aalto1 untyped-item.component.html

Data-driven optimization of exhaust valve geometry for wear reduction

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Insinööritieteiden korkeakoulu | Master's thesis

Department

Major/Subject

Mcode

Language

en

Pages

82+7

Series

Abstract

This thesis presents a data-driven approach to optimizing exhaust valve geometry for reduced wear and enhanced sealing effectiveness. Utilizing a parametric model with various geometry variables and the friction coefficient of the exhaust valve, the study evaluates their effects on wear and the sealing effectiveness ratio between the components to identify the optimal geometry. The friction coefficient emerged as the most influential factor due to its impact on sliding distance, while the angle and length of contact significantly influenced sealing. This study employs a data-driven approach, utilizing two machine learning models to address the objective functions of sealing effectiveness and volume wear loss. Optimization is achieved through a Multi-objective Genetic Algorithm, with the optimal model selected using the multi-criteria decision-making methods TOPSIS and VIKOR. The TOPSIS-selected model achieved notable improvements: a 9.24% and 24.34% increase in the sealing effectiveness ratio compared to two industrial exhaust valves. However, the optimized model showed a slight underperformance in wear volume loss, trailing by 3.27% compared to one of the industrial models. This discrepancy highlights the challenges posed by prediction errors, the selection of appropriate weighting factors in MCDM, and the choice of objective functions in optimization. Despite this, the study underscores the potential for significant advancements in exhaust valve design through data-driven optimization, combining improved performance with computational efficiency.

Description

Supervisor

St-Pierre, Luc

Thesis advisor

Hintikka, Jouko

Other note

Citation

Endorsement

Review

Supplemented By

Referenced By