化工过程可拓PCA故障诊断方法研究与应用

Research and application of extension PCA fault diagnosis method for chemical process

随着过程工业生产安全要求的不断提高,对其过程的故障诊断已成为当前的研究热点,其中主元分析(PCA)方法广泛应用于故障诊断领域。但在实际应用过程中,数据的量纲、噪声、分布等问题对统计量的计算造成了一定的影响,使得故障发生时所获得的系统信息不准确,因而无法获得较准确的故障诊断结果。本文提出了一种可拓PCA故障诊断方法,该方法运用基元模型描述所获得的数据信息,采用关联函数计算数据与基元的关联度,并根据关联度计算结果对不同的数据信息分别进行处理,使得主元分析中的统计量更能反映故障特征,同时还利用关联度确定故障源,提升了传统PCA方法的故障诊断能力。该方法在TE流程上的仿真结果证明了其在故障诊断中的实用性和有效性。

With the increasing requirement for production safety, process fault diagnosis has become current research focus, in which principal component analysis （PCA） method is widely used in the fault diagnosis field. However, in the actual application process, the dimension, the noise, and the distribution of the data have made a certain influence on the calculation of the statistical, which made the system information inaccurate. Thus, the result of fault diagnosis is not accurate enough. In the paper, an extension PCA is proposed. First, the basic-element model is used to describe the data information. Second, the extension distance is adopted to construct the correlation function for the correlation degree of data and domain. Third, according to the result of correlation degree, dispose the data information in different ways. Then the statistical calculation further reflect the fault characters and the correlation degree give the fault source, which enhance the capability of fault diagnosis for PCA. The simulation results on TE process proved its practicality and effectiveness in fault diagaosis.