Real-time tracking of structural stiffness reduction with unknown inputs, using self-adaptive recursive least-square and curvature-change techniques

In this paper, a new computationally efficient algorithm is developed for online and real-time identification of time, location, and severity of abrupt changes in structural stiffness as well as the unknown inputs such as earthquake signal. The proposed algorithm consists of three stages and is based on self-adaptive recursive least-square (RLS) and curvature-change approaches. In stage 1 (intact structure), a simple compact RLS is hired to estimate the unknown parameters and input of the structure such as stiffness and earthquake. Once the damage has occurred, its time and location are identified in stage 2, using two robust damage indices which are based on the structural jerk response and the error between measured and estimated responses of structure from RLS. Finally, the damage severity as well as the unknown excitations are identified in the third stage (damaged structure), using a self-adaptive multiple-forgetting-factor RLS. The method is validated through numerical and experimental case studies including linear and nonlinear buildings, a truss structure, and a three-story steel frame with different excitations and damage scenarios. Results show that the proposed algorithm can effectively identify the time-varying structural stiffness as well as unknown excitations with high computational efficiency, even when the measured data is contaminated with different levels of noise. In addition, as no optimization method is used here, it can be applied to real-time applications with computational efficiency.