Professor University of Illinois Chicago, Illinois, United States
Abstract Submission: With over half the municipal solid waste (MSW) generated being disposed of in landfills, these systems are vital for global waste management. However, the design and operation of landfill components, such as leachate recirculation systems aimed at optimizing waste stabilization, are hindered by limited understanding of the complex hydraulic, thermal, biochemical, and mechanical processes occurring within landfills. The variability in MSW properties and site conditions further complicates efforts, as findings from laboratory or field-scale studies cannot be generalized. Numerical models offer a way to make more generalizable predictions, however, often such models tend to simplify the underlying processes, and as they become more accurate, the computational burden rises, making them impractical for optimization. Hence, current study aims to apply artificial intelligence (AI) (machine learning (ML), and deep learning (DL)) based surrogate models to the coupled thermo-hydro-bio-mechanical (CTHBM) model developed at the University of Illinois Chicago. A comprehensive dataset was generated using the CTHBM model by varying relevant input parameters. This data was used to train various ML/DL models including support vector machine, tree based ensemble models, and neural networks (NN) to test their ability to accurately predict landfill performance indicators: stabilization period and methane generation. Results showed that NNs could predict both stabilization period and methane generation accurately with an R2 > 0.9. Current study preliminarily assessed the efficacy of AI/ML/DL models to predict landfill data. Future research should focus on more advanced AI techniques, such as physics-informed neural networks, to improve accuracy and generalizability of the predictions.
.jpg "Krishna R R. Reddy, Ph.D., PE, ENV SP photo")
