基于BP神经网络的复杂过程参数优化方法研究

Study on Parameter Optimization Method of Complex Process Based on BP Neural Network

针对生产过程中存在响应变量受多重复杂因素的影响,响应曲面模型不能解决的复杂因素影响的参数优化问题,提出一种基于BP神经网络的参数优化方法。BP神经网络具有较高的学习能力和映射能力,可以解决响应曲面模型的不足,建立良好的映射模型。利用BP神经网络模型的高泛化能力对最优参数组合寻优搜索,以较少的试验数据得到理想的参数设计。防眩玻璃制备工艺参数优化的分析结果表明,基于响应曲面模型的参数优化方法具有局限性,而BP神经网络的优化方法能够将透过率提高到92.22%,验证了该参数优化方法的有效性和可行性。

Response variables influenced by multiple complex factors in the real production process while the response surface methodology couldn’t solve the problem of parameter optimization under the impact of the complex factors.Thus,a parameter optimization method based on BP neural network was proposed.BP neural network has high learning ability and mapping ability that can solve the shortage of the response surface model to establish a good mapping model.Using high generalization ability of BP neural network model search the optimal parameter combination can get the ideal parameter design with less test data.To illustrate computational aspects of the proposed method,one realistic example about anti-glare glass preparation process parameters optimization have been conducted.The analysis result point out the shortage of the parameter optimization method based on response surface model.The BP neural network optimization method can increase transmittance of Anti-glare glass to92.22%that verify the validity and feasibility of proposed optimization method.