A neural network forecasting model for consumable parts in semiconductor manufacturing

Purpose – The purpose of this paper is to create a usable life forecast model for consumable parts using neural network approach. It focuses on a consumable probe card used in the semiconductor wafer testing operation. Referring to the relevant resources and the semiconductor testing operation, a fundamental concept is built to develop a probe card management system.Design/methodology/approach – A neural network analysis software package, Q‐net2000, is applied in this study. In this case, there is one hidden layer and the neural network learning rates and momentum are set to 0.1 and 0.7. Forecast the usable life by inputting the initial values of the neural network variables into a back‐propagation neural network.Findings – In this system, the first thing is to collect the production, maintenance and repair data, and then analyze those data by using a neural network methodology to effectively forecast a probe card's usable life. Those data are integrated to derive an optimum timing of placing a probe card...