A Structured Approach to Machine Learning Condition Monitoring

The aim of the chapter is to explain the basic concepts of Machine Learning applied to condition monitoring in Industry 4.0. Machine learning is a common term used today in different fields, mainly related to an automated and self-learning routine in a decisional process. This chapter details how a Machine Learning approach may be structured, starting from a distinction between Supervised and Unsupervised approaches. These two classes have different advantages and disadvantages that constrain their application to specific boundary conditions. Machine Learning techniques are the core part of a structured methodology for the condition monitoring, but other phases, such as the pre-processing of data, the feature extraction and the evaluation of performances, are equally important for the success of a condition monitoring system. Together with standard parameters used to assess the performances of the machine learning method, a particular emphasis will be given to the interpretability of the results that can be determinant in the choice and development of a specific tool for condition monitoring in an industrial environment.