Adaptive topology optimization of fail-safe truss structures

Avoidance of disproportionate and progressive collapse, often termed 'fail-safe design', is a key consideration in the design of buildings and infrastructure. This paper addresses the problem of fail-safe truss topology optimization in the setting of plastic design, where damage is defined as a moveable circular region in which members are considered to have zero strength for that particular load case. A rigorous and computationally efficient iterative solution strategy is employed in both the dual (member adding) and primal (damage-case adding) problems simultaneously, which allows cases of high complexity and many damage cases (maximum of 16290 potential members and 16291 damage cases) to be solved to the global optimum. Common member-based damage definitions (e.g. damage to any one member) are shown to be highly dependent on the nodal grid; in the limiting case completely negating the effect of the fail-safe constraints. The method proposed in this article does not have such limitations, enabling a more sophisticated and robust treatment of fail-safe design. Moreover, the global minimization and high resolutions create new benchmarks for the least-material designs of 'fail-safe' structures using rigid-plastic materials. A number of example structures are considered (short cantilever, square cantilever, multi-span truss), and the effects of damage radius, location, and structure rationalisation are discussed.