Edge-based intelligent fault diagnosis for centrifugal pumps in microbreweries

The beer sector is a significant market worldwide and the number of small breweries is increasing. Centrifugal pumps are essential components for the proper functioning of the production system. However, failures in these equipment can be detected early by Intelligent Fault Diagnosis (IFD) Systems. In this context, this article aims to develop an IFD capable of detecting cavitation and dry-running faults. The proposed method explored the use of data provided by centrifugal pump drives, such as current, torque, and power factor. It was investigated two approaches: using the Shapley value as a feature selector and the Support Vector Machine (SVM) as the classifier, and applying the raw signal data to 1D Convolutional Neural Networks (CNN). The SVM-based model presented better results, with an accuracy of 100% for dry running and 99.3% for cavitation. The CNN-based model presented 97.4% and 80.2% respectively. It is also identified that torque is the most relevant variable.