Huang2) and Ming Gu3)
Total Page:16
File Type:pdf, Size:1020Kb
The 2019 World Congress on Advances in Structural Engineering and Mechanics (ASEM19) Jeju Island, Korea, September 17 - 21, 2019 Comparison study of typhoon characteristics based on stationary and non-stationary models *Wen Xie1) , Peng Huang2) and Ming Gu3) 1), 2),3) State Key Laboratory of Disaster Reduction in Civil Engineering, Tongji University, Shanghai 200092, China 2) [email protected] ABSTRACT A comparison study of the wind characteristics of typhoon Jongdari based on stationary and non-stationary mode was presented. The original data was collected at the height of 40m in seaside (Shanghai Province, China) where typhoon passed through directly. First, the run-test method and discrete wavelet transform way were employed to evaluate the stationarity and extract the time-varying mean wind speed, respectively. Then the gust factor, turbulence intensity and turbulence integral scale were compared accordingly. The results demonstrate that the wind characteristics described by the non-stationary model are more centralized and more stable. In addition, the power spectral density and the evolutionary power spectral density were calculated and compared. the Von Karman spectra fits well with the measured spectra. 1. INTRODUCTION Accurately understanding and establishing the wind characteristics of typhoon have always been a hot research. With the improvement of sensors acquisition capability and stability (e.g., structural health monitoring system), more reliable measurement data are providing(Xu et al. 2017).In general, the wind characteristics were studied based on stationary assumption (Davenport 1961) which has adopted by various codes (ASCE/SEI 7–10. 2010, AIJ-RLB-2004. 2014, GB50009-2012. 2012). During this hypothesis, the statistical characteristics of wind speed were considered as constant and it is convenient to apply to wind-resistant design (Solari et al. 2015, Fenerci et al. 2018, Lin et al. 2018, He et al. 2019, Li et al. 2019).However, recent measurement wind records indicate that the boundary-layer wind speed induced by typhoons might not be stationary(Tao et al. 2017), which means it may be inappropriate to describe the wind speed as the ergodic random process. As such, the analysis of 1) Graduate Student 2),3) Professor The 2019 World Congress on Advances in Structural Engineering and Mechanics (ASEM19) Jeju Island, Korea, September 17 - 21, 2019 nonstationary typhoon wind characteristics is also gradually concerned (Huang et al. 2015, Hong 2016, Kim et al. 2018, Wang et al. 2019). This article was concentrated on the comparison study of the near-ground wind characteristics during typhoon Jongdari (No.1812) landing. As the result of the influence induced by different factors (e.g. terrain, wind attack angle and measurement situation), although some articles (Tao et al. 2016, Tao et al. 2017)have studied wind characteristics using both two models, the measured results are still insufficient to establish a database of non-stationary characteristics for wind designing. There were two differences between this paper and forementioned. The original data captured the overall process that typhoon passed through directly. The terrain around measurement site was flat which is the typical landform on the southeast coast of China. The remainder of this article is organized as follows: Section 2 describes the detail of measurement site and wind data. Section 3 introduces the theories of stationary and nonstationary models briefly. The run-test method and discrete wavelet transform way were employed to evaluate the stationarity and separate the time-varying mean wind speed, respectively. Section 4 and section 5 investigate and compare the wind characteristics in two models, including the gust factor, turbulence intensity, turbulence integral scale, power spectral density (PSD) and evolutionary power spectral density (EPSD). Section 6 presents the main findings and conclusions. 2. FIELD MEASUREMENT AND WIND DATA 2.1. Field measurement The meteorological tower is 40 m in height and situated on a flat area close to the Yangtze River's estuary near Shanghai Pudong International Airport. The surrounding terrain can be regarded as terrain B according to the China design code (GB50009- 2012. 2012) and the exposures around the facility are of inhomogeneous roughness situations. The type of the anemometer is R.M. Young 81000, and it has a sampling frequency of 4 Hz, which can directly measure the three-dimensional wind speed, horizontal wind direction, and vertical wind direction. The north wind is defined as a wind direction angle of 0°, and the wind angle increases clockwise. For more detail about this measurement site see Huang (Huang et al. 2012) . 