A Physics‐Guided Residual Neural Network for Small‐Sample Cross‐Material‐Source Prediction of Contact Fatigue Life in Plastic Gears

In this study, a physics‐guided residual neural network (PGRNN) was developed for small‐sample cross‐material‐source prediction of plastic gear contact fatigue life and evaluated using POM, PEEK, and PA66 literature datasets. Stress and temperature were normalized by material strength, stiffness, and characteristic temperatures to describe comparable mechanical and thermal states across polymers. A Basquin‐based physical backbone represented the main fatigue trend, while a residual network captured nonlinear thermo‐mechanical coupling. Leave‐one‐material‐out (LOMO) validation held out each material and its single, distinct literature source together; repeated within‐material validation was also performed. PGRNN achieved the lowest root‐mean‐square error in all three LOMO folds compared with artificial neural network, support vector regression, and random forest baselines, with most predictions within the factor‐of‐three error band. Ablation and contribution analyses further showed that the physical backbone dominated the response, whereas bounded residual corrections improved accuracy without reversing the expected degradation trends with increasing stress and temperature.

Authors

Institutions

Publication Details

Journal
Fatigue & Fracture of Engineering Materials & Structures
Published
2026-11-19
DOI
https://doi.org/10.1111/ffe.70538
Primary Topic
Fatigue and fracture mechanics
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

A Physics‐Guided Residual Neural Network for Small‐Sample Cross‐Material‐Source Prediction of Contact Fatigue Life in Plastic Gears

Baofeng He, Shuiyu Liu, Edward Wang, Zhaoyao Shi
Fatigue & Fracture of Engineering Materials & Structures
Fatigue and fracture mechanics
article

A Physics‐Guided Residual Neural Network for Small‐Sample Cross‐Material‐Source Prediction of Contact Fatigue Life in Plastic Gears

Baofeng He, Shuiyu Liu, Edward Wang, Zhaoyao Shi
article en

Abstract

In this study, a physics‐guided residual neural network (PGRNN) was developed for small‐sample cross‐material‐source prediction of plastic gear contact fatigue life and evaluated using POM, PEEK, and PA66 literature datasets. Stress and temperature were normalized by material strength, stiffness, and characteristic temperatures to describe comparable mechanical and thermal states across polymers. A Basquin‐based physical backbone represented the main fatigue trend, while a residual network captured nonlinear thermo‐mechanical coupling. Leave‐one‐material‐out (LOMO) validation held out each material and its single, distinct literature source together; repeated within‐material validation was also performed. PGRNN achieved the lowest root‐mean‐square error in all three LOMO folds compared with artificial neural network, support vector regression, and random forest baselines, with most predictions within the factor‐of‐three error band. Ablation and contribution analyses further showed that the physical backbone dominated the response, whereas bounded residual corrections improved accuracy without reversing the expected degradation trends with increasing stress and temperature.

Fatigue & Fracture of Engineering Materials & Structures
Beijing University of Technology (CN), Virginia Tech (US)
Fatigue and fracture mechanics
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