Using Sequential Sampling Method to Construct Surrogate Model for Optimization on Torrefaction of Biomass

Abstract

The purpose of this research is to use the Kriging surrogate model and sequential sampling to optimize the torrefaction conditions of biomass and to compare with the results of the Taguchi method The empty fruit bunch (EFB) and palm kernel shell (PKS) were selected to torrefy under four torrefaction conditions and the PA index was calculated The PA index is an index established by proximate analysis and its value reflects the flammability of the biochar or biomass Through the surrogate model and sequential sampling to find the torrefaction condition of the maximal PA index with fewer experiments The EFB case shows that the number of experiment surrogate model need for optimization can indeed be less than the number required by Taguchi method but because of the linear relationship between its PA index and parameters the extreme value is located at the upper and lower limits of the parameter range so Taguchi method is used to optimize EFB cases can also find the best conditions in the whole parameter domain In the case of PKS it can be seen that the optimization result of the surrogate model is guaranteed to be the global solution in the whole parameter domain If the Taguchi method is used to design factor levels with the same number of experiments it will result in the optimization result of the Taguchi method is the local solution The Taguchi method needs more experiments than the surrogate model to find the optimal conditions similar to the result of the surrogate model

Date of Award	2020
Original language	English
Supervisor	Yueh-Heng Li (Supervisor)