CN-122017403-A - Electromagnetic compatibility fault diagnosis method and system based on data driving

CN122017403ACN 122017403 ACN122017403 ACN 122017403ACN-122017403-A

Abstract

The invention relates to a data-driven electromagnetic compatibility fault diagnosis method and a data-driven electromagnetic compatibility fault diagnosis system, belonging to the technical field of electromagnetic compatibility (EMC); extracting the characteristics of the parameter data, including calculating the statistical characteristics in a time window and extracting the periodic characteristics through frequency domain transformation, carrying out sample equalization processing on the extracted characteristics to balance the number of fault samples and normal samples, constructing a fault diagnosis model by adopting a machine learning algorithm based on the equalized characteristic data, and carrying out electromagnetic compatibility fault prediction and positioning on equipment to be tested by utilizing the fault diagnosis model. The invention can realize accurate collection of multidimensional parameters, automatic extraction of fault characteristics and intelligent prediction of models, and remarkably improve the efficiency and accuracy of positioning electromagnetic compatibility problems.

Inventors

GUO WEI
WANG ZIYAN
LIU PENGFEI
LIN MIN
LEI YANG

Assignees

华东计算技术研究所（中国电子科技集团公司第三十二研究所）

Dates

Publication Date: 20260512
Application Date: 20260126

Claims (10)

1. The electromagnetic compatibility fault diagnosis method based on data driving is characterized by comprising the following steps of: Injecting an interference signal into the equipment to be tested, and collecting multidimensional operation parameter data of the equipment in an electromagnetic environment; extracting the characteristics of the parameter data, including calculating the statistical characteristics in a time window and extracting the periodic characteristics through frequency domain transformation; Sample equalization processing is carried out on the extracted characteristics, so that the number of fault samples is balanced with that of normal samples; Based on the equalized characteristic data, constructing a fault diagnosis model by adopting a machine learning algorithm; And carrying out electromagnetic compatibility fault prediction and positioning on the equipment to be tested by using the fault diagnosis model.
2. The data-driven electromagnetic compatibility failure diagnosis method according to claim 1, wherein the multi-dimensional operation parameters include electrical parameters including voltage, current and power consumption, signal integrity parameters including clock stability and data signal waveforms, electromagnetic interference parameters including common/differential mode current and conduction noise, and thermal parameters including temperature of critical components.
3. The method of claim 1, wherein the collection points of the parameter data include power rails, DC/DC converter areas, high-speed signal transmission lines, core components, clock oscillator areas, analog-digital mixed signal areas, and/or board interface boundary locations.
4. The data-driven electromagnetic compatibility failure diagnosis method according to claim 1, wherein the feature extraction includes: segmenting the time sequence data by adopting a time window with a fixed length; Calculating the average value, variance, maximum value and minimum value in each time window; performing discrete Fourier transform on the time sequence data to obtain a main frequency component, and fitting and extracting periodic characteristics by using a trigonometric function; And normalizing all the extracted features.
5. The method for diagnosing an electromagnetic compatibility failure based on data driving as recited in claim 4, wherein the periodic feature extraction includes converting a signal from a time domain to a frequency domain by a fast Fourier transform, fitting a sine function model by a nonlinear optimization method after identifying a principal frequency component, and extracting frequency information from the fitting parameters.
6. The method for diagnosing an electromagnetic compatibility failure based on data driving of claim 1, wherein said sample equalization processing includes oversampling said few classes of failure samples by using a SMOTE method based on K-Means clustering, and deleting said majority classes of normal samples from said class boundaries by using an undersampling method based on Euclidean distance.
7. The method for diagnosing the electromagnetic compatibility failure based on data driving according to claim 1 or 6, wherein the feature selection comprises the steps of firstly carrying out primary screening by calculating the correlation between the feature and a target variable through a statistical method, then carrying out secondary screening by utilizing feature importance evaluation based on a tree model, and determining an optimal feature subset through cross verification.
8. The method for diagnosing an electromagnetic compatibility failure based on data driving of claim 1, wherein the selection of the machine learning algorithm includes adopting a random forest or gradient lifting decision tree when the data size is smaller than a preset threshold, adopting a convolutional neural network, a cyclic neural network or a transducer model when the data size is larger than the preset threshold, and adopting a CNN-LSTM combination model for data with time-space correlation characteristics.
9. The data-driven electromagnetic compatibility failure diagnosis method according to claim 1, wherein the injection of the interference signal includes the steps of: electrifying the board to be tested and achieving a stable working state; collecting reference electromagnetic characteristic data; Connecting an interference signal generator to a power port of the board card through an injection clamp, and injecting an interference signal according to a specified strength until a sensitive phenomenon occurs; Recording electromagnetic characteristic data in an abnormal state, and comparing and calibrating the electromagnetic characteristic data with reference data; And gradually reducing the interference intensity until the board card is recovered to be normal, and repeating the test until no new fault is generated.
10. A system applying the data-driven electromagnetic compatibility failure diagnosis method according to any one of claims 1 to 9, characterized by comprising: The interference signal generation module is used for generating and amplifying an interference signal; The data acquisition module comprises a measuring sensor and an instrument and is used for acquiring multidimensional electromagnetic compatibility parameters of the equipment to be measured; the data processing module is used for carrying out feature extraction, sample equalization and feature selection on the acquired data; the model training and diagnosing module is used for constructing a fault diagnosing model and outputting diagnosing results.

