Fluidized Bed Agglomeration Diagnosis Based on Wavelet Packet Entropy and Gaussian Test

Abstract

Aiming at the synthesis process through the ethylene polymerization by virtue of fluidized bed reactors, this paper proposed a different approach for agglomeration monitoring and early alarming in polymerization reaction with Gaussian test of wavelet packet entropies from low-frequency audible acoustic signals. In comparison to high-frequency acoustic emission signals, audible acoustic signals can reduce the signal bandwidth, simplify the signal processing circuit, facilitate the real-time measurement, and fault detection. Compared with the approach based on regression modeling and pattern recognition, this approach can overcome the defects of false alarm and missing alarm caused by inadequate samples and weak generalization capability of diagnosis model and is easy to be implemented. Experiments in a pilot plant have verified the effectiveness of the approach, and earlier warning could be realized compared with the existing approaches on the production apparatus (material level detection and temperature measurement).