基于误差补偿模型的工艺参数优化研究

Study on Process Parameters Optimization Based on Error Compensation Model

复杂系统或过程参数优化问题往往采用建模发现其潜在规律,再通过优化方法利用该规律获取最佳工艺参数。而建模误差的存在,往往使优化解与实际最佳工艺参数存在差距,难以获得理想性能。为此提出一种基于误差补偿模型的优化决策方法,通过分析并选取影响建模误差的因素,构建误差补偿模型,修正模型,提高决策性能。首先,从数据挖掘角度建立复杂工艺近似模型,并分析影响建模误差的主要因素;其次,以训练误差为导师信号,利用BP网络建立影响因素与建模误差之间的函数关系,确定误差补偿函数;最后,将近似模型与补偿函数叠加作为最终的工艺模型。数学仿真与电路系统优化实验结果表明:误差补偿后,仿真模型得到的优化函数最优值相对误差降低9.63%,而电路系统中决策参数的超调量下降2.17%。可见,补偿模型优化参数控制效果优于近似模型,验证了所提方法对于提高工艺参数优化决策性能有效性。

Optimizing parameters of complex system or process is often based on system or process modeling. However, the modeling error often results in the discrepancy between the optimized solutions and the real optimal parameters and it＇s hard to achieve the ideal properties. For this reason the paper presents a novel method optimizing the decision parameters based on an error compensation model, which enhances the decision-making performance by analyzing the main factors of the modeling error and constructing an error compensation model. Firstly, the approximate model of the complex process based on data mining is established, and the main factors influencing the modeling error are analyzed. Secondly, the compensation function is obtained by modeling the relationship between the influence factors and the model error using BP networks. Finally, the final process model is obtained by adding compensation function and the approximate model. Mathematical simulation and circuit experiment results show that after error compensation, the relative error of the simulation model descends 9.63 %, and the overshoot in circuit system descends 2.17 %. In consequence, the optimization results based on compensation model are better than the approximate model, which verifies the effectiveness of the proposed method.