Advantages of modeling ABL properties to determine wind loads on structures

In present paper we introduce the main advantages of modeling atmospheric boundary layer (ABL) properties for determination of mean and peak wind loads on structures. Experimental tests were carried out using wind tunnel with uniform velocity profile and new Landscape Wind tunnel (LWT) of Krylov State Research Centre. General characteristics of simulated boundary layer (such as mean velocity profile and turbulence intensity profile) are presented. For wind loads measurements model of Silsoe cube in a scale of 1:15 was used. Comparison of the data obtained at both test rigs with data published in early studies is performed. New data about aerodynamic forces acting on the model in range of angles of oncoming flow are presented. Algorithm for determination of peak wind loads is discussed.


Introduction
In the age of computer algorithms correct physical tests still is one of the main sources of data almost in all aerodynamic tasks. Indeed quality of the numerical simulations can be proven only by comparison with experimental data. For modern applied build engineering tasks the most interest question is determination of full and local wind loads on the surface of the designed object, for example stadium, exhibition hall, dwelling house etc. For a long time wind tunnels with uniform velocity profile were used to perform necessary simulations. Even in building regulations documents (such as [1]) all formulas for determinations of wind loads are based on such type experimental investigations. However the growing level of difficulty and responsibility of design of the structures became the reason to develop new type of test rigsatmospheric boundary layer wind tunnels [2, 3]. The main aims of creation of such experimental stands are desire to correctly simulate of wind properties for different conditions, i.e. suburban conditions, open areas and seas, for different year seasons etc. The use of such stands allows obtaining more accurate data for objects with complex geometry, or buildings located in a densely built-up area.
Landscape Wind tunnel of Krylov State Research Centre is subsonic wind tunnel with closed-type large-size test section. Dimensions of this section provide possibilities to perform large-scale measurements of testing object with necessary number of surrounding development. The goal of present study is to compare results obtained in wind tunnel with uniform velocity profile with data obtained in LWT, i.e. summary and local wind loads. For this purpose model of the Silsoe cube [4 -6] was created in scale 1:15.

Methods
The scheme of the LWT is shown at the fig. 1. There are a lot of known approaches to create boundary layer in wind tunnel [7 -9], but the most verified one is use large-size spires for creation of main form of wind profile [10] and small-size blocks for creation of correct form of turbulent intensity profile [11]. Properties of modeling ABL was corresponded to "0" type profile in [12]. Photo of the model of the cube in LWT test section is shown at the fig. 2a. For determination of summary wind loads single dynamometer (with coordinate system which is shown at the fig. 2b) was used.  fig. 2c. Each tap was connected to the record system using tubes with length 200 mm and diameter 1 mm. Complete time of measurement for one angle α was 30 s and data sampling frequency was set as 100 Hz. According to the configuration of the used wind tunnels and sizes of the model blockage ratio in both experiments was less than 1 %. Reynolds numbers calculated using height of the model (H) and velocity at the height of the model (VH) were in range Re = [2…5]  10 5 for case I and 2.9  10 5 in case II. Strictly speaking Reynolds number almost in all natural-size objects is higher than 5  10 5 , but in experimental investigation it is impossible to make Reynolds number comparable with natural. That's why independence of some observed data from Reynolds number will be shown further.

Results and Discussion
Denote experimental investigation in wind tunnel with uniform velocity profile as "case I" and experimental investigation in LWT as "case II". All linear sizes are reduced to dimensionless form using cube height, all velocitiesusing velocity of the flow at the cube height. Mean velocity U and turbulence intensity It (determined as 100 The main results of the tests in wind tunnel that can be used to compute wind loads are pressure coefficientsmean and peak. Information about distribution of mean coefficients on the surface of the cube in ABL conditions is highly common in world literature [4 -6, 15 -18]. The pressure diagrams for both cases are presented at the fig. 5. The main difference between all data observed on the top surface of the cube. It should be noted that correct measure of roof pressure is still the big problem for engineers [19]. In present work reason of such difference can be difference in model Reynolds numbers or turbulence intensities. a b The most important and most interesting part of tests is determination of peak pressure, because algorithm of this procedure does not described in regulations documents. More than ten different techniques can be found in papers of different authors (for example [20,21]). Almost all of them recommend to use approaches of probability theory and try to find best fitted curve for experimental data. In present work we show how this algorithm can work applied to experimental data. Algorithm which will be described further is just an example appropriate for simple geometry. For real tasks we highly recommend to use more accurate algorithms, though the accuracy of algorithms are the separate topic for full research work.
Suggesting approach contains four steps: 1. separate input signal to small intervals; 2. find global maximums (or minimums for negative peak values) on each interval. Form a new "signal" using these maximums (or minimums); 3. calculate probability distribution function (or cumulative distribution function) for this new "signal"; 4. use this PDF (or CDF) to find limits of the data. In such case upper limit corresponds to positive peak value and lower limit corresponds to negative peak value (if lower limit is negative).
Step 3 can be seriously simplified using known PDF, for example, Gumbel function, which was specially developed for these purposes [22]. Demonstration of determination of the negative peak values at vertical symmetry line for angle 0 o is presented at fig. 6, where letter "d" denotes distance from bottom edge (as shown on a scheme near to the graph). It should be noted, that for windward side at the angle 0 o determination of negative peak values is impossible. Positive peak values (Cp peak+~1 .15) are in a good agreement with technical data (such as [1] where Cp peak+~1 .20).

Conclusions
In this paper, we present the main results that must be obtained in experimental study of aerodynamic loads on architectural objects. Two methods for conducting an experimental study are considered: an experiment performed in a uniform velocity profile and an experiment performed taking into account the unevenness of the velocity profile (the velocity profile of the Atmospheric Boundary Layer). Both studies were performed on the basis of the Krylov State Research Center, in particular using a new Landscape Wind Tunnel [23 -27].
As a result of the tests, data about the total and local aerodynamic loads on the model of Silsoe cube at a scale of 1:15 were obtained. The total aerodynamic loads are represented using aerodynamic coefficients Cx, Cy. It is established that qualitatively the dependences of the coefficients on the angle of the flow are similar in both sets, but quantitatively the data can differ by 20-30%.
Local aerodynamic loadsthe main result of the experimental testsare represented by distributions of mean and peak pressure coefficients. The average coefficients are in good agreement with the data presented in known works [4][5][6][16][17][18][19]. Research performed in the Landscape Wind Tunnel makes it possible to more accurately determine the average loads and reduce the influence of extra functions in determination of the loads (the algorithm described in [1]).
The paper presents a simple and intuitive way to determine peak loads, which is applicable for simple geometry or during methodological experiments. However, the data obtained using this approach are in good agreement with the data recommended by [1].