论文部分内容阅读
传统基于主元分析的故障检测方法大多假设工业过程只运行在1个稳定工况,数据服从单一的高斯分布。若这些方法直接用于多工况过程则将会产生大量的误检。为此,本文提出了1种基于高斯混合模型的多工况过程监测方法。首先利用PCA变换对过程数据集进行降维,在主元空间建立高斯混合模型对过程数据进行聚类,自动获取工况数和相关分布特性。然后对每个工况建立主元分析(principal component analysis,PCA)模型来描述整个运行过程数据分布的统计特性。最后在过程监测中,根据监测样本属于各个工况的概率构造综合统计量,实现对多工况过程的故障检测。TE过程的仿真结果表明,本文提出的方法与传统的PCA方法相比,能自动获取工况和精确估计各个工况的统计特性,从而能更准确及时地检测出多工况过程的各种故障。
Most traditional fault detection methods based on principal component analysis assume that the industrial process runs in only one stable condition and the data follows a single Gaussian distribution. If these methods are used directly in a multi-case process, a lot of false positives will result. To this end, this paper presents a method based on Gaussian mixture model of multi-condition process monitoring. Firstly, the PCA transform is used to reduce the dimension of the process dataset, the Gaussian mixture model is established in the principal component space to process the process data, and the number of working conditions and the correlation distribution are obtained automatically. Then, a principal component analysis (PCA) model is established for each working condition to describe the statistical characteristics of the data distribution throughout the running process. Finally, in process monitoring, the comprehensive statistics are constructed according to the probabilities of the monitoring samples belonging to each working condition, so as to realize the fault detection of the multi-working process. The simulation results of the TE process show that compared with the traditional PCA method, the proposed method can automatically obtain the working conditions and accurately estimate the statistical characteristics of each working condition so as to detect various faults in the multi-working process more accurately and timely .