Research about online failure prediction of rail vehicle core components (such as wheels, bearings, and bogies) based on big data and artificial intelligence (AI) has become popular in view of its role of improving rail vehicle operation safety. The recent vibration energy harvesting sensor network relieves sensor nodes' dependence on wired power, which provides a green and low-cost way of collecting data from rail vehicle core components. However, the integration of an energy harvesting sensor network and AI to provide online failure prediction for rail vehicle components still faces several challenges, such as weak energy harvesting power and unstable vehicle-ground communication data rate. In this article, EH-Edge, an energy harvesting-driven cloud-edge-end device collaborative Internet of Things (IoT) platform, is proposed to efficiently integrate energy harvesting and AI to solve these challenges. A two-level collaborative AI failure prediction is proposed and deployed in the EH-Edge platform to reduce energy consumption in terms of sensor node, amount of data upload, and time delay of failure prediction. Detailed software and hardware designs and real-world data sets are also published.
Read full abstract