Fuzzy regression decision systems for assessment of the potential vulnerability of bridge to earthquakes

Abstract

This paper presents an innovative fuzzy regression approach for the assessment of the potential vulnerability of bridge to earthquakes. Taking the Mao-Luo-Hsi Bridge in Nantou County of Taiwan as an example, structural models representing vulnerable sections of the bridge are established and updated by ambient vibration tests. Because of uncertainties in the structural models and ambient vibration data preprocessed by wavelet filtering, the fuzzy regression approach is adopted to synthesize the maximum displacements of the bridge sections so as to develop a bridge alert–action principle. The fuzzy regression model can be achieved by the fuzzy blending of each individual input–output realization for the critical sections D, E, and F of the bridge subjected to various earthquake intensity scales ranging from one to seven. Verification of the resulting fuzzy regression models along the longitudinal and transverse directions of the bridge reflects on the high R 2 values of the models being 0.7317 and 0.8401, respectively. The developed decision systems suggest that the action value can be determined for the location of the bridge site when the earthquake intensity reaches scale 6 corresponding to the maximum slopes of the resulting fuzzy regression curves, and the next one lower scale 5 refers the alert value.