Lye, Adolphus, Prinja, Nawal and Patelli, Edoardo 
ORCID: 0000-0002-5007-7247
(2022)
Probabilistic Artificial Intelligence Prediction of Material Properties for Nuclear Reactor Designs.
In: 32nd European Safety and Reliability Conference, 2022-8-28 - 2022-9-7, Dublin, Ireland.
|
Text
ESREL_Conference_2022 [Adolphus Lye].pdf - Author Accepted Manuscript Download (1MB) | Preview |
Abstract
This work presents the results of a feasibility study towards the development of Probabilistic Artificial Intelligence Prediction of Material Properties (PROMAP) for Nuclear reactor designs. Currently, Artificial Intelligence (AI) approaches are not largely adopted in the nuclear sectors compared to other sectors such as aerospace of manufacturing. One of the main challenges is the availability of a sparse data set to train AI models and the poor consideration of the uncertainty in the data and prediction. As such, the proposed work seeks to merge the AI tools with probabilistic methods. Specifically, probabilistic methods are used to increase the training data set while retaining the physical dependencies among variables. This allowed the provision of large data set to train a set of Artificial Neural Networks, accounting for model uncertainty, for the prediction of selected material properties relevant for nuclear industry. Using Adaptive Bayesian Model Selection method, the results are combined using Bayesian statistic to yield predictions with their associated confidence intervals. The results demonstrated the capability of the proposed approach to develop robust AI tools where the estimates are well-validated against the experimental data with improved accuracy.
| Item Type: | Conference or Workshop Item (Unspecified) |
|---|---|
| Uncontrolled Keywords: | Adaptive Bayesian Model Selection, Artificial Intelligence, Artificial Neural Network, Bayesian Statistics, Confidence Intervals, Material property, Model Uncertainty, Nuclear |
| Divisions: | Faculty of Science and Engineering > School of Engineering |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 06 Jul 2022 07:30 |
| Last Modified: | 17 Mar 2024 14:35 |
| DOI: | 10.3850/978-981-18-5183-4_s24-02-306-cd |
| Related URLs: | |
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3157744 |
Dimensions
Dimensions
