How to determine the value of n1, n2 and n3 in the 'common_vars.f90'

[yayihhu]

Hello,Thank you for sharing this source to the public .I'm using this model and I'm having some problems: when I simulated long time evolution from the case3-file , the wave elevations has been a marked decline and oblique rise, when I increase the n1 the phenomenon improved obviously but there are still a slight decline (Fig. 1 -2), I guess as the effect of n1 or n3 led to a decline or rise, but I did not find in your papers about how to determine the value of n1. I am looking forward to your reply

[gducrozet]

@yayihhu Sorry for my late reply.

At first, some questions: what is represented in your Figs. 1 & 2? Is it one direct output of the HOS-NWT model (for instance probes.dat)?

Then, can you provide the wave maker motion you input? It is possible to extract the temparl evolution of this wave maker motion as an output file (i_wmk in input file). It is possible that what you observe comes from a 'strange' wave maker motion.

Best regards,

Guillaume

[yayihhu]

Thank you for your sugestions. The figures are the ouput of the probes.dat, I think it's my problem ,little x-modes(n1=513) and high frequenciese(2Hz) in input file,which result in these phenomenons. Now I modified n1=1025, f=0.3~0.9Hz. But after days of trial and error, I have other questions about icase=3:

1. How to determine the n1(what is mean 'number of modes/points'? modes per points or modes=points? ), n3 and mHOS?
2. The parameters(like the ramp and typ_ramp) in wmk-cfg are all useful? and when the cfg don't agree with the input_HOS-NWT(Warning : Different basin width), which is decisive？ 3.Connected to Q2, the beam(m) of the wavetank is half of the width of the wavetank?

[yayihhu]

Sorry，I've had some problems lately. When I run HOS-NWT with the wave-making file is shown in the attachment, the informations in the terminal window is inconsistent with the cfg file (in this case, the n1=8193, n2=1, n3=33, mHOS=3). How can I avoid this situation？

Hs4_Tp10.zip

[gducrozet]

At first, I advise you to look at the HOS-NWT publication for details about the parameters (even if the naming of variables may be slightly different). Some elements of answer:

1. modes=points and n1 provides you the discretization of the domain in the x-direction. Similarly, n3 is the discretization used for the wave maker displacement. mHOS corresponds to the order of non-linearity for the solution of the wave propagation using HOS model. For the choice of parameters, usually one has to do a convergence study, since it depends on the level of accuracy you want to achieve. n1 has to be chosen with respect to the length of the domain in order to have around 30 points per wavelengths. n3 has to be chosen with respect to the shape of your wavemaker but usually n3=33 or 65 is sufficient. mHOS is chosen with resepct to the physical processes to be solved. It has to be chosen at least equal to 3 for non-linear simulations

2. The easiest way is to set-up the same values! In some parts the input file may be used and in other parts it may be the cfg file. Then, I strongly advise to adjust to the same correct values.

3. The beam of the tank is the total width

With respect to your last question, what is provided in the terminal are the dimensionless values of each parameter. In HOS-NWT the length scale is the water depth and the temporal scale is sqrt(L/g)

Best regards,

Guillaume.

[yayihhu]

Thanks for your detailed answers. I basically understand the meaning of n1, n2, n3, mHOS and the dimensionless. I'm sorry I have two more questions to ask you：

1. I have read the two references you recommended on the website，but I still can't figure out the relationship between wave components and modes, I always thought components(the first value x in .cfg correspond to 2^x) are modes. May I ask the differences between components and modes?

2. The orther question: n1≈30*Lx/Lmin (Lx: the lenght of the flume, Lmin: the minimun wave length in components) while I simulate the nonlinear evolutions of large scale (Lx~2000m), the n1 will be too high (~ 10^4) to run long time. In my case( Lx=2000, T_stop=1000, n1=4097), the elevation of probes see Fig.1(all probes), Fig.2(first probe) and Fig.3(fourth probe). Is the rise and fall of the wave elevation due to the small value of n1（4097<<10^4） or something else?

Best regards,

Liuyayi.

