loading page

A Global Sensitivity Analysis Methodology for Anaerobic Digestion Models through Functional Principal Components Projection
  • +5
  • Dhan Lord Fortela,
  • Alyssa DeLattre,
  • Spencer Kowalski,
  • Wayne Sharp,
  • Emmanuel Revellame,
  • Rafael Hernandez,
  • Daniel Gang,
  • Mark Zappi
Dhan Lord Fortela
University of Louisiana at Lafayette College of Engineering
Author Profile
Alyssa DeLattre
University of Louisiana at Lafayette
Author Profile
Spencer Kowalski
University of Louisiana at Lafayette
Author Profile
Wayne Sharp
University of Louisiana at Lafayette
Author Profile
Emmanuel Revellame
University of Louisiana at Lafayette
Author Profile
Rafael Hernandez
University of Louisiana at Lafayette
Author Profile
Daniel Gang
University of Louisiana at Lafayette
Author Profile
Mark Zappi
University of Louisiana at Lafayette
Author Profile

Abstract

Sensitivity analysis (SA) for the influence of model parametric constants has been integral in the use of mathematical kinetic models for design and operation of various anaerobic digestion applications. Using Anaerobic Digestion Model No. 1 (ADM1) as case study, this work aimed to broaden the approach for SA on the time-dependent model outputs of anaerobic digestion models by demonstrating the use of functional principal component analysis (fPCA) scores as input analysis variables into global SA (GSA) for the influence of stoichiometric parameters in ADM1. The methodology involved the following: Morris’ screening design as the GSA technique; ADM1 biomass yield and product yield coefficients as GSA parameters; and ADM1 outputs transformation via fPCA to generate principal component (PC) scores for GSA. Results indicate that 95-99% of the variations in the time-dependent outputs can be captured by the PCs after fPCA transformation, and that the first PC is sufficient to represent the model outputs. Ranked Morris sensitivity indices calculated from the first PC scores revealed the stoichiometric parameters that dominantly affect kinetic responses and those that are least sensitive. The ranking of stoichiometric sensitivities can be used for various purposes including driving mechanisms identification, and mathematical model modification.