The OpenSees Community

dasgovind wrote:
> Hello. Thanks for replying. I have applied a pulse of 100000 cycles "timeseries pulse
> 1 0 100000 -factor 70". Is there anything regarding Rayleigh damping. ? I have not
> applied any Rayleigh damping.

Hi,
It seems you have not specified any period for the applied pulse. Kindly see the Command manual http://opensees.berkeley.edu/wiki/index ... TimeSeries. The command is:

timeSeries Pulse $tag $tStart $tEnd $period <-width $pulseWidth> <-shift $shift> <-factor $cFactor>

Here $period is the period of the pulse and is a mandatory argument of the Pulse timeseries command.

Hope this helps.

Hi,
Kindly check the duration of your applied pulse. Does your applied cyclic load continue beyond 1000 sec? If no, then it is natural to have more or less constant value, since BNWF has dampers attached.

Hi,

I have not myself used ManzariDafalias model earlier and hence it is not possible for me to tell you something definitive. In that case I think it is better to contact Prof Arduino.

Good luck.

Hi,

Please refer to change log (http://opensees.berkeley.edu/OpenSees/changeLog.php) for the difference. It says:
"ManzariDafalias updated wth new staggering method and line search algorithms by P. Arduino (UW) ".

Hope this clarifies the issue.

Dear Zackchen,

Thanks a lot for the source code uploaded.

It would be of great value, to have the INSTALLATION instructions in English for Non-Chinese users like me.

Thank you once more for your very useful contribution to the community.

Best Wishes.

Dear Zackchen,

I searched GitHub for the source code of your beautiful program OSLite, as you informed earlier about its posting during June, 2017. But I could not find it there.

I am eagerly waiting for Linux version of your very useful software or for the source code with compile instructions.

Hope to get it asap.

Congratulations and thank you for your contribution.

mahsashamsaei wrote:
> thank you for your answering
> i use nd material when i analyzed stage 0 and dynamic analyze, converges
> can i ignore update stage 1 or ignore gravity analyze?
>
> i have another question
> i model pipe line with nonlinear beam column with unixial material steel 02
> when i analyze model with pilpe line and with out pipe line output strain in elements
> of soil that is In the vicinity of the pipeline is equal
> I think the pipe did not affect the hardness of the soil in that area
>
> The soil and pipe connection was performed with equal dof

Hi,

1. As it converges in elastic stage, model seems to be OK
2. Ignoring plastic analysis (material stage 1) will mean linear analysis only. You should decide whether it is appropriate or not, depending on your objectives of analysis.
3. Usually gravity analysis is done using elastic material (material-stage 0. See examples in wiki page) and then, before starting the dynamic analysis, material stage is updated.
4. It seems to me now that i) you may try to increase the number of analysis steps by reducing the dT in analysis. ii) You may also try to change the solution algorithm, iii) systems, and/or iv) increase maximum number of trials in a time step before convergence is achieved. Refer to command manual for these.

In nonlinear dynamic analysis, you need to try different combinations of these parameters to achieve convergence, and it is quite normal to have non-convergence in the first few trial analyses. Unfortunately such combinations depend on the model at hand and no general prescriptions can be made. One needs to learn from previous experience only.

It is better to open a logFile by using logFile command and check if the analysis converges at every time step or not.
I do not know about your dynamic load pattern (is it a load, or UniformExcitation, multiSupport Excitation etc.) So it is difficult to comment. Usually, if you are using recorded ground motion file then the time steps should not be an issue and you can try varying other parameters mentioned above.
Hope it helps. BTW, sorry for my delayed response.

Reg your pipe model more details are necessary to comment. However, due to equalDOF constraint, it is natural to have deformations in soil near the pipe same as that of the pipe.

Thanks and Best Wishes.

Does it converge without updating the material stage?
If it does not, then please check your model, particularly the BCs.

If you use InitialState Wrapper material, take care that you use the material tag corresponding to the initialState Wrapper and not the tag for original nDMaterial, while updating the material stage.

Hi,
This might not be an exact response to the question.

However, it is possible to determine the Damping ratio by measuring the area within a particular hysteretic loop. For this, (a) one may use an element recorder to record the (shear) stress and (shear) strain and (b) then calculate the area within the loop by numerical integration. (c) Then the damping ratio can be calculated for that loop by using standard procedure (e.g. Kramer, Geotechnical Earthquake Engineering).

