Robust design for etching process parameters of hard disk drive slider fabrication using data mining and multi response optimization

• Main Authors: Holimchayachotikul,P., Leksakul,K., Guizzi,G.
• Format: Article
• Published: World Scientific and Engineering Academy and Society 2015
• Subjects: Artificial Intelligence; Control and Systems Engineering; Mathematics (all)
• Online Access: http://www.scopus.com/inward/record.url?partnerID=HzOxMe3b&scp=79961044870&origin=inward; http://cmuir.cmu.ac.th/handle/6653943832/38982
• Institution: Chiang Mai University

This paper is to provide a proper insight into solving a multi-response optimization problem using hybrid response surface methodology (RSM) based on robust parameter design (RPD) concepts and data mining (DM) for the multi-response optimization of a RIE process. Over the recent years in many high precision manufacturing organizations, domain experience and engineering judgment have been used to handle multiple response optimization problems. These manners have led to the increase in uncertainty during the decision-making process. This situation has also happened in the hard disk drive fabrication based on reactive ion etching (RIE). This process is highly complicated in setting the parameters using the slider to the right customer specifications. Therefore, this paper presents a hybrid model to optimize the concerning responses of this process in terms of mean and variance. The silicon plates with a patterned wet film photo resistance as a base substrate are used to demonstrate this research. To begin the proposed approach, design of experiment (DOE), named central composite design (CCD), was employed to accumulate the process records and to specify the significant parameters of the process. Then, support vector regression (SVR) was brought into play to institute the nonlinear multivariate relationships between process parameters and responses. Data obtained from DOE were used in the training process. Last but not least, the regression decision tree and domain engineering knowledge were opted for the initial point of optimization algorithms as well. In conclusion, the reduced gradient search algorithm, a hill-climbing procedure and desirability function were adapted to the DOE model while grid search and desirability function were adapted to the SVR model to find the optimum parameter setting. SVR was found to be the technique with the highest prominent accuracy performance, so it was selected to construct a RIE process model. Consequently, the optimum condition from the final model has been efficiently enabled to be applied in real production based on its experiment confirmation.