Global sensitivity analysis and uncertainty quantification for a mathematical model of dry anaerobic digestion in plug-flow reactors.

Abstract

In many applications, complex biological phenomena can be reproduced via structured mathematical models, which depend on numerous biotic and abiotic input parameters, whose effect on model outputs can be of paramount importance. The calibration of model parameters is crucial to obtain the best fit between simulated and experimental data. Sensitivity analysis and uncertainty quantification constitute essential tools in the field of biological systems modeling. Despite the significant number of applications of sensitivity analysis in wet anaerobic digestion, there are no examples of global sensitivity analysis for mathematical models describing the dry anaerobic digestion in plug-flow reactors. For the first time, the present study explores the global sensitivity analysis and uncertainty quantification for a plug-flow reactor model. The investigated model accounts for the mass$ / $volume variation that takes place in these systems as a result of solid waste conversion in gaseous value-added compounds. A preliminary screening based on the Morris' method allowed for the definition of three different groups of parameters. A surrogate model was constructed to investigate the relation between input and output parameters without running demanding simulations from scratch. The obtained Sobol' indices allowed to perform the quantitative global sensitivity analysis. Finally, the uncertainty quantification results led to the definition of the probability density function related to the investigated quantity of interest. The study showed that the net methane production is mostly sensitive to the values of the conversion parameter related to the particulate biodegradable volatile solids in acetic acid $ k_1 $ and to the kinetic parameter describing the acetic acid uptake $ k_2 $. The application of these techniques led to helpful information for model calibration and validation.