Switching power supply test method, device, equipment and medium

Abstract

The invention discloses a switching power supply test method, device and equipment and a medium, and relates to the technical field of power supply management. The method comprises the following steps: performing signal acquisition on a to-be-tested circuit node in a switching power supply of a target topological structure by using a multi-channel oscilloscope to obtain real-time waveform data and basic electric quantity information of the to-be-tested circuit node; performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency map, and performing time-frequency analysis on the time-frequency map by using a pre-constructed convolutional neural network to output an analysis processing result; wherein the convolutional neural network is trained by using known correct normal waveform data of the switching power supply, and is used for identifying an abnormal working state of the switching power supply; and taking the analysis processing result and the basic electric quantity information as input parameters of the full-connection neural network, and outputting a final test result of the to-be-tested circuit node through the full-connection neural network. According to the technical scheme of the invention, comprehensive function testing can be carried out on the switching power supply.

Description

Technical Field

[0001] The present invention relates to the technical field of power management, and in particular to a switching power supply testing method, device, equipment and medium. Background Art

[0002] Switching power supplies play a key role in modern servers. Their complex structure and numerous test points make functional testing cumbersome and time-consuming. Traditional manual testing methods are not only inefficient, but also difficult to achieve high-precision automatic testing. Therefore, how to provide a solution to the above technical problems is a problem that technicians in this field currently need to solve. Summary of the invention

[0003] In view of this, the purpose of the present invention is to provide a switching power supply testing method, device, equipment and medium, which can automatically test and identify the abnormal working state of the power supply, thereby greatly improving the testing efficiency and enhancing the analysis accuracy. The specific scheme is as follows:

[0004] In a first aspect, the present application discloses a switching power supply testing method, comprising:

[0005] Using a multi-channel oscilloscope to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure, so as to obtain real-time waveform data and basic power information of the circuit nodes to be tested;

[0006] Performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and performing time-frequency analysis on the time-frequency spectrum using a pre-built convolutional neural network to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply;

[0007] The analysis and processing results and the basic electrical quantity information are used as input parameters of a fully connected neural network, and a final test result of the circuit node to be tested is output through the fully connected neural network.

[0008] Optionally, the method of using a multi-channel oscilloscope to collect signals from a circuit node to be tested in a switching power supply of a target topology structure to obtain real-time waveform data and basic power information of the circuit node to be tested includes:

[0009] Using a multi-channel oscilloscope to collect signals from a circuit node to be tested in a switching power supply of a target topology structure to obtain a collected signal, and transmitting the collected signal to a target device through a preset application program interface for communication between instruments;

[0010] The real-time waveform data and basic power information in the collected signal are extracted using a preset analysis software package running on the target device; wherein the real-time waveform data is a time domain waveform image, and the basic power information includes amplitude measurement data and time measurement data.

[0011] Optionally, the target topology structure is a power circuit structure implemented by a resonant network; accordingly, the method further includes:

[0012] Determine any one or more of a switch node, an oscillator signal, an output voltage, a feedback signal, and a compensation network signal in the switching power supply of the target topology as a circuit node to be tested;

[0013] Among them, the switching node is used to analyze the switching frequency, waveform distortion and switching noise of the switching power supply, the oscillator signal is used to control the switching frequency, the output voltage is used to observe voltage stability, ripple and noise, the feedback signal is used to evaluate the stability and response characteristics of the control loop of the target topology structure, and the compensation network signal is used to verify the stability and dynamic response of the control loop.

[0014] Optionally, performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum includes:

[0015] Performing denoising processing on the real-time waveform data by a preset noise processing method to remove high-frequency noise in the real-time waveform data to obtain first processed data;

[0016] Normalizing the first processed data based on a preset data scale to adjust the numerical range of the first processed data to a uniform interval, thereby obtaining second processed data;

[0017] Frame the signal in the second processed data, and perform windowing on each frame of the obtained signal using a preset windowing function to generate third processed data;

[0018] Performing short-time Fourier transform or continuous wavelet transform on the third processed data to convert the time domain representation of the signal in the third processed data into the frequency domain representation, and generating a corresponding spectrum result;

[0019] Performing a two-dimensional matrix conversion on the spectrum result to generate a corresponding time-frequency spectrum;

[0020] Accordingly, the time-frequency analysis of the time-frequency spectrum is performed using a pre-built convolutional neural network to output an analysis result, including:

[0021] Performing format conversion on the time-frequency spectrum based on a preset data format to generate target format data supporting the input of the convolutional neural network;

[0022] The convolutional neural network is used to perform time-frequency analysis on the target format data to output analysis and processing results.

