Improving the performance of evolutionary engine calibration algorithms with principal component analysis

Mohammad Tayarani, Adam Prugel-Bennett, Xin Yao, Hongming Xu

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

By studying the fitness landscape properties of engine calibration problem we propose a new Principal Component Analysis (PCA) based optimisation algorithm for the problem. The engine calibration problem in this paper is to minimise the fuel consumption, gas emission and particle emission of a Jaguar car engine. To evaluate the fuel consumption and emissions of the engine, a model of the engine that was developed in University of Birmingham was used. A strength Pareto method is used to convert the three objectives into one fitness value. Then a local search algorithm is used to find local optima. We then study these local optima to find the properties of good solutions in the landscape. Our studies on the good solutions show that the best solutions in the landscape show some patterns. We perform Principal Component Analysis (PCA) on the good solutions and show that these components present certain properties, which can be exploited to develop new exploration operators for evolutionary algorithms. We use the newly proposed operator on some well-known algorithms and show that the performance of the algorithms can be improved significantly.
Original languageEnglish
Title of host publication2016 IEEE Congress on Evolutionary Computation (CEC)
Publication statusPublished - 3 Sep 2016

Fingerprint

Dive into the research topics of 'Improving the performance of evolutionary engine calibration algorithms with principal component analysis'. Together they form a unique fingerprint.

Cite this