Version 7 beta

VictorVictor
edited April 2017
Mecway 7 beta is available at the link below. I welcome any feedback on bugs and the changes but probably won't add any new features. All licences that include version 5 or 6 are also valid for version 7. That is everyone who bought it after 2015-11-15. Just re-enter your product key to activate it.


General
Moment loads can be applied to faces (solids and shell 2D faces) even for elements without rotational DOF.
Direction of displacement constraints can be a function of position to allow cylindrical and other coordinate systems.
Opens .unv mesh files
New units - Gal (cm/s^2) and mm^2/N (1/MPa)
Stress linearization - stress classification line doesn’t have to pass through nodes and settings are saved with the model
Pressure XYZ (traction) can be a function of position.
Load cases replaced by configurations:
Load cases Configurations
Only 7 types of load All loads and constraints
Only static analysis type All analysis types
Only internal solver Both the internal and CCX solvers
A load can only be in one load case A load can be in multiple configurations
Solved simultaneously Solved one at a time
Table can include multiple load cases Table only shows the active configuration’s data
Factor of safety with Von Mises failure criterion for shell elements
Multiple gravity loads can coexist.
Multiple sets of node temperatures are allowed and can be suppressed independently.
Labs feature: Direct force per unit length and shear force per unit length outputs for shells
Write solution to .liml file includes the modeler model
Can read *BEAM GENERAL SECTION from CalculiX .inp files
Fixed bugs 9 and 10 about hydrostatic and position-dependent pressure on thick shell faces

CalculiX solver
Includes 64 bit CCX 2.11
Pressure XYZ (traction) for linear analysis
Pressure can be a function of position for linear analysis
Multiple time dependent force loads on the same node
Convection (*FILM)
Node temperature for thermal stress
Pyramid element faces for pressure, convection, and surfaces
Time dependent formulas in loads are evaluated at automatically chosen time points.

Meshing tools
Gmsh mesher (not included) for STEP files
Change shape option quad8 → 2 x tri6
Refine x2 refines line elements independently of their adjacent elements if they’re not fully connected to them.
Merge nearby nodes with a zero tolerance merges exactly coincident nodes.

UI
Local refinement spheres are shown graphically
Contour plot of values can have different colors for above and below the range
Edge detecting selection is more reasonable with curved quadratic element faces


Comments

  • VMHVMH
    Victor, can we have unit options instead of "SIunits" for the "FORC_1", "FORC_2", and "FORC_3" output when using ccx solver?  Thanks
  • SergioSergio
    edited March 2017
    Have tested 

    • Gmsh, only adjusting the element size and using local refinements (need to investigate what custom modifiers should I apply still). I felt that is slower than Netgen, but sometimes can produce better meshes, so is a good new feature to have this option. First picture

    • UNV import... awesome! Have tested with Salome and NX meshes, it recognize the node/element/surfaces sets in both cases better than UNICAL, so maybe is time to give a retirement to him :-) Second picture, is an NX .unv mesh

    • Local refinament spheres... nice bubles!, now is more clear. Can be defined selecting vertices on the geometry directly? Third picture

    • I'm triyng now the Configurations...very usefull, now we can have different scenarios, but why not keep calling it load cases as before? I have an issue, have made two configurations but can't make to have dfferent results as should, can you check the liml attached file?

  • SergioSergio
    This is my first try with configurations, I can't find a way to have two different results, and if I delete one result then the first gets corrupted (or at least the display).

    Wolud be great to be able to see the different configurations side by side on next versions.

    By the way, configurations works for CCX??? Maybe is my mistake :-)


  • SergioSergio
    edited March 2017
    Update, I open and recalculate both configurations, and then it display different results... great I thought. I delete again the results and rerun and again is showing the same result on both config.

    Have deleted all again, and compute the first config with the column unrestricted, but is showing the results as if were constrained (the second config).... could it be some problem reading the temporal input/results files? I have seted in Options to use the same folder as the liml for the CCX results.
  • VictorVictor
    Thanks for finding these bugs with configurations. It should show two different results. I've been able to reproduce both the corrupted display and the common solution for both configurations and will fix them for the final release of v7. For the second bug, it seems to be incorrectly solving the same .inp file for both configurations.

