Structural damage detection has been a significant research topic in the last two decades. Many of the damage detection methods that have been proposed do not work accurately for identification of location and severity of damage. This article presents an accurate damage identification method using stiffness matrix decomposition and modal identification method. The total stiffness matrix is decomposed to story stiffness matrices, and damage location and severity will be detected by solving a linear regression problem. Blind Source Separation–Sparse Component Analysis (BSS-SCA) and frequency domain decomposition (FDD) have been used as output-only modal identification methods for calculating structural modal parameters. Modal parameters such as natural frequencies and mode shapes are used to determining the total stiffness matrix of structure in the current state. These methods are output-only and do not need to measure excitation's inputs. Therefore, it is capable of online structural health monitoring. The proposed method is applied in a structure, modeled as a shear building. In order to simulate a real condition, up to ± 2.5% noise has been added to the calculated responses of the structure. Finally, results show that the presented method with both of the output-only modal identification methods can detect damages and their locations properly, but the detection algorithm with the BSS-SCA method is more accurate than the one with FDD method in diagnosing damage severity. However, the accuracy of the proposed method is decreased when a complicated excitation is applied to the structure and when the damping ratio increases.

在过去的二十年中，结构损伤的检测一直是一个重要的研究课题。已经提出的许多损坏检测方法不能准确地用于识别损坏的位置和严重性。本文提出了一种使用刚度矩阵分解和模态识别方法的精确损伤识别方法。总刚度矩阵分解为楼层刚度矩阵，并且可以通过解决线性回归问题来检测损坏的位置和严重程度。盲源分离-稀疏分量分析（BSS-SCA）和频域分解（FDD）已用作计算结构模态参数的仅输出模态识别方法。模态参数（例如固有频率和振型）用于确定当前状态下结构的总刚度矩阵。这些方法仅用于输出，不需要测量激励的输入。因此，它可以进行在线结构健康监视。所提出的方法适用于以剪力建筑物为模型的结构。为了模拟真实条件，已将高达±2.5％的噪声添加到所计算的结构响应中。最后，结果表明，提出的两种仅输出模态识别方法都能正确地检测出损伤及其位置，但是BSS-SCA方法的检测算法在损伤严重度诊断方面比FDD方法更为准确。然而，