为了提高半导体生产线产出率预测的效率与准确性,研究一种基于数学规划模型融合BP神经网络的方法,该方法考虑了已有预测方法所忽略的范围预测问题,以降低预测过程的复杂性.采用主元分析法选取影响产出率的关键性能指标,并借鉴特征加权思想,利用选定的性能指标构建产出率多元线性回归模型;将该回归模型代入线性规划算法中,通过与模糊算法的结合确定产出率的最小预测范围;利用预测范围参数构建非线性规划模型调节BP神经网络参数,从而改进神经网络模型,得到最终的产出率预测值.仿真实验表明,该方法的预测范围精确且过程简便,具有可行性.%To improve the efficiency and accuracy of semiconductor wafer fabrication's throughput rate,an approach which combined mathematical programming and Back Propagation Neural network (BPN) was proposed,which could reduce the complexity of prediction by considering the ignored predicted problem.The throughput rate multiple linear regression model was built by using Principal Component Analysis (PCA) to select key performance indicators,and the linear programming models were used to determine the minimum range of throughput rate.The prediction range parameters were took into nonlinear programming to adjust neural network parameters to get the final forecast result.According to the theoretic analysis and experiments,the proposed method was proved available,the prediction range was more accurate and the forecasting process was relatively simple.
展开▼
