This week is International Blog Delurking Week. If you read the blog but never comment, please say Hello in the Comments section below and let us know how you came across the blog. Don’t be shy. Frequent and infrequent commenters are also welcome to say Hello and give their back stories.

Now, on to important human subjects research.

I made this Twitter poll to see how people in the intersection of the Twitter and OpenSees communities would like to start off the new year.

While the question is straightforward, the options are open to interpretation:

loadConst – don’t give me any more sh!t to do this year

reset – I’d like to try that again

sendSelf – send me somewhere else, perhaps far away

wipe – I’d like to start over from square one

The poll was not designed well–sendSelf is not an actual OpenSees command, but rather, the C++ function that implements the save command for database and parallel processing. Likewise, the recvSelf function implements the restore command.

In hindsight, I should have included save and restore in the poll. The results could have been very different.

I have been involved in the development, maintenance, and growth of OpenSees since its early days. I am interested in learning Python and improving my academic writing.
View all posts by Michael H. Scott

Another regular reader here 🙂 I stumbled across the blog on someone’s webpage or github page, where different links to OpenSees resources were given. I don’t remember whose page though!

I would go for loadConst, in view of a new dynamic analysis.

This brings up several issues.
To start with, if you’re solving a dynamic problem, with explicit methods (like Newmark’s) you dispense with the global stiffness matrix altogether.
With static problems, the global stiffness matrix is really not part of the physics of the problem but simply a way of “pointing” the problem to a solution. I think this is better understood in CFD with their Jacobian (which is really what a stiffness matrix is.) This is why with nonlinear problems there are several way of altering the stiffness matrix during the solution, including leaving it fixed (which actually works in some cases, did with my PhD dissertation.)

Structural engineers definitely love their implicit methods! By holding the stiffness matrix constant do you mean Modified Newton, aka, the chord method or Shamanskii’s method.

Sorry for the delayed response. I used two methodologies for my dissertation because I was solving both a static and a dynamic problem. For the static problem, it was a Modified Newton’s method. For the dynamic problem, it was an explicit Newmark’s Method where the global stiffness matrix is not used. The whole thing is here: http://vulcanhammer.net/2017/03/21/improved-methods-for-forward-and-inverse-solution-of-the-wave-equation-for-piles/
As far as altering the stiffness matrix, I was thinking about the methods mostly associated with Fletcher. When I took Optimization, my professor (who went back to NASA shortly after the course) went through these methods, where Fletcher’s name appeared several times. My response to this was “Fletcher sure plays both sides of the street,” a sentiment to which he agreed.
These days you’re seeing static problems modelled as dynamic problems in slow motion with a lot of time steps to avoid the global stiffness matrix problem altogether.

All I want is for ok to be zero!

LikeLiked by 1 person

As in

set ok [analyze $N $dt]

or

ok = ops.analyze(N,dt)

Which one?

LikeLike

Hi,

I have already delurked in last year’s delurking week. Regarding twitter’s survey, I would like to reset and analyze.

FR

LikeLiked by 1 person

Hi there.

I didn’t know what delurking means, 🙂 I’m glad to let you know I’m a regular reader.

Happy new year to you all.

Luis.

LikeLiked by 1 person

Hi,

Another regular reader here 🙂 I stumbled across the blog on someone’s webpage or github page, where different links to OpenSees resources were given. I don’t remember whose page though!

I would go for loadConst, in view of a new dynamic analysis.

Thanks for all the good content.

Luigi

LikeLiked by 1 person

This brings up several issues.

To start with, if you’re solving a dynamic problem, with explicit methods (like Newmark’s) you dispense with the global stiffness matrix altogether.

With static problems, the global stiffness matrix is really not part of the physics of the problem but simply a way of “pointing” the problem to a solution. I think this is better understood in CFD with their Jacobian (which is really what a stiffness matrix is.) This is why with nonlinear problems there are several way of altering the stiffness matrix during the solution, including leaving it fixed (which actually works in some cases, did with my PhD dissertation.)

LikeLike

This was for the previous post, for some reason my computer went crazy between the two.

LikeLiked by 1 person

Structural engineers definitely love their implicit methods! By holding the stiffness matrix constant do you mean Modified Newton, aka, the chord method or Shamanskii’s method.

You can also use accelerated Newton methods, e.g., secant or Krylov acceleration.

LikeLike

Sorry for the delayed response. I used two methodologies for my dissertation because I was solving both a static and a dynamic problem. For the static problem, it was a Modified Newton’s method. For the dynamic problem, it was an explicit Newmark’s Method where the global stiffness matrix is not used. The whole thing is here: http://vulcanhammer.net/2017/03/21/improved-methods-for-forward-and-inverse-solution-of-the-wave-equation-for-piles/

As far as altering the stiffness matrix, I was thinking about the methods mostly associated with Fletcher. When I took Optimization, my professor (who went back to NASA shortly after the course) went through these methods, where Fletcher’s name appeared several times. My response to this was “Fletcher sure plays both sides of the street,” a sentiment to which he agreed.

These days you’re seeing static problems modelled as dynamic problems in slow motion with a lot of time steps to avoid the global stiffness matrix problem altogether.

LikeLike