Jieve
Mechanical
- Jul 16, 2011
- 131
Hello,
I am a mechanical engineer relatively new to design (<1 year), work in technical education and am working on an independent machine design project of a machine used to demonstrate machine vibration phenomena. I have been running FEA analyses on parts/subassemblies on the mostly finalized design to determine things like optimal rib locations and bolt clamping stresses using solidworks simulation, as most of the loaded parts are to be made of 6082-T6 Aluminum, and I was concerned that the class 8.8 M10 hex screws used to bolt down the caps of a pillow-block style bearing housing to 75% of proof could cause heavy embedding and local yielding issues. I have never taken a formal FEA course, but I’ve had the luxury of not being under time pressure on this project and have spent the last 2 months pouring over internet forums, technical documentation and practice problems. I have even purchased and worked through all of the exercises in two books specifically aimed at solidworks simulation (I’m using 2011), so I feel that I have a relatively solid understanding of the pitfalls and details necessary to produce relatively accurate analyses. However, I have a few questions for you guys out there that have been doing this for a long time.
1) As I understand from the solidworks forums, the software’s bolt connector does not give accurate results at the bolt head and nut/part interfaces. Since I am interested in the compressive stresses at the bolt hole locations under full bolt preload, I have simplified the problem by creating what are essentially 2 pucks, one representing the bolt head with washer face diameter, and one representing the nut (I think I remember reading a technical document from ANSYS where this was a simplified method of modeling bolts, but doesn’t take into account the joint constants). As these are class 8.8 screws, I created a material definition with a tensile strength of 800Mpa and Yield of 640 (these according to the Tabellenbuch Metall (Mechanical and Metal Trades Handbook), german version of the Machinery’s Handbook), and applied them to the pucks. Contact was node to surface (as otherwise surface to surface seems to take FOREVER) and I applied forces normal to each of the pucks to represent the preload, something around 25000N. I put small radii on all corners (normally I try for 3 rows of elements along each radii min, but this time I stuck with defaults) and ran an h-adaptive analysis. This brings me to my first question: When adding fillets to interior edges for the sake analysis, is there a rule of thumb as to what size they should be? I understand everything should be radiused and called on drawings. But let’s assume a part was simply milled or drilled without radiusing. For example, as drilled (not deburred) bolt holes, or as-milled 90 degree edges, with no radii called out. Or the outer edge of ground washers. Of course sharp corners are bad news = stress concentrations, but in real life these stresses are not infinite (obviously), whereas in software, the stress is infinite. With regard to software when setting up the problem, how small is minimum for an accurate representation of a real-life sharp corner? Using the method I just described, even with radii I seem to have convergence issues (stresses keep increasing).
2) This may seem like an ignorant question, but hand-calcs: assuming the washer is significantly harder than the clamped material, is it safe to assume a ballpark figure for the compressive stress to be nothing more than the bolt preload force divided by the washer contact area?
3) I setup another problem similar to (1) above except used a disc washer (used actual washer dimensions) between the pucks. The washer hardness is 300HV (tensile strength of around 1000Mpa) so I used one of the steels in the solidworks material library that somewhat resembled this (I couldn’t find any more data than this). In every modeling case, I found that the stresses around the internal edges even with fillet radii on every corner and a superfine mesh were excessively high (I’m only using an elastic model for the material though) but we’re talking 10x yield, which OF COURSE is not realistic since plasticity isn’t taken into account, but regardless this just seemed excessively high. I wasn’t able to run a convergence study just because the mesh was so fine that even with my quad core 8 gig ram comp it took over 24 hours to complete. I found that the inner edge of the washer was buckling inward. Any input, comments on this? A better way to do it maybe? I wanted a realistic model, and am also trying to hone my skills using some of these problems as exercises.
4) BTW I’ve been using the Direct Sparse solver in solidworks usually with large displacement checked.
