4 Presentations

• 
  03:30 PM - 03:55 PM

  A Matrix-Free Augmented Lagrangian Method for Structural Optimization

  • Andrew Lambe, presenter, University of Toronto
  • Sylvain Arreckx, Polytechnique Montréal

  We present how using a matrix-free optimizer for structural optimization allows us to handle a large number of design variables and failure constraints efficiently when the structural analysis is expensive. In particular, we show how to solve large structural design problems without aggregating the failure constraints.

• 
  03:55 PM - 04:20 PM

  Implementation of a Matrix-Free Augmented Lagrangian for Nonlinear Optimization

  • Sylvain Arreckx, presenter, Polytechnique Montréal
  • Andrew Lambe, University of Toronto

  In many applications, problems are so large that we cannot compute/store explicit Jacobians and don't have access to Hessian information. In this talk we outline a matrix-free algorithm for solving nonlinear problems with both equalities and inequalities. Our algorithm is based on an augmented Lagrangian approach and relies on matrix-vector products only.

• 
  04:20 PM - 04:45 PM

  Optimization of Composite Wing-Box Structures: An Engineering Perspective

  • Graeme Kennedy, presenter, University of Toronto

  Composite materials have favourable strength and stiffness to weight properties that can be exploited to obtain highly tailored structures for high-performance structural applications. For these reasons, composites are increasingly used in primary aricraft structures. However, the favourable structural properties of composites can only be fully exploited using structural design optimization. Realistic composite wing-box design optimization problems require thousands of design variables and hundreds to thousands of constraints. In this presentation we outline our structural optimization framework and illustrate the difficulty and complexity of realistic composite wing-box design problems.

• 
  04:45 PM - 05:10 PM

  An Hybrid Approach for Solving Non-Convex Quadratic Programs with Non-Convex Quadratic Constraints

  • Anthony Guillou, presenter, GERAD - HEC Montréal
  • Pierre Hansen, HEC Montréal
  • Sylvain Perron, GERAD, HEC Montréal

  We present a new hybrid method to solve exactly non-convex quadratic programs with non-convex quadratic constraints. The method combines interval arithmetic branch and bound approaches (Messine 2004) and an improved version of the branch and cut algorithm of Audet et al (2000).