Domain knowledge Informed Data Analytics Tools for Predictive Maintenance of High-Value Critical Assets

One of the key elements for a successful implementation of the PdM strategy is the usage of technologies that can effectively convert relevant measurement data into actionable information, such as the health condition, and/or the remaining useful life of critical assets.

Currently, Artificial Intelligence (AI) based big data analytics tools have been massively developed in the research community to address this challenge. These AI-based analytics tools are data-driven and black-box, so the interpretation of how predictions are made by such techniques is still an open problem. Although these techniques have been successfully demonstrated in laboratory environments, however, the barrier to deploying such pure-data-driven black-box techniques in the industry is still high.

One of the main reasons is that the performance of such techniques highly depends on a large amount of good-quality data. Unfortunately, the availability of good-quality data is typically limited for high-value critical assets.

To remedy this gap, this PhD project will focus on developing, evaluating, and demonstrating domain knowledge-informed data analytics tools for the predictive maintenance (PdM) strategy applications to high-value critical assets. Among others, the recently developed Physics-informed Neural Network (PINN) technique will be first explored, tailored, and extended into the PdM context of high-value critical assets. It is expected that combining the domain knowledge and the existing data analytics tools will help deploy these technologies in the industry context without the need for a large amount of datasets.

The successful candidate will have the opportunity to work with experts in the data analytics, condition monitoring, prognostics, and health management field, as well as be part of our strong and dynamic research centre at Cranfield University.