[0023] Optionally, the process of building a convolutional neural network includes:

[0024] Constructing a convolutional neural network model using a preset deep learning framework, and defining a corresponding network structure for the convolutional neural network model; the network structure includes an input layer, a convolution layer, an activation layer, a pooling layer, a data flattening layer, a fully connected layer, and an output layer;

[0025] Corresponding layer parameters are configured for the network structure, and the configured convolutional neural network model is trained using the normal waveform data to obtain a trained convolutional neural network.

[0026] Optionally, the using of a pre-built convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis result includes:

[0027] The sequence data corresponding to the time-frequency spectrum is divided into multiple windows according to a preset window length, and a pre-built convolutional neural network is used to perform time-frequency analysis on each window based on a preset overlapping window strategy to output the analysis and processing results.

[0028] Optionally, after outputting the final test result of the circuit node to be tested through the fully connected neural network, the method further includes:

[0029] A test report is automatically generated according to the final test result, and the test report is analyzed so as to provide corresponding fault generation reasons for the circuit node to be tested according to the test report.

[0030] In a second aspect, the present application discloses a switching power supply testing device, comprising:

[0031] A data acquisition module is used to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure using a multi-channel oscilloscope to obtain real-time waveform data and basic power information of the circuit nodes to be tested;

[0032] A convolutional neural network processing module is used to perform time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and use a pre-built convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply

[0033] A test result output module is used to use the analysis and processing results and the basic power information as input parameters of a fully connected neural network, and output the final test result of the circuit node to be tested through the fully connected neural network.

[0034] In a third aspect, the present application discloses an electronic device, comprising a processor and a memory; wherein the memory is used to store a computer program, and the computer program is loaded and executed by the processor to implement the switching power supply testing method as described above.

[0035] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the switching power supply testing method as described above.

[0036] The present application provides a switching power supply testing method, comprising: using a multi-channel oscilloscope to collect signals from a circuit node to be tested in a switching power supply of a target topology structure to obtain real-time waveform data and basic power information of the circuit node to be tested; performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and using a pre-constructed convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply; using the analysis and processing result and the basic power information as input parameters of a fully connected neural network, and outputting a final test result of the circuit node to be tested through the fully connected neural network.

[0037] The beneficial technical effects of the present application are as follows: first, the signal of the circuit node to be tested in the switching power supply of the target topology structure usually has high-frequency characteristics. By converting the real-time waveform data measured by the multi-channel oscilloscope, which is difficult to capture subtle waveform distortion, into a time-frequency spectrum, the time-frequency characteristics of the signal can be accurately displayed, which helps to identify high-frequency noise and waveform distortion in the switching process, thereby improving the analysis accuracy. Secondly, since the power supply circuit of the switching power supply of the target topology structure is very sensitive to the matching of the resonant frequency and the switching frequency, the time-frequency characteristics of different scales can be extracted by using a convolutional neural network to perform time-frequency analysis on the generated time-frequency spectrum, effectively capture different modes of the signal in time and frequency, and output preliminary analysis and processing results, so as to achieve a more accurate and efficient analysis of the switching power supply signal. By comparing with normal waveform data, the static and dynamic performance of the switching power supply can be accurately tested and analyzed, and the abnormal working state of the switching power supply can be identified. Finally, combined with the basic power information of the circuit node to be tested, the fully connected neural network is used to judge the comprehensive data and output the final test results. In this way, the process of fully automated identification of the working state of the switching power supply is realized, which significantly improves the test efficiency and comprehensiveness, reduces manual analysis time and errors, and improves the efficiency and accuracy of abnormal detection of the switching power supply.

[0038] In addition, the present application provides a switching power supply testing device, equipment and storage medium, which correspond to the above-mentioned switching power supply testing method and have the same effect as above. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0040] Figure 1 A flow chart of a switching power supply testing method disclosed in this application;

[0041] Figure 2 A block diagram of a switching power supply is provided as an example;

[0042] Figure 3 An ideal time-frequency diagram of different types of waveforms of an AC power input disclosed in the present application;

[0043] Figure 4 This is a waveform diagram of a switching power supply main control chip driving a transformer through a MOSFET disclosed in this application;

[0044] Figure 5 A switching power supply test program execution diagram disclosed in this application;

[0045] Figure 6 A schematic diagram of the structure of a switching power supply testing device disclosed in this application;

[0046] Figure 7 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] At present, the key nodes of the switching power supply are mainly measured by an oscilloscope, which can directly obtain the time domain waveform of the measurement point. At the same time, with the help of the measurement function of the oscilloscope, basic power parameters such as the root mean square value (RMS), average value (AVG), peak value (Peak) and frequency (Freq) of the measurement point can also be obtained. In-depth analysis of these waveforms and data can intuitively reveal the working status of the power supply. This method belongs to the traditional testing method.

