8 Replies Latest reply on Mar 13, 2015 9:54 AM by Bob Thesponge

    Static study of assemblies

    Bob Thesponge
    Bob Thesponge Mar 1, 2015 2:51 PM

      Hello,

       

      I have experience in SW simulation of parts (i.e. assemblies with no internal degrees of freedom, which I call kinematic sets. Another way to put this is to say kinematic sets are sets of parts rigidly linked together and which move together). Now I am willing to perform a simulation of an assembly with internal degrees of freedom (i.e. the system now includes different kinematic sets linked by prismatic and pivot linkages), this is all new to me. I have reviewed the different ways offered by SW simulation to solve such a system, but as this is new to me I do not feel confident on which approach I should use - i.e. which approach is the most standard.

       

      Here are the methods I could think about:

      1. Simulate each kinematic set individually: compute the efforts at all internal and external links (to get to know the interactions between kinematic sets) then simulate each kinematic set individually
        • This brings me back to the situation I know how to handle in SW
        • But I need an extra stage where I compute all the link efforts. Getting these link efforts can be done in several ways:
          • A static study by hand: this is the "back to school" way where you simplify your system, write the equations and solve them
            • I find this approach time consuming and error prone
          • Perform a motion study: motion studies allow to record the efforts in links (over time) and later import these in SW simulation
            • This is more than what I want to do (as you study the system for a range of time - which I do not really need)
            • You need the Motion plugin in order to do so
          • Simulate kinematic sets one after another: simulate a single kinematic set, probe the resultant efforts in the links (this is done is SW simulation) then simulate the contiguous kinematic set (re-applying the resultant efforts probed at the previous kinematic set)
            • This may not be very accurate as the resultant efforts over a whole link may arise from a very non-uniform distribution of the constraints inside the link (there can even be opposite constraints inside the same link)
            • Hence the resultant effort may not represent the behavior inside the link
      2. Simulate links using contacts: you use the geometry of the link to simulate its kinematic effect
        • Very resource intensive and does not always find a solution
        • Need to simplify the model in case of complex geometry (rather often)
        • Seems overkill (but I am not an expert)
      3. Parametric link: use the mates put in place in the modeler to set-up constraints on the mesh nodes
        • This "gut feeling solution" does not not seem to be possible for some (dark) theoretical reasons
        • More specifically: I find it weird to be able to specify parametric mechanical links for the fixtures (prismatic, pivot...) but not to be able to do so for connection
        • I guess there is a theoretical reason behind this which finds its roots in the FEM used by the simulation (but this goes beyond my knowledge of simulation)

       

      Anyone having advice on the technique I should use? What is the most standard approach used by the professionals? My gut feeling is that using contacts seems the easiest way, though is seems a bit overkill...

       

      Thanks for your help,

       

      Antoine.

        • Re: Static study of assemblies
          Mike Pogue
          Mike Pogue Mar 2, 2015 12:15 AM (in response to Bob Thesponge)

          I hope I understand you question correctly. If so, no FEA program is going to solve an assembly with a mechanism (internal degrees of freedom) in it. Mechanisms, correctly, cause numerical instability in the stiffness matrix, because certain "directions" have no stiffness. In some programs, it is possible to force past this instability to see the results. But this is done for debugging. You can't use the results. Dynamic studies involving vibration usually involve fully constrained assemblies, but even if not, it's always possible to eliminate rigid body modes by inspection.

           

          Simulating links using contacts has the disadvantage of being non-linear. However, if the contacts remove all degrees of freedom, then this method is correct.

           

          Parametric links are also correct--if and only if they remove all degrees of freedom.

           

          As far as analyzing mechanisms, I'm pretty sure you can perform a motion study to get a series of load cases, which you can then hand off to Simulation to solve, it I've never tried it in SolidWorks. The other option is to

          • Re: Static study of assemblies
            Jared Conway
            Jared Conway Mar 3, 2015 8:59 PM (in response to Bob Thesponge)

            what do you want to learn in your simulation?

             

            can you post an example of the type of geometry that you're going to work with?

             

            I think you're digging too deep in this one.

             

            your options are

            static analysis > small displacements, deformable bodies, no time effects

            nonlinear static > large displacements, deformable bodies, no time effects, long run times

            motion > rigid body motion

            linear dynamic, time history > time effects, linear materials, deformable bodies, long run times

            nonlinear dynamic > kitchen sink simulation, long run times

             

            motion > static can be done, but generally not enough if the deformation is important during the motion

              • Re: Static study of assemblies
                Bob Thesponge
                Bob Thesponge Mar 9, 2015 5:45 AM (in response to Jared Conway)

                Hi and thanks for your answers so far.

                 

                Here is a bit more information on what I want to achieve with my simulation. My basic design is an elevator tray. The tray is linearly guided thanks to 2 linear ball slides (shown in dark grey on the picture), these slides are linked to a column. The tray is actuated using a power screw (which nut appears in brown in the picture below). The following picture shows the tray, column and screw:

                 

                 

                WhatIWantToDo.jpg

                 

                Now I would like to optimize my design for weight (to get a lighter machine) and for deformation (to get as low deformation as possible at the tip of the tray). In order to do so I would like to statically simulate my system in action (i.e. apply an effort as shown below and see how the reaction). This optimization is aimed at focusing on 2 goals:

                • Knowning the areas in the parts which have less stress (Von Mises stress): this will allow me to remove matter in these locations and thus reduce the overall weight
                • I also want to estimate the deformation: this will allow me to add reinforcement parts where needed in order to limit the deformation

                 

                WhatIWantToDo2.jpg

                 

                Now simply simulating the tray without the column would be easy (there would be no movable link between the parts hence the static simulation would be straightforward) - unfortunately I also need to account for the behavior of the column behind the tray. The problematic thing in this simulation is that the simulated system includes mecanical links (between the column and the tray) which have a major impact on the behavior of the system. This means I cannot approximate the system by simply replacing the movable links with rigid links: I need to somehow simulate the effect of the movable links. And this is where my above question becomes relevant: which method should I use to simulate my movable links and make it able to optimize my design?

                 

                Thanks,

                 

                Antoine.