Industrial IoT Edge Architecture for Machine and Deep Learning
ecosystem including device abstraction to transport data to VMware Pulse IoT Centerâ„¢, which offers an end to end infrastructure management solution for secure and reliable data flow from devices to applications. The figure below shows the different ways that Pulse IoT Center connects with edge devices. See details at: www. vmware. com / products / pulse. html. Liota is part of Pulse IoT Center; however, it can be used separately to orchestrate data from sensors through gateways to other IoT services.
Figure 12: Platform Tier Architecture for Deep Learning
PLATFORM TIER ARCHITECTURE FOR DEEP LEARNING
The Platform tier receives real-time and batch data from the edge and performs Deep Learning model inference and training. It caches raw data as well as processed data after feature engineering. It also transfers the final model inference results to the enterprise tier for Business Intelligence( BI) reporting and visualization.
The Platform tier performs Deep Learning training as well as inference. The platform pulls raw data from the edge cache, as well as performs feature engineering on real-time and stored edge data. The feature data includes information from all edges to create aggregate information such as median temperature of devices at all edges.
Use Case for Edge and Platform Tier Learning
The example below shows a use case for separating Machine and Deep Learning at the edge and Platform tiers.
Figure 6 shows temperatures from four devices. The median temperature of all four devices( in red) is overlaid on the device temperatures in blue. The edge can detect the anomaly in Device 3 due to a sudden drop in temperature but the anomaly in Device 4 is
IIC Journal of Innovation- 65-