[0049] However, it is worth noting that traditional test methods have certain limitations in dynamic characteristics analysis. In switching power supplies, due to the use of a large number of inductance components, these inductance components will cause overshoot and ringing signals to be superimposed on the switching square wave pulse in the switching state. Therefore, the analysis of these signals is crucial for the development and testing of switching power supplies. However, traditional test methods mainly rely on manual observation for signal analysis, which is obviously not efficient and accurate enough.

[0050] To this end, the present application provides a switching power supply testing solution that can automatically test and identify abnormal working conditions of the power supply, thereby greatly improving testing efficiency and enhancing analysis accuracy.

[0051] The embodiment of the present invention discloses a switching power supply testing method, see Figure 1 As shown, the method includes:

[0052] Step S11: using a multi-channel oscilloscope to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure, so as to obtain real-time waveform data and basic power information of the circuit nodes to be tested.

[0053] In an embodiment of the present application, a multi-channel oscilloscope is used to collect waveform and power information of key nodes in the circuit of the switching power supply. Therefore, the circuit nodes to be tested are first determined. The selection and application of the circuit nodes to be tested are crucial. The performance and stability of the switching power supply under various working conditions are ensured by analyzing and testing the circuit nodes to be tested. Specifically, any one or more of the switch nodes, oscillator signals, output voltages, feedback signals, and compensation network signals in the switching power supply of the target topology are determined as the circuit nodes to be tested. By accurately analyzing the signals at these nodes, an important basis can be provided for the optimal design of the power supply system.

[0054] Among them, the switching node is used to analyze the switching frequency, waveform distortion and switching noise of the switching power supply, the oscillator signal is used to control the switching frequency, the output voltage is used to observe voltage stability, ripple and noise, the feedback signal is used to evaluate the stability and response characteristics of the control loop of the target topology structure, and the compensation network signal is used to verify the stability and dynamic response of the control loop.

[0055] In the embodiment of the present application, after the circuit node to be tested is determined, the real-time waveform data and basic power information of the circuit node to be tested are obtained. Specifically, a multi-channel oscilloscope is used to collect signals from the circuit node to be tested in the switching power supply of the target topology structure to obtain a collected signal, and it is necessary to ensure that the collected data has a sufficient sampling rate to capture the key features of the signal; then the collected signal is transmitted to the target device through a preset application program interface for communication between instruments; and the real-time waveform data and basic power information in the collected signal are extracted using a preset analysis software package running on the target device.

[0056] In a specific implementation, after acquiring the acquisition signal from the circuit node to be tested, the acquisition signal is transmitted to a PC (Personal Computer) through a network interface using the VISA (Virtual Instrument Software Architecture) protocol. The analysis software package running on the PC extracts the real-time waveform data and basic power information from the acquisition signal acquired by the oscilloscope. The real-time waveform data is a time domain waveform image; the basic power information includes amplitude measurement data (such as peak value, peak-to-peak value, root mean square value, mean value) and time measurement data (such as period, frequency, pulse width, rise time and fall time).

[0057] In addition, in the embodiment of the present application, the target topology structure may be a power circuit structure implemented by an LLC resonant network. Figure 2As shown in the figure, a block diagram of a switching power supply with power factor correction (PFC) is shown as an example. The power supply system is usually composed of multiple parts such as electromagnetic filtering and rectification circuit, auxiliary power supply circuit, PFC circuit, control circuit, synchronous rectification drive circuit and optocoupler feedback circuit. At present, high-power switching power supplies widely use the LLC resonant converter topology. Its working principle relies on the resonance characteristics and improves the conversion efficiency through the resonance effect between inductance and capacitance. Due to its high power density, the LLC converter can achieve high power output in a small volume.

[0058] It can be seen that the switch node signal in the power supply circuit in the embodiment of the present application has a high-frequency characteristic and is very sensitive to the matching of the resonant frequency and the switching frequency. Traditional measurement methods are difficult to capture subtle waveform distortions, and its complex structure puts higher requirements on the testing means. Traditional current and voltage measurement methods cannot meet the needs of comprehensive testing of such power supplies.

[0059] Step S12: performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and performing time-frequency analysis on the time-frequency spectrum using a pre-built convolutional neural network to output an analysis and processing result.

[0060] In view of the circuit characteristics of the switching power supply of the target topology, in the embodiment of the present application, the acquired real-time waveform data is converted into a corresponding time-frequency (TF) spectrum using a time-frequency transformation method. The time-frequency diagram uses time and frequency as two dimensions, and can effectively analyze the spectral characteristics of the signal and its changes over time. This image shows the intensity changes of each frequency component in the signal, and the signal is fully characterized through the time-frequency diagram, which helps to identify transient characteristics and reveal the periodicity, burstiness and noise characteristics of the signal.

