Jump to content


  • Content Count

  • Joined

  • Last visited

  • Days Won


Andy last won the day on June 17

Andy had the most liked content!

About Andy

  • Rank
    Advanced Member

Profile Information

  • Gender
  • Country
    United States
  • Are you University user?

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

  1. Hi Ayman, It is hard to tell the issue without more details. Is this Linux or Windows? queuing software? submission command used? The easiest method way to submit a Radioss job is to use the radioss script provided in the installation. This script will run the starter and then the engine and should look like one job to your queuing system. Linux: <install_dir>/altair/scripts/radioss "filename" -option argument To run from a Windows DOS prompt, type the following: <install_dir>\hwsolvers\scripts\radioss.bat "filename" -option argument An example on Linux which uses some additional option would be: /altair/hw/2018/altair/scripts/radioss -v 2019 phone_0000.rad -mpi i -np 24 -nt 1 -hostfile /var/spool/PBS/aux/92305.admin -mpiargs -genv KMP_AFFINITY=scatter -genv I_MPI_PIN_DOMAIN=auto -genv I_MPI_ADJUST_BCAST=1 -genv I_MPI_ADJUST_REDUCE=2 -genv I_MPI_MPIRUN_CLEANUP=1 -genv KMP_STACKSIZE=400m -genv I_MPI_FABRICS=shm:dapl -noh3d Thanks, Andy
  2. Hi Pritish, If your global hourglass energy is high, then switch your shell elements to the /PROP/SHELL, Ishell=24 and solid elements to /PROP/SOLID, Isolid=24. Your analysis will take a little longer to run but you should get lower global hourglass energy. Thanks, Andy
  3. Hi, Adding some more information, if you could share the global energy plots for your model and some information about the type of simulation and loads that exist in the model. Please plot kinetic energy, Internal, Energy, Translational total energy (TTE) , Hourglass, energy, Contact energy, and External Work. If the contact energy is large relative to TTE, then you can get a large energy error. The energy due to friction is stored in the contact energy and this is sometimes the reason why you get the negative energy error. If the energy error is caused by Friction, then it is OK to ignore the energy error. Thanks, Andy
  4. Hi, With LAW2, Iflag=1 the UTS and strain at UTS are engineering strain values. But the output by Radioss is true stress and strain. If you look in the Radioss starter you can see the Iflag=0, a, b, n values that are calculated using your input. JOHNSON COOK MODEL : YIELD STRESS . . . . . . . . . . . . .= 355.0000000000 ULTIMATE STRESS (UTS) . . . . . . . . .= 470.0000000000 STRAIN AT UTS (Ag). . . . . . . . . . .= 9.0000000000000E-02 YIELD COEFFICIENT A . . . . . . . . . .= 355.0000000000 YIELD COEFFICIENT B . . . . . . . . . .= 312.9879880695 YIELD COEFFICIENT N . . . . . . . . . .= 0.2806664576242 EPS-MAX . . . . . . . . . . . . . . . .= 0.2100000000000 SIG-MAX . . . . . . . . . . . . . . . .= 1.0000000000000E+30 ISO-KINEMATIC HARDENING FACTOR. . . . .= 0.000000000000 You can input the a, b, n value into the equation in the comments of LAW2. Note this equation is true stress as a function of plastic strain and I have left off the temperature and strain rate terms. Or the pre-processor HyperCrash lets you plot a material when you use a, b, n in LAW2. Note that this is a plot of True stress versus true strain (not plastic strain). Please check if your output makes better sense using these stress strain plots. It is good that you are thinking about the results with respect to your material input to see if they make sense. Not everyone does this but they should! Thanks, Andy
  5. Hi, Please review the User Guide, Explicit Structural Finite Element Analysis, Finite Elements, Beam Elements (/PROP/BEAM, /PROP/INT_BEAM) Also the comments in Reference guide for /PROP/INT_BEAM. The /PROP/INT_BEAM works fine for bending. The use of several integration points in the section allows to get an elasto-plastic model in which von Mises criteria is written on each integration point and the section can be partially plastified unlike the classical beam element (/PROP/BEAM). For lattice structures, you need to decide if you need the extra accuracy of the /PROP/INT_BEAM beam or if /PROP/BEAM will be accurate enough. Thanks, Andy
  6. Hi, The energy error in Radioss is calculated as follows. Note there is no contact energy or hourglass energy terms included in this formula. If contact energy or hourglass energy become large usually that energy ends up increasing the internal or kinetic energy in the model and this causes a larger % energy error. The contact energy or hourglass energy are numerical (not part of the physical problem) and thus for a good simulation they should remain small. Typically if you have a large energy error you should try to understand why by plotting the energies in the model. In simulations with no or small initial energy, a small numerical difference in energy causes a large percent error which can be ignored. After additional loads are applied the energy error will usually decrease. In models where contact friction causes a large amount of contact energy, the energy error can be higher.
  7. Hi Vishruth, The initial kinetic energy has to get absorbed by the material deforming plastically. In your LAW2 material are your yield stress and plastic hardening parameters correct? You can define inertia of the pole in the /RBODY. By default, the /RBODY uses the inertia of the slave nodes of the mesh that are part of the /RBODY so it is important to specify a material and thickness to the mesh. For /RBODY any mass and inertia values are added to the values from the mesh slave nodes. Or if using /RWALL you can specify the mass directly. Thanks, Andy
