Analog Circuit Fault Diagnosis Method Based On EMD And Composite Multiscale Entropy

In the era of rapid development of electronic technology and circuit integration technology,electronic products can be said to be everywhere in people's lives.Once the electronic equipment fails,it will bring disastrous impact on people's life.How to ensure the safety and stability of electronic equipment work is a work that people need to pay attention to.However,due to the continuity,nonlinearity and tolerance of analog circuits,the development of fault diagnosis technology has been in a bottleneck state.Traditional methods can not meet the needs of today's high integration and high complexity circuits.Therefore,how to use the tide of artificial intelligence to seek a new breakthrough for analog circuit fault diagnosis technology is the direction of today's researchers.In analog circuit fault diagnosis technology,fault feature extraction and pattern recognition are particularly important,How to extract the key information that can reflect the fault characteristics and construct an efficient and accurate fault pattern classifier are the main research directions.In this paper,the fault feature extraction of circuit is the key research direction,and the main contents are as follows:Firstly,this paper summarizes the research background and development process of analog circuit fault diagnosis technology,and introduces some common fault diagnosis methods.A fault feature extraction method combined with time-frequency statistical information is designed.The working principle and related algorithms of neural network are introduced.The simulation model is established and the corresponding experiments are designed to verify the effectiveness of the time-frequency parameter fusion method.Secondly,the theoretical knowledge of wavelet transform and wavelet packet transform is described in detail.Wavelet packet decomposition and multi-resolution analysis are applied to the fault feature extraction.After wavelet packet decomposition,each frequency band component is reconstructed,and the energy of each reconstructed signal is further calculated.As the feature vector set of the original signal,it is input into BP neural network for training and testing,and the result is satisfactory The results show that the method can effectively improve the accuracy of fault diagnosis.Finally,the method of EMD(empirical mode decomposition)combined with composite multi-scale entropy to extract circuit eigenvectors is explored.EMD algorithm does not need any preprocessing and analysis of the original signal.It can decompose a very complex original signal into multiple single frequency component signals according to the time scale characteristics of the data itself,which is suitable for processing non-linear and non-stationary signals.Multi scale entropy reflects the complexity and autocorrelation of the original sequence at different scales,and has strong anti-interference ability.Aiming at the problem of inaccurate entropy calculation caused by too short coarse-grained sequence,The improved composite multi-scale entropy algorithm is further studied.In this paper,EMD algorithm and composite multi-scale entropy are combined to extract the fault feature vector of the circuit.The simulation experiment is designed to verify it.The final diagnosis accuracy reaches 99.44%.Compared with other methods in the literature,the results show that the EMD algorithm combined with composite multi-scale entropy has obvious advantages.