LAUSR

The objective of this study was to develop a technique for vibration-based damage detection through data reduction and a visualization technique known as Sammon mapping, which directly uses sensing signals. Based on the instrumented accelerometer, damage assessment of the structure is done using either ambient vibration or seismic response for centralized and sensor-level data analyses in the time or frequency domain. For the centralized data analysis, the embedded data matrix [X]m × n was constructed, with m as the number of sensing nodes and n as the number for discrete Fourier amplitude/time series data. Incorporated with the principal component analysis, data compression and pattern recognition were used for damage detection and localization. First, the matrix of mutual distances of data sets from each sensing node on the high-dimensional space was calculated. The two-dimensional space was then generated to create a Sammon mapping configuration. Damage detection and localization could be evaluated through the change in two-dimensional visualization on a Sammon map between different damage states. The proposed two-dimensional visualization technique, which uses multivariate signal-processing algorithms to identify structural damage, was verified using experimental test data of two different damage types of structures (damages occurring either in lower modes or in non-linearity). Using the proposed technique and based on the building’s seismic response data, we can systematically detect the damage occurrence (Level 1 damage detection) and localization (Level 2 damage detection).