基于有效分类的多模态过程故障检测及应用

多模态的故障检测作为复杂的实际问题,得到越来越多的重视.围绕多模态的故障检测问题展开相应关键问题研究,首先提出一种基于K均值聚类算法结合聚类有效性指标求解出最佳模态数方法,通过数值仿真和带钢热连轧生产过程数据进行验证;然后利用模糊C均值算法对训练数据进行模态划分,针对不同模态,利用主成分回归方法建立相应的监测模型,实现对故障的有效检测;最后将该故障检测方法应用到带钢热连轧生产过程.仿真结果表明,不仅实现合理模态划分和识别,而且取得良好的检测效果.

Fault Detection and Application Based on the Effective Classification for Multimode Processes

Multimode process fault detection, a complicated practical problem, has attracted increasing attention. Some key issues in multimode process fault detection were researched. Firstly, a way to solve the optimal mode number based on k-means clustering algorithm combining with clustering validity index was put forward, which was verified by numerical simulation and the data of hot strip mill process. Then, the fuzzy C-means algorithm was utilized to classify the training data. According to different modes, the principal component regression method was employed to establish the corresponding monitoring models, in order to implement effective fault monitoring. At last, the fault detection method was applied in hot strip mill process. The simulation results showed that it could not only realize the reasonable mode classification and recognition, but also achieve good detection effect.