9 Replies Latest reply on Apr 1, 2015 2:09 PM by Jared Conway

    Simulation study advice needed

    Aaron Hayden

      I am working on studying a pivoting door. The door is made of 1/2" thick aluminum plates with low carbon steel ribs bolted to it. There are 5 ribs keyed to the shaft and held in place with setscrews. There are two crank arm weldments (one on each end) mounted on the shaft in a similar manner. There are also two large counterweights (one on each end of the shaft) also keyed with setscrews. The counterweights are mounted outboard of the outermost bearing (cantilevered). The door itself is cantilevered and during operation will be supported only by 4 pillow block bearings. I want to analyze the deflection in the door since it will be largely unsupported. We know it will sag in the middle, but how much? I am also interested in the shaft to verify sizing and material. For the first study I used split lines to create the bearing surfaces. I then added bearing supports to those surfaces. I fixed the clevis pin hole in the crank arm where the hydraulic cylinders will attach. I used global bonded contact. The study resulted in a max stress of about 26K psi concentrated on the corner of the keyway in the crank arm.  I then decided to add no penetration to the shaft and keyway surfaces that interface with the crank arm and reran the study. Now the max stress is 582k psi located in the crank arm weldment. So my question is when is it appropriate to use bonded versus non contact? Nothing will be welded to the shaft so I feel like bonded isn't exactly representative, but 582K psi seems very unrealistic too.

       

      Also, as I stated above the door panel is a bolted construction, but I am using bonded here as well. Is bonded going to make the door seem much more rigid than it is? Is this something where I need to added bolted connects at every bolt location to get realistic results?

       

      Regards,

      Aaron

       

      door_study.png

        • Re: Simulation study advice needed
          Mike Pogue

          No penetration is the correct boundary condition for bearing stress. But it is touchy and difficult to get it to converge. 582 ksi is probably not physical, but the result of a singularity in the model. You should start with a bearing stress hand calc to figure out the neighborhood you should be in.

          • Re: Simulation study advice needed
            Aaron Hayden

            What steps are needed to utilize the no penetration conditions and get a more accurate representation from the study?

             

            -Edit: After more investigation I reduced the mesh size and reran the study. The result was a max stress of 39K psi concentrated at the end of the keyway. Then I reduced the mesh size all the way down to finest and the results was 1.65 million psi. Is this something where I need to use mesh control on these parts to avoid nodes that result in singularities?

            • Re: Simulation study advice needed
              Aaron Hayden

              With the help of a coworker we suppressed some features (holes and chamfers) to simplify the model a bit. I also added some mesh to control to two of the components. I can now mesh the part with varying levels of element size and get somewhat consistent results. Roughly 10 to 15 percent lets say. Hey, it's within the same magnitude! I tried running P adaptive but I keep getting an adaptive mesh error. Depending on what changes I make to the study it may run three iterations before the failure, or not make it through the first. Any idea how to deal with this issue?

              • Re: Simulation study advice needed
                Aaron Hayden

                My mistake, I meant H adaptive. H adaptive gives me the mesh adaptive error. P adaptive won't run. It says it is not supported for gap/contact and a bunch of other reasons.

                 

                This morning I simplified the model further. I removed the keys, keyways, set screw holes and chamfers that aren't significant. Using no penetration without the keys & keyways there is now a degree of freedom (rotation) that I need to prevent. What is the best way to do this? There doesn't seem to be much in the way of diagnostics to figure out why I am getting this mesh adaptive error. I can mesh the study with varying mesh sizes and the study will run fine. That is why I removed the keys and keyways to simplify in an effort to get past this error. So far, no dice.

                 

                -Edit: Perhaps a pin connection with key is what I need here? Can I use a pin connection and no penetration together?

                  • Re: Simulation study advice needed
                    Jared Conway

                    mesh adaptive error tells me you've got either a bad mesh to start with or geometry that it can't adapt

                     

                    based on your other questions, I would recommend you take a step back from your analysis and really simplify it down to the basic elements to work on the basics of simulation and slowly add complexity as you add skill.

                     

                    I think you're getting bogged down with high level features that aren't adding a lot of value to your analysis.

                     

                    When we mentor customers, we spend a LOT of time on this planning part because then your setup is really fast, accurate and solves quickly.

                     

                    And regarding a pin, it basically replaces the need to have the components or contact.

                  • Re: Simulation study advice needed
                    Aaron Hayden

                    I created a study using pin connectors w/ retaining ring. It doesn't appear that you can take advantage of hole sets with pin connectors like you can with bolt connectors. So, if you have a lot of holes, as I do, setting up the pin connectors is very tedious. However, the study ran in about half the time, which is a good thing. If you plan on running the study several times, the investment of adding the pins is easily offset by the solve time. However, in the case of my study the deflection result from the pin connector is about 30% less than when bolted connectors were used. Depending on what you need out of the study, 30% can be very significant.

                     

                    I also, ran a study with pin connectors and used symmetry. The symmetry fixture feature would not work. I believe due to the fact that the ends of the two large plates meet at the plane of symmetry. I am guessing the symmetry fixture needs to cut through all solids and cannot be coincident with the edge of a solid. So, I added a roller/slider to the cut faces and the end of the large plate. The results between the pin connector study and pin connector w/ symmetry using roller/slider study varied. Pin connector w/out symmetry: stress 28 ksi, deflection .224". Pin connector w/ symmetry: stress 17.5 ksi, deflection .204". Perhaps there is a better constraint method that roller/slider for this?

                     

                    I agree that simpler is better in terms of encountering mesh/solve issues and also dramatically improves solve time. As I am sure you are aware, you have to be cautious that the sacrifice of accuracy isn't greater than what the application can accept.

                      • Re: Simulation study advice needed
                        Jared Conway

                        Perhaps there is a better constraint method that roller/slider for this?

                         

                        >>symmetry and roller slider are exactly the same.

                         

                        I agree that simpler is better in terms of encountering mesh/solve issues and also dramatically improves solve time. As I am sure you are aware, you have to be cautious that the sacrifice of accuracy isn't greater than what the application can accept.

                         

                        >>the methodology i'm recommending is a crawl, walk, run approach so that if you run into issues you can identify them easier and spend less time troubleshooting. you can always add complexity and as you gain experience, you can jump right to the end. but what you'll also learn through doing this is that you gain minimal improvements of accuracy just by adding more features to your analysis setup. and the tradeoff is that you spend longer times setting them up and running them.

                         

                        Here's an article I recently wrote about the topic: What level of detail should be included in my model for analysis?

                         

                        And something to consider is just adding more features to the model doesn't necessarily mean more accuracy. To determine accuracy you either need to test, compare against hand calcs or use your experience. A great example is your question about the high stress in a singularity. Without that knowledge, you may have redesigned that part without knowing it was only numerical.