Framework for real-time predictive maintenance supported by big data technologies

Abstract

Industry 4.0 boosted the generation of large volumes of sensor data in manufacturing production lines. When adequately mined, this information can anticipate failures and launch maintenance actions increasing quality and productivity. This paper explores the integration of real-time big data techniques in industry. Specifically, this work contributes with a framework for real-time predictive maintenance supported by big data technologies. The proposed framework is composed of: (i) Apache Kafka as messaging system to manage sensor data; (ii) Spark as Machine Learning engine for large-scale data processing; and (iii) Cassandra as NoSQL distributed database. We showcase the synergy of these cutting-edge technologies in a predictive maintenance system tailored for the request. By leveraging advanced data analysis methods, we reveal hidden patterns and insights valuable for researchers across various disciplines. The experiments were performed with the NASA turbofan jet engine dataset, which includes run-to-failure simulated data from turbo fan jet engines.