This paper proposes a method to support decisions to be taken within a concurrent approach for the space system preliminary design: the defined architecture is based on a Multi-Criteria Decision Making approach mixed with methodologies coming from the Approximate Reasoning domain. The method here presented is focused on saving analysts' time and effort by addressing the decisions they have to make during the preliminary design process to solve inconsistency and bottlenecks risen from the parallel design of several subsystems From a theoretical point of view, revisited Genetic Algorithms are applied, within each single subsystem design domain, in order to obtain a non-dominated solution set to be considered for solving conflicting design at system level; the Analytical Hierarchical Process - supported by dedicated blocks implemented by the Fuzzy Logic approach - has been selected as the fittest tool to simulate the causal relationships between variables and objectives, normally prerogative of the analysts' experience in the spacecraft design domain, within the system level point of view. Simulations showed the ability of the algorithm to find conflicts and suggest a set of subsystem parameters to be tuned to converge -consistently with a user defined cost functions vectors - to a final spacecraft configuration; the tool runs in real-time with the on-going space system design process, in order to support the team leader in making decisions. The mission analysis is completely involved in the process ending having predominance on the overall system design. A comparison with a completely transparent optimization process, implemented by applying a Multi-criteria optimization based on a revised Genetic Algorithm approach, highlighted the capability of the proposed approach to move towards the final Pareto front solution.
|Titolo:||Preliminary Spacecraft Design: Genetic Algorithms and AHP to Support the Concurrent Process Approach|
|Data di pubblicazione:||2003|
|Appare nelle tipologie:||04.1 Contributo in Atti di convegno|