A Business Workflow Architecture for Predictive Maintenance using Real-Time Anomaly Prediction On Streaming IoT Data
Emrullah Gultekin, Mehmet S. Aktaş · 2022 IEEE International Conference on Big Data (Big Data) · 2022
The Internet of Things (IoT) usually consists of many fully synchronized devices. Since these devices sense data from outside, they are usually outside and sometimes in places that are difficult to access in case of failure. Sometimes, late-recognized errors cause the system not to work and cause significant damage to the system. For this reason, it is crucial to avoid faulty situations with minor damage or even prevent these situations before they occur. Although there are studies on the topic in the literature, there is an emerging need to examine this problem comprehensively and target predictive maintenance and self-healing systems utilizing IoT systems. In this study, we propose a business workflow architecture using streaming-based machine learning algorithms to provide predictive maintenance utilizing IoT systems. In the proposed predictive maintenance workflow, we utilized various machine learning algorithms such as Adaptive Random Forest, Hoeffding Tree, Leveraging Bagging, SPegasos, and Single Drift Classifier classification algorithms. To show the usability of the proposed business workflow, we provide a prototype implementation. We provide an experimental study on the prototype to investigate the prediction success. We also investigate how fast the system can learn from the streaming data. We conduct this evaluation with unbalanced and balanced data. In this manuscript, we report the results of our experimental study.