[0061] Furthermore, a CNN network (Convolutional Neural Network) is constructed to analyze the TF map of the circuit nodes to be tested in the switching power supply. The TF map shows in detail the time domain and frequency domain characteristics of the circuit node signals to be tested in the switching power supply, providing effective input data for CNN analysis. When the TF map is input into the CNN network, the network focuses on identifying and classifying abnormal features and patterns in the TF map, further improving the accuracy of fault detection.

[0062] It should be noted that the convolutional neural network is trained using the known correct normal waveform data of the switching power supply to identify the abnormal working state of the switching power supply. Since the CNN convolutional neural network can effectively extract the time domain waveform features in the TF spectrum, the time domain waveform data of the circuit node to be tested is compared with the normal waveform through the special convolution kernel in the CNN network to find the abnormal points of the power supply waveform.

[0063] Step S13: using the analysis and processing result and the basic power information as input parameters of a fully connected neural network, and outputting a final test result of the circuit node to be tested through the fully connected neural network.

[0064] In the embodiment of the present application, the analysis and processing results output by the CNN network are data of an array structure that can be used to characterize the current waveform anomaly of the circuit node to be tested, and the analysis and processing structure is not easy to understand. Therefore, a fully connected neural network is used to combine the analysis and processing results of the output of CNN with basic electrical information for comprehensive abnormality detection and judgment, and output the final test results of the circuit node to be tested. The final test result is a result that can directly represent the abnormal waveform of the current circuit node to be tested, so it is more intuitive and can more comprehensively identify and mark the abnormal working state of the switching power supply. It can be seen that the combination of CNN and the fully connected neural network takes advantage of their respective advantages, retaining the advantages of CNN in feature extraction and spatial information retention, and enhancing the generalization ability of the model and the accuracy of the final output through the fully connected neural network.

[0065] In a feasible implementation, an intelligent maintenance and optimization working mode is provided. Specifically, a test report is automatically generated for the final test result, and the test report is analyzed so as to provide the corresponding fault generation cause for the circuit node to be tested according to the test report. It can be seen that in this working mode, a test report can be automatically generated and possible fault causes can be provided, which is crucial for the maintenance and optimization of the power supply system. Intelligent analysis means can quickly locate problems and take improvement measures.

[0066] In addition, for each circuit node to be tested in the switching power supply, a corresponding switching power supply protection circuit can be set. If there is a target circuit node with an abnormal power waveform in the final test result, the switching power supply protection circuit can be started and corresponding protection measures can be taken to protect the target circuit node, such as ensuring the normal operation of the switching power supply, or quickly cutting off the switching power supply to avoid excessive device loss.

[0067] The present application provides a switching power supply testing method, comprising: using a multi-channel oscilloscope to collect signals from a circuit node to be tested in a switching power supply of a target topology structure to obtain real-time waveform data and basic power information of the circuit node to be tested; performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and using a pre-constructed convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply; using the analysis and processing result and the basic power information as input parameters of a fully connected neural network, and outputting a final test result of the circuit node to be tested through the fully connected neural network.

[0068] The beneficial technical effects of the present application are as follows: first, the signal of the circuit node to be tested in the switching power supply of the target topology structure usually has high-frequency characteristics. By converting the real-time waveform data measured by the multi-channel oscilloscope, which is difficult to capture subtle waveform distortion, into a time-frequency spectrum, the time-frequency characteristics of the signal can be accurately displayed, which helps to identify high-frequency noise and waveform distortion in the switching process, thereby improving the analysis accuracy. Secondly, since the power supply circuit of the switching power supply of the target topology structure is very sensitive to the matching of the resonant frequency and the switching frequency, the time-frequency characteristics of different scales can be extracted by using a convolutional neural network to perform time-frequency analysis on the generated time-frequency spectrum, effectively capture different modes of the signal in time and frequency, and output preliminary analysis and processing results, so as to achieve a more accurate and efficient analysis of the switching power supply signal. By comparing with normal waveform data, the static and dynamic performance of the switching power supply can be accurately tested and analyzed, and the abnormal working state of the switching power supply can be identified. Finally, combined with the basic power information of the circuit node to be tested, the fully connected neural network is used to judge the comprehensive data and output the final test results. In this way, the process of fully automated identification of the working state of the switching power supply is realized, which significantly improves the test efficiency and comprehensiveness, reduces manual analysis time and errors, and improves the efficiency and accuracy of abnormal detection of the switching power supply.