The attachments

case_example.zip

[gducrozet]

Some elements of answer:

1. There are two completely different notions: n1 corresponds to the horizontal discretization of the spatial domain. Since we are using pseudo-spectral method this is associated to the modal representation (number of modes) of the spatial domain and is consequently linked to the wavenumbers solved. On the other hand, the cfg file describes the temporal wave maker motion with the use of components understood as the different frequencies used to represent the wave maker motion. There is consequently not a direct link between those parameters. The link is actually obtained through the appropriate choice of horizontal discretization (i.e. the wave lengths you want to generate (which are link to the wave frequencies)).

2. You need a sufficient number of points to get accurate results... Either you reduce the spatial and temporal extent of your study or you need to wait for the results to come out! Reducing too much the number n1 will results in wroing calculations so this is not a good option... You may be able to optimize the choice of parameters (n3, order of non-linearity of wave maker) to increase the efficiency but since it is case-dependent, you need to do small studies prior to do your complete calculations. By the way, what you observe on the wave probes is probably associated to the wave-maker motion you give as input (and not the choice of n1). This is low frequency components that you can try to filter before creating your .dat file (or you can use in the input file the nuc_low parameter once i_cut parameter is set to 1).

Best regards,

Guillaume

[gducrozet]

Edit, I have just seen that you sent the input file used and you only have three wave components, so this is actually another issue. What you observe at the start of the wavemaker, you probably generated a long wave, propagating at speed sqrt(g*h) and this is this wave that is recorded on your probes signals. So this is a physical effect that some authors have studied in the literature. I do not know if your set-up (wave-maker shape, etc.) is fixed. Otherwise, what may reduce this effect is:

• change of wave-maker shape (a flap wave maker may generate lower amplitude long wave)
• change in the length of the time-ramp at the beginning of the signal (longer)

Best regards,

Guillaume.

[yayihhu]

Through your answers，I basicly undertand the mean of components and modes/nodes. As the instability of the wave elevation, by means of longer time-ramp, I reduced the effect of a spurious long wave and the elevations are smoother. I think I need to study your articles carefully about the HOS-NWT. Thanks for your detailed answers.

Best regards,

Liuyayi.

[yayihhu]

Dear Guillaume: Thanks for your detailed answers before. I think I have basically mastered the customed wave making method from the .cfg file. But I encountered two new problems when I studied the kinematics of wave motion.

• Firstly, how is the wave speed (when iprobe=2) calculated in the HOS-NWT? I did not find the calculation formula in the Velocity Module-code.

• Second problem is about the Post-Processing, I don't understand the differences between i_card=1 or 2. Mostly, the wave simulation should use i_card=2 (with free-surface boundary), so what is i_card=1 (constant z=no wave motion) used for? And I am questioned how to draw velocity field like the Fig.15-17 in your article "On the equivalence of unidrectional rogue waves detected in periodic simulations and reproduced in numerical wave tanks".

Best regards,

Liuyayi.

[gducrozet]

Dear Liuyayi,

Sorry for the delay, I thought I answered to those questions. Regarding your specific questions :

• When iprobe=2, the velocity is evaluated through velpress routine, which is located in the source file IO/velocities.f90
• the difference between i_card=1 and 2 is regarding the vertical distribution of points. i_card=1 provides velocity/pressure card with constant z for all horizontal location. It is consequently a rectangular grid which is very efficient to compute. On the other hand, i_card=2 provides a free-surface surface following velocity/pressure card. This is what corresponds to the figures you mention.

Best regards,

Guillaume

[yayihhu]

It's okay. Thank you for your specific answers. I have read the velpress routine, but there're still some issues I don't understand , which results from I'm poor at numerical analysis.

• Which theory/references the velocity calculation is based on? I don't know how these variables about modes are calculated , and I don't see the water depth parameters (It could be dimensionless, but I'm not sure).

• As i_card=2 which provides a free-surface surface following velocity/pressure card, how do we calculate the velocity/pressure above the still water? Extension method or others?

