论文部分内容阅读
针对化学机械研磨(CMP)过程非线性、时变和不易在线测量的特性,提出了基于径向基函数(RBF)神经网络和微粒群(PSO)算法的CMP过程run-to-run(R2R)预测控制器NNPR2R。首先通过样本数据用减聚类算法和最小二乘法构建CMP过程的RBF神经网络预测模型,解决了复杂CMP过程难以建立精确数学模型的难题和提高了预测模型的精度。然后通过PSO算法滚动优化求取控制律,解决了基于导数的优化技术易于陷入局部最优的问题并提高了控制精度。仿真结果表明,CMP过程NNPR2R控制器的性能优于常规的EWMA方法,有效抑制了过程漂移和减小了不同批次间产品的差异,显著降低了材料去除率(MRR)的均方根误差。
Aiming at the characteristics of non-linear, time-varying and difficult to measure on-line during chemical mechanical polishing (CMP) process, a run-to-run (R2R) CMP process based on Radial Basis Function (RBF) neural network and Particle Swarm Optimization (PSO) Predictive Controller NNPR2R. Firstly, RBF neural network prediction model of CMP process is constructed by using clustering algorithm and least square method through sample data, which solves the difficult problem of complex mathematical model in complex CMP process and improves the accuracy of prediction model. Then the algorithm is scrolled and optimized by the PSO algorithm to solve the problem that the derivative-based optimization technique is apt to fall into the local optimum and improve the control accuracy. The simulation results show that the performance of NNPR2R controller is better than that of the conventional EWMA method in CMP process, which can effectively suppress the process drift and reduce the difference between the products in different batches, and significantly reduce the root mean square error of material removal rate (MRR).