Politecnico di Torino (logo)

Prediction of bearing quality nonconformities in manufacturing processes: a case study

Giulia Fiore

Prediction of bearing quality nonconformities in manufacturing processes: a case study.

Rel. Franco Lombardi, Giulia Bruno, Emiliano Traini. Politecnico di Torino, Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica), 2020


Today the industrial ecosystem is experiencing a real transformation. With Industry 4.0 and its enabling technologies we are facing a transition from the old factory model to the new Smart Factory. In this context, quality standards (e.g. ISO 9001:2015) promote the application of predictive maintenance systems in Companies, to the point this has become a certification criterion. Predicting and consequently scheduling the appropriate maintenance action, through online or offline condition monitoring, is crucial, not only in terms of cost-savings over routine preventive maintenance but also for ensuring higher quality products. In this frame, one wonders whether it is possible to highlight a deviation of process quality using the same information and data collected for predictive maintenance purpose. The present study addresses the concept of quality monitoring with an application on bearing manufacturing process. More in detail, real data, currently collected on a SKF production line for condition monitoring of machines, and information, coming from the related Digital Twin, are processed through Machine Learning and Artificial Intelligence algorithms. Main aim of this research is to study the possibility of predicting the amount of bearing quality nonconformities by monitoring mechanical vibrations of grinding machine. This is initially examined via analysis of variance through Ordinary Least Squares (OLS) and Feasible Generalized Least Squares (FGLS) models. Then more flexible Machine Learning models are trained. Finally, the time-dependence of project variables is demonstrated and a Nonlinear Autoregressive Exogenous (NARX) model is used to map the relationship between product quality and vibrations of manufacturing machine. Results validate the investigation path selected in this work. However, they highlight overall vibration values are not enough accurate to predict the amount of quality nonconformities, even if they could be useful to predict their occurrence. Suggestions are also provided to develop a quality monitoring system based on IIoT sensor network.

Relators: Franco Lombardi, Giulia Bruno, Emiliano Traini
Academic year: 2019/20
Publication type: Electronic
Number of Pages: 110
Additional Information: Tesi secretata. Fulltext non presente
Corso di laurea: Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica)
Classe di laurea: New organization > Master science > LM-25 - AUTOMATION ENGINEERING
Aziende collaboratrici: UNSPECIFIED
URI: http://webthesis.biblio.polito.it/id/eprint/15349
Modify record (reserved for operators) Modify record (reserved for operators)