms of frequency resolution and leakage; It is the core part of Fourier transform and represents the complex exponential function, where is an imaginary unit, is the angular frequency, which represents the frequency component to be analyzed in the signal. This expression is used to convert the time domain signal to the frequency domain; is the angular frequency, which is directly related to the frequency content in the signal and is used to detect different frequency components in the Fourier transform; is the center time of the window function, representing the location where local signal analysis is performed at a specific time. The results of the short-time Fourier transform are converted into a frequency domain signal spectrogram. The Fourier transform results at each time point are mapped into a graph of frequency versus amplitude, usually represented as a spectrogram, which is an important tool in circuit analysis for identifying various frequency components in the signal. The initial shape of the spectrogram may cause spectrum leakage due to edge effects and the choice of window functions, that is, energy "leaks" from one frequency component to nearby frequency components. To solve this problem, the initial circuit spectrum is processed by a window function. By selecting a suitable window function, such as a Hamming window or a Blackman window, the resolution and leakage characteristics of the spectrum are optimized. After window function processing, the energy of the spectrum is more concentrated and the frequency components are clearer. The optimized circuit spectrum is filtered and the main frequency component analysis is performed. Filtering is intended to eliminate background noise and unwanted frequency components and highlight useful signal information. Filtering can be achieved through low-pass, high-pass or band-pass filters, depending on the needs of the analysis. Main frequency component analysis involves identifying and extracting the main energy peaks in the spectrum, which represent the main operating frequencies or possible fault frequencies of the circuit.

[0046] In a specific embodiment, the process of performing distribution mapping on the thermal effect data to obtain a thermal effect distribution map may specifically include the following steps:

[0047] (1) Calculate the heat flux of multiple target points in the digital circuit model based on the thermal effect data to obtain the heat flux data of each target point;

[0048] (2) Obtaining the three-dimensional coordinate information of multiple target points, and constructing an initial temperature distribution map based on the three-dimensional coordinate information and the heat flux data of each target point;

[0049] (3) The Kriging interpolation function is used to smooth the initial temperature distribution map to obtain the thermal effect distribution map.

[0050] Specifically, heat flux calculations are performed on multiple target points in the digital circuit model based on the thermal effect data to obtain heat flux data for each target point. Heat flux, also known as heat flow rate, is the amount of heat transferred through a unit area per unit time, usually expressed in watts per square meter (W / m²). After the heat flux calculation is completed, the three-dimensional coordinate information of these target points is collected. This information can be obtained from the CAD model or design file of the circuit. An initial temperature distribution graph is constructed based on the three-dimensional coordinate information and the heat flux data of each target point. The graph is constructed by associating the heat flux value of each point with its coordinates in three-dimensional space to form a spatial model that describes the thermal characteristics of the circuit. Since the distribution of the measurement points may be uneven, the temperature distribution graph obtained directly from these points may show discontinuous or rough features. The Kriging interpolation method is used to smooth the temperature distribution. Kriging interpolation is an optimal linear unbiased estimation method based on geostatistics and is widely used for spatial data interpolation. It not only considers the distance between data points, but also the spatial relationship between these points. The basic expression of Kriging interpolation is: .in, It is the prediction point The estimated value of is the value of a known data point, are weights, which are optimized by solving the Kriging system to minimize the variance of the estimation error. By constructing a variogram to describe the structural characteristics of spatial data, the Kriging method can effectively perform spatial interpolation to smooth the entire temperature distribution map. The thermal effect distribution map helps the system identify possible thermal problem areas, such as hot spots or thermally uneven areas, so as to optimize the circuit design or heat dissipation of specific components.

[0051] In a specific embodiment, the step of extracting response features from the two-dimensional circuit spectrum diagram using a discrete Volterra series to obtain a response feature set may specifically include the following steps:

[0052] (1) Extract features from the two-dimensional circuit spectrum to obtain multiple circuit response features;

[0053] (2) Performing feature vector transformation on multiple circuit response features to obtain an initial feature vector;

[0054] (3) The initial feature vector is deeply extracted through the discrete Volterra series to obtain multiple nonlinear response features;

[0055] (4) Performing feature dimensionality reduction on multiple nonlinear response features to obtain multiple core features, and performing cluster analysis on the multiple core features to obtain multiple feature groups;

[0056] (5) Perform set mapping on multiple feature groups to obtain a response feature set.

