基于自适应动态时间规整(DTW)的GA-FCM多阶段间歇过程故障诊断

多时段是间歇过程的固有特征，对间歇过程划分阶段可以提高故障诊断的精度。采用模糊C-均值聚类（FCM）算法划分阶段存在对初始聚类中心敏感、易于陷入局部极优值的问题。提出遗传算法与FCM算法相结合的方法（GA-FCM），用于克服FCM易于陷入局部极优值的问题，以达到全局最优。同时，针对间歇过程数据不等长问题，提出自适应动态时间规整（DTW）算法。随后，用GA-FCM方法完成阶段划分，再建立多向核主元分析（MKPCA）模型完成故障检测。最后将此算法应用于青霉素发酵过程，仿真结果验证了所提方法的可行性和有效性。

Fault diagnosis of GA-FCM multi-stage batch processes based on adaptive dynamic time warping algorithm

Multi-stage is an important characteristic of many batch processes. The division of batch processes could improve the accuracy of fault diagnosis. The fuzzy C-means clustering (FCM) algorithm is sensitive to initial cluster center and easy to fall into local excellent value, which may not search the global cluster center. The combination of genetic algorithm and FCM algorithm (GA-FCM) is proposed in this paper to overcome the problem that the fuzzy C-means clustering algorithm is easy to fall into local optimal value, which could achieve global optimization. In addition, an adaptive dynamic time warping (DTW) algorithm is presented for the batch process data with different time length. After that, the GA-FCM method is employed to accomplish the process of stage division and the multi-way kernel principal component analysis (MKPCA) model is established to accomplish the fault detection. Finally, the proposed method is utilized to the fermentation process of penicillin and the obtained results clearly demonstrate the power and advantages of the proposed method.