Artificial intelligence methods for predictive maintenance

Abstract

The rapid evolution of Artificial Intelligence over the past few years has revolutionised applications in several fields. Such field is the Predictive Maintenance (PdM) concept. In contrast to reactive or scheduled maintenance, it aims to predict when machinery or infrastracture is likely to fail. This approach is proven to be cost-efficient, preventing unplanned downtime and optimising operational efficiency. Therefore, it was decided to explore this interesting concept using Machine Learning algorithms on a benchmark dataset. Two problems were approached: • Regression problem - Predict Remaining Useful Life (RUL) • Classification problem - Classify the condition of the machine (No fault or type of fault) The process began with performing an Exploratory Data Analysis (EDA) on the provided dataset. After the necessary manipulation of the data and graphical evaluation, new Features were engineered to serve the purpose of the tasks. Several Models were tested in order to select the best performing one. All of the above will be analysed in the following chapters.