General Mechanical

General Mechanical

Topics related to Mechanical Enterprise, Motion, Additive Print and more.

Does solution converge after memory error?

    • krada
      Subscriber
    • peteroznewman
      Subscriber
      nI seems that the solution stopped due to insufficient RAM. You should be able to tell from the convergence history where it shows the progress through the substeps.nRestart your computer to free up any extra RAM. Set the solve options to Distributed and use Out-of-core memory option and try again.n
    • krada
      Subscriber
      Thanks Array. i think my solution starts and then nothing is happening i mean in force convergence plot there is no residual force or moment is showing. there is only plot of criteria. but however i got message of solution is done and in error file there is also no mentioning of unconvergence. i have attavhed picture below of that.ni do not get your point of  You should be able to tell from the convergence history where it shows the progress through the substeps. do you mean i need to increase substeps or else? i will try with distributed memory as you suggested.nn n
    • peteroznewman
      Subscriber
      nI was thinking about the plot I see in Workbench, which below the plot you show from APDL, it also has a Time vs Iteration plot that shows where the solution is. If you asked for 10 initial substeps, the first point is at 0.1 and you can see it go to 0.2 etc so if the last point is at 1.0 you can see that the solution has finished.n
    • krada
      Subscriber
      nYes. i am using APDL. i have tried distributed memory option but it takes too much time after two hours of waiting ANSYS is still calculating to data it needed to simulate. in my model i have approx. 0.9 mililion nodes. how can i make it faster in distributed memory and out-of core option? n
    • peteroznewman
      Subscriber
      nWith your current computer configuration, one option is to use the iterative (PCG) solver, which uses less memory than the direct (sparse) solver. However, some models cannot be solved in the iterative solver because some entities are not supported by the iterative solver.nAnother option with your current computer configuration is to reduce the number of nodes.nThe best approach is to install the maximum amount of RAM that your motherboard supports.n
    • krada
      Subscriber
      nReducing the nodes is not possible. i have 64GB RAM in my computer. isn't it enough to solve this? and how can i use maximum amount of RAM in ANSYS APDL, i will try with PCG solver if it is possible or not. n
    • peteroznewman
      Subscriber
      No, 64GB is not enough RAM for this model. ANSYS manages the memory usage automatically. You just need more RAM to run this model.nOne approach is to have a coarse mesh for the system model, and a breakout model of a small portion of the system to evaluate detailed stress. This is called Substructure in ANSYS. In this way, you will have two models that can run in the RAM you have, that together provide the information you need. nhttps://www.youtube.com/watch?v=ctheCGB7Urs&ab_channel=GraspEngineeringnYou could replace some subsystems in the system model with a Component Mode Synthesis (CMS) superelement of a subsystem where you don't need to know detailed stress.nhttps://www.ansys.com/-/media/ansys/corporate/resourcelibrary/article/aa-v1-i3-component-mode-synthesis.pdfn
    • krada
      Subscriber

      i am modelling bearing and i want to apply preload in model. my rolling elements are modelled with combination of MPC184 and traction only combin39 element. i have preload force value about 8KN. and i want to apply thermal loads to MPC184 element so it elongates and hence my spring element get force. Does anyone have idea how to do that in ANSYS APDL? i know one way of doing that which is i need to set initial temperature of MPC element and then TREF = 0 in solution module but this is not working Although i have set Langrange method for MPC element which is necessary for thermal condition in addition to thermal co-efficient of expansion. If someone has new idea or any error in my thought please let me know. below is material parameter.ThermExpanCoefficient_MPC184 = -12.000E-06t![1/K]InitialTemperature_MPC184 = 2940 !294ttt![K]*get, HighestMat, Mat,,NUM,MAXMatModel_MPC_Elements = HighestMat+1mp, dens, MatModel_MPC_Elements, 0.0ttmp, alpx, MatModel_MPC_Elements, ThermExpanCoefficient_MPC184mp, reft, MatModel_MPC_Elements, InitialTemperature_MPC184keyopt,ETypeNum_MPC184,2,1t/SOLU ALLSEL,ALLTREF, 0tSOLVE/forum/discussion/22504/thermal-loading-mpc184

      @peteroznewman, nActually this is my original problem. model is showed in post above you can consider model as cross section view of one of raceways of roller bearing. there is two more rows of rolling element. i think to use of cyclic symmetry of the model but the model piece-wise symmetric i mean Outer ring and and inner ring sections are symmetric and raceways above and below of this raceway is symmetric. can i use cyclic symmetry although? i want to find force generated in COMBIN39 element due to thermal loading and consequently expansion of MPC elements. what do you think about this?n
    • krada
      Subscriber
      hii in my model i am not getting memory error when i applied force in the model the problem only lies is when i apply thermal strain ( temparature gradient) load to MPC184 elements. is it possible that for same model this kind of error? below is picture from product launcher where is it mentioning memory required to solve model is 20GB and i have 64GB in my computer. i have tried to allocate more memory to solver by DSPOPTION command but solver is not taking it into consideration. value in second picture is not increasing from 4096MB. how can i do that in ANSYS APDL?nn
    • peteroznewman
      Subscriber
      nThe screen shots shown above indicate that this model with an applied force can solve in 20 GB so you do have enough RAM to solve this model. You don't need to do anything, ANSYS is managing the memory.
Viewing 10 reply threads
  • The topic ‘Does solution converge after memory error?’ is closed to new replies.