f.perrone2

Hi Morten, I am myself a load engineer, and at least to me "faking" time series sounds as a rather important issue: especially if I wanna benchmark the correctness of my simulated loads with those coming from measurement campaigns. I do understand that, for WT certification needs, TI should reproduce the input and expected values, but I still have the feeling that simulations should crunch the gap with reality and not the other way around. Besides, as you properly pointed out, in reality wind does not exactly match the theoretical spectra, but it will always be filtered in the high frequency range. Btw, since I'm in the process of coding my own wind modeler, I will introduce kinda flag for scaling wind time series or not. Finally, would you mind to pinpoint some conspicous bibliographic reference accounting for the terrain effects on wind components standard deviation? What I deduce from your post is that WEng embodies this effect within the theoretical spectra wind field stemms from. I thank a lot for the poignant hints you gave me. Kindest regards, Francesco

Hi everybody, I have got a theoretical question regarding the turbulent wind field. Generally the wind field stemms from a Gaussian process with 0 mean value and unitary standard deviation. Now, because of being a random process, as well as integrating the power spectrum for only a limited range of frequency/wave numbers and because of smoothing effects due to the discretization of the space, the "rough" time series (those coming out from the fft) will not exactly match the expected standard deviations. Thus, a scaling process is usually required in order to get std_u = 1.0; std_v = (0.7 - 0.8)*std_u; std_w = 0.5*std_w. The question is: what are the implications drawn by such a scaling factor? Is it reasonable the all the u-components have exactly std = 1 (of course the same for v and w components respecting the formula above)? What do we lose in terms of correlation in time and space? I thank you all in advance for any support. Best regards, Francesco Perrone

Hi Morten, I warmly thank you for the support. I was interested in understanding a bit more how the compensation works, and you fulfilled my need. Kindest regards from Germany, Francesco

Hi Morten, I thank you for the reference to Jakob Mann paper: I already read it more than once, but I did not find any hint on how to overcome or compensate for high frequency losses. I was thinking of this method/procedure (which is physically meaningless): once you generate the time series, you estimate where the power spectrum certainly deviates from the ideal one; from that point you fit a polynomial curve through the simulated points and get the square root of the ratio spectrum_ideal/spectrum_sim. Applying this factor to the spetrum amplitudes, it should be possible to "stretch" the simulated spetrum in order to match the target one. But maybe, in the IEC Turbulence Simulator, the high frequency loss are recovered by means of another method. I was more interested on the compensating approach from a mathematical aspect. I look forward to hearing from you again. Kindet regards, Francesco

Hi everybody, I would like to ask you what the high frequency compensation consists of, when I go generating wind fields with the Mann model. I know that because of the discretization (especially in the L2 and L3 directions) there is a loss in the wind power spectrum in the area of high frequencies. To recover this abrupt reduction, one way is to normalize the variance of each time series to 1 by means of a safety factor defined as the square root of the ratio between the target variance and the variance in the center of the simulation box. Then, this factor is to be applied to each of the time series. Nevertheless, I was wondering is there is any other way to achieve a better track of the ideal power spectrum. Any literary reference or hint will be welcome. I thank you all in advance and look forward to hearing from you. Best regards, Francesco Perrone

Hallo again, the Mann model performs a final fft from the wave number space to the spatial domain.Let's say that: N1 = 1024; N2 = 32; N3 = 32; L1 = 6140; L2 = 140; L3 = 140. Regard to those values,we have the below wave number vectors: k1 = i*2*pi/L1; i = -N1/2,...,N1/2 (the 0-th order is elided) k2 = j*2*pi/L2; j = -N2/2,...,N2/2 k3 = k*2*pi/L3; k = -N3/2,...,N3/2 but there's clearly a size mismatch between [k2,k3](they have N2=N3 elements) and k1;so the question is: how can I evaluate k = sqrt(k1.^2+k2.^2+k3.^2)?? Do I have to fill k2,k3 with zeros for N2/2+1:N1 or what?? The same question arise with the coordinates vector definition. Could please shed a light? I will be really grateful. Thanks in advance and best regards, Francesco

Dear Morten, thanks for the early answer;I will have a look at the advised paper. I read both the Manns articles and I noticed that the code is somewhat complex to be implemented;however,it seems more efficient and less computing-heavy than the Veers method. Best regards, Francesco

Hallo everybody, I'm an italian mechanical engineering student and now I'm carrying out my Msc thesis. My task is to model a 3d turbulent wind-field model in Matlab. At the moment I've already implemented the Kaimal and von Karman models by means of the Sandia/Veers method. My supervisor told me about the Mann model,which better simulate the real behaviour of wind turbulence.Therefore, I would like to implement the Mann model in Matlab in order to compare the results. So I'm wondering if I could find somewhere a piece of code regarding the implementation,but in the IEC turbulence simulator I did not find anything. My thesis has no commercial interests,hence I definitively don't mean to handle tha source code in a unpolite way. I thank you in advance and I look forward to getting an answer whenever possible. Kind regards, Francesco