自动编码器与典型相关分析方法联合驱动的工业过程质量监测

本文将自动编码器(AE)特征提取方法和典型相关分析方法(CCA)有机结合,提出了一种联合驱动的质量监测模型及其质量相关的故障检测方法.首先,利用AE算法对输入样本进行无监督自动学习和重构,实现数据的特征提取和降维;其次,利用CCA算法实现特征与质量变量关联最大化,建立质量变量与特征变量的关系模型;根据监测模型的潜结构投影,构建T2统计量和SPE统计量及其相应控制限.将提出的方法用于分析带钢热连轧过程现场实际数据,结果表明,基于自动编码器-典型相关分析方法(AE-CCA)的质量监测方法能够准确的检测出故障,并且检测效果优于传统的核典型相关分析(KCCA)算法.

Industrial process quality monitoring method and application jointdriven by automatic encoder and canonical correlation analysis method

Abstract: With the expansion of industrial production scale and the increase of complexity, the quality and safety of the production process are increasingly valued. Quality monitoring is an important method which ensures product safety and quality. Feature extraction, which is an effective dimension reduction method, is applied to industrial processes gradually to analyze and utilize the high-dimensional industrial big data. In this paper, the feature extraction method of automatic encoder (AE) and canonical correlation analysis method (CCA) are organically combined, and a joint-driven quality monitoring model and quality-related fault detection method are proposed. Firstly, AE algorithm is used to automatically learn and reconstruct the input samples to complete the feature extraction and dimensionality reduction of the data. Secondly, CCA algorithm is used to maximize the correlation between the feature and the quality variables to establish the monitoring model of quality variables and characteristics. According to the latent structure projection of the monitoring model, T2statistics and SPE statistics and their control limits are constructed. The proposed method was applied to the actual data of hot strip mill process (HSMP). The result shows that the quality monitoring method based on AE-CCA can detect faults accurately, and the effect of detection is significantly better than that of traditional kernel canonical correlation analysis (KCCA) algorithm.