中文摘要：

为解决半导体制造业中小批量、多品种的生产特性和设备老化造成的控制器性能下降的问题，针对最常用的指数加权移动平均控制器，提出一种基于性能评估的在线优化方法。在性能评估过程中，以时间序列分析中定义的输出均方差作为评判基准，提出一种新的性能评价指标。通过评估EWMA控制器的性能得到控制器当前的最优参数并自动整定，同时结合移动窗口技术实现在线优化，使控制器性能始终保持在最佳状态。为验证方法的可行性和有效性，以化学机械抛光过程的在线优化问题为实例进行了仿真实验。

英文摘要：

To solve the manufacturing characteristic of multi-type, small-batch and the quality reduction of controller caused by aging equipments in semiconductor manufacturing industries, an online optimization method based on per- formanee assessment was proposed for the most common used Exponentially Weighted Moving Average （EWMA） controller. In the process of performance evaluation, a new index was proposed by taking the mean square error de- fined in the time series analysis as the evaluation benchmark. The performance of EWMA controller was evaluated to obtain the current optimal parameters of controller and to auto-tuning. Meanwhile, online optimization was realized by combining with the moving window technology, which could make the control system maintain the optimum con- dition all the time. In addition, simulation of chemical mechanical polishing process was carried out to verify the fea- sibility and effectiveness of proposed method.