Affiliation:
1. aDepartment of Chemical Engineering, Imperial College London, UK
2. bCentre for Process Systems Engineering, Department of Chemical Engineering, Imperial College London, UK
Abstract
Identifying an appropriate model structure for a particular system, such as the kinetics of a chemical reaction, is of the utmost importance both industrially and academically. In this chapter, we introduce different performance metrics, called information criteria, that are capable of ranking candidate models in a model class from best to worst. The presented information criteria are scrutinized and their performance is examined on a simple case study, exploring how various variables (e.g. amount of additive noise, number of datapoints) can influence the behaviour of the criteria and their ability to identify correct model structures. This chapter also introduces different model-based design of experiments techniques for model discrimination techniques, where optimal discriminatory experiments are determined based on a specific objective function.
Publisher
Royal Society of Chemistry
Reference18 articles.
1. Model-order selection;Stoica;IEEE Signal Process. Mag.,2004
2. Model selection techniques: An overview;Ding;IEEE Signal Process. Mag.,2018
3. Some properties of the order of an autoregressive model selected by a generalization of akaikes EPF criterion;Bhansali;Biometrika,1977
4. A Corrected Akaike Information Criterion for Vector Autoregressive Model Selection;Hurvich;J. Time Ser. Anal.,1993