Monitoring Agitation Intensity in Fluidized Beds Containing Inert Particles via Acoustic Emissions and Neural Networks

This study utilized passive acoustic emissions from a fluidized bed containing spherical inert ABS particles, captured by an external piezoelectric microphone, to monitor fluidization agitation intensity. Acoustic signals were recorded during fluidization profiles achieved under air velocities ranging from 0.5 to 3.0 m/s and during the drying of water or maltodextrin aqueous solution (1:5 w/w) introduced as droplets. Analyzing audio features like waveforms, the Discrete Fourier Transform (DFT), and Mel Frequency Cepstral Coefficients (MFCCs) revealed changes corresponding to the agitation intensity of the particles. The MFCC coefficients were input into a three-layer artificial neural network (ANN) to predict fluidization dynamics based on air velocity, liquid flow rate, and drying time. The ANN efficiently learned from the data, achieving high predictive accuracy (R2 > 0.8) after 15 epochs of training, showcasing the robustness of MFCC coefficients for modeling. This approach highlights that the application of passive acoustic signals and neural networks allows for real-time monitoring of fluidization behavior during drying processes.