关键词: 间歇过程 , 主元分析 , 隐马尔可夫树 , 在线监控

Abstract: In order to overcome the misclassification problems， multiway batch process monitoring based on the waveletbased multihidden Markov model tree (MHMT) was developed. MHMT can capture the clustering and persistence of the statistical characteristics for practical measured data. This approach provides less signal distortion and better understanding of the principal source of the system variability affecting the process. It solved the problem that discrete wavelet transform(DWT)can not obtain shiftinvariance，and then made simple modifications of the expansion structure， the method using stochastic model analysis the time domain expanded to timefrequency domain and extracted the main characteristics of the historical data， it is a good way to solve the multiway batch process monitoring problem. The proposed method was used to evaluate the industrial penicillin fermentation process data，the results clearly demonstrated the power and advantages of the proposed method in comparison with conventional MPCA method．

Key words: Batch process , Principal component analysis ,  Hidden Markov tree , Online monitoring