Industrial processes become more and more complex. Questions for the operator can be

  • “can the specifications be reached ?”
  • “how large is the optimization potential ?”
  • “which components are overspecified ?”
  • “what are the driving parameters ?”
  • “what are my parameter settings for optimum performance in all scenarios?”

and others.

The answer of such questions is almost always obtained with deeper system understanding. A detailed knowlege of the component of a system and their mode of interaction is mandatory. A very good approach to such knowlege is the setup of a parameterized system model. Such a model should be designed such that it incorporates the underlying physics of the process under investigation, leaves however heuristic parameters to be fitted to the real measured data. By carefully building and tuning such models the system response to deterministic system inputs can be predicted. Disrepancies between measured and predicted model input-output relations can be used to adapt model parameters and by such detect the most probable cause for such discrepancies.


From your real system which already might have internal sensors, a controller software and possibly first live production data we arrange our system modelling building blocks as needed.

Based on physics, we devise and implement time dependent low-dimensional models of your processes. By doing so, the deterministic part of the system is covered based on physical principles. We formulate the governing differential equations for your system and implement them in an efficient environment, such that the resulting model can be used side by side of the with the life process data at a later step. By parameter fitting of the model with measurement data (such as temperatures, time scales, efficiencies, mass flows, burning rates, fuel efficiencies, energy consumption, flow rates, etc.) a very good understanding of the process can be established.

Depending on the identified use-cases the models can be refined to deliver the required accuracy. Additional aspects can be added anytime. As digital twins such models connect time-series data of the real existing process with performance parameters that cannot be determined directly.

In case your system needs further optimization and insight in a particular fluid dynamic effect we are well able to assess these with Computational Fluid Dynamics methods. The insight obtained from this possible additional step leads to improved reduced order models, slim enough to be used as digital twins for life data processing.