Research On Analog Circuit Fault Diagnosis Based On SVM

Electronic designs and manufacturing develops rapidly, however, fault detection anddiagnosis technology are relatively slow, which lead to the existing maintenance technology andtesting ability far couldn’t keep up with the demands of the complex electronic equipments; Inthe maintenance and fault diagnosis of circuits, for the characteristics of correspondingcontinuity, circuit nonlinear, device tolerance and fault diversity etc, the analog circuit faultdiagnosis started later and developed very slowly. The fault probability is very high and the costof testing and diagnosing is extremely high. According to the statistics, in the modern electronicequipments, analog circuit is accounted for the20%of the circuit, but its faults’ rate is accountedfor80%of the total circuit faults, and in90%of mixed-signal circuits, analog circuit is althoughaccounted for only5%of the chip’s area, the cost of testing and maintenance is accounted for theentire chip’s cost of test and maintenance by95%;The testing and diagnosing of the analogcircuit has become the bottleneck problems in the modern electronic circuit testing and diagnosisfield.This paper puts forward "the LUPI-SVM Analog Circuit Fault Diagnosis Method Based onDual-space Feature Extraction Algorithm". According to the complementary of the principalcomponent analysis and independent component analysis in feature extraction in the analoguecircuit fault diagnosis, the paper analyzes them from the practical improvement, and putsforward "the double space feature extraction algorithm". According to the existing problems ofsupport vector machine algorithms in analog circuit fault diagnosis, the paper puts forward "theanalog circuit fault diagnosis method based on the LUPI-SVM (SVM of Learning UsingPrivileged Information)",and combining the double space feature extraction algorithm with theLUPI-SVM, the paper puts forward "the LUPI-SVM Analog Circuit Fault Diagnosis MethodBased on Dual-Space Feature Extraction Algorithm ".the main contents include:1.This paper puts forward the adopted and improved algorithms in the feature extractionand fault diagnosis, and then charts the flow chart of analog circuit intelligent fault diagnosis. Itfurther analyzes the theory of the analog circuit faults, and Seriously studies the pretreatmentprocess and the methods of classifying diagnosis, and then puts forward the improved methods infeature extraction, classification and diagnostic on the basis of comparing and analyzing thecommon methods of feature extraction, classification and diagnostic, and then charts the flowchart of analog circuit intelligent fault diagnosis.2.This paper puts forward “the Double-space Feature Extraction Algorithm”. According tothe higher dimension of the original characteristic and complicated computation problems in theanalog circuit fault diagnosis, the paper analyzes the algorithm performance of PCA and ICA,and combined with the complementary of the principal component analysis and independentcomponent analysis in analog circuit fault diagnosis, the paper puts forward“the Double-space Feature Extraction Algorithm” fit for the analog circuit fault diagnosis; and then this paper doesthe simulation analysis in the chapter4quarter4and the measured data validation in the chapter5,the results show that the algorithm could be used in the analog circuit fault diagnosis and gainhigher rate of identification.3.This paper puts forward “the LUPI-SVM Analog Circuit Fault Diagnosis Method Basedon the Dual-space Feature Extraction Algorithm”. It seriously studies the SVM theories andalgorithms, and deeply analyzes and compares the advantages and disadvantages of the basicclassifying algorithms, multi-classification algorithms and other forms’ algorithms of SVM;According to the advantages and disadvantages of the common SVM classification algorithms, ituses and improves the classification algorithm based on the binary tree SVM and analog circuitsfault diagnosis method based on LUPI-SVM; and then it puts forward “the LUPI-SVM AnalogCircuit Fault Diagnosis Method Based on the Dual-space Feature Extraction Algorithm”.Through the typical circuit simulation and analysis, it verifies the feasibility and effectiveness ofthe method.4.This paper combines the proposed method with he automatic comprehensive testdiagnosis system to develop the TPS(Test Program sets) in the project, and then automaticallytests and diagnoses the WJ8615ultrashort wave receiver with the TPS from the machine circuitto the unit circuit, from the unit circuit to the function circuit, from the function circuit to thecomponents, the results of diagnostics demonstrate the validity and rationality of the proposedmethod.