Damage Detection for Underground Arch Structure Considering Errors in Baseline Model

Abstract

Underground arch structure is an important structure form of civil air defense engineering, and its damage detection is an important link in safety evaluation of civil air defense engineering. In this study, a neural networks based damage detection method using the modal properties is presented. Based on the analysis of the dynamic properties of underground arch structure before and after damage, it was proved that the ratio of mode shape was insensitive to finite element model errors. The variation rate of mode shape was taken as the signature for damage detection, and it was proved that the variation of mode shape caused by multiple damage state was transformed into superposition of single damage state. It is not needed to constitute multiple location damage patterns when using the neural networks. A typical underground arch structure is analyzed to demonstrate the effectiveness of the proposed method. When 10%’s modeling errors exists, the damage location and extent can be recognized well.