论文部分内容阅读
主元分析作为一种常用的多变量统计监控算法,在工业过程中已有广泛的应用.然而,现代复杂工业过程中基于主元分析的故障检测往往存在一些问题.针对工业过程中的微小故障,本文利用滑动窗口的策略,选择合适的窗口宽度,合并窗口内的采样数据,以实现误差的累加,从而使得故障样本数据与非故障样本数据的分化更加明显,实现对微小故障的放大,能够达到及早检测出微小故障的目的.通过给出的两类微小故障的数值仿真,以及青霉素发酵过程的仿真,验证了该方法的可行性与有效性.
As a commonly used multivariate statistical monitoring algorithm, principal component analysis (PCA) has been widely used in industrial processes.However, there are some problems in fault detection based on principal component analysis (PCA) in modern complex industrial processes.Aiming at the minor faults in industrial processes In this paper, we use the sliding window strategy to select the appropriate window width and merge the sampling data in the window to achieve the accumulation of errors, so that the differentiation of the fault sample data and the non-fault sample data is more obvious, and the micro fault amplification can be realized And achieve the purpose of early detection of minor faults.Through the numerical simulation of two kinds of micro faults and the simulation of penicillin fermentation process, the feasibility and effectiveness of the method are verified.