Metamodel-Assisted Optimization of Connecting Rod Big-End Bearings

Abstract

From a very general point of view, optimization involves numerous calculations and therefore a high computational cost. In the fields where a single calculation is long and the optimization is crucial, specific techniques, devoted to this task, have been developed. First, the surrogate-based models are introduced and a short review of optimization in tribology is presented. The aim of the present work is to combine both. To demonstrate the power of the methodology on a lubricated bearing, the theoretical background is first outlined. Then, the two aforementioned processes are described: the construction of the surrogate, based on the Finite Element Method well-chosen computations, and the Multiobjective Optimization, thanks to a Nondominated Sorting Genetic Algorithm. Both are utilized on a connecting rod big-end bearing. As a result, the power loss and the functioning severity are simultaneously minimized upon a set of ten input parameters. The user is then provided with simple analytical expressions of the input variables, for which the bearing behavior is optimal.