| Summary: | 碩士 === 雲林科技大學 === 資訊管理系碩士班 === 96 === In the manufacturing process of semiconductors, not only it can increase productivity, but also can decrease the unnecessary cost of human and resource and increase its competitiveness if a company can appropriately adjust its product control and find out its manufacturing problem. In this study, we apply the back-propagation neural networks (BPN)to train and test different product manufacturing, adopts the steepest descent method to train and update weight values. and try to find out inappropriate product control mechanisms. we create the simulated process data of chemical mechanical polishing (CMP) process and use the data to validate and verify this process control system.
In the manufacturing process many disturbances exist which usually cause the process output to draft or shift from the design output. In this research an on-line process prediction and control for multiple-input and multiple-output (MIMO) systems, based on back-propagation neural networks, is developed. Moreover, a comparison of estimated performance between the proposed neural network predictor and a Box-Jenkins predictor is also given. A simulation study based on the chemical mechanical polishing process is used to verify the proposed controller. The simulation results show that the proposed neural network controller can effectively reduce the variance of output.
|
|---|
