Аннотация:Digital shadows in industrial IT environments are virtual
copies of production or manufacturing processes based on historical or
real-time data obtained from physical sensors, control or automation systems. There is only one-way interaction between the shadowed process
and its virtual copy, which differentiates a digital shadow from a digital twin exchanging the data in both directions. For many industrial
applications, however, building a digital shadow using historical data is
a sufficient, but quite challenging, task requiring the deployment of the
entire data analytics pipeline.
The presented paper demonstrates how machine learning and some
related AI-based approaches can assist in developing and effective usage
of intelligent IT applications. Thermal spray coating has been chosen as
a use-case for demonstrating the applicability and feasibility of the chosen methodology for enhancing the operation of an industrial IT system
supporting the coating process. The outcome of a comparative experimental studies demonstrated that artificial neural networks provide the
most robust, versatile and generalisable technique for engineering data
analytics in the chosen problem domain.