Integrated with Multi-Criteria Optimization |

Corresponding
Author: Pokojski, Jerzy |

Author(s):
Pokojski, J. |

Organisation(s):
Institute of Machine Design Fundamentals, Warsaw University of Technology
(Poland) |

In many cases one large design problem consists of a number of sub-problems
which are solved in some order by the designer or a team of designers.
The designers look at the sub-problems as problems which have to be solved
as single activities.
The system of sub-problems can have various structures. The particular structure can reflect the designer's knowledge about the plan of a particular design process. It is not easy to depict what knowledge is needed to solve particular problems of design. In any case we need a theoretical basis, experience and the abilities to model successfully, to make correct observations and to create a suitable hypothesis. Not in every design process are processed the same components. It is very important to see during the design process all important functions of a product. If we think about computer environment supporting design process we should think about making designer free from schematic, standard activities. The designer should be enabled to build and analyze the best models of the considered phenomena, according to his merit knowledge. He should have the possibility to break the limitations of the existing methods, algorithms, computer systems, and the way on which they are integrated. The maze model of the design process fulfills the above claims. With the maze model we can initiate a design process from different nodes. We can conduct this process in different ways and finish in different nodes. The maze model of the design process reflects a multi-sequential aspect of the design process. Parallel to the process of creating a path in the maze model the designer can try to select the best parameters for his solutions and as a consequence create a sequence of problems belonging to the category of optimization problems. We can assume that with every node of the maze model a multi-criteria optimization problem created by the designer can be connected via selecting decision variables, constraints, criteria function. The sequence of optimization problems - accompanying a path in the maze model can be mutually interacted via decision variables, constraints or criteria. The designer solving his problem in the particular node should have the chance to create and solve an optimization problem (connected with this node) as well. Optimization problems from different nodes can interfere. If we want to protect our optimization results from getting game solutions we should use the multi -criteria decomposed multi-level optimization approach. The designer should have the possibility to change the path in the maze model, repeat some parts, consider different structure solutions. And optimization problem should always follow him and be automatically generated according to his actual position in the maze model. |