Agglomeration Detection in Horizontal Stirred Bed Reactor Based on Autoregression Model by Acoustic Emission Signals

Abstract

Agglomeration occurring in horizontal stirred bed reactors (HSBR) for polyolefin production has negative impacts on the efficiency of the reactor operation and may sometimes lead to unscheduled shutdown of the plant. In this paper, an autoregression (AR) model based on acoustic emission (AE) technique has been proposed to establish the qualitative relationship between AE signals and agglomeration in the HSBR. In this method, the frequency of AE signal varies with particles of different sizes striking the reactor walls. From the cold model experiments, it was found that AR power spectrum became fluctuant after the addition of agglomerations into laboratorial scale HSBR, and meanwhile the low frequency band energy ratio and the variance of AE signals kept rising. Furthermore, this AE-based AR model was also successfully applied to detect the agglomeration in an industrial HSBR unit, showing that the method could monitor agglomerations in an environmentally friendly manner and with fairly good accuracy.