基于深度学习和偏最小二乘法的4-CBA含量软测量  

Soft Sensor of 4-CBA Concentration Based on the Deep Learning and Partial Least Square Method

作　　者：毛佳敏 赵厚田 赵亚 高逸煊 MAO Jiamin;ZHAO Houtian;ZHAO Ya;GAO Yixuan(Nanjing Normal University Zhongbei College,Danyang Jiangsu 212300,China)

机构地区：[1]南京师范大学中北学院,江苏丹阳212300

出　　处：《信息与电脑》2022年第20期47-50,66,共5页Information & Computer

摘　　要：4-CBA含量是精对苯二甲酸(Pure Terephthalic Acid,PTA)工业产品质量的重要指标,因此必须重视4-CBA含量的软测量精度。文章将深度学习算法引入软测量建模,提出基于深度学习和偏最小二乘法的4-CBA含量软测量建模方法,并用该模型处理实际化工生产中的数据。实验结果表明,基于深度学习和偏最小二乘法的4-CBA含量软测量模型能够对4-CBA含量进行估计,结合了深度学习算法的优势,其预测精度比单纯的偏最小二乘法模型和神经网络模型更高。4-CBA concentration is an important index of Pure Terephthalic Acid(PTA) industry product quality, so we must pay attention to the soft measurement accuracy of 4-CBA content. This paper introduced the deep learning algorithm into soft sensor modeling, presenting soft sensor modeling method of 4-CBA concentration based on the deep learning and partial least quares. At the same time, the model was used to deal with the actual data of chemical production. Experimental results show that the soft sensor model of 4-CBA concentration based on the deep learning and partial least quares can estimate 4-CBA concentration. Combined with feature learning advantage of deep learning algorithm, the prediction accuracy compared with simple partial least squares model and neural network model is higher.

关 键 词：机器学习 深度学习 偏最小二乘法 神经网络 软测量 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]