A remarkable non-Gaussian distribution of local acceleration responses of the cooling tower is indicated through field measured data analysis

and extreme estimation method based on Gaussian assumption of wind-induced responses are not fully applicable. To study the extreme considering non-Gaussian characteristics

a 179 m high large cooling tower in northwest was selected

and acceleration vibration signals at representative positions of the tower were obtained under ambient excitation. The vibration signals were preprocessed combining random decrement technique and natural excitation technique. Three pattern recognition methods (ARMA

ITD and STD) were applied to analyze the frequencies and damping ratios for the first ten order modes

and comparison between measured values and the finite element calculation were carried out. Following modal combination

equivalent synthetic damping ratio of the cooling tower was derived. The peak factors and acceleration response extreme of samples were calculated by two kinds of extreme estimation methods (peak factor method and Sadek-Simiu method)

and a single value of the measured response was given. The wind-induced response analysis was carried out under the measured damping ratio 2% and the standard damping ratio 5% based on the total transient time domain method

and error analysis between measured values and the finite element calculation was carried out. The comparative study shows that the results of the measured and finite element analysis are consistent with the results of the first 10 frequencies

with the maximum difference of 9%. The maximum difference between the measured acceleration responses of the two extreme estimation methods is 32.02%

and local points of wind vibration response extreme of finite element under measured damping ratio and the measured extreme are consistent.