Anomaly detection in IIoT: a case study using machine learning

In this paper, we explore multiple machine learning techniques applied for anomaly detection in IIoT data from engine-based machines. We evaluate sensor data on different engine characteristics such as fuel usage, engine load, and oil pressure to gauge when a particular engine shows anomalous behavior and may experience a failure. In particular, we use methods such as multi-variate linear regression, Gaussian mixture models, and time-series data analysis to detect outliers in the machine behavior. Timely detection of such anomalies helps the maintenance staff to perform preventive maintenance and ensure maximum up-time for the machines. In addition, anomalous behavior may not always indicate failure but simply inefficient usage of the machines; we also try to optimize these inefficiencies.