Research Of Optimal Run-to-run Control Methods For High-mix Semiconductor Manufacturing

With increasing human demands and emerging new technologies, the high-mix production feature which means that one equipment is used to produce different kinds of products has been more popular in semiconductor manufacturing. The exponentially weighted moving average(EWMA) controller is the most widely used run-to-run control scheme; however, most traditional EWMA controllers do not apply to the high-mix semiconductor manufacturing since they are developed for the single-product one. To address the issue, the research of optimal run-to-run control methods for high-mix semiconductor manufacturing is put forward in this paper.Firstly, a more accurate mathematic model of high-mix semiconductor manufacturing is established in this paper. Based on this model, the performance of the two main controllers(the tool-based EWMA controller and the product-based EWMA controller) for high-mix semiconductor manufacturing is analyzed in depth. The new analysis results extend and supplement the previous research results. The accuracy of these conclusions is validated by simulation experiments.According to the above studies, a new EWMA controller which can overcome defects both of the tool-based EWMA controller and the product-based EWMA controller is proposed for high-mix semiconductor manufacturing in this paper. For this new controller, when the product category of a run is the same with that of the last run, the noise of this run is estimated by EWMA estimator according to data of the last run. Otherwise, the noise of this run is derived by combining that of the latest run in which the same product is processed with drifts during the two runs when other products are processed. Meanwhile, an algorithm to estimate the values of these drifts which are needed in the new controller is also developed. The effectiveness of the new EWMA controller is validated both through theoretical analysis and simulation experiments.Furthermore, a new performance evaluator is developed for the new EWMA controller, which can monitor the issue when the controller performance becomes poor and then can provide adjusting plan. The mean square error of output is chosen as the performance evaluation benchmark after comparing advantages and disadvantages of minimum variance, variance and mean square error, and thus the controller performance is quantified. On the basis, the performance index which is based on the mean square error is derived with respect to the new EWMA controller. Meanwhile, algorithms to identify and estimate model parameters that are relevant with the performance index are proposed. The effectiveness and advantages of the proposed performance evaluator are validated through simulation experiments.At the end of this paper, the research work is summarized and some prospects are put forward.