Conferences and supporting programme
Localizing Analytics for Speed, Reliability and Reduced Power Consumption
New tools make possible physics-based analytics in an embedded environment. By computing locally – performing predictive and prescriptive analytics at the edge of the IoT – significantly less data must be directed to the cloud. Further, the data sent are more informative and they are available in server-less situations. This improves reliability, speeds computation time and reduces power consumption. In addition, physics-based models have the ability to assess the internal state of an observed system. This makes their predictions more accurate. By enabling physics-based models to operate in real time in small footprint embedded devices, the resultant robust predictive ability can lead to a reduction of needed, and often expensive, system monitoring sensors. To illustrate how embedded model-driven analytics can be implemented, two real world examples will be demonstrated: an electric motor health monitor and a high voltage safety system. Each step in the implementation process will be shown, from model design to the utilization of embedded scientific computing tools, final real-time model optimization, and system predictions.
--- Date: 27.02.2018 Time: 4:30 PM - 5:00 PM Location: Conference Counter NCC Ost