Design of Integrated Circuit Chip Fault Diagnosis System Based on Neural Network

Abstract

This paper focused on the application of neural network in fault diagnosis and its implementation on FPGA. The function of the feature parameter processing module is to process the feature parameters into a form suitable for the input of the neural network model. The feature parameter processing module includes a receiving algorithm, a digital signal processing algorithm, a Kalman filtering algorithm, and a dispersion normalization algorithm, all of which are designed using Verilog language and implemented on an FPGA. The function of the neural network diagnosis module is to analyze the feature parameters and predict the failure state of the system to be tested; the neural network diagnosis module includes a neural network training platform and a feedforward neural network model, wherein the neural network training platform is designed using Python language and implemented by software; The feedforward neural network model is designed using Verilog language and implemented on an FPGA. The test results show that when the number of training exceeds 2000 times, the failure state diagnosis is more than 97% stable for the high-temperature failure diagnosis accuracy of the CMOS static memory cell circuit and the JFM4VSX55RT FPGA.