Data collection is one of the first steps to consider when implementing an IoT solution. Manufacturing machines can be equipped with small sensors to measure all sorts of parameters, e.g. temperature, number of products produced, tension, etc. In a next step, this data needs to be uploaded to a central IoT Hub, typically running in a public cloud, such as Microsoft Azure.
A very simple architecture to implement this, is by establishing direct connections from each of your sensors to the central hub:
Once the raw data is in the central IoT hub, the analytics service can do its magic, and provide the users with a clear interface, showing all sorts of views on the production process.
However, this simple architecture does spark some concerns, including:
- Availability: what happens if the Internet connectivity of your production plant fails? Will production data be lost? Can the IoT devices buffer a large amount of data?
- Bandwidth usage : if hundreds of devices start uploading a continuous stream of very detailed raw data, a lot of bandwidth will be needed;
- Speed of detection: a device anomaly can only be identified when the data has been uploaded and analyzed by the IoT hub, although the sensor itself is much faster aware of the problem.
The concerns mentioned above can be tackled by implementing an on-premise “edge” component, the so called IoT Edge. The architecture still remains straight-forward:
The picture above shows that the IoT devices in the production plant no longer directly communicate to the IoT Hub, but connect to the local IoT Edge component. This component assembles all data, and establishes one single connection to the central IoT Hub running in the public cloud. If the Internet connection would fail, the IoT edge component will buffer all data, and send it to the IoT Hub once connectivity has been restored.
However, IoT Edge is much more than a simple fall-back mechanism to cope with connectivity loss, as it can also take care of :
- Data aggregation: the edge component can analyse the collected raw data, perform a first data aggregation or clean-up, and send only the relevant data to the central IoT hub. This will substantially decrease the amount of data sent to the public cloud;
- Low Latency communication between all sensors and the on-premise IoT Edge, when nearly real-time communication is needed;
- Artificial Intelligence: by using Edge Modules, you can simply deploy complex event processing, machine learning, image recognition or other high value Artificial Intelligence modules, without developing a single line of code yourself.
- Bring your own code: IoT Edge allows you to deploy your own code to the IoT devices.
- Privacy of data: only the necessary aggregated data is transferred to the central IoT Hub, other data remains locally.
IoT Edge can run on all sorts of devices, on different operating systems. For industrial applications, Inimco proposes to work with Advantech equipment, including different models from:
- Advantech UNO series
- Advantech ADAM series
- Advantech WISE series
- Inimco has been using the IoT Edge principles in many of its customer implementations; we are a strong advocate of this concept because of the:
- Possibility to connect with local data sources, such as data historians, OPC Servers, … and process data locally and centrally;
- Control centrally, but run locally principle;
- flexibility of centrally deploying code (using Docker containers);
- possibilities towards (IoT) device management;
- fact that data can be processed and anonymized locally;
- local detection of anomalies in real-time scenarios;
- Azure IoT Edge, currently in public preview, is making it possible to achieve these architecture principles in an easier and more controlled manner. Inimco is currently already testing multiple industrial customer scenarios with Azure IoT Edge.
Contact us if you need more information on how Azure IoT Edge can accelerate your IoT plans.