[0069] Based on the above embodiment, this embodiment will specifically explain S12 in the above embodiment. Among them, the real-time waveform data is converted into a TF spectrum using a time-frequency conversion method, which can show in detail the time domain and frequency domain characteristics of the node signal of the circuit to be tested of the switching power supply, and provide effective input data for CNN analysis. Specifically, the process of performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum may include the following steps:

[0070] Performing denoising processing on the real-time waveform data by a preset noise processing method to remove high-frequency noise in the real-time waveform data to obtain first processed data;

[0071] Normalizing the first processed data based on a preset data scale to adjust the numerical range of the first processed data to a uniform interval, thereby obtaining second processed data;

[0072] Frame the signal in the second processed data, and perform windowing on each frame of the obtained signal using a preset windowing function to generate third processed data;

[0073] Performing short-time Fourier transform or continuous wavelet transform on the third processed data to convert the time domain representation of the signal in the third processed data into the frequency domain representation, and generating a corresponding spectrum result;

[0074] Performing a two-dimensional matrix conversion on the spectrum result to generate a corresponding time-frequency spectrum;

[0075] Accordingly, the time-frequency analysis of the time-frequency spectrum is performed using a pre-built convolutional neural network to output an analysis result, including:

[0076] Performing format conversion on the time-frequency spectrum based on a preset data format to generate target format data supporting the input of the convolutional neural network;

[0077] The convolutional neural network is used to perform time-frequency analysis on the target format data to output analysis and processing results.

[0078] In the embodiment of the present application, the purpose of generating the TF spectrum is to convert the time domain waveform image of the switching power supply into a time-frequency spectrum to reveal the changing characteristics of the signal in time and frequency, which involves signal preprocessing, time-frequency conversion, spectrum generation and optimization.

[0079] First, signal preprocessing is performed, which includes three steps: denoising, normalization, and windowing. Denoising is the process of denoising the collected waveform data, including the use of noise processing methods such as low-pass filtering and Gaussian filtering to remove high-frequency noise. Normalization can normalize the waveform data to a specific range to unify the data scale and ensure the consistency of time-frequency spectrum generation. Apply a preset windowing function to perform windowing on the signal to reduce boundary effects and spectral leakage. Specifically, the continuous signal is first divided into several short time frames, and the length of each frame is usually determined according to the specific application; then a window function is applied to each frame of the signal. Taking a commonly used Hamming window function as an example, it is defined as:

[0080] ;

[0081] Where: w(n) is the value of the window function; N is the length of the window (that is, the total number of samples in the window); n is the index of the sample.

[0082] Secondly, for time-frequency conversion, short-time Fourier transform (STFT) or continuous wavelet transform (CWT) can be used to convert the time domain signal into a frequency domain signal. It can be understood that the waveform data in the time domain spectrum is in vector form, which is essentially time series data. The time-frequency conversion process is to realize short-time Fourier transform or continuous wavelet transform through windowing to generate a time-frequency spectrum.

[0083] Among them, short-time Fourier transform (STFT):

[0084] ; is the window function, is the center of the window, It's the frequency.

[0085] Continuous Wavelet Transform (CWT):

[0086] ; is the wavelet basis function, is the scale parameter, is the complex conjugate of the wavelet.

[0087] The time domain signal can be converted into a detailed time-frequency spectrum through STFT or CWT, which provides rich feature information for subsequent convolutional neural network analysis. Further, the spectrum is generated to convert the waveform data into a two-dimensional matrix, one dimension of which represents the time series and the other dimension represents different channels or signal features. Finally, optimization is performed and the formed TF graph is used as the input of the CNN network to ensure that this information is provided to the CNN in a suitable format to support in-depth analysis and anomaly detection of power supply signals.

[0088] like Figure 3 As shown, the time domain and time-frequency diagrams of ideal waveform signals of several important nodes of the switching power supply are displayed. Figure 3 The ideal TF spectra of the AC (Alternating Current) power input sine wave, the 50% duty cycle square wave, and the triangular wave and sawtooth wave generated by the PWM (Pulse Width Modulation) comparator are shown in Figure 1. Figure 4 The waveform of the transformer driven by the switching power supply main control chip through the MOSFET (Metal-Oxide-SemiconductorField-Effect Transistor). Taking the ideal square wave drive signal as an example, compared with the ideal square wave drive signal, Figure 4Signal overshoot and oscillation attenuation wave appear on the rising edge of the square wave, reflecting the dynamic characteristics of the switching power supply. Factors such as the speed of MOSFET, dead time and leakage magnetic flux of the high-frequency transformer will affect the oscillation attenuation wave. As can be seen from the figure, the square wave with overshoot oscillation is significantly different from the standard square wave on the TF graph. Traditional analysis methods rely on the experience of testers, are relatively arbitrary, and cannot quantitatively analyze the waveforms of key nodes of the power supply. Therefore, in the embodiment of the present application, a convolutional neural network is used to accurately quantitatively analyze these waveforms, reflecting its unique advantages.

