改进的MPCA方法及其在批过程故障诊断中的应用

Batch Process Monitoring and Fault Detection Based on an Improved MPCA

基于传统的多方向主元分析(MPCA)常会导致误诊断,且对批生产过程难以保证在线状态监测和故障诊断的实时性,提出了一种改进的MPCA方法,该方法采用多模型非线性结构代替传统MPCA单模型线性化结构,克服了后者不能处理非线性过程和实时性问题,并避免了MPCA在线应用时预报未来测量值带来的误差,提高了批过程性能监测和故障诊断的准确性.

An improved multiway principle component analysis (MPCA) for on-line batch process monitoring and fault detection was proposed. It integrates the time-delay windows of process dynamic behavior with the MPCA. Using multi-model instead of single model, the improved MPCA approach emphasizes on-line process performance monitoring and exactly fault detecting which results in extraordinary behavior of batch process.