Modelling and optimization of sewage sludge composting using biomass ash via deep neural network and genetic...

Area of Science:

• Environmental Science
• Biotechnology
• Computational Science

Background:

• Co-composting dewatered sewage sludge and biomass fly ash (BFA) presents complex linear and non-linear challenges.
• Accurate modeling of composting parameters is crucial for efficient waste management and resource recovery.

Purpose of the Study:

• To investigate the efficacy of Deep Cascade Forward Neural Network (DCFNN) in modeling co-composting processes.
• To compare DCFNN performance against other neural network models and Response Surface Methodology (RSM).
• To optimize composting process parameters using Genetic Algorithm (GA).

Main Methods:

• Development and application of a Deep Cascade Forward Neural Network (DCFNN) model.
• Comparative analysis with Feed Forward Neural Network (FFNN), Feed Back Neural Network (FBNN), Cascade Forward Neural Network (CFNN), and RSM.
• Optimization of decision variables using Genetic Algorithm (GA).

Main Results:

• DCFNN demonstrated high predictive accuracy with Mean Absolute Percentage Error (MAPE) below 1% for most datasets.
• DCFNN outperformed other evaluated models in predicting co-composting parameters.
• GA optimization achieved desirability levels of 100% in several designs and above 95% in others.

Conclusions:

• DCFNN is a reliable and consistent tool for modeling complex composting processes.
• GA is an effective tool for optimizing input parameters to achieve desired composting outputs.
• The combined DCFNN and GA approach offers a robust solution for composting process management.