On Demand Explainable AI driven Optimization of Critical Nuclear Safety Parameters using “No Code” AI tool

Chowdhury, Khairul; Alam, Syed Bahauddin

Permalink

https://hdl.handle.net/2142/121820 Copy

Description

Title

On Demand Explainable AI driven Optimization of Critical Nuclear Safety Parameters using “No Code” AI tool

Author(s)

• Chowdhury, Khairul
• Alam, Syed Bahauddin

Issue Date

2023

Keyword(s)

• Explainable AI
• Optimization
• Critical nuclear safety parameters
• No Code AI tool

Abstract

Nuclear energy is a part of the silver lining along with renewables and energy efficient systems. However, due to its inefficient and high operations and maintenance (O&M) costs, nuclear energy is more expensive to produce than other forms of energy generation. The modernization or digitalization through artificial intelligence (AI) and machine learning (ML) of other industry provides an unprecedented opportunities. However, the efforts are failing in 70% of cases, fundamentally due to the lack of actionable interpretation of AI driven digitalization within the industry, which require on demand parametric study followed by optimization. Such challenges are quite general throughout the energy industry, not to mention about nuclear, where AI driven digitalization just began. To overcome such the critical challenges of digitalization, a robotized AI driven Autonomous technology centric Artificial Intelligence/Machine Learning (AIML) solutions builder platform is built to improve business process for efficient operation and predictive maintenance (PdM) in the nuclear power plant’s (NPP) light water reactor (LWR) fleet’s operating model. The “Builder” platform offers easy AI model building for predictive solutions, automatic exploratory statistical and machine learning interpretation of data, individual variable contribution and local variable importance for each data points for an explainable data driven solutions. In addition, the tool offers an on spot what if scenario analysis to understand extreme scenarios with added parameter optimization features to maximize, minimize or set a target outcome. The no code platform is tested with numerous use cases pertinent to offshore-onshore oil gas, solar, wind and nuclear predictive asset integrity and production generation. One of the use cases is the reproduction of our previous work on risk assessment of a small modular reactor (SMR) systems. The predictive solutions turnout time has proven to be significantly shorter (3hrs) than the manual analysis time (3 weeks). In addition, the automatic creation of on-demand what-if scenario and optimization application will allow optimization in real time industry environments. The novelty of the technology is, end to end complex predictive solution creation is possible without being an AI expert and adopting complex digitalization technology, while ensuring the desired accuracy in predictive solutions.

Type of Resource

text

Language

eng

Handle URL

https://hdl.handle.net/2142/121820

Owning Collections