Faster EHL FEM Simulations Based on AI-Driven Initial Guesses: Evaluation of Hybrid Modeling Approaches

Numerical elastohydrodynamic lubrication (EHL) simulations provide physically consistent, locally resolved solutions but may require substantial nonlinear iteration effort. This study investigates a hybrid modeling approach in which artificial neural network (ANN) predictions are used as informed initial guesses for a fully coupled finite element method (FEM) solver for smooth isothermal Newtonian EHL line contacts. Different initialization strategies are compared across 58 cases with different operating conditions. This includes two conventional approaches that are based on Hertzian contact theory and employ either a constant or an analytically estimated central film thickness in initializing the rigid-body separation. Moreover, two ANN-supported initialization strategies are investigated: one based on the predicted pressure distribution, and another more comprehensive one using the predicted pressure and film thickness distributions. Performance is evaluated using the nonlinear iteration count, the dimensionless initial residual, and the dimensionless root mean square (RMS) difference from the converged solution. Both ANN-based strategies substantially improve the initial field proximity and generally accelerate convergence. The first one, using only the ANN-predicted pressure distribution in the initialization process, achieves the largest reduction in computational effort, decreasing the median iteration count and overall CPU time by approximately 55% and 61%, respectively, relative to the conventional approach that uses an analytically estimated central film thickness, which is taken as a reference. The second one, using both the ANN-predicted pressure and film thickness distribution in the initialization process, produces the lowest initial residual but achieves smaller median iteration and overall CPU time reductions of approximately 40% and 58%, respectively. The results demonstrate that transferring additional predicted distributions does not necessarily improve nonlinear convergence. The proposed hybrid approach preserves the physical consistency of the converged FEM solution while substantially reducing numerical effort.

Authors

Institutions

Publication Details

Journal
Lubricants
Published
2026-11-18
DOI
https://doi.org/10.3390/lubricants14110440
Primary Topic
Tribology and Lubrication Engineering
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Faster EHL FEM Simulations Based on AI-Driven Initial Guesses: Evaluation of Hybrid Modeling Approaches

Volker Schneider, Josephine Kelley, Max Marian, Thomas Lohner et al.
3 links
Lubricants
Tribology and Lubrication Engineering
article

Faster EHL FEM Simulations Based on AI-Driven Initial Guesses: Evaluation of Hybrid Modeling Approaches

Volker Schneider, Josephine Kelley, Max Marian, Thomas Lohner, Marko Tošić, Wassim Habchi, Chongyu Zhang
article en

Abstract

Numerical elastohydrodynamic lubrication (EHL) simulations provide physically consistent, locally resolved solutions but may require substantial nonlinear iteration effort. This study investigates a hybrid modeling approach in which artificial neural network (ANN) predictions are used as informed initial guesses for a fully coupled finite element method (FEM) solver for smooth isothermal Newtonian EHL line contacts. Different initialization strategies are compared across 58 cases with different operating conditions. This includes two conventional approaches that are based on Hertzian contact theory and employ either a constant or an analytically estimated central film thickness in initializing the rigid-body separation. Moreover, two ANN-supported initialization strategies are investigated: one based on the predicted pressure distribution, and another more comprehensive one using the predicted pressure and film thickness distributions. Performance is evaluated using the nonlinear iteration count, the dimensionless initial residual, and the dimensionless root mean square (RMS) difference from the converged solution. Both ANN-based strategies substantially improve the initial field proximity and generally accelerate convergence. The first one, using only the ANN-predicted pressure distribution in the initialization process, achieves the largest reduction in computational effort, decreasing the median iteration count and overall CPU time by approximately 55% and 61%, respectively, relative to the conventional approach that uses an analytically estimated central film thickness, which is taken as a reference. The second one, using both the ANN-predicted pressure and film thickness distribution in the initialization process, produces the lowest initial residual but achieves smaller median iteration and overall CPU time reductions of approximately 40% and 58%, respectively. The results demonstrate that transferring additional predicted distributions does not necessarily improve nonlinear convergence. The proposed hybrid approach preserves the physical consistency of the converged FEM solution while substantially reducing numerical effort.

LubricantsVol. 14(11)
Leibniz University Hannover (DE), Pontificia Universidad Católica de Chile (CL), Arts et Métiers (FR), Laboratoire Procédés et Ingénierie en Mécanique et Matériaux (FR), HESAM Université (FR), Lebanese American University (LB), Technical University of Munich (DE)
Deutsche Forschungsgemeinschaft, Volkswagen Foundation, Gottfried Wilhelm Leibniz Universität Hannover
Tribology and Lubrication Engineering
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon