基于回归神经网络的氧乐果合成过程建模与仿真

氧乐果合成过程具有非线性、时变和不确定性的特点，难以采用常规的建模方法建立模型。BP网络可用于非线性系统的建模，但不能很好的反映实际系统的动态特性。提出了一种静态BP神经网络加多分头延时(TDL)环节构成回归神经网络建立系统动态模型的方法，用于氧乐果合成反应温度控制过程的建模与仿真研究。结果表明，该方法建立的模型误差较小、网络泛化能力较强，能较好的反应实际系统的动态特性，从而为培训操作人员的控制水平提供了低成本的仿真手段，也为自动控制算法的研究提供了仿真模型。

Recurrent Neural Network Based Modeling and Simulation for the Omethoate Synthesis Process

Omethoate synthesis process has the characteristic of the time-variant, nonlinear and uncertainties. It is difficult to model using the conventional modeling methods. BP network can be used to modeling the nonlinear system, but it can't be good at reflecting the dynamic characteristic of the system. The method of the dynamic modeling based on the recurrent neural network, which is made up of static BP network and TDL block, is presented. With this method, it carries on an investigation in the modeling and simulation for the process of omethoate synthesis temperature control. The result indicates the model based on this method has smaller error, stronger ability of network generalization and better dynamic performance of practice system. So it supplies the low cost simulation method for training the operator and presents the simulation model for research on auto-control algorithms.