Structural damage detection using a computer vision–based measurement and finite element modeling updating method

A novel structural damage identification method based on the noncontact measurement technique and Jaya algorithm is presented. The “Features from Accelerated Segment Test” (FAST) algorithm and Kanade–Lucas–Tomasi algorithm are used to identify the displacements from the video clips synergistically. Since the fixed threshold value for the FAST algorithm may not meet the pixel requirements for different images, an adaptive threshold value strategy is applied in this study. The natural frequencies extracted from the acquired displacements are used to formulate a multisample objective function, which is defined on the basis of the Bayesian inference. The K-means clustering strategy, the Hooke–Jeeves pattern search mode, and the colony reduction mechanism are integrated into the Jaya algorithm to improve the convergence of the Jaya technique. Structural damage location and severity can be identified through minimizing the multisample objective function using this enhanced Jaya algorithm. The proposed method is then applied to a laboratory space frame. The damaged elements of the five verification cases (cases 1–5) are detected correctly. Finally, the damaged elements in cases 6–10 are identified and reported.