The event, annual meeting under the international Master’s degree in Production Engineering and Management introduced by both the University of Trieste (Italy) and of Ostwestfalen-Lippe (Germany) University of Applied Sciences and Arts, offers the opportunity for researchers, experts and professionals to meet and exchange information, discuss experience, specific practices, technical solutions for planning, design, and management of manufacturing, service systems and processes.
This year’s special focus was on green production and digital manufacturing in the context of Industry 4.0, that are currently major topics of discussion since past editions of the PEM, in which sustainability and efficiency emerged as key factors.
With further study and development of direct digital manufacturing technologies in connection with new management practices and product lifecycle management, the 9th edition of the PEM conference aimed to offer new and interesting scientific contributions, with a rich program including more then 30 speeches organized in several sessions, dedicated to Direct Digital Manufacturing in the context of Industry 4.0, Product Life-Cycle and Supply Chain Management, Industrial Engineering and Lean Management, Wood Processing Technologies and Furniture Production e Innovative Management Techniques and Methodologies.
SARTEC, together with the Department of Mechanical, Chemical and Material Engineering at University of Cagliari (Italy), presented a speech entitled Fault prediction of a centrifugal pump using machine learning methods techniques.
In the paper has been pointed out that the demand for cost-effective, reliable and safe machinery operation requires accurate fault detection and classification to achieve an efficient maintenance strategy and increase performance. In this scenario, fault prediction plays a key role to extend the lifetime of components and to reduce unplanned equipment thus preventing costly breakdowns and plant shutdowns.
SARTEC presented the preliminary development of an algorithm for failure prediction of a centrifugal pump which operates in the production line at a refinery plant operating in Italy. This component is an industrial machine of critical importance and its correct functioning affects the efficiency of a production line and possible failures would penalize the company’s profitability. The study uses Machine Learning (ML) and Data Science techniques applied to data collected from the measurement sensor databases of the pump. A binary classification technique based on Artificial Neural Networks (ANN) is used to predict pump failures a week in advance, an interval of time considered optimal to allow an adequate maintenance intervention. The results show that the model works correctly and predicts some of the pump’s failures resulting in a significant maintenance decision support system for operatives.
SARTEC speech has been included and published in official Proceedings of the Conference.