Research Of Fault Diagnosis Method Of Analog Circuit Based On Improved Support Vector Machines

The reliability of analog circuit determines the reliability of the electronic system, so fault diagnosis level of analog circuit is extremely important. With the ever-increasing intensity and complexity of analog circuit, fault diagnosis has put forward higher requirements to ensure reliability of the circuit. Now, the fault diagnosis method based on support vector machine is one hot current study. It has the important value of academic theory and practical application meaning.Based on the Liaoning province nature science fund project'New methods based on support vector machine study for fault diagnosis and forecast of electric and electron system', the thesis researches and realizes fault diagnosis method of analog circuit based on support vector machines.The thesis access time domain and frequency domain responses of the circuit from PSpice, extract the voltage characteristic feature both in the normal state and in the fault state of the circuit.Based on the learning of statistical learning theory,support vector machine algorithm,the performance of commonly used kernel function and parameter optimization method, considering the advantages and disadvantages of existing support vector machine multi-classification algorithm and the specific classification precision requirements of analog circuit fault diagnosis, the thesis propose improved support vector machine algorithm. The algorithm include preprocessing the sample training set,improvement of the binary tree classification algorithm and incremental sample learning algorithm. The thesis integrates the three algorithms and achieved good results.The thesis develops the analog circuit fault diagnosis platform based on improved support vector machine algorithm, does a lot of analog circuit fault diagnosis experiments including the application of different support vector machine algorithm for fault diagnosis of four analog circuits and two complex industrial process. The thesis compares the improved algorithm with other support vector machine algorithms and analyze fault diagnosis effect of the improved algorithm with different kernel function. The simulation of different analog circuits demonstrate that the improved algorithm get higher classification precision and faster diagnosis speed compared to traditional support vector machine algorithm.