基于子空间方法和散度测度的大型天然气液化装置微小故障检测与诊断研究

Large-scale liquefied natural gas (LNG) production process can be taken as a typical complex dynamic industrial process which have high safety and reliability requirements. Incipient fault detection and diagnosis based on subspace approach and divergence measure is studied. The fault detection and diagnosis based on subspace approach can derive the residual generator directly from subspace identification in data driven approach. However, due to the fault symptom is not obvious, the T2 or X2 statistics of residuals reduce the sensitivity to incipient fault. Focusing on the incipient fault detection and diagnosis, the research adopts moving window strategy and transforms the single sample monitoring problem into distribution test between normal sample set and subset of test samples. Based on these ideas, some incipient fault detection and diagnosis problems will be studied and solved: (i) Using the divergence measure theory, the methodologies of statistics construction, control limit determination and the sensitivity analysis to incipient fault; (ii) The incipient fault detection and diagnosis schemes and algorithms towards quality-related dynamic process; (iii) The incipient fault detection and diagnosis schemes and algorithms towards and multi-mode dynamic process. Based on the above work, the related theory and methods will be tested and verified using complex LNG production process as a typical industrial process. Through the research work in this project, we will provide an effective and the critical basis in improving process safety and product quality of complex dynamic industrial processes.

面向以大型天然气液化装置为代表的一类具有高运行安全和高产品质量要求的复杂工业动态过程，本项目研究基于子空间方法(Subspace Aided Approach)和散度测度(Divergence Measure)的微小故障检测与诊断方法。现有基于子空间方法的故障检测与诊断所采用的残差T^2或X^2统计量，降低了对微小故障的检测灵敏度。围绕此问题，本项目拟采用滑窗策略，将单样本检验问题转为分布检验问题，研究和解决以下几方面问题：(1)基于散度测度的统计量构造、控制限确定以及灵敏度分析；(2)质量相关动态过程的微小故障检测与诊断方法；(3)多工况动态过程的微小故障检测与诊断方法。同时以大型天然气液化装置作为典型范例和试验对象，对相关的理论方法进行试验验证。项目研究成果将为提高复杂工业动态过程的运行安全与产品质量，提供有效的方法和理论依据。

针对以大型天然气液化装置为代表的一类具有高运行安全和高产品质量要求的复杂工业动态过程，微小故障诊断对于提高工业过程的可靠性及安全性，具有重要实际意义。为此，本项目在理论上拟研究：1）基于数据驱动的微小故障检测；2）基于散度测度工业过程监测；3）数据驱动的复杂非线性过程微小故障诊断方法。在工程上，以具有大型天然气液化过程和高炉冶炼过程为应用对象，对本项目的理论研究成果进行应用验证。通过本项目的研究，可以大型天然气液化过程和高炉冶炼过程这一类复杂工业过程的微小故障诊断策略提供一种新的研究思路，进而促进工业过程高质量安全生产的发展。