[0057] Specifically, feature extraction is performed on the two-dimensional circuit spectrum diagram to identify key information in the diagram, such as frequency peaks, energy density areas, and shape characteristics of the spectrum. These features directly reflect the response of the circuit under different operating conditions, such as resonance peaks or energy loss anomalies at a specific frequency. These significant spectrum features are detected and recorded by spectrum analysis software. The extracted features are converted into feature vectors, and the feature data is standardized and formatted into a form that can be processed by the mathematical model. For example, each feature (such as peak frequency, peak amplitude, etc.) is assigned a numerical value to form a vector. After obtaining the initial feature vector, the discrete Volterra series is used to deeply extract the vector to reveal the nonlinear dynamic characteristics of the circuit. The discrete Volterra series is a mathematical tool used to analyze the nonlinear response of the system to the input. Its basic form can be expressed as: .in, is the output of the system, is the input signal, is the nonlinear response coefficient of the system, is the order of the series. Through analysis, a set of features describing the nonlinear behavior of the circuit is obtained. These features can reveal the actual working condition of the circuit in more detail than simple linear response. The nonlinear response features are subjected to feature dimensionality reduction processing, such as principal component analysis or singular value decomposition, to identify the most important core features from a large number of features. Feature dimensionality reduction not only reduces the complexity of data processing, but also helps to improve the efficiency and accuracy of subsequent analysis steps. The core features are grouped into several feature groups through cluster analysis, such as K-means or hierarchical clustering, and each group represents a certain aspect of the behavior or characteristics of the circuit. The feature groups are set mapped to integrate them into a comprehensive set of response features. The feature set describes the comprehensive performance of the circuit under various test conditions.

[0058] In a specific embodiment, the process of performing feature extraction on the thermal effect distribution map to obtain a thermal effect feature set may specifically include the following steps:

[0059] (1) The thermal interaction relationship between regions in the thermal effect distribution map is captured through the spatiotemporal graph convolutional network to obtain the thermal interaction relationship matrix;

[0060] (2) The dynamic change features of the thermal effect distribution map in the time dimension are extracted through the spatiotemporal graph convolutional network to obtain the time-varying feature matrix;

[0061] (3) Normalize and project the thermal interaction matrix and the time-varying feature matrix into feature subspaces to obtain the target interaction matrix and the target time-varying matrix;

[0062] (4) Perform spatial attention mechanism analysis and feature fusion on the target interaction relationship matrix and target time change matrix respectively to obtain the spatiotemporal fusion feature matrix;

[0063] (5) Calculate the corresponding multiple thermal effect features according to the spatiotemporal fusion feature matrix, and perform set transformation on the multiple thermal effect features to obtain a thermal effect feature set.

[0064] Specifically, the thermal effect distribution map is analyzed by the spatiotemporal graph convolutional network to obtain the thermal interaction relationship between the regions inside the device. The network not only learns the characteristics of a single region, but also the interaction between regions by applying convolution operations on the nodes of the graph. In this process, each region is regarded as a node, and the edges between nodes represent the intensity of thermal interaction between regions. Through network training, a thermal interaction relationship matrix is ​​obtained, in which each element quantifies the degree of thermal interaction between two regions and reflects how they affect each other through heat transfer and thermal radiation. The dynamic change feature of the thermal effect distribution map in the time dimension is extracted by the spatiotemporal graph convolutional network. The network not only considers the thermal distribution state at each time point, but also analyzes the change of the state over time to generate a time-varying feature matrix. The matrix records the thermal characteristics of each region that change over time, such as the rapid increase or decrease of temperature, which helps to predict the performance of the device under long-term operation or high load. The thermal interaction relationship matrix and the time-varying feature matrix are normalized respectively to eliminate the influence of different dimensions and scales and ensure the consistency of the data. After normalization, feature subspace projections such as principal component analysis or autoencoders are applied to extract and highlight the most critical features to form a target interaction relationship matrix and a target time change matrix. The target interaction relationship matrix and the target time change matrix are analyzed using a spatial attention mechanism to improve the accuracy and efficiency of feature extraction by focusing on the most important parts of the data. The spatial attention mechanism enables the network to prioritize more influential regions or time points during analysis, thereby performing effective feature fusion. Feature fusion combines data in spatial and temporal dimensions to generate a spatiotemporal fusion feature matrix that provides a comprehensive view of the thermal behavior of the entire device under different conditions. Based on the spatiotemporal fusion feature matrix, the corresponding multiple thermal effect features are calculated and transformed into a set to form a comprehensive thermal effect feature set. The feature set contains all key thermal characteristic information, such as key hotspot locations, areas with the most sensitive thermal response, and possible thermal risk areas, providing data support for the formulation and optimization of thermal management strategies.

