An innovative hybrid strategy for structural health monitoring by modal flexibility and clustering methods

Structural health monitoring is usually implemented by model-driven or data-driven methods. Both of them have their advantages and disadvantages. This article proposes an innovative hybrid strategy as a combination of model-driven and data-driven approaches to detecting and locating damage in civil structures. In this regard, modal flexibility matrices of the undamaged and damaged conditions are initially derived from their modal frequencies and mode shapes. Subsequently, the discrepancy between these matrices is proposed as a damage-sensitive feature. To increase damage detectability and localizability, the modal flexibility discrepancy matrix is expanded by the Kronecker product and then converted into a vector by a simple vectorization algorithm yielding vector-style feature samples. To detect and locate damage, this article introduces the k-medoids and density-based spatial clustering of applications with noise techniques. The vector-style feature samples are incorporated into these clustering methods to obtain two different damage indices including the direct clustering outputs and their Frobenius norms. The great novelty of this article is to develop an innovative hybrid strategy for damage detection and localization under noise-free and noisy conditions so that the damage-sensitive feature is obtained from a model-driven scheme and the decision-making is carried out by a data-driven strategy. A shear-building frame and the numerical model of the ASCE benchmark structure are used to validate the accuracy and performance of the proposed methods. Results demonstrate that the hybrid strategy presented here is influentially able to detect and locate damage in the presence of noisy modal data.