PLS方法应用于PTA脱水塔的软测量建模

论文编号:HG131  论文字数:19938,页数:45

摘 要：及时测定化工过程变量，对确保生产过程稳定、有效控制产品质量具有重要意义。目前基于实验测定的分析，时间滞后明显，不能实现产品质量精确控制的要求，而基于过程参数的软测量模型，通过采集实时样本数据，构建其与性质参数间的经验数学模型，可实时用于产品质量的在线分析和控制。本论文将基于PTA装置溶剂脱水塔化工过程采集的时序样本数据，并考虑数据属高维强相关的小样本，拟采用有偏估计的偏最小二乘回归方法，为塔顶排出的醋酸含量建立软测量模型，并将用其它经典软测量建模方法，以作比较。

关键词：软测量；偏最小二乘；PTA溶剂脱水塔；回归；主成份回归

Abstract: It is well known that to measure and estimate the chemical process variables in time has vital significance in ensuring process stabilization and effectively controlling its product quality.Based on the experimental determination of the present analysis, the time lag obviously, can not be achieved precise control of the product quality requirements, and process parameters based on the soft measurement model, sample real-time data acquisition and building parameters and the nature of their experience mathematical model can be used for real-time On-line analysis of product quality and control. The internship will be based on the PTA plant solvent dehydration tower chemical process data collected samples of the timing, and taking into account the high-dimensional data-related small sample, to be adopted biased estimate of the partial least squares regression method for the solution of acetic acid from the top in the establishment of soft measurement model, and with other classical soft measurement modeling method for comparison.
Keywords：Soft-sensing; Partial least squares regression; PTA solvent dehydration tower; Regression; Principle component regression

目  录
中文摘要 I
Abstract II
目录 III
1. 绪论 1
 1.1 研究背景 1
 1.2  软测量建模方法研究现状 2
 1.3 PTA溶剂脱水塔工艺概述 6
 1.4 论文内容安排 8
2. 实验部分 9
 2.1 模型说明与样本数据分析 9
 2.2 试验方式与性能评价指标 9
 2.3 样本数据特征分析 10
 2.4 多元线性回归建模（MLR） 12
 2.5 PCA_NN神经网络的软测量建模 14
 2.6 偏最小二乘回归建模（PLSR） 18
 2.7 PLSR_NN神经网络建模 21
     2.7.3 小结 26
 2.8 不同因素对PLSR_NN模型影响 27
3. 结果与讨论 30
 3.1 实验样本数据分析及讨论   30
 3.2 多元线性回归方法建模结果及讨论 30
 3.3 PCA_NN方法建模及讨论 31
 3.4 偏最小二乘回归方法建模结果及讨论 32
 3.5 PLSR_NN方法建模结果及讨论 33
 3.6 不同主元数对PLSR_NN模型影响的讨论 34
 3.7 节点数对PLSR_NN模型影响的讨论 35
 3.8 四种方法建模效果讨论 37
4．总结与展望 38
致谢 40
参考文献 41