Research On Fault Diagnosis Technology Of Analog Circuits Based On SVDD

In Mechatronics Systems and products,in view of the nonlinearity,tolerance and complexity of the analog circuits,the existing testing technology,testing instruments and diagnostic theory cannot accurately locate,classify and analyze the fault of analog circuits.In this paper,the feature extraction and fault diagnosis method of analog circuits are studied,which is difficult to extract valid faults of analog circuits,and so on,the specific research work is as follows:Firstly,the time-frequency domain Analysis method: Wigner-Ville distribution,wavelet transform and wavelet packet transform,HHT are discussed in detail.Combining with the characteristic of analog circuit and Leapfrog datum circuit illustration,the characteristic data extracted by HHT algorithm is more effective than wavelet technology and Wigner-Ville distribution,and it can accurately reflect the symptom of fault,so it finally chooses HHT as the feature extraction method of analog circuit fault diagnosis technology.Secondly,the SVDD theory method and kernel function,and the effect of different kernel function,penalty coefficient and kernel parameter on the performance of SVDD classifier are studied,and according to the output response characteristics of analog circuits,the training parameters of SVDD are optimized to achieve the goal of training the smallest ultra-sphere.The HHT algorithm is combined with the SVDD classifier of parameter optimization as a fault diagnosis method for analog circuits.Through the simulation of Leapfrog Datum circuit,the accuracy of the final fault diagnosis of SVDD classifier is averaged about 92%,the result shows that the simulated circuit fault diagnosis method is better.Finally,the capture CIS and PSpice combine the triangle wave occurrence circuit principle,its fault data as the computer simulation fault sample;The hardware part of the voice recorder and the signal acquisition system is constructed,the hardware debugging is qualified,and the data acquisition of speech signal is realized by using the LabVIEW software platform and the Advantech DAQ component,which is used as a kind of fault sample in the physical test;All of data is introduced into MATLAB software for data preprocessing and feature extraction,some feature samples are fed into the SVDD classifier to train a minimal hyper-sphere,and the remainder of the feature samples are tested.The validity and accuracy of the analog circuit fault diagnosis method based on SVDD is proved by the two kinds of verification methods of computer simulation,voice recorder fault data physical test.