About this Abstract |
Meeting |
2020 TMS Annual Meeting & Exhibition
|
Symposium
|
General Poster Session
|
Presentation Title |
L-34: Prediction of Aluminum Alloy Mechanical Properties with Bayesian Neural Network |
Author(s) |
Shimpei Takemoto, Yoshishige Okuno, Kenji Nagata, Junya Inoue, Manabu Enoki |
On-Site Speaker (Planned) |
Shimpei Takemoto |
Abstract Scope |
Understanding the Process-Structure-Property-Performance (PSPP) relationship is one of the goals of the computational materials design. We have constructed a Bayesian neural network for predicting multiple mechanical properties of aluminum alloys. The Markov Chain Monte Carlo (MCMC) method is widely used for simulating multi-dimensional posterior distribution in Bayesian Statistics. We have applied the Replica Exchange Monte Carlo method based on the Metropolis algorithm, an improved MCMC method, to estimate the neural network architecture and its hyperparameters. From the obtained neural network, we have discussed the PSPP relationship in aluminum alloys such as dominant factors that affect their mechanical properties. We have also addressed an inverse problem for optimizing the process for a desired set of properties. |
Proceedings Inclusion? |
Undecided |