Predictive maintenance with sensor data analytics on a Raspberry Pi-based experimental platform

Abstract

Predictive maintenance techniques can determine the conditions of equipment in order to evaluate when maintenance should be performed. Thus, it minimizes the unexpected device downtime, lowers the maintenance costs, extends equipment lifecycle, etc. Therefore, this article developed a predictive maintenance mechanism with the construction of a test platform and data analysis along with machine learning. The information transmission of sensors was based on Raspberry Pi via the TCP/IP (Transmission Control Protocol/Internet Protocol) communication protocol. The sensors used for environmental sensing were implemented on the programmable interface controller and the data were stored in time sequence. A statistical analysis software platform was adopted for data preprocessing, modelling, and prediction to provide necessary maintenance decision. Using multivariate analysis users can obtain more information about the equipment’s status, and the administrator can also determine the operational situation before unexpected device anomalies. The developed modules are decisively helpful in preventing unpredictable losses, thus improving the quality of services.