Quantifying seismic response uncertainty of electrical substation structures using endurance time method

A new procedure is proposed to quantify uncertainties involved in the seismic response of structures including ground motion and modeling variability. The advent of performance-based earthquake engineering (PBEE) and probabilistic seismic risk analysis necessitates consideration of uncertainties in predicting seismic responses. However, the high computational cost of conventional methods restricts their application to only simple numerical models. The present study employs the endurance time (ET) method concept to predict the response distribution which requires much less computational efforts than conventional methods. The proposed procedure for the quantification of record-to-record variability uses adaptation of the ET excitation function?s response spectrum with those of ground motions in a range of structural natural period. The proposed procedure for the quantification of modeling variability is a sampling-based approach in which ET analysis is carried out on randomly generated structural models. The application of the methods is investigated in predicting the response distribution of two electrical substation structures. The results are compared with the benchmark methods such as incremental dynamic analysis (IDA) in order to evaluate the accuracy level of the method. It is found that the proposed method can predict the distribution of responses with less than 15% discrepancy relative to the benchmark method at the majority of intensity levels, while the computational effort reduces more than 15 times compared to the benchmark method. Thus, the proposed method provides a strong tool extending the application of the performance-based design in structures with a reasonable computational cost.

Automated summary

A new procedure is proposed to quantify uncertainties involved in the seismic response of structures, using the endurance time method concept and a sampling-based approach on randomly generated structural models. This method can accurately predict response distribution with reduced computational effort, extending the application of performance-based design in structures.