The OpenSees Community

Linguan:

Yes, it makes difference for our element, which uses gradient nonlocal theories to provide objective responses for the softening part. You may find it here, if interested:

http://0-ascelibrary.org.libraries.colo ... 89.0001083

Sometimes OpenSees does not complete printing the .out files before you close the program. Instead of closing the program, you may also give a try to the command "wipe" at the end of your .tcl file. If your .out files are of the zero size, it means that either the element for which you are recording the responses does not exist in the domain, or the response argument that you are recording is invalid. According to your script, however, none of these apply.

By the way, I am wondering why your recorder commands have # symbol at their beginnings (as you know, it makes them comment).

Hello Dr. McKenna,

Thanks for your quick reply. It is a force-based beam element that I have developed and it uses the OpenSees beamIntegration classes. Actually, I could resolve this problem myself.

For those who may encounter similar issue, looking at some of the beamIntegration classes, I found out that some of them, such as Gauss-Lobatto and Newton-Cotes, provide the IP locations and weights for up to only 10 IPs. As a result, I modified those classes and now I can get the locations and the weights for up to 20 IPs for Newton-Cotes method and unlimited number of IPs for Gauss-Lobatto. I also had to replace the fixed array secX in the function setResponse() with a dynamically allocated array with the arbitrary size of numSections.

Hello everyone,

I have developed a force-based beam-column element which needs large number of integration points to yield objective response. When I use more than 20 integration points, however, OpenSees stops working, without displaying any error. May I possibly know what should be modified, please? Thanks in advance.

I found out what the problem was and solved it... the problem was existence of a number of gap elements in the model which were of very large stiffness; these elements would cause the numerical solution mess up. I reduced those elements' stiffness value and now there is no problem in terms of trial displacements.

Whenever you use imposedMotion, you need to use "Transformation" type of constraints, otherwise you won't get any result. I hope it helps.

Thanks for your reply, it gave me an insight into the potential issue, but my model is correct and its axial strength is way larger to be yielded at the start of lateral loads' application.

Now I am trying another model which is more complex and comprises my own implemented element, which is a force-based, finite-length, 3-integration point, joint element and includes a friction hysteretic model at the middle integration point. When I try applying a lateral load to the top of a short cantilever including my element which connects two elastic beam-column elements, after a short time that sliding happens in the friction model, again I see that the trial basic displacements which are related to the lateral deformation of my element (two end rotations in the basic reference system) get too large (in the order of 10 Rad or larger)... By the way, in the model there also exist two post-tensioning tendons connecting the top elastic segment to the base and now I am using HHT integrator to numerically add damping to the system... Is the reason behind this the too small friction stiffness as sliding initiates (1.0E-4)? How may I tackle this issue? Does it have to do with the integrator? It is worth noting that my element's stiffness matrix is reasonable.

Thanks so much in advance.

Now I am using the version 2.2.2 and it does not have those problems. You may download it from http://opensees.berkeley.edu/OpenSees/d ... wnload.php

I have the same problem with that page... one option is using the free software called TortoiseSVN and downloading the updated files to your desired folder, as instructed in https://www.youtube.com/watch?v=BaxFLF6A8Rg

I found out how to do this... for those who may concern, you need to modify the FullGenEigenSolver class or similar EigenSolver classes to get a Matrix object without calling analysis classes and needing Graph objects to set the size given to the EigenSolver; you also need to utilize a couple of functions of FullGenEigenSOE to add matrices to the class in the form of pointers, i.e. FullGenEigenSOE::addA() and addM()... By the way, note that while FullGenEigenSolver refers to a Fortran code which can compute all the N eigenvalues of a N-by-N matrix, Arpack solver can only compute the first N-1.

Hello everyone,

I am working on an element in whose formulation I need to determine the eigenvalues and the eigenvectors of a matrix... could anybody let me know how I may do eigen analysis in OpenSees? I tried using the following lines, but I could not set the size for theEigenSOE object:

static BandGenLinSolver *theSolver = new BandGenLinLapackSolver();
static LinearSOE *theArpackSOE = new BandGenLinSOE(*theSolver);
static EigenSOE *theEigenSOE = new ArpackSOE(*theArpackSOE, 0.0);

theEigenSOE->zeroA();
theEigenSOE->zeroM();

theEigenSOE->addA(K_ms, eleID); // "K_ms" is the matrix I need its eigenvalues
theEigenSOE->addM(I, eleID); // "I" is identity matrix

theEigenSOE->solve(9, true);
eigenValues(i) = theEigenSOE->getEigenvalue(i);
eigenVectors[i] = theEigenSOE->getEigenvector(i);

Also, I am wondering if any of the solvers other than Arpack can compute all the N eigenvalues of a N-by-N matrix, instead of the first N-1. I am looking forward to hearing from you and thanks in advance.

Hello Frank,

I have developed a force-based finite-length joint element and built a whole OpenSees project having it implemented in. In order to verify the element's response, I am trying to perform a displacement-control pushover analysis on a model including only my element as a cantilever beam-column with the length of 80 mm and square section of 400 mm depth. The problem is that after the axial static load is applied (say as gravity load), the axial component of the trial displacements got from the top node (theNodes[1]->getTrialDisp()) is very large (about 10000 times larger than the last committed step's counterpart), although pushover displacement increment is laterally imposed. I tried to see if I face the same problem while using forceBeamColumn with the same length and sections, and it led to the same issue. Could I possibly know why it happens and how I may resolve this problem? I tried to figure out how displacementControl integrator works but I could not understand it completely.

Thanks so much in advance...

Thanks Frank, I tried building the whole OpenSees project and it worked after resolving some errors. By the way, I could not download the zip file of the new version and I had to build the 2.2.2 version; is this version fine? I tried to get the newest files from SVN using TortoiseSVN, but I faced too many errors while debugging, saying that some .lib files are not found.

I have the same problem, too... Could you let me know if you have found a solution for this? Thanks!

Hello everyone,

I have written a C++ code for an element which needs the function Matrix::Solve to multiply the inverse of a matrix by a vector in its formulation. Now I have managed to make the dll file, but when I try to run an example model, I get this error which says this function is not implemented for dll. I took a look at Matrix.cpp and it seems that for the dll projects this function returns -1. Could anybody let me know if I have an option other than using the function Matrix::Solve to do my operation? Do I have to build the whole OpenSees instead of the dll for my element?

Thanks so much in advance.