Measuring Average Particle Size for Fluidized Bed Reactors by Employing Acoustic Emission Signals and Neural Networks

Abstract

AbstractAverage particle size is one of the key parameters of fluidized bed reactors. A novel method to measure the average particle size in fluidized bed reactors by acoustic emission (AE) signal is proposed. The measurement of the average particle size by AE is superior to other methods since it is inherently safe to use and it is also a non‐invasive method. AE signals originating from different particle sizes were 8‐level decomposed by a Daubechies wavelet of order 3 after being denoised with a sym8 wavelet filter combined with the rigrsure threshold method. Principal component analysis (PCA) was applied to overcome the complex collinearity and reduce the number of input variables of the neural network. A feed‐forward back propagation neural network with two hidden layers was used to predict the average particle size according to the principal components. The results show that this soft‐measuring model is suitable for measuring the average particle size in the fluidized bed reactors by online AE. It achieved high accuracy when applied to a model‐scale fluidized bed.