精馏塔神经元网络推断控制

精馏塔馏出物产品的质量控制是精馏过程中非常重要的一部分,根据精馏塔的特点,提出了一种精馏塔产品质量神经网络推断控制方法.为了增加控制的精度,采用神经元网络作为模型预估器.首先给出了作为模型预估器的RBF神经元网络结构及算法实现,隐层节点的作用函数采用高斯核函数,采用K-均值聚类算法,调整中心,利用递推最小二乘,求取输出层的权值;然后,给出了精馏塔产品质量神经网络推断控制框图,并以某16块塔板精馏塔作为仿真对象,对系统进行仿真研究,结果表明,在进料成分改变时,该系统具有良好的抗干扰能力并具有较高的控制精度.

The Neural Network Inferential Control Strategy for Distillation Column

The quality control of product in distillation column is an important part in rectification process. According to the characteristics of the distillation column, a new neural network inferential control strategy of the product is put forward. For increasing the precision of control, the neural network has been used as the model prediction and the Gaussian kernel function has been taken as the network activate function. The K -means algorithm was used to adjust center, and the recursive least square method was used to calculate the output weights of middle layers. Then flowcharts of quality control by neural network for distillation column has been given. The simulation result of a distillation column with 16 boards shows that the system has higher control precision and good anti-interference ability.