In this lecture, I will introduce methodologies and practical case studies of the application of machine learning techniques in advanced manufacturing, with a strong focus on the condition monitoring approaches.
In the first part, model-based, signal-based and knowledge-based algorithms and schemes for fault diagnosis and monitoring will be presented. In model-based approach a model of the system under monitoring is developed, and any deviation of measured signals from the model predictions is considered as a symptom of fault. Signal-based approaches rely on specific indicators extracted from specific signals. It is however in knowledge-based approaches that machine learning methods are more vastly employed.
The second part of the lecture is devoted to the presentation of industrial cases studies on the application of learning tools for condition monitoring in manufacturing. Three case studies will be presented: the monitoring of a flight electro-mechanical actuator, the monitoring of medium-voltage circuit breakers and the residual useful life (RUL) estimation of electrical valves.
Finally, I will present an overview of envisaged applications of the artificial intelligence in manufacturing with a higher-level perspective, that envisages the role of smart products, smart factories, and smart value chains.
Mirko Mazzoleni is assistant professor (RTD-A) at University of Bergamo, Italy, where he teaches the courses about system identification and data science. His research interests are on nonparametric system identification and fault diagnosis methods with the Control and Automation Laboratory at University of Bergamo. He is author of several scientific papers on the aforementioned topics and a Springer book titled “Electro-Mechanical Actuators for the More Electric Aircraft”. He actively contributed for european research projects on fault diagnosis methods for electro-mechanical actuators (FP7 CleanSky JTI HOLMES project and H2020 CleanSky2 JTI REPRISE project. He founded AISent srl, a start-up that engineers artificial intelligence solutions.