The paper presents a global optimization method to compute the minimum limit load
factor of trusses subjected to unknown but bounded loads. We assume that the
external forces consist of a part proportional to a load factor and a part that
is uncertain around its nominal value. The worst-case limit load factor is
introduced as the smallest limit load factor realized with some uncertain
parameters. In order to detect the worst case, we have to find the global
optimal solution of a nonconvex optimization problem, which is the major
difficulty of the worst-case limit analysis. By reformulating the worst-case
determination problem as a mixed 0-1 programming problem, we propose
a global optimization algorithm as a combination of a branch-and-bound
method based on the linear programming relaxations and a cutting plane
method based on the disjunctive or lift-and-project cuts. The worst-case limit
loads, as well as the corresponding critical loading patterns, are computed
to demonstrate that our method converges to the global optimal solutions
successfully.
Keywords
data uncertainty, limit analysis, integer programming,
cutting plane, branch-and-bound, global optimization