Improving an IoT-Based Motor Health Predictive Maintenance System Through Edge-Cloud Computing

Abstract

One of the most prominent use case of Industry 4.0 in manufacturing is predictive maintenance (PdM), which can be used to maintain and manage motor equipment. PdM is a condition-based maintenance technique that predicts when an equipment might fail by monitoring the performance of the equipment. The predictions, in turn, will let the users perform maintenance on the equipment before it fails. Through these predictions done by PdM systems, equipment health can be monitored, equipment can be proactively maintained, and usage of maintenance resources can be optimized. Most of the current PdM systems use a pure cloud setup where the processes are being executed in the cloud, relying heavily on Internet connectivity. The main focus of this paper is to improve the base pure cloud system by incorporating edge computing to create an edge-cloud setup, wherein major processes will be executed in the edge device. System prototypes using pure cloud setups and edge-cloud setups are evaluated through experiments, comparing the results with regards to timing breakdown of the processes, and the CPU and memory usage. Experiments show that while an edge-cloud setup can certainly perform better than a pure cloud setup, it will only be able to do so under certain circumstances. If the edge device involved has very low computing power and therefore takes a significant amount of time to perform the computations, it may still be better to just incur the network delays and send the data to a fast cloud server for computation. Put another way, the specifications of the edge device and of the cloud instance that will be used must be considered when deciding whether to go with a pure cloud setup or a cloud-edge one.