Skip to main content

5 replies

V
ValeryOchkov24-Ruby IVAuthor
24-Ruby IV
June 1, 2016

Fine?

R
RichardJ19-Tanzanite
19-Tanzanite
June 1, 2016

The new solver is the "Knitro" solver. They (Knitro) of course claim it's better, but I have seen it fail where good old Levenberg-Marquardt worked, so I would just say that it's different.

    V
    ValeryOchkov24-Ruby IVAuthor
    24-Ruby IV
    June 3, 2016

    Richard Jackson написал(а):

    The new solver is the "Knitro" solver

    Yes - see Artelys | Optimization solutions - Page d'accueil

    n=25! What is the max n for Knitro ? Work please with attach!

    V
    ValeryOchkov24-Ruby IVAuthor
    24-Ruby IV
    June 3, 2016

    Verify for n=25 masses

    -
    -MFra-21-Topaz II
    21-Topaz II
    June 4, 2016

    Why, then, let pass variables with subscript, for elements of a vector?

    origin.jpg

    V
    ValeryOchkov24-Ruby IVAuthor
    24-Ruby IV
    June 4, 2016

    Sorry, but all is OK!

    x0, y0, xn+1 and yn+1 are x.0, y.0, x.n+1 and y.n+1 not x[0, y[0, x[n+1 and y[n+1

    A
    AlvaroDíaz12-Amethyst
    12-Amethyst
    September 29, 2016

    Hi. Attached seems to work fine when masses are uniformly distributed, but I'm not convinced enough when the chains are varying.

    Best regards.

    Alvaro.

    FIG.gif

    A
    AlvaroDíaz12-Amethyst
    12-Amethyst
    September 30, 2016

    Hi Valery.

    Some other few modifications. It's a hard problem, have a lot of variations. Can't found my Elgotz to check what he says about this variational calculus.

    Best regards.

    Alvaro.

    FIG.gif

    V
    ValeryOchkov24-Ruby IVAuthor
    24-Ruby IV
    September 30, 2016

    Thanks, Alvaro!

    The problem is so.

    I have in the Solver

    but I want to have so

    But it is not possible in Mathcad 15 and Prime.

    A
    AlvaroDíaz12-Amethyst
    12-Amethyst
    September 30, 2016

    Yep, I know.

    In the attached, instead of the summatorie, which give scalar result, in the solve block there are vectorial restrictions, introducing a Delta function, which is the diff function in matlab. So, you have in some way what you ask: a lot of equations in the solve block, but without writing them one by one.

    Notice that the poor results, reflected in the Potential Energy values, are problem of the Mathcad solver, maybe in prime, with the knitro solver can get better results.

    Also shows that in a Given-Find block, you can use range variables, but with some limitations. At the end an example of that.

    Best regards.

    Alvaro.

    Terms & ConditionsAccessibility statement