PCA-SVM-based Analog Signal Conversion Board Fault Diagnosis System Design And FPGA Implementation

With the development of intelligent electronic industry technology,some analog circuits in traditional electronic equipment are gradually replaced by digital circuits,resulting in electronic products with higher accuracy,faster system operation speed,and lower cost.But the objective world is full of continuous,non-linear natural quantities that need to be processed by analog circuits,so analog circuits cannot be completely replaced by digital circuits.According to relevant data,about 80% of the faults in the equipment come from the analog circuit part,and the tolerance and nonlinearity of the analog circuit itself make it difficult to carry out fault diagnosis.At present,the fault dictionary method is applied to the fault diagnosis of analog circuits,which can only diagnose single faults and hard faults.In recent years,some scientific research projects at home and abroad have also proposed a method based on neural networks,but this method is easy to fall into a local optimal solution,it is difficult to choose a model structure and it cannot solve the problem of small samples.This thesis mainly applies the method based on statistical learning theory-Support Vector Machine(SVM)to fault diagnosis of analog circuits,and the method of sample feature dimension reduction-Principal Component Analysis(PCA)to speed up data convergence.It has been proved in practical applications that support vector machines have the characteristics of simple structure,global optimization,strong generalization ability,and good robustness in solving small sample,nonlinear and high-dimensional fault identification.The MTALAB auxiliary tool was used to find the optimal values of the parameters such as the kernel width and penalty factor of the support vector machine,and the “one-to-one” method was used to improve the accuracy of multi-fault classification to above 95%.This thesis designs a set of fault diagnosis system for the distributed control system(DCS)simulation board of power station,analyzes the typical simulation board,and elaborates the design principle of the system.The core control part of the system uses FPGA to implement the fault diagnosis algorithm,and according to the Avalon bus principle,the on-chip soft core processor Nios II is used to design various peripheral interface controllers.In terms of system hardware,hardware circuit designs such as NAND Flash memory module,high-speed AD acquisition module,LCD display module,and SDRAM cache module have been completed,so that the system has data storage,analog-to-digital signal conversion,result display and other functions.This fault diagnosis system combines the high-speed parallel computing capability of FPGA and the fault recognition capability of support vector machines,so that the system exhibits high speed and high accuracy when performing fault diagnosis on analog circuits.This system is not only used for fault diagnosis of DCS analog boards in power stations,but also provides a reference for the implementation of other analog circuit fault diagnosis systems,and has certain development potential.