General Mechanical

General Mechanical

Difference between NR-Residuals and the magenta line in convergence plots?

    • Rameez_ul_Haq
      Subscriber

      Can anyone please tell me if there is any difference in the basic calculation of Newton-Raphson residuals and the magenta line shown in the convergence plots such as force convergence? I know that the NR-residuals, if opened, will show me the residuals at each and every location of the structure while magenta line only shows me a scalar overall value. But they both are residuals, right? I mean the way these both are calculated is the same. One is for illustrating the residuals at each and every location in the whole body while other is an overall value (to be shown on the convergence plots).

      Moreover, if displacements or moments are not converging while the force has converged, so will still the NR-Residual plots be able to correctly detect where the imbalance is occurring i.e. at which locations within the body the equilibrium is not being achieved? What I mean to say is that we can use the NR-residuals as a means of detecting the problematic locations within the geometry (if the solution is not converging) for any type of non-convergence, it be displacement or moments or force, right?

    • Rameez_ul_Haq
      Subscriber
      ,.
      I would be glad if anyone could provide an answer to this query. Thank you.
    • John Doyle
      Ansys Employee
      The Newton-Raphson residual (imbalance) is calculated at each node. This can be plotted on the geometry to determine where the highest source of imbalance. The "magenta line" plotted under Solution Information represents what the doc refers to as the "L2 norm" which is the square root of the sum of the squares (SRSS) of all the calculated nodal imbalances. This is what is compared to the criteria as the test for overall convergence. Please refer to Section 14.10.2 of the R2021-R2 Theory Manual for more details.
    • Rameez_ul_Haq
      Subscriber
      thank you for reply. And sorry by my side for taking a little long to reply back.
      Can you also please answer this question, "Moreover, if displacements or moments are not converging while the force has converged, so will still the NR-Residual plots be able to correctly detect where the imbalance is occurring i.e. at which locations within the body the equilibrium is not being achieved? What I mean to say is that we can use the NR-residuals as a means of detecting the problematic locations within the geometry (if the solution is not converging) for any type of non-convergence, it be displacement or moments or force, right?"
    • Rameez_ul_Haq
      Subscriber
      ,your thoughts on this will be very much appreciated :)
    • Rameez_ul_Haq
      Subscriber
      ,a reply from you this one will also be much applauded :)
    • Rameez_ul_Haq
      Subscriber
      ,can you kindly provide your views on the second paragraph of the question I have asked in this thread?
    • peteroznewman
      Subscriber
      I don't know the answer. I was hoping someone who does would reply.
    • Rameez_ul_Haq
      Subscriber
      ,ohh okay. lets hope then someone with this knowledge could help us. Although I have tagged all of the ANSYS staff I know here, maybe someone having the answer could reply.
    • Rameez_ul_Haq
      Subscriber
      ,maybe you could help us?
    • Rameez_ul_Haq
      Subscriber
      ,if you could also shared your views on this query, it would help us to a great extent :)
    • Erik Kostson
      Ansys Employee
      Hi - I would bet my 5 cents that it is NR residual forces plots we provide - perhaps see this which might help:
      Thank you

      Erik
    • Rameez_ul_Haq
      Subscriber
      thank you for replying.
      So I what I am getting from this is that the NR-Residuals are only useful for observing the regions within the model where force imbalance is occuring. I mean if the displacements are not converging, then most probably the force convergence also won't happen since internal forces are essentially based on the calculated nodal displacements. But, for moment unconvergence, we can't still be dependent on the force imbalance regions. We don't know how to locate the regions where moment is not converging inside our model. Or if moment convergence or criterion is somehow dependent on the forces itself, then we can trust the regions with high NR-Residual forces that these are where the moment is not converging.
      What would you say, sir?
    • Rameez_ul_Haq
      Subscriber
      I am struggling with a problem in static structural, something similar to this:
      I already tried increasing the number of substeps (and also the max number of iterations for a single substep) but still it doesn't solve the problem of moment convergence since after some certain number of substeps, the 'moment convergence' graph just keeps on oscillating, with the same mean value. However, as far as the force convergence plot is concerned, it smoothly and easily converges for this region where moment convergence is not found. My model is includes shell and solid, both elements. Now, since the force is already converging, when I access the NR-Residual plots, unfortunately, I cannot see any region within my model which is not finding convergence.
      Therefore, I need to know how can I determine that region where moment convergence locally is not met? NR-Residuals, as already discussed in this thread, shows where the force convergence is not met. But for moment convergence, I don't know what to do or where to look, in order to spot those regions where moment convergence criteria is not being met.


    • Rameez_ul_Haq
      Subscriber
      I would extremely thankful if anyone could provide some insight onto this matter, please :)
    • peteroznewman
      Subscriber
      One way to accidentally create a moment convergence problem is to use Duplicate on a Remote Displacement and edit the scoped geometry to something at the other end of the structure, but forget to also update the Location, which is why I never use Duplicate on Remote Displacements.
      Here is the originally created Remote Displacement:
      Here is the duplicated Remote Displacement that has the Vertex scoped at the other end of the cable, but note that the the coordinates are still at the 40 m end of the cable. This introduces a massive Moment into the model that should not be there.
      There is a Location line item that allows you to change it to the correct end, but if you fail to do that, you have created a defect in your model. If you just made the second Remote Displacement from scratch, it would have the correct Location.
    • Rameez_ul_Haq
      Subscriber
      ,well thankyou for providing an example here.
      But still I am eager to know how to spot those locations where moment convergence criteria is not being met, for complex structures comprising of shell bodies.
Viewing 16 reply threads
  • You must be logged in to reply to this topic.