System Modeling

In the past only trial-and-error approach was used in the design process: given a problem, it was solved directly, without knowing the indirect consequences onto other aspects. A solution was achieved only after several iterations which was impossible to quantify previously. The time-to-market needs becoming more strict, this approach is effective no more. Since the origins and alongside the design path the Customer must think about system in its entirety: the must is to consider every single component not only as a single entity, but also as a player which interacts with other players mutually. The System Modeling idea is born.

DOFWARE has a deep and certified expertise concerning modeling of systems involving several different physical aspects. Dymola is the reference platform, based on Modelica standard language. The typical approach to the System Modeling could be subdivided into the following steps:

  • system process analysis in its entirety;
  • identification and isolation of every single component which can be modeled virtually by means of ordinary differential equations, algebraic relations and/or appropriate algorithms;
  • appropriate model construction for every single component: either by exploiting what Dymola library already offers or by implementing from scratch mathematical algorithms based on Modelica language;
  • entire virtual system assembly: single component models are made communicating by the Drag&Drop technique; the multi-physical nature of the solver enables components belonging to different world to interact without any problem.

System Modeling can be used for several purposes:

  • pre-design: high-level models implementation for analyzing and comparing preliminary configurations during feasibility studies;
  • design: collaborative platform creation where engineering domains interact, each by inserting its pertaining models and by recovering constraints coming from other domains;
  • on-design calibration: once a real prototype has been built, an experimental campaign is performed which allows the calibration of the virtual system model and the evaluation of margins of error;
  • off-design evaluation: by exploiting the calibrated model, the system is made working in critical operating conditions (e.g. dictated by regulations) and possible malfunctioning are highlighted;
  • control logics validation: the calibrated model allows the validation of control algorithms, even implemented in different environments;
  • optimization: by means of appropriate mathematical strategies (ranging from gradient-based to evolutionary algorithms), the design variable optimal parameters are identified which minimize objective functions chosen by designers. By exploiting the computational lightness of system models, DOE activities can be performed beforehand for considering most influencing parameters exclusively. DOFWARE uses several tools to this aim, beloging both to commercial (Isight®, Matlab®, Mathematica®, Dymola®) and open-source (Dakota, Scilab, Octave, Python, JModelica) worlds;
  • MiL-HiL: by means of hard real-time techniques, the developed models are made communicating among themselves in order to certify the cohesion to the theoretic models. The same check is then performed by building a co-simulation system between the plant and the real hardware;
  • virtual realities: mathematical models are put inside immersive 3D graphical engines in order to reproduce the behavior of the real world. The main applications pertain to the technical personnel training tools and to monitoring systems of hostile places.