Modeling and solving composite structure design with constraint programming

Details

Supervisors

Schaus, Pierre

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] en sciences informatiques, à finalité spécialisée

Abstract

enComposite Structures are widely used due to their numerous desirable mechanical, cost and weight properties. They are composed of plies (layers) of carbon fibers. For each ply, one must decide its orientation from the set of possible angles: -45°, 0°, 45°, and 90°. The stack of plies must follow strict constraints on the chosen orientations to achieve mechanical properties of the composite, such as sufficient buckling load. The design problem becomes more complex when determining the stack of plies for a complete surface material, that does not require the same number of plies in every region of the surface. Not only must the orientations be selected in each region, but it is also necessary to decide which plies are discontinued between adjacent regions. Thanks to its declarative nature, Constraint Programming (CP) offers an elegant modeling of the constraints, making it easy for designers to activate or deactivate them as needed. This thesis proposes a CP model to solve the composite structure problem. The performance of this model on synthetic yet realistic instances when solved by different exact solvers, including Mixed Integer Programming (MIP) solvers, demonstrates the superiority of CP over MIP and over a commercial solution implemented by an industrial partner. We then propose a few search strategies, using MaxiCP, that can enhance the solving of this problem and study how to find solutions to unfeasible instances of the problem by relaxing some constraints. It opens up the adoption of CP as an efficient building block of Computer-Aided Design tools for composite structures.