AIoT

Artificial Intelligence of Things (AIoT) is the merging of Artificial Intelligence (AI) technologies with the Internet of Things (IoT) infrastructure to realise more efficient IoT operations, revolutionise human-machine interaction and optimise data management and analysis. AIoT will contribute to efficiency gains especially in the areas of wearables, smart home, smart city and smart industry.

Application example

You may already be familiar with the operation of lamps or heating by your smartphone or voice assistant. These examples are often referred to as smart lighting control and smart heating control. However, this control is usually not as smart as one would first think. The devices send, for example, the status of the device "on" or "off", or sensor values such as the set and actual room temperature. However, this is mostly rule-based if-then programming, which cannot really be called "smart".

Example Predictive Maintenance

A real "smart" application can be seen in the area of predictive maintenance in an industrial context. Here, the machine is always serviced shortly before the actual failure of the machine and exactly those parts are replaced that will actually break down shortly.

An example of a ball bearing can be used to illustrate the different types of maintenance. When designing bearings, the bearing service life is typically calculated. However, this calculation is not really precise. For example, if the bearing life is calculated with 100,000 revolutions and the machine has already completed these 100,000 revolutions, the bearing is not necessarily functionally limited. However, the bearing is usually replaced as a preventive measure, which means that the actual service life is not fully utilised. The alternative would be to wait until the bearing breaks down and causes problems in the machine. In this case, however, the failure of the machine is guaranteed, and in addition, spare parts have to be procured, which means that the failure of the machine can last longer.

Predictive maintenance takes a different approach. The sensors installed in the machines or retrofitted (retrofit) can record all kinds of values, such as the vibration of the machine, the heat development, the pressure in the interior, the humidity and much more. The data becomes most interesting when it is recorded up to the point of a malfunction. By using machine learning algorithms, correlations among the data can be determined (features), which have led to a failure of the warehouse. The more often a failure is registered, the better patterns can be recognised. The advantage is now obvious. Machine operators now know exactly when which components of the machine need to be serviced. Maintenance is thus carried out at the ideal time (e.g. outside operating hours) and the service life of the components is fully utilised. In addition, spare parts can be anticipated and ordered at the optimal time.