Detecting Sensor Faults and Outliers in Industrial Internet of Things

Abstract

The Internet of Things (IoT) is rapidly growing in nearly every industry. People are adopting and using IoT technology in every facet of their lives, ranging from civilian use (e.g., smart homes, hobbies, fitness, and health monitoring), to military (e.g., battlefield surveillance systems) to industrial (i.e., industrial automation, plant and equipment monitoring) etc. More and more companies are introducing IoT technology innovations to increase efficiency. In the resources industry specifically, IoT can technically be used to imitate human behaviour and separate or even eliminate human and operational equipment interaction, hence keeping people safer. Another vital IoT offering in the mining space is equipment reliability. Maintenance and engineering teams are switching from the traditional time-based maintenance approach to predictive maintenance using IoT. Therefore, a key deliverable from IoT technology in mining is to have accurate and reliable data to assist management with making informed decisions. In the IoT process chain, there are several influences that can cause sensor faults and result in erroneous data. The data can also contain anomalies or outliers. This paper aims to describe IoT sensor faults, anomalies or outliers, fault detection using Deep anomaly detection techniques and mitigation or correction strategies specific to the Industrial Internet of Things within the resources industry.