Description

Electromagnetic compatibility fault diagnosis method and system based on data driving Technical Field The invention relates to the technical field of electromagnetic compatibility (EMC), in particular to a data-driven electromagnetic compatibility sensitive diagnosis method and a data-driven electromagnetic compatibility sensitive diagnosis system, which are suitable for sensitive phenomenon analysis and optimal design of a complex electronic system and a digital circuit board card. Background With the improvement of the integration level of an electronic system and the continuous complicating of an electromagnetic environment, the problem of electromagnetic compatibility of the electronic system becomes more prominent, and aiming at the frequent problem of electromagnetic compatibility sensitivity phenomenon caused by the easy interference of a digital circuit board card, the realization of effective electromagnetic compatibility fault diagnosis becomes important. Based on a conduction sensitive excitation test, a data-driven electromagnetic compatibility data acquisition and characteristic engineering method is explored, and classification and prediction of EMC sensitive phenomena are realized by using an intelligent algorithm, so that means and methods are provided for rapidly and accurately positioning and diagnosing electromagnetic compatibility problems. Through analyzing and representing the electromagnetic characteristic parameter of the high relativity of the fault phenomenon, can provide targeted optimization and improvement thinking for the electromagnetic compatibility forward design of the digital circuit board card, and solve the problems of the traditional empirical type and the error-testing type that the electromagnetic compatibility fault rectification work is too low in efficiency and complicated. The electromagnetic compatibility standard provides a test method for the electromagnetic compatibility problem of equipment and systems, and is a main way for solving the problems. By performing electromagnetic compatibility tests, electromagnetic compatibility weaknesses existing in the system can be accurately analyzed, so that improvement and modification can be performed in a targeted manner to ensure the stability and reliability of the system. However, due to the mechanical complexity and the phenomenon instability of the electromagnetic compatibility problem, the most common electromagnetic compatibility problem can only be avoided by following the related standard, and no radical treatment approach is available for the electromagnetic compatibility failure occurring under external interference and other complex conditions. Conventional electromagnetic compatibility problem troubleshooting is mainly solved by following established standards and empirical tests. However, in complex environments, this approach has limited efficiency and effectiveness, making it difficult to accurately diagnose faults and effectively optimize designs. In the research of fault diagnosis problems, the main focus is on the system software and hardware body at present, and systematic solutions are provided in different subdivision fields and application scenes, but the related research aiming at the electromagnetic compatibility field is relatively less, and the technical development is still in the exploration and test stage. In the early years, related researches explored an electromagnetic compatibility fault diagnosis method based on a fault tree, and a fault tree model is constructed by researching three elements of an electromagnetic interference source, an interference coupling path and sensitive equipment so as to realize EMC fault diagnosis. On the basis, subsequent researches further develop on-line EMC fault diagnosis platform software, and the practical degree of the technology is greatly improved. However, such methods generally rely on prior modeling and correlation analysis, and cannot mine analysis mechanism rules from a large amount of actual data, so that the method is only suitable for application in specific scenes. Further, researches are carried out to establish a nonlinear relation between fault data and fault sources, an electromagnetic compatibility fault diagnosis method based on SVM (Support Vector Machine ) is provided, and an SVM model is trained to predict faults. Meanwhile, there have been studies on the introduction of PNN neural networks based on training of learning fault diagnosis models. The method preliminarily verifies the applicability and feasibility of the intelligent algorithm in solving the EMC fault diagnosis problem. Disclosure of Invention The invention aims to solve the following technical problems: 1. designing a data acquisition method for collecting electromagnetic compatibility sensitive data, conducting sensitive injection by adopting a specific signal, exciting electromagnetic compatibility sensitive phenomenon of an electronic board card, and collecting oper