Bridge damage analysis under joint environmental and operational variability

Abstract

One of the main challenges for bridge damage identification using monitoring data is to acquire sensitive damage features but insensitive to operational and environmental effects as well as noise. Specifically, the temperature as part of environmental variability can mask structural damages in bridges. Inspired by the capabilities of machine learning methods, Principal Component Analysis (PCA) has been applied here as a well-known and robust technique for removing environmental variability. As a first purpose, PCA is used considering only ambient vibrations and the natural frequencies are considered as damage indicators. As a second objective, PCA in conjunction with Hilbert Huang Transform (HHT) and Variational Mode Decomposition (VMD) are applied to eliminate the environmental influence in transient vibrations due to traffic. The combined methodology is applied to the case of a numerical benchmark by using the Instantaneous Phase Difference (IPD) as novel vibration damage feature in the case of non-stationary vibrations. The results show that the proposed strategy to use the non-stationary vibration due to traffic instead of ambient vibration seems a promising tool for damage identification and, therefore, its capabilities in real bridge applications are worth exploring further when experimental data from real bridges will become available.