Remote Anomaly Detection in Industry 4.0 Using Resource-Constrained\n Devices

A central use case for the Internet of Things (IoT) is the adoption of\nsensors to monitor physical processes, such as the environment and industrial\nmanufacturing processes, where they provide data for predictive maintenance,\nanomaly detection, or similar. The sensor devices are typically\nresource-constrained in terms of computation and power, and need to rely on\ncloud or edge computing for data processing. However, the capacity of the\nwireless link and their power constraints limit the amount of data that can be\ntransmitted to the cloud. While this is not problematic for the monitoring of\nslowly varying processes such as temperature, it is more problematic for\ncomplex signals such as those captured by vibration and acoustic sensors. In\nthis paper, we consider the specific problem of remote anomaly detection based\non signals that fall into the latter category over wireless channels with\nresource-constrained sensors. We study the impact of source coding on the\ndetection accuracy with both an anomaly detector based on Principal Component\nAnalysis (PCA) and one based on an autoencoder. We show that the coded\ntransmission is beneficial when the signal-to-noise ratio (SNR) of the channel\nis low, while uncoded transmission performs best in the high SNR regime.\n