Using Machine Learning Algorithms to Predict Coating Thickness of Medical Screws

Abstract

With the rapid development of semiconductor technology the storage capacity and running speed of the computer are increasing rapidly The researches and applications of machine learning have been more and more common For the manufacturing fields machine learning is often used for determining the optimal parameters to improve processes In this paper we introduce the process of using machine learning algorithms in real case and four algorithms containing support vector regression (SVR) back propagation neural network (BPNN) M5’ model tree (M5’) and multiple linear regression (MLR) are applied in dealing with a real case The real case is a traditional screw factory that begin to develop medical screws for business transformation Considering the environmental protection and medical material safety the company developed a coated X film that is a new process whose parameter values are very different from the old process In the experiments we collect data of the last one year In order to achieve the best coating thickness machine learning algorithms are applied to predict the coating thickness with the history data And we find that the SVR has the best performances compared with BPNN M5’ and MLR with evaluation metrics mean absolute percent error and root-mean-square error So that the SVR will be used to find the best values of production parameters in the future

Date of Award	2019
Original language	English
Supervisor	Der-Chiang Li (Supervisor)