Identifying Empirically Important Variables in IC Engine Operation Through Redundancy Analysis

Abstract

Computational studies incur engineering costs. While direct numerical simulations can provide detailed solutions, they cannot deliver quick and convenient solutions which are pragmatic for industry applications such as fault detection and diagnosis. But in the study of internal combustion engines, there are no such unified models that can completely capture the engine operation and hence, computational methods still are of great value. In this pursuit, an attempt had been made in this study to evaluate the empirical redundancy amongst the engine variables. Using Pearson correlation coefficient to quantify the linear dependencies among a set of variables, a representation score was developed to measure how effectively various variables can represent other variables. With the suggested methodology, those empirically important variables can be identified and they can be used to develop empirically reduced models for its possible employment in further computational studies.