基于混合KPLS-FDA的过程监控和质量预报方法

Process monitoring and quality prediction method based on hybrid KPLS-FDA

提出一种基于核偏最小二乘(KPLS)与费舍尔判别分析(FDA)相结合的过程监控和质量预报方法—–混合KPLS-FDA方法.首先,利用KPLS提取过程数据的非线性特征,使用FDA建立KPLS的内部模型;然后,求出满足最大分离度的核Fisher特征向量和判别向量来实现状态监测,若系统运行正常,则根据KPLS回归模型预报产品的质量,否则利用Fisher相似度系数确定故障类型;最后,通过轧钢过程的仿真研究验证了混合KPLS-FDA方法的有效性.

A process monitoring and quality prediction method based on combining kernel partial least squares （KPLS） with Fisher discriminant analysis （FDA）, hybrid KPLS-FDA, is proposed. Firstly, the nonlinear feature of process data is extracted by using KPLS, the internal model of KPLS is established by using FDA, and the optimal feature vector and the discriminant vector which satisfies maximal separation degree are obtained for condition monitoring. If the process is under normal condition, the regression model of KPLS is further used for quality prediction. Otherwise, the similar degree in the fault feature direction is used for fault diagnosis. Finally, the simulation research for steel rolling process is performed to show its accuracy and effectiveness in fault diagnosis and quality prediction.