Research On Fault Feature Extraction And Intelligent Fusion Diagnosis Of Switched Current Circuits

Switched current(SI)technology is another new analog sampled data signal processing technology after switched capacitor(SC)technology.The difference between the two technologies is that the sampled data of SC technology is through voltage mode and SI Technology current mode.Switched current filters are an important development direction of switched current technology,because unlike switched capacitor filters,switched current filters do not use operational amplifiers or floating capacitors.This not only greatly saves the chip area,but also injects new impetus and provides research direction into the mixed-signal Very Large Scale Integration(VLSI)that can be realized in standard digital CMOS(Complementary Metal Oxide Semiconductor)technology.Due to the non-Gaussian,nonlinear and component parameter tolerance of the response of the switch current circuit,some traditional analog circuit fault diagnosis theories and methods are difficult to achieve the expected results in the fault diagnosis of the switch current circuit.Therefore,it is of particular importance to study the relevant fault diagnosis theory and methods that are efficient and adapt to the characteristics of the switching current circuit.This paper,while adopting analog circuit fault diagnosis methods,it has conducted research and exploration on the fault feature extraction and fault diagnosis and recognition of the switch current circuit based on the inherent characteristics of the switch current filter circuit itself.The main research contents include:1.In order to improve the accuracy of fault diagnosis,from the perspective of fault feature extraction,three feature extraction methods suitable for switching current circuits are proposed,which are the fault feature extraction method based on wavelet packet optimization,the fault feature extraction method based on wavelet fractal,and the information entropy and kurtosis Feature extraction method.Based on the wavelet packet optimal switching current circuit fault feature extraction method,the original response signal of the switching current circuit is first subjected to multi-level wavelet packet decomposition,and then the normalized energy value after layer decomposition is calculated,and the optimal deviation is selected as the evaluation to select the best Wavelet packet basis to construct the optimal fault feature vector.Based on wavelet fractal and kernel principal component analysis method,the fractal dimension is calculated after wavelet decomposition of the node response signal,and then the kernel principal component analysis is used to reduce the eigenvalue dimension to realize the extraction of the optimal fault feature.The feature extraction method based on entropy and kurtosis is a new method of fault feature extraction based on independent component analysis technology.This method utilizes the characteristic that the parameter changes of each element in the switching current circuit are independent of each other,using independence and non-Gaussian as the basis of extraction,information entropy and kurtosis as two non-Gaussian measures,gives a specific fault feature extraction algorithm.2.A fault diagnosis method for switching current circuits based on the integrated radius multiple kernel learning support vector machine(MKL-SVM)and particle swarm optimization(Particle Swarm Optimization,PSO)algorithm is proposed.On the basis of the aforementioned feature extraction,a feature-based integrated radius MKL-SVM method is established to establish a fault diagnosis model,and the integrated radius multi-core learning algorithm learns to obtain the optimal base core combination coefficient through the minimum generalized error bound.At the same time,the PSO algorithm generates kernel parameters for the MKL-SVM method.Through the fault diagnosis simulation of the two fault types of the 6th-order Chebyshev switch current low-pass filter circuit,it is verified that the proposed diagnostic method has good diagnostic accuracy,and the diagnostic performance is better than traditional BPNN,fault dictionary,PSO-SVM and other diagnostic methods.3.A heuristic fusion algorithm combining the characteristics of the Whale Optimization Algorithm(WOA)algorithm and the Genetic Algorithm(GA)algorithm is proposed,and the structural parameters of MKL-SVM are optimized based on this.The proposed algorithm improves the performance of each step of the whale algorithm and genetic algorithm by selecting the optimal value of the whale algorithm and genetic algorithm in each iteration,thereby improving the performance of the fusion algorithm and reducing the premature convergence of the whale algorithm,overcoming the shortcomings of low convergence accuracy of genetic algorithm.By testing 5 benchmark test functions and filter circuit,the experiment shows that the WOA＿GA algorithm proposed in this paper has higher convergence accuracy,faster convergence speed and better robustness.4.Considering the response data of the collected switching current circuit can only represent part of the information of the original data,and cannot reflect the overall appearance of the original data,this paper adopted the Gaussian mixture model(Gaussian mixture model,GMM)method to extract the continuous timing curve of the switching circuit,and combined with the hidden Markov model(hidden Markov model,HMM)theory,a GMM-HMM Theoretical fault diagnosis model was established and tested and verified.