Variable Selection and Fault Detection Using a Hybrid Intelligent Water Drop Algorithm

TLDR: A recently proposed swarm intelligence-based hybrid intelligent water drop (IWD) optimization algorithm in combination with support vector machines and an information gain heuristic for selecting a subset of relevant fault indicators demonstrates its viability as a strong candidate for complex classification and prediction tasks.

Abstract: Process fault detection concerns itself with monitoring process variables and identifying when a fault has occurred in the process workflow. Sophisticated learning algorithms may be used to select the relevant process state variables out of a massive search space and can be used to build more efficient and robust fault detection models. In this study, we present a recently proposed swarm intelligence-based hybrid intelligent water drop (IWD) optimization algorithm in combination with support vector machines and an information gain heuristic for selecting a subset of relevant fault indicators. In the process, we demonstrate the successful application and effectiveness of this swarm intelligence-based method to variable selection and fault identification. Moreover, performance testing on standard machine learning benchmark datasets also indicates its viability as a strong candidate for complex classification and prediction tasks.