Add a Sense of Reality to Your Simulation

I have seen many simulation results which do not make sense to a normal engineer. It is usually not because the theories or corresponding simulation codes are wrong, but because the inputs are not real, or not in consistent units. For example, I have seen students obtain a strain of 0.5 from a linear elastic model and stress of a few Pascals in a loading-bearing solid. There are a few tips a simulator should keep in mind:

  1. Make your inputs as real as possible. Don't just give a random number. For example, if you are analyzing a composite laminate with each layer assumed to be homogeneous, find the lamina constants from a trustworthy database such as the NIAR' AGATE database at s And also not only lamina constants, also use realistic layer thickness. A layer thickness of 0.1 meter is laughable. 
  2. Make sure that you use a consistent unit system. Most software does not care which unit system you uses, For example, if you have Young's modulus E=73 GPa. You can input Young's modulus as 73E9, geometry of your model in terms of meter, and load in terms of Newton. However, if your structure is very small, such as in the size of milimeter, and you would like to use milimeter as the unit for the geometry, and Newton for the load, then you need to input E=73e3 MPa, and the stress results will be in MPa for a elastic analysis. For multifunctional analysis which involves different physics, then using consistent units become even more important. 
  3. Make sure you indicate the units for your tables and plots so that others don't have to ask. And also ask yourself whether the results make realistic sense or not. For example, if the results has stresses in the order of hundreds of GPa, these should cause to re-think your results as most materials cannot sustain such a high stress. For a linear elastic model, if you have a strain of 0.5, it does not make sense either. Even a strain of 0.1 should cause a read flag for a linear elastic problem. 

Comments on this entry

  1. Banghua Zhao

    Thanks for the tips, Prof. Yu! During my study and work, I also meet the situations that the simulation result seems odd to me. There tips could solve most of my problem:

    1. The inputs directly determine the material models you input in your simulation. And the accuracy of your simulation is greatly influenced by the material models. For example, the ductile material and brittle material is pretty different. For ductile material, the yield strength (YS), ultimate tensile strength (UTS), uniform elongation (UE) are other key parameters beside the Young's modulue and Poisson's ratio. For brittle material, it is reasonable to use linear elastic model. Other good material databases are MatWeb ( and MakeItFrom (

    2. For small scale geometry (phone, watch), it is convenient to use "mm" so the unit system is MPa, mm, ms, N. For large scale geometry (car, airplane), it is convenient to use "m" so the unit system is GPa, m, s, N. 

    3. Indeed, it is always a good practice to indicate the unit and other interest information in the result contour plots so that design engineers can easily understand the result. Especially for design engineers without FEA experience, we need to make the result simple, precise and reasonable.

    And I have some experience to double check the reality of the simulation result:

    1. Check the quality of elements in your FEA model. Bad quality elements will give unreasonable result. For 3D mesh, make sure Tet Collapse > 0.05 and Jacobian > 0.4

    2. Make sure to use appropriate element type. During the simulation, we may meet a lot of numerical issues caused by element formulation, like shear locking and hourglass. For bending problem, if you use linear element, avoid to use full integration. And if you use reduced integration, make use to have at least 4 elements in the thickness direction or use incompatible mode element.

    3. Understand the result. The true stress and strain is different from engineering stress and strain.

    4. Understand the physics. Make sure the boundary conditions (BCs) is appropriate.

    Reply Report abuse

    Replying to Banghua Zhao

Post a comment

You must log in to post comments.

Please keep comments relevant to this entry. Comments deemed offensive or inappropriate may be removed.