基于KPLS数据重构的非线性过程监测与故障辨识

发布时间：2018-03-26 11:24

本文选题：故障辨识　切入点：核偏最小二乘　出处：《中国安全生产科学技术》2015年12期

【摘要】：为适应快速变化的化工产品需求,过程工业逐步向柔性生产发展,使得间歇过程的应用日益广泛。这一类工艺过程具有动态和非线性的特征,过程故障带来的工艺波动和安全风险是较为突出的挑战。采用基于核函数的偏最小二乘方法,在高维特征空间提取特征变量,这些变量包含了生产过程的非线性结构特征,也反应了过程工况的模式特征。针对传统线性方法存在的故障漏报等问题,利用核函数技巧,在特征空间进行数据重构,进而计算统计监控指标SPE,并通过对SPE的在线监测实现更加有效地故障辨识。本方法针对标准非线性测试对象进行了过程监测,实现结果充分说明了方法的有效性。
[Abstract]:In order to adapt to the rapidly changing demand of chemical products, the process industry is gradually developing to flexible production, which makes the batch process more and more widely used. This kind of process has the characteristics of dynamic and nonlinear. The process fluctuation and safety risk caused by process failure are more prominent challenges. Using the kernel function based partial least squares method, the feature variables are extracted in the high dimensional feature space, which contain the nonlinear structural features of the production process. It also reflects the mode characteristics of the process conditions. Aiming at the problems of the traditional linear method, such as missing the fault report, the kernel function technique is used to reconstruct the data in the feature space. Then the statistical monitoring index SPE is calculated, and the fault identification is realized more effectively by on-line monitoring of SPE. The process monitoring of standard nonlinear test object is carried out in this method, and the results fully demonstrate the effectiveness of the method.
【作者单位】：中国安全生产科学研究院;重大危险源监控与事故应急技术国家安全监管总局安全生产重点实验室;
【基金】：国家科技支撑计划项目(2015BAK16B04)
【分类号】：TQ086

【参考文献】