₪... Digital twin of a heavy machinery plant

  The requirement resulting from the monitoring and analysis of technical processes to create a digital image of a plant in order to obtain predictions about optimal maintenance intervals or failure scenarios through sensor parameter modifications, the use of previous history or simulated plant states has become an important research topic in the manufacturing industry in recent years.

  Above all, the realization of the two sub-topics: the use of previous history and simulated plant states can usually only be realized using CI-driven categorizers. This is due on the one hand to the fact that machines in particular are subject to a natural aging process, secondly, especially in the heavy machinery sector, not all materials are the same and therefore the sensors monitoring the plant produce a wide range of differing sensor data values ​​even in normal conditions, and thirdly, it is only the interaction of the various machine components that defines the overall condition of the machines. And this interaction can be very different. Strangely enough, it is relatively easy for trained personnel to evaluate the machine in terms of its running safety without having access to the sensors that are placed on the various units of the machine. In this case, one speaks of a mental model that the personnel have.

   Ideally, a CI-driven twin should map a corresponding model, which suggests the use of CwA structures that act in a particularly blurred manner.

   The following is a simple digital twin of a production plant created by IngB RT&S and currently in operation. Before the actual intelligence of the twin comes into play, it must be ensured that the S/N ratio of the sensor data flowing into it is as high as possible. In the case of data with a high level of interference, this should be done using DLS operators.