By recording the dT used in analysis, the duration of the loop can be checked and thus, any particular loop can be selected which is relevant in the analysis.

The whole process can be automated using MATLAB or SCILAB script or any other language of choice for post processing.

Hi all,

Finally I could locate the problem. I was using the unncessary input argument 'material' in the command.

An example of correct command would be:

recorder pvd $fileName <disp> <vel> <accel> <eleResponse stress >
recorder pvd $fileName1 <disp> <vel> <accel> <eleResponse strain>

To repeat, note that (i) the eleResponse must be the last argument,
(ii) directories named as $fileName and $fileName1 must preexist in the working directory.

However, I could not get desired results by combining both stress and strain in the same command.

Hope this is helpful for some one.

Hi,
As I understand your question, you are interested in frequency domain analysis of soil-structure interaction which is valid only in Linear soil-structure system. On the other hand OpenSees is designed/tailored for time domain analysis of non-linear soil-structure system, to be useful in performance based earthquake engineering.

You can use constant spring and damper parallel material to represent (very very approximately, provided these constants are chosen calling for great engineering judgement and experience) the far field soil in an OpenSees (or many other commercial softwares) model, which essentially remain elastic, provided your near-field soil domain is sufficiently large (with increased computational 'cost'). These 'constant' spring-damping parameters may be taken judiciously from appropriate models (such as BEM, SBFEM, Cone-models etc., Literature is quite voluminous). Out of these, cone models are the easiest conceptually and also at implementational level; MATLAB or Scilab may be a better option to use in such case to find the impedance functions.

For homogeneous half space, simple expressions given by Gazetas (1991, ASCE Geotech Jnl, 117(9), 1991) may be used, which are also adopted by FEMA later. However, in principle, frequency dependance cannot be handled directly in time-domain, unless the structure-soil system is perfectly elastic (so-called elasto-dynamic solution, using IFFT).

On the other hand, for direct analysis of the nonlinear soil-structure system in time domain, fulfilling the radiation condition exactly is a tremendously difficult problem, more so in case of seismic excitation because of many other issues involved. Thus, the present state of the art is to have a coupled FEM-BEM/ FEM-SBFEM analysis using sub-structuring.

To keep it straight, the problem you mentioned, I believe, is not conducive for modelling directly, using OpenSees only, irrespective of whether the structure is SDOF or MDOF.

Hope this clarifies.

Regards.

Dear fmk,

Thanks for your reply.

Warm regards.

Dear All,
I wonder if there is a way to mark a previous topic as "[SOLVED]" by editing the title.
I think, once solved, it is useful to mark so. This will be useful in avoiding posts partially repeated/related to earlier posts. Such provision exists in many other forums.

Regards.

Hi all,

May I have some advice from my learned forum members?

Any comment depicting the strength/weaknesses of the proposal and suggestions thereon will be highly appreciated.

Thanks and regards.

Dear fmk and all,

I would like to implement Scaled Boundary Finite Element method (SBFEM)[Wolf, 2003; Wolf and Song, 2000; Song and Wolf, 1996] in OpenSees.

As is well known, SBFEM combines the advantages of FEM and Boundary Element Method (BEM), while disadvantages of both are avoided. This method is particularly suited for DSSI analysis. In coupled FEM-SBFEM approach, The non-linearity in the near field may be handled by FEM while radiation condition in the far-field may be handled using SBFEM. This is the motivation of my proposal.

I sincerely believe, OpenSees is one of the best platforms for this purpose. This will enhance the capability of OpenSees for more realistic DSSI analysis, compared to the presently available approximate spring-dashpot models. Also the method may be extended for fracture-mechanics applications later.

My tentative plan is to
1. have some kind of new model builder (say "model SbfeBuilder" command or something of that sort) which would be a part of overall model.
2. It will generate its own SBFE model by accessing the normal model builder objects (viz. nodes/elements/materials) and generate the [K], [C] and [M] matrix for the SBFE part of the model
3. The near-field part of the overall model will be generated as usual in OpenSees
4. Both will be coupled using equalDOF constraint of the dofs at the interface of the two model regions (near-field and far-field).
5. The SBFE matrices will be finally assembled with the corresponding FE matrices and solved by usual OpenSees solvers.

I would like to know your comments on the feasibility of such implementation in OpenSees. I invite any other suggestions on the matter.

Looking forward to your expert comments.

Regards.