Not stated in source 2023 Needs review

Machine learning for predictive maintenance applied in a teaching plant

Abstract

The new technologies from Industry 4.0 allow factories to become smart. It is understood that material, human and financial resources are better used and consequently there is an increase in efficiency and sustainability of processes. It is essential that an intelligent factory be able to detect failures in its process and even prevent them when possible. In this sense, the scenario that is promising due to the results already validated in real applications is the application of Artificial Intelligence. This technology considered to enable Industry 4.0 refers to a large area that is made up of several others. Among them, Machine Learning has shown interesting results when intended for industrial applications. Based on the context presented and the importance of fault detection for modern industry, this work aims to detect anomalous behavior patterns that can identify faults through the analysis of temperature and level curves from an industrial didactic plant. In order to be able to train the intelligent system used, anomalous (simulating failures) and normal situations will be applied to the plant. The collection of variables inherent to the simulated process was transmitted to a dedicated computer through OPC communication, where they were applied to the algorithm capable of learning and approximating the temperature and level curves. The expected and actual behavior were compared, and the divergence between the two curves indicated or not the presence of some anomalous event to the industrial process.

Keywords

Not available

Specialisations

Similar theses