Font Size: a A A

Research On Phase Partition Method Based On Multi-batch Data Tensor Feature Analysis For Batch Process

Posted on:2019-10-29Degree:MasterType:Thesis
Country:ChinaCandidate:X CaoFull Text:PDF
GTID:2428330551457175Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Batch process can obtain a little products in the prescribed order by limited materials,meet the requirements of customized,complex,volume products,adapt to the business model of multi-variety and small batch,it has been widely used in the area of pharmacy,biochemical,food industry and so on.Real batch process operating status of batch process varies over times and steps,showing multi-phase characteristics.For obtaining the accurate and reliable process monitoring model,batch process should be divided into several phases firstly,and model each phase separately,so that the unique process characteristics can be described accurately.Thus,the method of phase partition has very important significance for improving statistical modeling's accuracy and fault detection's reliability.The paper has studied the phase partition method for batch process,on the basis of analyzing multi-phase and three-way characteristics.Firstly,a normalized batch process three-way data set was constructed,the PARAFAC2-based batch process time-slice models and time-block models were built,taking advantage of PARAFAC2-based three-way unfolding modeling,SPE statistics was chosen to describe the process characteristics of the above models;and control limits was obtained by kernel density estimation.then a phase partition rule was given to find position of phase partition point by evaluating the difference of SPE control limits between time-slice models and time-block models;finally,the PPCI index was adopted to to evaluate the results of phase partition for obtaining the most optimal one adaptively.The proposed method preserved three-way structural characteristics and data integrity of the batch process,took the timing of the actual operation of the batch process into consideration,and improved the accuracy of the phase partition.The proposed phase partition method for batch process based on PARAFAC2 was attested by a penicillin fermentation process.The experiment results show that the proposed method can realize the phase partition for batch process effectively,decrease the possibility of mismatch model because the loss of structure characteristics when the three-way data is folded,show higher accuracy compared with other two-way phase partition methods in terms of indicators.The fault detection of batch process was realized based on the proposed phase partition method,showing higher comprehensive detection rate compared with other phase partition methods data-folding based,proved that the proposed method has greater application value in the actual production field of batch process.
Keywords/Search Tags:batch process, multi-phase, PARAFAC2, fault detection
PDF Full Text Request
Related items