Gas-Solid Fluidized Bed Agglomeration Warning Based On Wavelet Packet Entropy And Gauss Test

In the process of fluidized bed gas phase ethylene polymerization, small particles of polymer have great possibilities to gather together and form agglomeration, which will affect the fluidization state, melt the bed, eventually lead to unexpected shutdown of the plant. In order to reduce the agglomeration problems, proper process monitoring is of essential importance. Among current investigations and researches, methods based on electrostatic, radiation, pressure fluctuation and attractor comparison are put forward to monitor the plant process. However, in some cases, these methods have more or less disadvantages. Therefore, more attention is paid to acoustic methods due to the low cost, convenient installation and reliable monitoring.Acoustic method is applied in this article to detect the agglomeration. In terms of the hardware, a piezoelectric acoustic sensor is stuck to the outer wall of the fluidized bed. The signal besides the bed sensed by the piezoelectric is transmitted to the PC by a long shielded cable, then sound card in the PC acquire the signal in frequency of 44.1K and store them in the database. In order to reduce the difficulty of real-time signal processing, down sampling method is explored in this article, besides, power frequency filter is used to reduce the power frequency interference.In this article, acoustic feature extraction based on wavelet package entropy is proposed. The decomposition scale of the wavelet packet decomposition is determined by the offset of power spectrum gravity center. Following this, wavelet package entropy is calculated as acoustic signal characteristic. After that, agglomeration diagnosis method based on gauss test is used to monitor the condition of fluidized bed. Furthermore, agglomeration monitoring system software based on Qt is written to realize human-computer interaction.Pilot polyethylene fluidized bed experimental results show that the method can realize the agglomerate warning from 21 to 79 minutes ahead and overcomes the problem of time lag of the traditional method based on material level height and temperature signals which enables the workers to adjust the process parameters and avoids accidents.