Hi everyone,

I need your valuable support to solve a non linear problem.
I tried the functions solve, roots, Findroot, al_nleqsolve, bisection but I obtained only errors.
I red a lot of topic and tried different approach but without result.

I extracted a part of the entire file. The procedure goal is to evaluate the moment Vs curvature diagram in a reinforced concrete section. The section is discretized in fibers (10x10 in the examples).
In the attached file you will find one iteration (a fixed value of curvature) where the non linear solver have to find the position of neutral axis (x) that lead to the equilibrium of the section (between compression and tension forces).

I also tried (in another file) to avoid units but the result was always the same.

Thank in advance to anyone who can dedicate some time to my problem

Lorenzo




RC_forum_nonlinear.sm (82 KiB) downloaded 37 time(s).

Here's a transparent, home grown bisection routine that does the job:



RC_forum_nonlinear_b.sm (79 KiB) downloaded 30 time(s).

Many thanks for the quick solution.
I'll include the code in the inner loop of the procedure.

Lorenzo

Just noticed that, strictly, my bisection routine should be as follows:

Thanks. Yes I've already correct the code.
My initial question was about the possibility to use built-in function, as it normally takes less time.
If there isn't this possibility, no problem. I can write alternative solution method.
I've implemented the quasi-Newton method (Broyden method - starting from Davide Carpi) to evaluate the best strategy regarding time and number of iterations.

Advice is always welcome.

Regards,

Lorenzo

Wrote

Thanks. Yes I've already correct the code.
My initial question was about the possibility to use built-in function, as it normally takes less time.

Regards,

Lorenzo

It isn't the bisection routine that's slowing the calculation; it's your melprod routine. Try the following instead:

Thank you! With your improvement the speed is significantly increased
Wonderful