中文摘要：

为结合参数模型与非参数模型各自的优势,提高建模精度,将非线性半参数模型引入到工业过程建模中.提出基于遗传算法和神经网络的非线性半参数模型的建模方法及结构方案,并给出同时估计参数模型部分和非参数模型部分的交叉循环迭代的算法步骤.对算法中的神经网络的设计和遗传算法进行了改进研究,重点讨论了在增加精英保留策略、增加算法的记忆功能、提出新的适应度计算方法和交叉变异策略等方面的改进措施.采用聚乙烯装置的现场工业数据对方法的有效性进行了验证.结果表明：半参数模型比传统的参数模型有更好的预测精度,并能够较好地跟踪过程变化.

英文摘要：

Nonlinear se-mi-parametric models are introduced for industrial process modeling to improve the modeling accuracy by taking the advantages of both parameter and non-parameter models.The modeling methodology and structure of nonlinear semi-parametric modeling are proposed based on the genetic algorithm and the neural network,and the cross-loop iterative algorithm procedures are also introduced for estimating the parameters of both the parametric and non-parametric parts.Then,the design of neural network and the genetic algorithm are investigated,which increase the elite preserving strategy,enhance the memory function,propose an innovative fitness calculation method,and improve the crossover and mutation strategy.The on-site industrial data of polyethylene plant is used to demonstrate the effective of this method.The result shows that the proposed approach is more accurate in prediction than the conventional parametric models and can better track the variation of the process.