    I decided not to call them load cases because they work so differently and I hope to expand configurations in future to include other things like analysis settings, materials and maybe parts of meshes.

    No picking local refinement points on the geometry yet but you can still do it by right clicking mesh nodes after a mesh has been made.

    VMH: Thanks for the suggestion about FORC_1, etc units. Hopefully for the next version but I can't promise.

  • VictorVictor
    This is beta 2 which fixes those two bugs:


  • SergioSergio
    Great, tested and now works ok.
  • VMHVMH
    edited March 2017
    Moment load on solid and shell is such a great feature! Thanks!
  • VMHVMH
    edited March 2017
    Also, I did a quick comparsion with the built-in CCX v2.11 and CCX v2.11 complied by Kwip. The results were the same for linear static analysis.  CCX v2.11 complied by Kwip solved almost two time faster for larger models (if anyone are interested).
  • VMHVMH
    edited March 2017
    Victor, for the configuration, can we run one configuration and work on the another active configuration and then run that one while the first configuration still running?  Seems like it stopped my 1st run when I run the 2nd configuration using CCX solver.
  • VictorVictor
    Thanks for the feedback. It currently doesn't let you solve two configurations at the same time. I'll look into it stopping the first solve because that doesn't sound like very good behavior.

    I've also compiled the multithreaded version (ccx_MT.exe) and included it with Mecway, so you can change to that (Tools -> Options -> CalculiX) and it should be faster if you also set the environment variable for the number of threads. Not sure if Kwip's one uses a faster matrix solver or optimization settings though.

  • VictorVictor
    Beta 3 fixes the problem of stopping the solver to start a new solve. Instead it just won't start a new solve until you cancel the running one.
  • AndreaAndrea
    edited April 2017
    I found that CCX_MT.exe could give inconsistent results for buckling analysis.
    Solving the same model several times the results are different. For simple model is possible to understand that the solution is wrong but for complex model is better to avoid this solver.
    See attached pictures.


  • VictorVictor
    Is that the ccx_MT.exe with Mecway or from Kwip? I can't reproduce the problem. Do you mind sending me a .liml or .inp file that does it? If it's going to be unpredictably wrong like this, I'd better remove it from Mecway.
  • AndreaAndrea
    edited April 2017
    Inside Mecway....using ccx.exe is all right. Using ccx_MT.exe I had issues. If your test will be positive depends from the solver instead could depend from the OS (I use W7)
    I attach liml file.

    I will try the ccx.exe from kwip.....same problem

    Run several times the solver
  • VictorVictor
    I see it happening now by solving several times, on both your model and one I made. Thanks for discovering this and isolating it to the multithreaded version! I'll try compiling with lower optimization or remove it if it's still going wrong.

  • jvrajadelljvrajadell
    edited April 2017
    My test

    steel ancorage
    concrete foundation pad

    NL contact analysis

    image

    image

    principal concrete stresses (Mohr Coulomb)

    (1) red points: MECWAY 7 Beta 3, ccx_MT, steel NL, only compression concrete
    (2) green points:MECWAY 7 Beta 3, ccx_MT, steel NL, NL concrete
    (3) blue crosses: MECWAY 7 Beta 3, calculix 2.10 bConverged, steel NL, NL concrete

    I've obtained the same results in (2) and (3),  (1) ok

     
  • VMHVMH
    Andrea/Victor, I tried and got the same results for both Victor and Kwip complied ccx v2.11 MT using Windows 10 Pro 64-bit Operating System.  See quick video below.


  • AndreaAndrea
    Look at this.
    Try to check also high eigenvalues


    https://youtu.be/EjBBlOg1QKE
  • VMHVMH
    Confirmed different values when using higher numbers of modes (attached).
  • kwipkwip
    CCX 2.12 is out, I will compile a non-optimized version and an optimized one, let's see if there is a difference.
  • SergioSergio
    Victor, have tried the new colors for the contour plots,  and there is one color for "no results". I have selected one color for them, but when I display no lineal results, at the time 0 still the model is showing with transparence. Should it show with the "no result" selected color?


  • VictorVictor
    Yea that would make sense, but really it should not even make time=0 available at all when there's no data there. That's more of another problem for a future version.

Sign In or Register to comment.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Sign In Register