GUIDES - Generalized Support and Investigation Design for Nested Systems

Smart Factory is a central concept in Industry 4.0, national high-tech development strategy. Physical processes, which occur at such a factory are monitored and are translated to a virtual model. Size and complexity of this model makes an analysis, both manual and automated, extremely costly. The model can be presented as a number of high-dimensional elements, e.g., tasks that should be finished by specific deadlines, machines that are required to perform specific tasks, employees with various shifts and access to a limited amount of machines, etc.

The goal of Generalized Support and Investigation Design for Nested Systems~(GUIDES) project is an effective enterprise resource planning, which presumes construction of a schedule, i.e., allocation of tasks on respective machines at the same time minimizing computation time, operational cost, etc. From the research point of view such a scheduling introduces the following challenges: treatment of data's high-dimensionality and avoidance of a local optima, which is a frequent result for various optimization heuristics.

To achieve this goal, we tackle the aforementioned problems separately. First, we reduce the model's dimensionality by the means of Multi-Dimensional Scaling, whose task is to transpose a high number features of a respective element to a low-dimensional equivalent (e.g., position on a plane) based on user-defined comparison operators, whilst still preserving dependencies between the elements. Significantly simplified model, in contrast to high-dimensional one, can be analyzed by global optimum search approaches such as Cluster Analysis or Nearest Neighbor Search to couple similar elements. Such a coupling, in turn, can be treated as either a near-optimal scheduling recommendation or can be further refined in order to find a global optima.

  • Gefördert von: SAB
  • Kontaktperson: Karsten Wendt
  • Projektlaufzeit: 09/2015 bis 11/2017

Zu dieser Seite

Sebastian Götz
Letzte Änderung: 12.07.2016