• Finally, do you have any suggestions(Tecplot or the Grid2Grid Program) for post-processing field drawing, because I use matlab to read datas very slowly

Best regards,

Liuyayi

[gducrozet]

1. Inside the program, everything is indeed done dimensionless (with respect to water depth and gravity). The velocity calculation has been detailed here: https://tel.archives-ouvertes.fr/tel-00263596 (in french) and the preliminary results are available here http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.660.3794&rep=rep1&type=pdf (see references herein)
2. The method used for velocity calculation is not restricted to the mean free surface position, but can evaluate the kinematics up to the exact free surface location.
3. Tecplot was initially used for graphical outputs (and the output files are currently designed for this program). Note that Grid2Grid is not a graphical program, it is just providing output files, that are to the VTK format and can possibly be used with Paraview.

Best regards,

Guillaume

[yayihhu]

 Thank you for your answers. I will read the references carefully (due to the network, the literature download has encountered some problems). 

• However, I think there is one problem with velocity calculation: it is still possible to calculate the velocity clearly beyond the the wave surface, so it is difficult to obtain the speed process of a single point above the still water surface through the probe file.

• In addition, could you please share Tecplot layout file of velocity/pressure field to me? Because I am still studying the reconstruction experiment wave of HOS-NWT, I have no time to learn the image processing method of Tecplot.

• When it comes to wave reconstruction, I used linear theory to reversely deduce the spectrum at the wavemaker as the constituent wave according to the elevation at the target position, but the results are not good. What do you suggest?

Best regards,

Liuyayi

[gducrozet]

@yayihhu

• I do not understand exactly the problem you mention. If you have an example, you can share it so that it is clearer to me
• The layout should be quite straightforward. I have done quickly an example, which is attached VP_card.zip
• For wave reconstruction, when waves are quite non-linear this linear procedure do not work well indeed. What I suggest is to use Time-Reversal procedure (I have an accepted paper about this that will come out in the coming days/weeks).

Best regards,

Guillaume

[yayihhu]

@gducrozet

• The problem I mentioned can be seen in attached example: the probe18 which clearly exceeds the maximum of the regular wave elevation, but the procedures can still calculate the velocity and save to the probes.dat file. bin_H=0.04.zip

• As wave reconstruction, when I reproduced the irregular wave elevations simulated by the HOS-NWT itself, I found it difficult to match the components above 32. But I want to simulate some quite nonlinear examples, I'll study your new paper carefully.

Thanks for your attached layout file. It's worked.

Best regards,

Liuyayi

[gducrozet]

@yayihhu

• OK, I see your concern with respect to the evaluation of velocities using the dedicated output (iprobes=2). Actually, since the pressure and velocity calculations are based on the use of a set of analytic basis functions (with specific modal amplitudes), they can be evaluated at whatever vertical and horizontal coordinate (even outside the domain of definition). For the post-processing, I have done a switch to set the values to zero when it is not in water, switch which is probably not applied here. You can apply it easily once you have the data by simply putting the values to zero when z>eta.

Best regards,

Guillaume

[yayihhu]

• I have studied the post-process results, and I think the nodes in the z direction should be evenly distributed from z_min to the free surface which vary with x. In this case, the wave face value at the top nodes（Line 2398~Line 2435 in VP_card_fitted_50-51s included in attached files, I deleted big files for convenience）in the z direction should be the same as eta, but the output file is not equal and he value of the top nodes in the z direction is equal to the wave surface recorded by the corresponding position probes (x≈20m). What is the reason? case3.zip

• In addition, I need the measured datas to verify the reconstruction method. I wonder if you can send me(yayiliu@hhu.edu.cn) the measured datas of the New Year wave and Yura wave in your new paper.

Best regards,

Liuyayi

[yayihhu]

@gducrozet

• I used the Post-Processing program to analyze the velocity and pressure profile of the dispersion focused wave, but it was not continuous near the free surface (there was a mutation) . I encrypted the nodes in the z direction, there was no improvement. I don't know if you have encountered similar problems before，if you have any suggestions i would be very grateful.

• I packed and compressed the original simulation files (NxNz=102564) and the figures of profiles. QS-profiles.zip