  8. Hi, Please check that the physics of your model is correct as defined. If you take the mass of the particles in, /PROP/SPH, mp*#particles does that equal the expected weight of the material in the car? Also check material properties and units used. I assume the units are kg, m, s. If yes then, the wall of your car is 0.00635 m which is 6.35 mm or 0.25 inches which I am sure is to thin and most likely the issue. What is the loading condition your are simulating using 5 m/s initial velocity. What about gravity? Please set /INTER/TYPE7, GapMin= the 1/2 the pitch you used to generate the SPH particles + 1/2 the thickness of the walls. The pitch is the distance between the particles so 1/2 the pitch is the radius. Then set Inacti=5 to remove any initial penetrations due the meshing. We normally recommend face centered cubic for SPH meshes if you redo the SPH mesh you can use that option in HyperMesh by the pitch: setting. Last for improved speedup, please check the number of core your have in your CPU and use the run option, -np #cores to improve run times. Thanks, Andy
  9. Hi, If your part should be absorbing more energy then I would suggest you check material input parameters or thickness for mistakes. Is the model mass correct? Is the initial energy in the system correct? Also, run the model checks in HyperMesh or HyperCrash to look for errors. You can plot other energies, such as global contact, hourglass, TTE, total transnational energy, external work. If you add all your parts to the time history output (/TH/PART) you can check each parts energy to look for strange results. How are you modeling your pole? Using /RWALL or with a mesh and /RBODY? Thanks, Andy
  10. Hi, Lot of good advice given by Ivan. As he mentioned you need to understand why your timestep is so low. It would be useful to understand what type of analysis you are doing. Is it a drop test, impact crash, quasi-static imposed force or displacement loading? For quasi-static loadings, you can shorten the run time by applying the load faster but you have to make sure the load isn't so fast it causes dynamic effects by checking the kinetic energy in the model. Typically it is important to see what your energy error is at the end of the simulation. Last, please find the number of cores in your CPU by searching for it. For example, mine is an Intel i7-4910MQ. https://ark.intel.com/content/www/us/en/ark/products/78939/intel-core-i7-4910mq-processor-8m-cache-up-to-3-90-ghz.html Note that this page says my CPU has # of Cores = 4. Then take the # of Cores and use it with the -np option when you submit the job. This will cause Radioss to use 4 cores when it solves. So if you submit from the command line you can use radioss -np 4 test_0000.rad or put "-np 4" in the solver run manager window. While running the "Remaining Time" is an estimate that will change depending on the model and how much you use your PC for other things while it is running. Thanks, Andy
  11. Hi, First some definitions. From https://superuser.com/questions/1105654/logical-vs-physical-cpu-performance In Radioss, we typically recommend using only physical cores where you set the number of physical cores = -np option. So if you have a machine with 4 physical cores using -np 4. You should get a small amount of speed up by using the extract Hyperthreading cores by using, -np 4 -nt 2. Or for your other machine with 36 physical cores use, -np 36. or you can try -np 36 -nt 2 if you want to use all 72 logical cores but the speedup will be not very large. Meaning it won't run 2x faster when using all 72 logical cores. Thanks, Andy
  12. Hi, Unfortunately, GPU calculation is not supported for Radioss explicit. In older versions of Radioss implicit, the GPU could be used but it is no longer supported. It is best to use multiple CPU cores to reduce the run time of your simulation. The Intel MPI is included in Radioss 2017 installation so you should be able to use the <install directory>/altair/scripts/radioss -np # input_0000.rad where # is the number of cores you want to use to run your model. Thanks, Andy
  13. Hi, I have attached an example model of using two spring in series to model the failure and pretension. For pretension one type 32 spring with a second duplicated element with type13 spring for DOF 2-6. A second TYPE13 spring includes the failure properties. As far as modeling the washer as rigid or not it depends on how detailed of a model you want to create. Thanks, Andy model_001a_0000.rad model_001a_0001.rad
  14. Hi, Note that /PROP/SPR_PRE (TYPE32) only includes stiffness in the axial direction. If you use TYPE32 you will need to include a second TYPE13 with stiffness in DOF 2-6 to model the other bolt stiffness. Note that RD-T: 3510 tutorial doesn't include the TYPE13 spring. To model the shear failure, you can define failure values in the TYPE13 spring. Please read the comments in the TYPE32 spring about the pretension force curve needed to go to zero to lock the pretension force. If you don't lock the spring then the pretension force is always applied in the model. Thanks, Andy
  15. Hi @fcolomb Sorry for the late reply. Concerning run time,, you can try applying the load faster then the actual forming. This is often still valid if the kinetic energy doesn't get too high. Next, to increase the time step you could use /DT/NODA/CST and define a higher timestep which will add mass in the model. Alternatively, you can artificially increase the density of the whole part to increase the time step. Last instead of using /DT/NODA/CST or increasing the density, you could try using advanced mass scaling. /DT/AMS in the engine and /AMS in the starter. Using AMS you can normally increase the timestep by a factor of 10 - 20 time your original timestep. The AMS method requires more computations so your simulation will run approximately 4-10 time faster. With all these methods, I would recommend running the model a short amount of time with and without these options and compare the results to make sure these options give good results. The amount of mass add using /DT/NODA/CST or using the AMS is model dependent so you have to validate using these options. Last concerning the energy error. I suspect that is coming from the friction which is included in the contact energy. If you have high contact energy relative to your total energy you can get a larger error. Plot the contact energy and compare to the TTE: Translational total energy. If this is true then it is OK to ignore the energy error. Thanks, Andy
  • Create New...