[0089] Based on the above embodiment, by comparing CNN with normal power supply data, the static and dynamic performance of the power supply can be accurately tested and analyzed, thereby realizing the automation and intelligence of dynamic performance testing.

[0090] In an embodiment of the present application, after the TF spectrum is formed, it is sent to a convolutional neural network for further analysis, and the sequence data corresponding to the time-frequency spectrum is divided into multiple windows according to the preset window length, and a pre-built convolutional neural network is used to perform time-frequency analysis on each window based on a preset overlapping window strategy to output the analysis and processing results. For example, the TF sequence is divided into windows of length 512 so that CNN processing can be performed on each window. It can be understood that the use of overlapping window strategies can ensure the continuity of data and the integrity of information.

[0091] Specifically, the process of building a convolutional neural network can include the following steps:

[0092] Constructing a convolutional neural network model using a preset deep learning framework, and defining a corresponding network structure for the convolutional neural network model;

[0093] Corresponding layer parameters are configured for the network structure, and the configured convolutional neural network model is trained using the normal waveform data to obtain a trained convolutional neural network.

[0094] In the embodiment of the present application, a deep learning framework is used to build a CNN model. In a specific embodiment, a library such as Keras or PyTorch can be used to build a CNN model. Among them, Keras is an open source neural network library that supports fast and convenient neural network construction; PyTorch is an open source machine learning library that is widely used in computer vision and natural language processing.

[0095] After the CNN model is built, the corresponding network structure is defined for it, where the network structure includes input layer, convolution layer, activation layer, pooling layer, data flattening layer (Flatten), fully connected layer and output layer. Furthermore, appropriate parameters are configured for each layer in the network, such as the number and size of convolution kernels and the selection of activation functions.

[0096] When defining a convolution layer, in a specific implementation, two convolution kernels may be set, for example, and The size of the matrix is ​​used to extract features of different scales. Convolution kernel: used to capture features that are shorter in time and longer in frequency; Convolution kernel: It is used to capture features that are longer in time and shorter in frequency. These two complementary characteristics are more conducive to analyzing the key points of the power supply.

[0097] When performing a convolution operation, given the input feature map and convolution kernel, the formula for the convolution operation is:

[0098] ;

[0099] Where I is the input image, K is the convolution kernel, H and W are the height and width of the convolution kernel respectively, and x and y are the output positions of the convolution operation.

[0100] The convolution kernel is applied to the TF spectrum. Through these convolution operations, time-frequency features of different scales can be extracted, thereby achieving more accurate and efficient analysis of the switching power supply signal. The different sizes of the convolution kernel can effectively capture the different patterns of the signal in time and frequency.

[0101] like Figure 5 The figure shows a flowchart of the overall switching power supply test executed by the program. First, the acquired real-time waveform data is converted into a TF graph. In this process, the Tensor Flow machine learning library and the scipy.signal.stft function can be used to perform short-time Fourier transform to convert the waveform data into a TF graph. After the conversion, a two-dimensional array tf_image is obtained, which represents the amplitude of time and frequency. Further, a convolutional neural network is used to perform convolution processing on the TF graph. During this period, the TF graph will be normalized, visualized, reshaped to adapt to CNN input, and increase channel dimensions. The normalization of the TF graph can eliminate the numerical differences between different signals; Reshaping and increasing channel dimensions are used to convert the generated time-frequency graph into a format suitable for CNN input. Secondly, to build a CNN model, you can use Keras to build a simple CNN model, which contains two convolutional layers (using and The convolution kernel is then followed by the Flatten layer and the fully connected layer. The output layer is set to sigmoid, which is suitable for binary classification problems and can be adjusted as needed. The CNN model uses the training data set to iteratively train the model, calculates the prediction results through forward propagation, and adjusts the model parameters through back propagation. The final prediction output is data of an array structure that can be used to characterize the abnormal waveform of the current circuit node under test. Finally, the output of the CNN is combined with basic power information, such as amplitude information, time measurement information, and edge measurement information, and input into the fully connected neural network. Classification or regression is performed through the fully connected neural network, and activation functions, maximum pooling, and random discarding are combined to further extract and optimize features. Finally, the softmax layer outputs the judgment result of whether the power waveform is normal.

[0102] It can be seen that through TF graph and CNN analysis, we can deeply understand the resonant behavior and power conversion characteristics of LLC circuits, optimize power supply design, improve overall system efficiency and performance, and reduce energy loss and electromagnetic interference. It can provide more accurate signal analysis, more efficient anomaly detection and more optimized power supply design solutions, realize the full automation process from data acquisition to anomaly detection, realize more accurate and efficient switching power supply analysis, and significantly improve test efficiency and comprehensiveness, which is especially important in complex power supply systems.

