Current Research
Project duration: 2023 - present (on-going)
The research works embarked under this project contribute to the following United Nations Sustainable Development Goals:
The pursuit of nuclear power as an alternative source of energy is motivated by two key factors: 1) the increase in global energy demand; and 2) the issue of global warming and climate change. While such option serves as a viable solution to the two aforementioned key issues, it comes with considerable risk given that history has seen the devasting impact of a nuclear accident on life-forms and the environment at large - e.g., the Three-Mile Island accident (1979); the Chernobyl accident (1986); and the Fukushima-Daiichi accident (2011). For this reason, the question of nuclear risk and safety assessment becomes of paramount importance when the nuclear option is considered. As the saying goes: "Safety is freedom, freedom from unaffordable harm, and, thus, a human right" (Zio 2018).
The current research topic of interest, under the R2UQ framework, is Probabilistic Risk Analysis (PRA) for nuclear safety which is being investigated within the Singapore Nuclear Research and Safety Institute (SNRSI). The research seeks to address the following three grand challenges within the field:
To address the aforementioned challenges, the focus of the research are the following:
Reviewing the engineering design of advanced nuclear reactors, and advanced machine-learning approaches for PRA as part of a literature survey;
Developing state-of-the-art machine-learning strategies towards risk and reliability analyses for nuclear safety; and
Assessing the verification and validation performance of machine-learning strategies for risk and reliability analyses through well-defined benchmark problems.
As such, the aims and objectives of the research are the following:
To review and understand the state-of-the-art methods and standards towards PRA on nuclear reactors;
To provide data-centric solutions towards risk and reliability analysis for nuclear safety under uncertainty; and
To engage in regional and international discussions on research developments regarding PRA in nuclear safety.
In achieving the above aims and objectives, the research focus is classified into four distinct themes:
Theme 1: Learning from data, physics, and contextual knowledge
Research focus: Developing supervised machine-learning strategies to predict the reactor system performance towards performing a reliability analysis, resilience assessment, and system health monitoring – e.g., Artificial Neural Network, Physics-informed Neural Network, and Reinforced-learning.
Key publications (since 2023):
Adolphus Lye, W. Vechgama, M. Sallak, S. Destercke, Scott Ferson, and Sicong Xiao (2025). Advances in the Reliability Analysis of Coherent Systems under Limited Data with Confidence Boxes. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A: Civil Engineering, 11, 04024074. doi: 10.1061/AJRUA6/RUENG-1380
Theme 2: Information extraction & Data interpretation
Research focus: Developing unsupervised machine-learning strategies to extract key features and information from the measurements, social media platforms, noisy monitoring data, and expert opinions – e.g., Principle Component Analysis, Reduced-order modelling, and K-Nearest Neighbours.
Key publications (since 2023):
Theme 3: Uncertainty quantification & Variability characterization
Research focus: Developing inverse methods to quantify the uncertainty and variability of key physical model input parameters given the limited data, as well as robust computational approaches for forward methods towards uncertainty propagation – e.g., Markov Chain Monte Carlo, Probability bounds analysis, and Dependency-tracking.
Key publications (since 2023):
Adolphus Lye, Scott Ferson, and Sicong Xiao (2024). Distribution-free stochastic model updating for the Physics-guided reliability analysis of a material thermal property under limited data. In Proceedings of the 17th Probabilistic Safety Assessment and Management and Asian Symposium on Risk Assessment and Management 2024, Sendai. Link to paper: Click here
Adolphus Lye, Scott Ferson, and Sicong Xiao (2024). Comparison between distance functions for Approximate Bayesian Computation towards Stochastic model updating and Model validation under limited data. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems Part A: Civil Engineering, 10, 03124001. doi: 10.1061/AJRUA6.RUENG-1223
Theme 4: Accident modelling & Emergency response
Research focus: Developing modelling strategies towards characterizing complex accident sequences, identifying the most significant contributing factor to the accident event, and deriving a risk-informed emergency response plan – e.g., Fault-tree analysis, Multi-unit PRA, and Bayesian network.
Key publications (since 2023):
Adolphus Lye, Jathniel Chang, Sicong Xiao, and Keng Yeow Chung (2024). An Overview of Probabilistic Safety Assessment for Nuclear Safety: What Has Been Done, and Where Do We Go from Here? Journal of Nuclear Engineering, 5, 456-485. doi: 10.3390/jne5040029
Funding body:
Collaborators:
