Comparison of stationary and non-stationary wind-induced responses of a super-large cooling tower based on field measurements

Recent measurements made by our research group of a super-large cooling tower indicate that its fluctuations in wind-induced responses exhibit a degree of nonstationarity, echoing similar observations reported for other structures. This might cause errors in estimates of extreme responses and misunderstanding of wind-induced effects if nonstationarity is neglected. In this study, the wind-induced response signals of a super-large cooling tower (height 190 m) in a coastal region were measured for the first time under real Reynolds number and turbulent flow conditions. After noise reduction had been performed, nonstationarity of the signals was identified within various time intervals. The mean wind effect, pulsating wind effect, probability density distribution, dynamic amplification factor, and extreme responses of the super-large cooling tower were studied based on stationary and non-stationary models. Finally, the power spectral density (PSD) and evolutionary power spectral density (EPSD) of the wind-induced response signal were analyzed. The resonance spectral expression of wind-induced responses at resonance excitation points, which is applicable to super-large cooling towers, is summarized. The wind-induced responses presented strong nonstationarity. Non-stationary models that consider response nonstationarity are important in the authentic assessment of the extreme responses of super-large cooling towers. The extreme constant calculated by the stationary model cannot provide an adequate assurance rate and reduces the economic efficiency of extreme response estimates. The vibration energy distributions of resonance excitation points in different regions of the cooling tower were similar, but the PSD functions at quasi-static points were dramatically different from each other. The energy distribution of the resonant excitation points showed a phased trend, and the proposed resonance spectral expression considers three stages of variation in the PSD function of the responses to achieve high predictive accuracy.