[0103] Correspondingly, the present application also discloses a switching power supply testing device, see Figure 6 As shown, the device comprises:

[0104] The data acquisition module 11 is used to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure using a multi-channel oscilloscope to obtain real-time waveform data and basic power information of the circuit nodes to be tested;

[0105] The convolutional neural network processing module 12 is used to perform time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and use a pre-built convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply

[0106] The test result output module 13 is used to use the analysis and processing results and the basic power information as input parameters of a fully connected neural network, and output the final test result of the circuit node to be tested through the fully connected neural network.

[0107] Among them, for more specific working processes of the above-mentioned modules, please refer to the corresponding contents disclosed in the aforementioned embodiments, which will not be repeated here.

[0108] It can be seen that the above scheme of this embodiment includes: using a multi-channel oscilloscope to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure to obtain real-time waveform data and basic power information of the circuit nodes to be tested; performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and using a pre-built convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply; using the analysis and processing results and the basic power information as input parameters of a fully connected neural network, and outputting the final test result of the circuit node to be tested through the fully connected neural network.

[0109] The beneficial technical effects of the present application are as follows: first, the signal of the circuit node to be tested in the switching power supply of the target topology structure usually has high-frequency characteristics. By converting the real-time waveform data measured by the multi-channel oscilloscope, which is difficult to capture subtle waveform distortion, into a time-frequency spectrum, the time-frequency characteristics of the signal can be accurately displayed, which helps to identify high-frequency noise and waveform distortion in the switching process, thereby improving the analysis accuracy. Secondly, since the power supply circuit of the switching power supply of the target topology structure is very sensitive to the matching of the resonant frequency and the switching frequency, the time-frequency characteristics of different scales can be extracted by using a convolutional neural network to perform time-frequency analysis on the generated time-frequency spectrum, effectively capture different modes of the signal in time and frequency, and output preliminary analysis and processing results, so as to achieve a more accurate and efficient analysis of the switching power supply signal. By comparing with normal waveform data, the static and dynamic performance of the switching power supply can be accurately tested and analyzed, and the abnormal working state of the switching power supply can be identified. Finally, combined with the basic power information of the circuit node to be tested, the fully connected neural network is used to judge the comprehensive data and output the final test results. In this way, the process of fully automated identification of the working state of the switching power supply is realized, which significantly improves the test efficiency and comprehensiveness, reduces manual analysis time and errors, and improves the efficiency and accuracy of abnormal detection of the switching power supply.

[0110] Furthermore, the present application also discloses an electronic device. Figure 7 This is a structural diagram of an electronic device 20 according to an exemplary embodiment, and the content in the diagram cannot be considered as any limitation on the scope of use of the present application.

[0111] Figure 7The present invention provides a schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the switching power supply test method disclosed in any of the above embodiments. In addition, the electronic device 20 in this embodiment may specifically be a computer.

[0112] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0113] In addition, the memory 22 as a carrier for resource storage may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222 and data 223, etc. The data 223 may include various data. The storage method may be temporary storage or permanent storage.

[0114] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, which can be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program that can be used to complete the switching power supply test method performed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include a computer program that can be used to complete other specific tasks.

[0115] Furthermore, the embodiment of the present application also discloses a computer-readable storage medium, wherein the computer-readable storage medium mentioned here includes a random access memory (Random Access Memory, RAM), a memory, a read-only memory (Read-Only Memory, ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a magnetic disk or an optical disk, or any other form of storage medium known in the technical field. Wherein, when the computer program is executed by the processor, the aforementioned switching power supply test method is implemented. For the specific steps of the method, reference can be made to the corresponding contents disclosed in the aforementioned embodiment, and no further description will be given here.

[0116] Furthermore, an embodiment of the present application also provides a computer program product, including a computer program / instruction, which implements any one of the above-mentioned switching power supply testing methods when executed by a processor.

[0117] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0118] The steps of the switching power supply test method or algorithm described in conjunction with the embodiments disclosed herein can be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module can be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0119] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0120] The above is a detailed introduction to a switching power supply testing method, device, equipment and medium provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for a general technician in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A switching power supply testing method, characterized in that: include: Using a multi-channel oscilloscope to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure, so as to obtain real-time waveform data and basic power information of the circuit nodes to be tested; Performing time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and performing time-frequency analysis on the time-frequency spectrum using a pre-built convolutional neural network to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply; The analysis and processing results and the basic electrical quantity information are used as input parameters of a fully connected neural network, and a final test result of the circuit node to be tested is output through the fully connected neural network.

