An Approach to Develop Compound Intensity Measures for Prediction of Damage Potential of Earthquake Records Using Canonical Correlation Analysis

A ground motion intensity measure (IM) well correlated with structural demand plays an important role in predicting the damage potential of earthquake records. This study presents a new approach to develop a compound IM through canonical correlation analysis (CCA). This method utilizes the ability of CCA to search for two linear combinations to maximize the correlation between the corresponding random vectors. To illustrate, the nonlinear responses of single-degree-of-freedom (SDOF) systems due to 98 real ground motion records are used. In a log-log scale, a compound IM in terms of a linear combination of six non-structure-specific IMs is developed with a maximum dependence on the linearly combined structural demand of both peak displacement and dissipated energy. Through a comprehensive discussion, the feasibility and performance of the presented approach is investigated for SDOF systems over the entire period range. It is found that the developed compound IM shows significantly high-level correlation with structural demand, and the corresponding correlation coefficients are greater than 0.95 at most periods. This validates the effectiveness of the presented method in improving prediction accuracy for earthquake damage potential through considering multiple IMs.