2.2. wind data The typhoon Jongdari was a strong, long-lived and erratic tropical cyclone that impacted Japan and East China in late July and early August 2018. Typhoon Jongdari was born on the northwest Pacific Ocean on July 25, 2018. On July 26, as Jongdari started to interact with an upper-level cold-core low to the north which significantly enhanced poleward outflow, it intensified to a typhoon in the afternoon despite increasingly unfavorable vertical wind shear. At around 01:00 JST on July 29 (16:00 UTC July 28), Typhoon Jongdari made landfall over Ise, Mie Prefecture with ten-minute maximum sustained winds at 120 km/h (75 mph) and the central pressure at 975 hPa (28.79 inHg). The storm weakened rapidly inland and made its second landfall over Buzen, Fukuoka Prefecture. At around 10:30 CST (02:30 UTC) on August 3, Jongdari made landfall over Jinshan District, Shanghai. It rapidly weakened after landfall and dissipated on the next day. The wind data recorded during Jongdari effected the The 2019 World Congress on Advances in Structural Engineering and Mechanics (ASEM19) Jeju Island, Korea, September 17 - 21, 2019 measurement site around 18:00 CST on August 2 to 18:00 CST on August 3.The typhoon rote and the measurement site are shown in Fig. 1 and Fig. 2. Fig. 1 typhoon rote (from the Japan Meteorological Agency website) Fig. 2 measurement site 16 8 original measured data wind speed (m/s) data after reconstructing 0 150 300 450 600 time (s) Fig. 3 comparison of original measured data and data after reconstructing The 2019 World Congress on Advances in Structural Engineering and Mechanics (ASEM19) Jeju Island, Korea, September 17 - 21, 2019 It should be noted that the accuracy, integrity and reliability of data are critical to the analysis result. Some data loss occurred in the original wind record inevitably, which might induce by sensor failures. So the compressive sensing method (Comerford et al. 2016), which is shown to estimate successfully the essential features of the stochastic process power spectrum, was employed to reconstruct the original signal. A sample is given in Fig. 3, Comparing with the original data, the data after reconstructing is more reasonable without data mutation. 3. THEORY BACKGROUND 3.1. stationary and non-stationary models For the traditional stationary wind model, the wind speed Ut() was regarded as a constant mean U plus a zero-mean turbulence component ut(), detailed as U()() t U u t (1) Based on the stationary random process hypothesis, the wind characteristics (e.g. gust factor, turbulence intensity, turbulence integral scale and power spectral density) can be calculated by using U and ut(). For the non-stationary wind model, the wind speed was regarded as a deterministic time-varying mean Ut*() plus a zero-mean turbulence component ut*() (Chen et al. 2004), detailed as U()()*() t U* t u t (2) Where the fluctuation ut*() of typhoon is regarded to be stationary. For simplicity,all the parameters in non-stationary model are distinguished with others in stationary model by asterisk. 3.2. Stationary test Several methods are applied to evaluate the stationarity of data series. Such as the run test method (Levitan 1988) and the reverse arrangement method(McCullough et al. 2013). Due to the simplicity of the run test method which is non-parametric and detect the existence of underlying trends of a signal in the view of hypothesis testing, it was adopted in this paper. For the detail see Levitan (Levitan 1988). The sample duration and the desired level of significance were taken as 10min and 5%, respectively. The stationary test result is shown in Fig. 4, the proportion of nonstationary segments is 55%. It shows that the data has strong nonstationary characteristics. 3.3. Extract the time-varying wind speed As mentioned before, extract the time-varying wind speed is critical to non- stationary analysis. A variety of techniques, including the moving average (MA)(Lombardo et al. 2014), empirical model decomposition (EMD) (Xu et al. 2004, Cheng et al. 2017) and discrete wavelet transform (DWT) (Chen et al. 2005, McCullough et al. 2013, Su et al. 2015) are using to extract the time-varying component by researchers. But MA has limited resolution and may lead to unsmoothed mean, it could not capture the rapidly varying mean (Chen and Letchford 2005). Comparing with MA, EMD and DWT perform better. Also known that EMD face the problem of boundary The 2019 World Congress on Advances in Structural Engineering and Mechanics (ASEM19) Jeju Island, Korea, September 17 - 21, 2019 effect and mode fixing and the choose of decomposition level in DWT are still critical to the estimation (Huang et al. 2015). Because DWT is capable to decompose both univariate and multivariate data, it will be used for deriving the mean component, detailed as J U()()() t DjJ t A t (3) j1 Where J is the decomposition levels, Dtj () an AtJ () are the detail component at level j and the approximation component, respectively. For more detail see Ye(Ye et al. 2017). Here Daubechies’ wavelets of order 10 (Db 10) was chosen and the wavelet-based self-adaptive method by Tao (Tao et al. 2017) was employed to calculate the reasonable decomposition level.