A fundamental Bayesian two-stage model with the generalized fused lasso penalty for structural damage identification

The existing fundamental Bayesian two-stage damage identification model with L1 regularization (FBT-L1) combines the most probable value of modal parameters with the associated posterior uncertainty to consider the impact of uncertainty on damage identification. However, the FBT-L1 model neglects the use of similarity information among multiple measurement data, which can reduce the identification accuracy. The generalized fused lasso penalty can achieve the fusion of similarity information from multiple measurement data. To enhance the identification accuracy, this paper proposes a fundamental Bayesian two-stage model with the generalized fused lasso penalty (FBT-GFLP). The primary goal of the FBT-GFLP model proposed in this paper is to improve the accuracy of damage identification by using the similarity information from multiple measurement data on the basis of considering the uncertainty of modal parameters. Numerical and experimental results indicate that the FBT-GFLP model incorporating the generalized fused lasso penalty improves identification accuracy by 8.03% and 18.26% than the two existing models, respectively. •A Bayesian model with fused lasso is proposed for damage identification.•An improved cuckoo search algorithm with ranking-based mutation is presented.•The method improves identification accuracy by utilizing measurement similarity.•Numerical and experimental results confirm the effectiveness of the proposed method.