Data Driven Models as A Powerful Tool to Simulate Emerging Bioprocesses: An Artificial Neural Network Model to Describe Methanotrophic Microbial Activity

Abstract

The vision for sewage treatment plants is being revised and they are no longer considered as pollutant removing facilities but rather as water resources recovery facilities (WRRFs). However, the newly adopted bioprocesses in WRRFs are not fully understood from the microbiological and kinetic perspectives. Thus, large variations in the outputs of the kinetics-based numerical models are evident. In this research, data driven models (DDM) are proposed as a robust alternative towards modelling emerging bio- processes. Methanotrophs are multi-use bacterium that can play key role in revalorizing the biogas in WRRFs, and thus, a Multi-Layer Perceptron Artificial Neural Network (ANN) model was developed and optimized to simulate the cultivation of mixed methanotrophic culture considering multiple environmental conditions. The influence of the input variables on the outputs was assessed through de- veloping and analyzing several different ANN model configurations. The constructed ANN models demonstrate that the indirect and complex relationships between the inputs and outputs can be accurately considered prior to the full understanding of the physical or mathematical processes. Furthermore, it was found that ANN models can be used to better understand and rank the influence of dif- ferent input variables (i.e., the physical parameters that influence methanotrophs) on the microbial activity. Methanotrophic-based bio- processes are complex due to the interactions between the gaseous, liquid and solid phases. Yet, for the first time, this study success- fully utilized DDM to model methanotrophic-based bioprocesses. The findings of this research suggest that DDM are a powerful, al- ternative modeling tool that can be used to model emerging bioprocesses towards their implementation in WRRFs.