Canonical Correlation Analysis of Selecting Optimal Ground Motion Intensity Measures for Bridges

The correlation between ground motion intensity measures (IMs) and that between component seismic demand parameters (DPs) have significant effects on the seismic demand of bridges. A canonical correlation analysis method was introduced to consider the two types of correlations in the optimal selection of IMs for probabilistic seismic demand analysis of bridges. A group of 100 ground motion-bridge samples were generated with the bin approach and Latin hypercube sampling method, to quantify the uncertainties of earthquakes and bridges. With the nonlinear time history analysis, the overall correlation between IMs and DPs was established (IMs and DPs were respectively treated as a whole), meanwhile the correlation between IMs and that between DPs were considered. The optimal IM was selected and verified with the criteria of efficiency, proficiency and sufficiency. The canonical correlation analysis reduced the complexity and limitation of simple correlation analysis used for all pairs of correlations between each of the IMs and each of the DPs. The combined peak IM of PGA (the peak ground acceleration), PGV (the peak ground velocity), PGD (the peak ground displacement) and combined root mean square IM of a(RMS), v(RMS), and d(RMS) were proved to be the optimal IMs and more appropriate in the probabilistic seismic demand analysis of normal girder bridges.