Abstract: Structure damage detection is often converted into a constrained optimization problem. An improved Particle Swarm Optimization (PSO) algorithm has been proposed to avoid the premature convergence of traditional PSO algorithm in solving the constrained optimization problem by increasing the diversity of the particles in the later phase of traditional PSO algorithm. The numerical results of detecting damages at one or multiple sites for a two-storey rigid frame and the convergence curves of PSO algorithms show that the proposed improved PSO algorithm is more effective to solve the constrained optimization problem than the traditional PSO algorithm with an inertia factor. Moreover, the experimental study of damage detection in 4 cases for a three-storey building model demonstrates the feasibility of the proposed improved PSO in structural damage detection.