Research On Abnormal Modal Analysis Method Of Batch Process Data In Wavelet Domain

Batch process is widely used in the production and manufacture of small batch and high value-added products such as medicine and semiconductor devices,which occupies an immeasurable position in modern industrial production.Establishing a monitoring model is an important measure to ensure the batch process running safely and smoothly.However,batch process has nonlinear and multi-modal characteristics.The dynamic characteristics and the correlation of process variables of batch process are different under different modes.The abnormal modal data,which is caused by accidental factors such as abnormal production conditions and whose duration is not negligible,has a greater impact on the complexity and accuracy of modeling.Therefore,dividing modes and analyzing abnormal modes are important issues that need to be solved first in the process of establishing a monitoring model.The traditional phase division method is only based on the statistical correlation of process variables,which is more sensitive to the sudden changes of input and output data,and has poor robustness.It also does not make further analysis on the mode of time period and abnormal mode.The instantaneous frequency response function in wavelet domain is used to characterize the instantaneous dynamic characteristics of batch process,which can replace the statistical analysis of process data for mode division and abnormal mode analysis.It is of great significance to study the mode division method of batch process based on wavelet domain and analyze the abnormal mode.In this paper,based on the analysis of the frequency response function estimation method and the multi-modal characteristics of the batch process,the instantaneous frequency response function estimation method based on the wavelet domain is studied,and the instantaneous frequency response function estimation of the nonlinear time-varying system is realized based on the wavelet power spectrum method;The modal division method of the batch process based on the instantaneous frequency response function is studied.Based on the principle that the dynamic characteristics in the same mode are similar and the instantaneous frequency response function has little change,the instantaneous frequency response function of the batch process is analyzed by clustering,and the mode division of the batch process is realized;On this basis,the abnormal modal analysis method of batch process based on statistical analysis theory is studied.This method uses clustering validity criteria to determine the number of clusters,statistics the frequency of each mode,and uses statistical analysis to detect the abnormal modes,so as to realize the abnormal modal analysis of batch process.The model division method and abnormal mode analysis method proposed in this paper are studied by using the Pensim simulation platform.The results show that the method based on instantaneous frequency response function in wavelet domain can reduce the impact of input variable mutation on the modal division results,and the proposed method has better robustness than the method based on process variable data;The abnormal modal analysis method of batch process based on statistical analysis theory can detect the occasional abnormal modes and realize the modal analysis of batch process.