A method of fatigue life prediction for rubber materials based on convolutional neural network
Studies show that mechanical load, ambient temperature, and material hardness significantly influence the fatigue life of rubber materials. The current physical models based on crack initiation methods face challenges in comprehensively addressing the combined effects of these factors on the fatigue...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IOP Publishing
2025-01-01
|
| Series: | Materials Research Express |
| Subjects: | |
| Online Access: | https://doi.org/10.1088/2053-1591/adda92 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Studies show that mechanical load, ambient temperature, and material hardness significantly influence the fatigue life of rubber materials. The current physical models based on crack initiation methods face challenges in comprehensively addressing the combined effects of these factors on the fatigue life of rubber materials. Meanwhile, data-driven models such as back-propagation neural networks (BPNN) or support vector machines (SVM) are often constrained by limited dataset sizes and uncertainties in model parameters, leading to low prediction accuracy for fatigue life. Aiming at resolving these limitations, a novel fatigue life prediction approach for rubber materials utilizing convolutional neural networks (CNN) was introduced. Dumbbell-shaped rubber specimens, vulcanized from rubber materials, were subjected to uniaxial tensile fatigue tests under varying temperatures, hardness levels, and load conditions. Using the obtained fatigue test data, a CNN model was developed, incorporating environmental temperature, material hardness, and peak engineering strain as input features, with the corresponding measured fatigue life of rubber materials as the output. The prediction performance of the CNN model was then evaluated against physical models, BPNN, and SVM models. Comparative analysis using the measured fatigue life as a reference revealed that the CNN model achieved superior predictive accuracy. The predicted fatigue lives consistently fell within the 1.5 times dispersion range of the experimental values. |
|---|---|
| ISSN: | 2053-1591 |