Industrial production is choking on data. It comes from everywhere: production lines, sensors, devices, call centers and all kinds of maintenance reports. It is tempting to just discard it, but we all know that it would be so valuable if only you could extract information from it! The question, however, is where and how. On the whole, extraction of value is best done at the source, such as the sensor itself.
The magic phrase is analytics on edge. In my opinion, IoT is no longer conceivable without edge analytics. The great thing about it is that the data is analyzed immediately, during the transfer, in the data stream and before it is saved. Think of technology as an intelligent filter. The data transport is reduced, and the analytical system can detect and avert problems, or trigger alert functions and any action that needs to be taken. If you have to store your data before using it, none of this is possible.
In my opinion, IoT is no longer conceivable without edge analytics.
Sounds good, but do you really need it? This is a central question. I think edge analytics offers many options for IoT, but it is not necessary for everyone. If the answer to at least one of the following questions is no, then it is probably worth thinking about edge analytics. Otherwise, the end does not really justify the effort.
- Is latency acceptable for an edge-to-cloud round-trip?
- Do your machines / devices always connect to the network?
- Can you transfer all the necessary data to the data center without any barriers?
If you are still not sure, you may be inspired by a customer example: GE Transportation. The company analyses sensor data for around 1,200 locomotives using edge analytics, drawing on SAS Event Stream Processing, directly on board and in real time. This allows it to optimize, among many other things, the energy consumption of its engines. Navistar use the technology for more than 15,000 trucks in the US, to find out when important parts are likely to fail. If the system sounds an alarm, an instruction is sent to the driver’s cab to drive to the workshop, where an appointment is already reserved.
How does it work?
Sensor data allows quite reliable predictions of the health of equipment such as turbochargers. Indicators include static data such as engine type and year of construction, and dynamic data such as total engine mileage or average oil temperature. This makes it possible to develop early indicators in a statistical model that indicates likely early failure of the turbocharger. With the analytical model, the information can then be given to the system so that, for example, an alert is only generated if the oil temperature rises in combination with other specific indicators. If the model is “hit”, an error message is sent to a central service point.
My advice on this is that IoT technology allows you to use state, position and motion data to make quick decisions. It follows, though, that the data should be analyzed promptly, which means edge analytics. Otherwise, it makes little sense to jump on the digitization bandwagon, because it realistically provides no proper value.
This article has given you a quick summary but has barely scratched the surface of this subject.
If you are interested in the possibilities of using IoT data in manufacturing, you might want to read this white paper: 5 Steps for Turning Industrial IoT Data into a Competitive Advantage.
Discover more insights on SAS Industrial IoT Information Hub