An enhanced method to determine the best-fit auto-regressive model (AR model) for structural damage identification is proposed in this paper. Whereby, two parameters of the model, including the number of model order and the window size of data, are analyzed simultaneously in order to accomplish the optimized values by means of Akaike’s Information Criterion (AIC) algorithm. The damage condition of structures can be detected by defined damage indicators obtained from the first three AR coefficients of the best-fit AR models. The ability of the proposed damage identification method is compared with the process that only utilizes conventional AR models without concern of parameter selection. The proposed method is verified using experimental data previously collected from a large-size bridge structure in the Structural Laboratory at the University of Central Florida. The results indicate that this method can detect and locate damage more effectively.

中文翻译：

---

基于最佳拟合自回归模型的桥梁结构损伤识别非参数方法

本文提出了一种确定用于结构损伤识别的最佳拟合自回归模型（AR模型）的增强方法。从而通过Akaike的信息准则(AIC)算法同时分析模型阶数和数据窗口大小两个模型参数以达到优化值。结构的损伤状况可以通过从最佳拟合 AR 模型的前三个 AR 系数中获得的定义损伤指标来检测。将所提出的损伤识别方法的能力与仅使用传统 AR 模型而不考虑参数选择的过程进行了比较。使用先前从中央佛罗里达大学结构实验室的大型桥梁结构收集的实验数据验证了所提出的方法。结果表明，该方法可以更有效地检测和定位损伤。