AIDA. Artificial Intelligence supported conceptual Design of Aircraft

AIDA. Artificial Intelligence supported conceptual Design of Aircraft

Rentema, D.W.E.

Jansen, F.W. (promotor)
Torenbeek, E. (promotor)

Aerospace Engineering

2004-11-15

This thesis describes the development of a computer support tool that supports the initial, conceptual design process. In this first design phase one or more concepts are defined which are assumed to be able to comply with the design specifications. These concepts can be elaborated in more detail with the aid of the conventional CAD tools that have been developed in the last decades and are currently widely available. Due to the non-deterministic character of conceptual design, however, such conventional computer techniques are inadequate to support this phase. Therefore the use of Artificial Intelligence techniques has been investigated in this project. The initial application was chosen in the area of aircraft design, because of its complexity. The tool and methodology are, however, also applicable to support the design of ships, buildings and industrial appliances. The developed design tool and methodology is based on the well-known "design cycle", which in our version consists of four steps: 1. Suggesting a concept: this task is supported with Case-Based Reasoning (CBR) techniques. From a case-base that is filled with data of existing artefacts, one artefact is selected whose performances best match the design specifications. This case is used as a starting point and delivers the configuration or topology of the design, and the initial parameter settings. Possibly two or more cases can be combined, resulting in an adapted case. 2. Simulating the artefact's performances: this task is supported with Rule-Based Reasoning (RBR) and Geometric Modelling (GM) techniques. With the rules representing the domain knowledge in an explicit, algebraic format, a network of rules and relations is built that relates the sizing parameters with performance parameters. GM is used to build a geometric model and to handle the geometric constraints. 3. Evaluating these performances: using the RBR network of relations and the GM constraints, the related solvers allow quickly and easy changing of parameter values to modify the design. 4. Proposing modifications: the design tool does not support this task explicitly. However, as is mentioned in step 1, a first modification of the case's topology (adaptation) is supported with CBR. Then the cycle is repeated until the design concept is satisfactory and can be used as input for further elaboration and evaluation with existing CAD/CAE tools. Each of these three techniques, CBR, RBR and GM, has been implemented in a separate module. Existing programs have been used and adapted for this study. To test the set-up and viability of the design tool, a case-study has been performed, concerning the conceptual design of an 80 passenger commercial aircraft. The study has proven that the design approach is useful for "configuration design" type of design tasks, but is less suitable for innovative and creative design. Although each technique has shown to be suitable for its task, there are also some difficulties. For example, each technique requires substantial effort in the preprocessing phase, when filling the data or knowledge bases. And experience is needed to operate the CBR and RBR modules properly. Many apparently minor practical issues can have a serious effect on the efficiency of the techniques. Issues which need much attention in following development studies are case adaptation and integration of the three applied techniques.

http://resolver.tudelft.nl/uuid:ef473d71-e384-4f2f-b9c2-881eb2fb9918

Institutional Repository

doctoral thesis

(c) 2004 D.W.E. Rentema