| About this Abstract |
| Meeting |
6th World Congress on Integrated Computational Materials Engineering (ICME 2022)
|
| Symposium
|
6th World Congress on Integrated Computational Materials Engineering (ICME 2022)
|
| Presentation Title |
Fatigue Strength Prediction and High-Throughput Design by Mechanics Theory Guided Transfer Learning for Extremely Small Sample Database of Steels |
| Author(s) |
Xiaolu Wei, Chenchong Wang, Chunguang Shen, Wei Xu |
| On-Site Speaker (Planned) |
Chenchong Wang |
| Abstract Scope |
Fatigue strength (FS) is one of most important properties. Traditional trial-and-error method for new materials is costly and time-consuming due to complicated fatigue tests. Machine learning (ML) has been widely employed in material science but adequate data is required. In order to reduce costs of data accumulation for FS, based on transfer learning and correlation between properties of steels, a transfer framework (TR) for FS prediction was proposed. The TR aims to predict FS based on small fatigue data and low-cost big data of tensile properties. In the TR, ML models were first trained to estimate tensile properties. Then, TR models were trained to estimate FS. The resulting TR performs great predictive capability. Further, genetic algorithm was applied to search new low-alloy steels with high FS. The results exhibit high reliability and effectiveness. This research provides inspiring guidance for prediction of properties difficult to accumulate data due to high cost. |
| Proceedings Inclusion? |
Definite: Other |