基于RBF神经网络预测模型的VLSI生产线智能控制算法

Intelligent Control Algorithm for VLSI Manufacturing Line Based on RBFNN Predictive Model

为了对半导体生产线进行预测控制,首先研究了利用RBF神经网络对半导体生产线建立预测模型的方法;然后利用马氏决策过程递阶方法的思路,将系统状态分成若干互不相交的区间,系统状态所属区间改变时,根据预测模型的输出,即预测的系统状态,利用所提出的智能控制算法确定将要采取的投料策略和调度策略组合对生产线进行控制.实验表明,该算法的总体性能优于若干其他常用算法.

A predictive control problem is discussed via radial basis functions neural network （RBFNN）. First RBFNN is introduced to build a predictive model for semiconductor manufacturing system. And then the hierarchical solution of Markov decision processes is used to divide state space into several independent units. When the state unit transfers, the optimized combination of release control and scheduling policy is achieved by the proposed intelligent control algorithm. The emulations show that the predictive control algorithm outgoes some other heuristic algorithms.