Reliability and Maintenance Optimisation in the Age of Data-Centric Engineering

Abstract

Data-centric engineering is drastically impacting all areas of engineering and industry. The goal of data-centric engineering is to make data science, mathematics, statistics, and machine learning fundamental to engineering practice, leading to engineered products that are more intelligently designed, monitored and maintained, more reliable, more cost e cient, and safer to use. Reliability and maintenance optimisation are two important research areas being impacted by data-centric engineering. The large amounts of dynamic data being collected by datadriven technology has the potential to provide accurate real-time information about the state of products, allowing for the health (or reliability) of products to be continuously monitored. The continuous monitoring of the reliability of products allows us to more accurately plan maintenance, reducing costs and increasing safety. This thesis has two main contributions. One is in the eld of reliability for hard-disk drives with automatic data-collecting devices and one is in the eld of condition-based maintenance for complex continuously monitored multi-component systems with dependencies. In Part I of this thesis, we propose a novel way to model the survival probabilities and failure times of hard drives, using data collected by SMART (an automatic data-collecting device). We de ne critical states for hard drives using data collected by SMART and model hard drive failure times using multi-state models. Using the proposed multi-state models, we seek to concretely de ne the impact of critical attributes on the failure time of a hard drive. In Part II of this thesis, we propose a novel condition-based maintenance policy for continuously monitored multi-component systems subject to economic and stochastic dependence. More speci cally, we propose a novel loss-based utility function, that is incorporated in a Bayesian sequential decision framework, to decide which components are to be maintained at maintenance opportunities for continuously monitored multi-component systems that are subject to economic and stochastic dependence.