滑动模型MPCA在间歇过程故障检测与诊断中的应用

Slipping-model MPCA Applicauion on Fault Detection and Diagnosis in Batch Process

本文针对间歇生产过程复杂非线性的特点及控制系统的实时监测要求,提出了滑动模型MPCA方法。这种方法首先将三维原始数据空间按不同测量时刻切割成一系列子数据空间,然后根据选取的模型大小将各个时间段中子数据块整合成具有相同数据结构的一系列新的子数据空间,分别建立各个子MPCA模型。滑动模型MPCA方法能够在尽可能小的时间段中监测系统运行状况,能够较好解决间歇生产过程中的非线性问题,并且更好地保证数据信息抽取的完整性。仿真实例验证了该方法的有效性。

Based on complexly nonlinear characteristic of batch trol system, Slipping - model MPCA is proposed. First, original process and on - time detection for condata space of three dimensions is incised into a series of child data space according to different measurement time. Then child data spaces are trimmed into a new series of child data space with same data structure according to the model length chosen, and each child MPCA model is established. Slipping - model MPCA can detect the function for system in small time as possible as, solves nonlinear problem in batch process better, and assures the integrality of data information extract. Simulations verify the effectiveness of the method.