A new approach for designing thin-walled fibre reinforced composite structures is presented. The goal of this work is to generate a pseudo-optimal preliminary design for composite structures, satisfying all specifications for design, production, maintenance and operation of the structure. A central meta-level design strategy is developed to control the design and discrete optimisation. By making efficient use of artificial intelligence techniques, such as constraint-, case- and rule-based reasoning techniques, all known empirical knowledge is used to full advantage, thereby strongly reducing the design space around a practical, feasible, optimal solution. Numerical optimisations and structural analyses are postponed until later stages of the design, and the number of numerical operations is strongly reduced.