A Combined Approach of Industrial Edge Computing and Machine Learning for Predictive Maintenance

Recent advances in edge computing have enabled latency-critical industrial applications to increase their capa-bilities by transferring computation data to the network edge, where data processing and analysis occur. However, the major challenge of a smart industry is to minimize downtime and maximize machine lifetime by intelligently estimating predictive maintenance. These demands can be solved by combining the benefits of Industrial Edge Computing (IEC) and Machine Learning (ML) paradigms. Therefore, in this work, we propose an IEC-enabled industrial task offloading and predictive maintenance framework for estimating downtime and future failures of industrial machines. At first, we classify industry-generated tasks into IEC executables and cloud executables using a heuristic technique. Then, tasks are offloaded to suitable computing devices using a queue allocation strategy. The remaining life of industrial machines is also accurately estimated with the help of several ML-based techniques. We carry out performance study with a practical dataset NASA Turbofan Jet Engine Dataset collected from 3 different types of industrial machines. Numerical results demonstrate that our proposed strategy is effective in reducing the end-to-end delay and energy consumption rates while estimating scheduled maintenance at the network edge, generating near-optimal solutions.