[0065] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0066] (1) Performing feature vector conversion on the response feature set and the thermal effect feature set respectively to obtain the response feature vector and the thermal effect feature vector;

[0067] (2) performing vector concatenation on the response feature vector and the thermal effect feature vector to obtain a concatenated feature vector;

[0068] (3) Inputting the concatenated feature vector into a preset hidden Markov model, which includes: a state transition probability matrix and a transmission probability matrix;

[0069] (4) Calculate the state transition probability of the concatenated feature vector through the forward algorithm;

[0070] (5) The most likely state sequence is solved by the Viterbi algorithm, and the state sequence is transformed to identify the potential fault type of the target circuit.

[0071] Specifically, the response feature set and the thermal effect feature set are transformed into feature vectors to obtain the response feature vector and the thermal effect feature vector. Each feature is quantified into a numerical value to form a vector. The response feature set may include the frequency distribution, peak value, signal attenuation rate, etc. of the signal, while the thermal effect feature set may include the highest temperature point, average temperature, hot spot distribution, etc. The quantification of each feature can determine its weight based on the physical meaning of the feature and the degree of influence on the system performance. For example, a temperature feature may be given a higher weight due to its importance in fault prediction. The response feature vector and the thermal effect feature vector are spliced ​​to form a comprehensive spliced ​​feature vector. The spliced ​​vector will comprehensively consider the electrical and thermal behaviors of the circuit and provide data support for a more comprehensive analysis of the circuit state. The spliced ​​feature vector is input into the preset hidden Markov model. The hidden Markov model is a statistical model that assumes that there are some internal states of the system that cannot be directly observed and that the state transition depends only on the previous state, which is consistent with the characteristics of many physical and engineering systems. The model includes the state transition probability matrix, which describes the transition probability from one state to another; the emission probability matrix, which describes the probability of observing a certain feature vector in any specific state. Describes the state Transfer to state The probability of emission probability matrix Describes the state Observed to a specific observation To evaluate the likelihood of each state in the model, the forward algorithm is used to calculate the probability distribution of the concatenated feature vector in each state. The forward algorithm is a dynamic programming technique that can effectively calculate the probability of each part of the sequence ending in each possible state under the condition of a given observation sequence. This calculation not only provides a good match between the model and the data, but also helps predict the trend of future state changes. The forward algorithm is expressed as: .in, is the number of states, Yes Status Transfer to state The probability of is in state Observed The probability of failure. In order to determine the most likely sequence of states of the system, that is, to determine the state changes that the circuit may experience during the entire observation period, the Viterbi algorithm is applied. The Viterbi algorithm is a special dynamic programming algorithm that is used to find the most likely sequence of states that best explains the observed data. In this way, the potential fault types and development processes of the circuit are identified, for example, a state of normal operation may gradually change to a state of overheating or performance degradation.

