Batch process on-line monitoring based on fisher discriminant analysis and kernel density estimation
Yingwang Xiao,Dongjie Liu,Jun Yang,Chenzhong Zhang,Meiying Yao
DOI: https://doi.org/10.11719/com.app.chem20140912
2014-01-01
Abstract:When multiway principal component analysis (MPCA) is used for on-line batch monitoring, the future behavior of each new batch must be inferred up to the end of the batch operation at each time and false alarms are produced easily, and in view of limitations of confirmation of statistical control limits which assumes that principal component scores are subjected to multivariate normal distribution, combining fisher discriminant analysis (FDA) which has the advantage in data classification and nonparametric statistics method of kernel density estimation (KDE) which has the advantage in computing probability density function, A FDA-KDE method for batch process on-line monitoring is proposed. Firstly, in order to get similarity degree about fisher eigenvector, FDA is used to obtain fisher eigenvector and discriminant vector of normal working conditions data and fault data; Secondly, on the basis of using biased mean integrated squared error (BMISE) cross validation method proposed to define KDE bandwidth, therefore, getting similarity degree statistical control limits, the proposed approach uses present data and overcomes pre-estimating the unknown part of process variable trajectory in MPCA. Finally, weight contribution plot about the fisher discriminant vector is used to perform fault diagnosis. Application results on a penicillin batch fermentation process demonstrate that, in comparison to the MPCA method, the proposed method is more accurate and efficient to detect and diagnose the malfunctions.