基于BPANN的4-CBA软测量模型研究  

The Research of 4-CBA Soft-sensor Model Based on BPANN

作　　者：胡永有[1] 古勇[1] 苏宏业[1] 王朝辉[1] 褚健[1] 

机构地区：[1]工业控制技术国家重点实验室浙江大学先进控制研究所,杭州310027

出　　处：《仪器仪表学报》2003年第3期226-230,240,共6页Chinese Journal of Scientific Instrument

基　　金：国家杰出青年科学基金 ( NSFC:60 0 2 5 3 0 8)资助项目

摘　　要：本文提出了一个基于 3层 BP人工神经网络 (BPANN)新的 4 - CBA软测量模型。在 MATL AB软件平台上 ,利用两种改进 BP算法以及由正交试验和机理模型仿真所产生的大量数据样本对不同网络结构模型进行了学习训练和验证 ,并对算法的训练效果进行了比较研究。仿真结果表明基于 L evenberg- Marquardt学习规则的合适结构 BPANN软测量模型 ,不仅学习快速 ,预测精度也远高于经验回归模型 ,为实现精对苯二甲酸 (PTA)产品中 4 - CBA含量的实时。A new 4 carboxybenzaldehyde (4 CBA) soft sensor model based on three layers Back Propagation artificial neural network is developed. Some kinds of network models constructed by different neuron numbers of the hidden layer are compared in two respects:learning effect and prediction precision of test samples. Two kinds of modified BP learning algorithms are used to train the BP models with enough data samples derived from orthogonal experiments and simulation by the p Xylene oxidation kinetic model. All training and testing procedures of these BP network models are performed by programming on MATLAB. The simulation results prove the 4 CBA soft sensor model we built is far more precise and efficient than the empirical regress model (ERM) and it is possible to realize the quality control of purified terephthalic acid (PTA) product promptly in the commercial reactor.

关 键 词：精对苯二甲酸 PTA 4-CBA 软测量 BP神经网络 改进BP算法 回归模型 

分 类 号：TQ245.12[化学工程—有机化工] TP274[自动化与计算机技术—检测技术与自动化装置]