[0072] The electrical performance test method in the embodiment of the present application is described above. The electrical performance test platform in the embodiment of the present application is described below. Figure 2 In the embodiment of the present application, an electrical performance test platform includes:

[0073] Modeling module 201, used to digitally model the target circuit to obtain a digital circuit model, and perform finite element analysis on the digital circuit model based on preset working condition parameters to obtain time domain analog signal data and thermal effect data;

[0074] The transformation module 202 is used to perform short-time Fourier transformation on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram, and to perform distribution mapping on the thermal effect data to obtain a thermal effect distribution diagram;

[0075] An extraction module 203 is used to extract response features of the two-dimensional circuit spectrum diagram through discrete Volterra series to obtain a response feature set, and to extract features of the thermal effect distribution diagram to obtain a thermal effect feature set;

[0076] The classification module 204 is used to input the response feature set and the thermal effect feature set into a preset hidden Markov model to classify the fault types, and obtain the potential fault types and state transition probabilities of the target circuit.

[0077] Through the synergy of the above components, by introducing digital modeling and finite element analysis, complex circuit responses can be simulated in a virtual environment, rather than relying solely on physical testing. In addition, the use of short-time Fourier transform and discrete Volterra series analysis, combined with hidden Markov models, can not only more accurately predict the performance of circuits under different working conditions, but also effectively identify and classify potential fault types, thereby significantly improving the efficiency and accuracy of testing.

[0078] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0079] 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 and other media that can store program codes.

[0080] 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. An electrical performance testing method, characterized in that: The electrical performance testing method comprises: Digitally modeling the target circuit to obtain a digital circuit model, and performing finite element analysis on the digital circuit model based on preset working condition parameters to obtain time domain analog signal data and thermal effect data; Performing short-time Fourier transform on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram, and performing distribution mapping on the thermal effect data to obtain a thermal effect distribution diagram; The method extracts response features of the two-dimensional circuit spectrum diagram by discrete Volterra series to obtain a response feature set, and extracts features of the thermal effect distribution diagram to obtain a thermal effect feature set; specifically includes: extracting features of the two-dimensional circuit spectrum diagram to obtain multiple circuit response features; performing feature vector conversion on the multiple circuit response features to obtain an initial feature vector; performing deep extraction on the initial feature vector by discrete Volterra series to obtain multiple nonlinear response features; performing feature dimension reduction on the multiple nonlinear response features to obtain multiple core features, and performing cluster analysis on the multiple core features to obtain multiple feature groups; performing set mapping on the multiple feature groups to obtain a response feature set; The spatial graph convolution network captures the thermal interaction relationship between regions of the thermal effect distribution map to obtain a thermal interaction relationship matrix; the spatiotemporal graph convolution network extracts the dynamic change features of the thermal effect distribution map in the time dimension to obtain a time change feature matrix; the thermal interaction relationship matrix and the time change feature matrix are normalized and projected into feature subspaces to obtain a target interaction relationship matrix and a target time change matrix; the target interaction relationship matrix and the target time change matrix are analyzed by spatial attention mechanism and feature fusion to obtain a spatiotemporal fusion feature matrix; multiple corresponding thermal effect features are calculated according to the spatiotemporal fusion feature matrix, and the multiple thermal effect features are converted into a set to obtain a thermal effect feature set; The response feature set and the thermal effect feature set are input into a preset hidden Markov model to classify the fault types, so as to obtain the potential fault types and state transition probabilities of the target circuit.

2. The electrical performance testing method according to claim 1, characterized in that: The digital modeling of the target circuit is performed to obtain a digital circuit model, and the finite element analysis of the digital circuit model is performed based on preset working condition parameters to obtain time domain simulation signal data and thermal effect data, including: Obtaining a circuit design file of a target circuit, and parsing the circuit components and connection relationships of the circuit design file to obtain circuit structure data; Inputting the circuit structure data into circuit modeling software for three-dimensional modeling processing to obtain an initial three-dimensional circuit model; Assigning material resistivity and dielectric constant properties to the initial three-dimensional circuit model to obtain a digital circuit model; Setting boundary conditions and initial electrical conditions of the digital circuit model, and performing simulation environment configuration to obtain working condition parameters; Based on the working condition parameters, a circuit signal analysis is performed on the digital circuit model to obtain time domain analog signal data, and a thermal simulation analysis is performed on the digital circuit model to obtain thermal effect data.

