Reliability Estimation of Electric Parts by Colour Image Processing and Neural Network

Abstract

This paper describes reliability estimation of electric parts using image processing and a neural network. Efficiency of carbon film resistors in reliability test is estimated from colour image features by feed-forward network. In general, image processing techniques play important roles in many engineering applications and developments. Thus, intelligent image processing techniques are required in the reliability fields in order to establish more efficient quality control schemes. This paper describes some experimental results on reliability estimation using a neural network and image processing for reliability test. Performance degradations of carbon film resistors are estimated from their colour images. We attempt to find appropriate estimated value of resistance at every testing hours, in order to establish a convinient method for reliability estimation. We first discuss some essential issues to be considered in reliability problems. It is shown that the image processing and neural network approarch give good results compared with multiple linear regression and real measured results. Furthermore, the efficiency of the method is indicated by applying it to image data for calculation of MTTF.