CIESC Journal ›› 2018, Vol. 69 ›› Issue (12): 5130-5138.DOI: 10.11949/j.issn.0438-1157.20180365

Previous Articles     Next Articles

Quality-related process monitoring approach based on semi-supervised orthogonal factor analysis

CUI Xiaohui, YANG Jian, SHI Hongbo   

  1. Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education, East China University of Science and Technology, Shanghai 200237, China
  • Received:2018-04-03 Revised:2018-08-13 Online:2018-12-05 Published:2018-12-05
  • Supported by:

    supported by the National Natural Science Foundation of China (61374140, 61673173) and the Fundamental Research Funds for the Central Universities (222201714031, 222201717006).

基于半监督正交因子分析的过程质量监控方法

崔晓惠, 杨健, 侍洪波   

  1. 华东理工大学化工过程先进控制和优化技术教育部重点实验室, 上海 200237
  • 通讯作者: 侍洪波
  • 基金资助:

    国家自然科学基金项目(61374140,61673173);中央高校基本科研业务费专项资金项目(222201714031);中央高校基本科研业务费重点科研基地创新基金项目(222201717006)。

Abstract:

In the process of industry, the most of the observation samples are polluted by random noises, which makes the probability model owning assumption of noises has been widely used. However, the factors obtained from the model may contain quality-unrelated information in the process monitoring. It is hard to identify the happened fault influencing the quality of product or not for traditional methods which monitor the factors directly. Meanwhile, aiming at the problem previously mentioned and the situation of unequal sample sizes of quality variables and process variables, a semi-supervised orthogonal factor analysis (Semi-SOFA) model was proposed. Firstly, it is formulated by the complete data collected in different sampling rates. Then, apply quality-related orthogonal decomposition to the factors, and construct statistics T2. At the same time, calculate the corresponding statistics SPE according to whether the new sample is labeled by quality variables or not. The proposed Semi-SOFA can effectively identify the period of quality influenced by fault. Finally, numerical simulation and Tennessee Eastman (TE) process simulation demonstrated effectiveness of the proposed approach.

Key words: orthogonal decomposition of factor, semi-supervised, quality-related fault detection, parameter estimation, process control, process systems

摘要:

实际工业过程中的观测样本大多会受到随机噪声的污染,因此带有噪声假设的概率模型得到广泛应用。传统方法直接对模型的因子进行监控,但由于建模所得因子中可能包含质量无关的信息,因此会增加质量相关故障的误报率,这对主要关心产品质量的生产过程是无益的。同时,针对实际过程与质量样本采样率不同导致的难以精确建模的问题,提出一种半监督正交因子分析(semi-supervised orthogonal factor analysis,Semi-SOFA)方法,建立概率模型,并对因子进行质量相关的正交分解,分别构造T2统计量;根据新样本是否含质量标签的数据性质计算相应的SPE统计量。提出的Semi-SOFA可有效检测出发生的故障是否影响质量,最后通过数值例子和Tennessee Eastman(TE)过程仿真验证了所提方法的有效性。

关键词: 因子正交分解, 半监督, 质量相关故障检测, 参数估值, 过程控制, 过程系统

CLC Number: