HOW THE TECHNOLOGY WORKS The Machine Learning software features GE’s SmartSignal® patented capabilities for mathematically compact modeling techniques to analyze data in real time. The solution creates and monitors unique, personalized, empirical models for each piece of equipment at the serial-number level. The models correlate groups of sensors based on actual performance data of each machine in its unique operating context, with dynamic bands across all known loads, ambient conditions, and varying operating contexts. The software will detect patterns of deviation from the expected behavior; anomaly notifications are triggered when real-time values deviate and persist from predicted band values. Notifications with apparent causes and suggestions are generated with a range of prioritization, depending on the criticality of the pattern identified. The analysis is supported by the visualization of the related charts, so reliability and maintenance experts can track and diagnose the developing failures. Alerts Dynamic band Sensor tag data Model Expected value Fixed alarm threshold Predict Detect any deviation pattern from expected behavior Early Stages of Damage Alarm Forecasting Normal Operation Diagnose Dynamic empirical analysis with high level of accuracy and minimal data science expertise required Forecast Linear regression modeling enables forecasts to reveal when an alarm limit will be reached Early Detection Based on Operation History Associated parameters from monitored assets are modeled together. In this figure, Blue is True (actual reading) and Green is Good (model predicted reading). It is the overall behavior correlation that determines what each individual sensor is “expected” to indicate.