Applied Machine Learning in Fitness-For-Service Assessments of CANDU Reactors

Abstract

Machine learning is a branch of artificial intelligence where computers are programmed with algorithms to learn a specific task from experience (i.e. data). In recent years, the increase in the collection of data,access to massive amounts of computing power and development of powerful algorithms implemented in accessible software led to the rapid deployment of machine learning in a variety of industries,including: automotive, banking, social media and cybersecurity to name a few. The nuclear industry in Canada is also looking to adopt machine learning. One example is the desire of utilities to improve the monitoring of the nuclear reactors and transition from time-based maintenance practices to condition based maintenance. Other examples are related to screening and processing of inspection data with the goals of increasing accuracy, reducing human analysis time, and reducing critical path impacts to outages. While these projects are large-scale and long-term developments, there are areas in the nuclear industry where comparatively smaller scale machine learning can be rapidly deployed at relatively low costs. This paper provides three examples where small-scale machine learning was implemented, or is currently in development, by Kinectrics, to facilitate fitness-for-service assessments of CANDU reactors in Ontario.