FaultBit: Generic and Efficient Wireless Fault Detection Using the Internet of Things

The timely monitoring and maintenance of industrial machines is critical to prevent expensive disruptions and down-time. The Internet of Things (IoT) offers a solution to instrument production processes at lower cost and complexity. However, wireless IoT networks are bandwidth constrained, which precludes the transmission of high frequency signals such as vibration or electrical current. Prior approaches to tackling this problem require a high degree of application-specific re-engineering. In this paper, we argue for a new approach to fault detection that is accurate, efficient and applicable to a large class of fault detection problems. We propose FaultBit, an application independent toolkit for fault classification that can gather, compress and classify data using IoT networks. We evaluate FaultBit in two representative scenarios using current and vibration data. In both cases, FaultBit offers classification accuracy 99%, which is very close to application-specific classifiers, while requiring 512 \(\times \) less bandwidth, enabling a battery life of several years.