engleski

Software Sensors for Monitoring of the Solid Waste Composting Process

This article deals with solid waste aerobic management process identification using empirical kinetic mathematical and neural network-based models. Composting of tobacco solid waste is carried within aerobic closed thermally insulated column reactor of 25 L effective volume. The experiments have been carried under adiabatic condition at different constant air flow rates of 0.3, 0.6 and 0.9 dm3 min-1 of volatile substance under the adiabatic conditions. Three soft sensors using kinetic and neural network-based models have been developed aiming to estimate conversion that cannot be measure in the continuously manner. Neural network models based on NFIR (Nonlinear Finite Impulse Response) and NARX (Nonlinear AutoRegressive model with eXogenous inputs) identification methods have been used. The neural networks have been trained by the adaptive gradient method using cascade learning. After the process models have been developed, the conversion during the course of the process is estimated. Results obtained by experiment, kinetic model, and neural network models have been compared. It was found that the selected models describe aerobic composting fairly well and confirm the hypothesis that the released heat is proportional to the biodegradation process. The developed neural-network models show that the neural networks are capable to be applied as intelligent software sensors giving the possibility of continuous process monitoring. The models have potential to be used for inferential control of composting process.

aerobic composting; process modeling and identification; neural network; soft sensor

nije evidentirano