3. The electrical performance testing method according to claim 1, characterized in that: The step of performing a short-time Fourier transform on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram includes: Performing digital signal conversion on the time domain analog signal data to obtain digital signal data; Perform short-time Fourier transform on the digital signal data to obtain time-frequency domain signal data: Performing frequency domain signal spectrum conversion on the time-frequency domain signal data to obtain an initial circuit spectrum diagram; Performing window function processing on the initial circuit spectrum diagram to obtain an optimized circuit spectrum diagram; The optimized circuit spectrum diagram is subjected to filtering processing and main frequency component analysis to obtain a two-dimensional circuit spectrum diagram.

4. The electrical performance testing method according to claim 1, characterized in that: The step of performing distribution mapping on the thermal effect data to obtain a thermal effect distribution map includes: Performing heat flux calculation on a plurality of target points in the digital circuit model according to the thermal effect data to obtain heat flux data for each target point; Acquire three-dimensional coordinate information of the plurality of target points, and construct an initial temperature distribution map according to the three-dimensional coordinate information and heat flux data of each target point; The Kriging interpolation function is used to perform temperature distribution smoothing processing on the initial temperature distribution map to obtain a thermal effect distribution map.

5. The electrical performance testing method according to claim 1, characterized in that: The step of inputting the response feature set and the thermal effect feature set into a preset hidden Markov model to classify the fault types and obtain the potential fault types and state transition probabilities of the target circuit includes: Performing feature vector conversion on the response feature set and the thermal effect feature set respectively to obtain a response feature vector and a thermal effect feature vector; Performing vector concatenation on the response characteristic vector and the thermal effect characteristic vector to obtain a concatenated characteristic vector; Inputting the concatenated feature vector into a preset hidden Markov model, wherein the hidden Markov model includes: a state transition probability matrix and a transmission probability matrix; Calculate the state transition probability of the concatenated feature vector by a forward algorithm; The most likely state sequence is solved by using the Viterbi algorithm, and the state sequence is converted to identify the potential fault type of the target circuit.

6. An electrical performance testing platform, characterized in that: The electrical performance testing platform is used to perform the electrical performance testing method according to any one of claims 1 to 5, and the electrical performance testing platform comprises: A modeling module, used to digitally model the target circuit to obtain a digital circuit model, and perform finite element analysis on the digital circuit model based on preset working condition parameters to obtain time domain analog signal data and thermal effect data; A transformation module, used for performing short-time Fourier transformation on the time-domain analog signal data to obtain a two-dimensional circuit spectrum diagram, and performing distribution mapping on the thermal effect data to obtain a thermal effect distribution diagram; The extraction module is used to extract response features of the two-dimensional circuit spectrum diagram through discrete Volterra series to obtain a response feature set, and to extract features of the thermal effect distribution diagram to obtain a thermal effect feature set; specifically comprising: extracting features of the two-dimensional circuit spectrum diagram to obtain multiple circuit response features; performing feature vector conversion on the multiple circuit response features to obtain an initial feature vector; performing deep extraction on the initial feature vector through discrete Volterra series to obtain multiple nonlinear response features; performing feature dimension reduction on the multiple nonlinear response features to obtain multiple core features, and performing cluster analysis on the multiple core features to obtain multiple feature groups; performing set mapping on the multiple feature groups to obtain a response feature set; The thermal interaction relationship between regions of the thermal effect distribution map is captured by a spatiotemporal graph convolution network to obtain a thermal interaction relationship matrix; the dynamic change features of the time dimension of the thermal effect distribution map are extracted by a spatiotemporal graph convolution network to obtain a time change feature matrix; the thermal interaction relationship matrix and the time change feature matrix are normalized and projected into feature subspaces to obtain a target interaction relationship matrix and a target time change matrix; the target interaction relationship matrix and the target time change matrix are analyzed by spatial attention mechanism and feature fusion to obtain a spatiotemporal fusion feature matrix; multiple corresponding thermal effect features are calculated according to the spatiotemporal fusion feature matrix, and the multiple thermal effect features are converted into a set to obtain a thermal effect feature set; The classification module is used to input the response feature set and the thermal effect feature set into a preset hidden Markov model to classify the fault types, so as to obtain the potential fault types and state transition probabilities of the target circuit.

Images