Incorporating fuzzy knowledge into fitness: multiobjective evolutionary 3D design of process plants

Designing technical plants is a complex and demanding process. It has been shown that the optimization of the simple facility placement problem is already NP-hard. Optimization of plant designs must obey a number of criteria derived from several fields of process engineering. We discuss an expansion of the simple facility placement problem with nonregular floor spaces and more than one layer. Additionally, we allow forbidden zones and predefined ways. In contrast to other approaches our system can cope with competitive criteria. These can be defined by a plant designer in an intuitive way according to concepts from fuzzy logic. This leads to the multiobjective optimization of costs and fulfillment of weighted design rules. We describe an evolutionary algorithm to construct Pareto-optimal blueprints of chemical plants. The smart indexing of rules and assignment of conclusions to components allows an efficient calculation of the rule fulfillment as part of the fitness function. O...

• Type: Conference
• Year: 2005
• Where: GECCO
• Authors: Ingo Mierswa
• Keywords: Chemical Plant, Facility Placement Problem, GECCO 2005, Simple Facility Placement