2. The switching power supply testing method according to claim 1, characterized in that: The method of using a multi-channel oscilloscope to collect signals from a circuit node to be tested in a switching power supply of a target topology structure to obtain real-time waveform data and basic power information of the circuit node to be tested includes: Using a multi-channel oscilloscope to collect signals from a circuit node to be tested in a switching power supply of a target topology structure to obtain a collected signal, and transmitting the collected signal to a target device through a preset application program interface for communication between instruments; The real-time waveform data and basic power information in the collected signal are extracted using a preset analysis software package running on the target device; wherein the real-time waveform data is a time domain waveform image, and the basic power information includes amplitude measurement data and time measurement data.

3. The switching power supply testing method according to claim 1, characterized in that: The target topology is a power circuit structure implemented by a resonant network; accordingly, the method further includes: Determine any one or more of a switch node, an oscillator signal, an output voltage, a feedback signal, and a compensation network signal in the switching power supply of the target topology as a circuit node to be tested; Among them, the switching node is used to analyze the switching frequency, waveform distortion and switching noise of the switching power supply, the oscillator signal is used to control the switching frequency, the output voltage is used to observe voltage stability, ripple and noise, the feedback signal is used to evaluate the stability and response characteristics of the control loop of the target topology structure, and the compensation network signal is used to verify the stability and dynamic response of the control loop.

4. The switching power supply testing method according to claim 1, characterized in that: The time-frequency conversion of the real-time waveform data to generate a corresponding time-frequency spectrum includes: Performing denoising processing on the real-time waveform data by a preset noise processing method to remove high-frequency noise in the real-time waveform data to obtain first processed data; Normalizing the first processed data based on a preset data scale to adjust the numerical range of the first processed data to a uniform interval to obtain second processed data; Frame the signal in the second processed data, and perform windowing on each frame of the obtained signal using a preset windowing function to generate third processed data; Performing short-time Fourier transform or continuous wavelet transform on the third processed data to convert the time domain representation of the signal in the third processed data into the frequency domain representation, and generating a corresponding spectrum result; Performing a two-dimensional matrix conversion on the spectrum result to generate a corresponding time-frequency spectrum; Accordingly, the time-frequency analysis of the time-frequency spectrum is performed using a pre-built convolutional neural network to output an analysis result, including: Performing format conversion on the time-frequency spectrum based on a preset data format to generate target format data supporting the convolutional neural network input; The convolutional neural network is used to perform time-frequency analysis on the target format data to output analysis and processing results.

5. The switching power supply testing method according to claim 1, characterized in that: The process of building a convolutional neural network includes: Constructing a convolutional neural network model using a preset deep learning framework, and defining a corresponding network structure for the convolutional neural network model; the network structure includes an input layer, a convolution layer, an activation layer, a pooling layer, a data flattening layer, a fully connected layer, and an output layer; The corresponding layer parameters are configured for the network structure, and the configured convolutional neural network model is trained using the normal waveform data to obtain a trained convolutional neural network.

6. The switching power supply testing method according to claim 1, characterized in that: The using of a pre-built convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output analysis and processing results includes: The sequence data corresponding to the time-frequency spectrum is divided into multiple windows according to a preset window length, and a pre-built convolutional neural network is used to perform time-frequency analysis on each window based on a preset overlapping window strategy to output the analysis and processing results.

7. The switching power supply testing method according to any one of claims 1 to 6, characterized in that: After the final test result of the circuit node to be tested is outputted by the fully connected neural network, the method further includes: A test report is automatically generated according to the final test result, and the test report is analyzed so as to provide corresponding fault generation reasons for the circuit node to be tested according to the test report.

8. A switching power supply testing device, characterized in that: include: A data acquisition module is used to collect signals from the circuit nodes to be tested in the switching power supply of the target topology structure using a multi-channel oscilloscope to obtain real-time waveform data and basic power information of the circuit nodes to be tested; A convolutional neural network processing module is used to perform time-frequency conversion on the real-time waveform data to generate a corresponding time-frequency spectrum, and use a pre-built convolutional neural network to perform time-frequency analysis on the time-frequency spectrum to output an analysis and processing result; wherein the convolutional neural network is trained using known correct normal waveform data of the switching power supply to identify abnormal working conditions of the switching power supply A test result output module is used to use the analysis and processing results and the basic power information as input parameters of a fully connected neural network, and output the final test result of the circuit node to be tested through the fully connected neural network.

9. An electronic device, characterized in that: The electronic device comprises a processor and a memory; wherein the memory is used to store a computer program, and the computer program is loaded and executed by the processor to implement the switching power supply testing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer programs; wherein when the computer program is executed by the processor, the switching power supply testing method according to any one of claims 1 to 7 is implemented.

Images