I’ve tried to attach some pics to give an idea of what I’m doing, hopefully both of them worked. Thanks for any info.
I am a mechanical engineer relatively new to design (<1 year), work in technical education and am working on an independent machine design project of a machine used to demonstrate machine vibration phenomena. I have been running FEA analyses on parts/subassemblies on the mostly finalized design to determine things like optimal rib locations and bolt clamping stresses using solidworks simulation, as most of the loaded parts are to be made of 6082-T6 Aluminum, and I was concerned that the class 8.8 M10 hex screws used to bolt down the caps of a pillow-block style bearing housing to 75% of proof could cause heavy embedding and local yielding issues. I have never taken a formal FEA course, but I’ve had the luxury of not being under time pressure on this project and have spent the last 2 months pouring over internet forums, technical documentation and practice problems. I have even purchased and worked through all of the exercises in two books specifically aimed at solidworks simulation (I’m using 2011), so I feel that I have a relatively solid understanding of the pitfalls and details necessary to produce relatively accurate analyses. However, I have a few questions for you guys out there that have been doing this for a long time.
1) As I understand from the solidworks forums, the software’s bolt connector does not give accurate results at the bolt head and nut/part interfaces. Since I am interested in the compressive stresses at the bolt hole locations under full bolt preload, I have simplified the problem by creating what are essentially 2 pucks, one representing the bolt head with washer face diameter, and one representing the nut (I think I remember reading a technical document from ANSYS where this was a simplified method of modeling bolts, but doesn’t take into account the joint constants). As these are class 8.8 screws, I created a material definition with a tensile strength of 800Mpa and Yield of 640 (these according to the Tabellenbuch Metall (Mechanical and Metal Trades Handbook), german version of the Machinery’s Handbook), and applied them to the pucks. Contact was node to surface (as otherwise surface to surface seems to take FOREVER) and I applied forces normal to each of the pucks to represent the preload, something around 25000N. I put small radii on all corners (normally I try for 3 rows of elements along each radii min, but this time I stuck with defaults) and ran an h-adaptive analysis. This brings me to my first question: When adding fillets to interior edges for the sake analysis, is there a rule of thumb as to what size they should be? I understand everything should be radiused and called on drawings. But let’s assume a part was simply milled or drilled without radiusing. For example, as drilled (not deburred) bolt holes, or as-milled 90 degree edges, with no radii called out. Or the outer edge of ground washers. Of course sharp corners are bad news = stress concentrations, but in real life these stresses are not infinite (obviously), whereas in software, the stress is infinite. With regard to software when setting up the problem, how small is minimum for an accurate representation of a real-life sharp corner? Using the method I just described, even with radii I seem to have convergence issues (stresses keep increasing).
2) This may seem like an ignorant question, but hand-calcs: assuming the washer is significantly harder than the clamped material, is it safe to assume a ballpark figure for the compressive stress to be nothing more than the bolt preload force divided by the washer contact area?
3) I setup another problem similar to (1) above except used a disc washer (used actual washer dimensions) between the pucks. The washer hardness is 300HV (tensile strength of around 1000Mpa) so I used one of the steels in the solidworks material library that somewhat resembled this (I couldn’t find any more data than this). In every modeling case, I found that the stresses around the internal edges even with fillet radii on every corner and a superfine mesh were excessively high (I’m only using an elastic model for the material though) but we’re talking 10x yield, which OF COURSE is not realistic since plasticity isn’t taken into account, but regardless this just seemed excessively high. I wasn’t able to run a convergence study just because the mesh was so fine that even with my quad core 8 gig ram comp it took over 24 hours to complete. I found that the inner edge of the washer was buckling inward. Any input, comments on this? A better way to do it maybe? I wanted a realistic model, and am also trying to hone my skills using some of these problems as exercises.
4) BTW I’ve been using the Direct Sparse solver in solidworks usually with large displacement checked.
I’ve tried to attach some pics to give an idea of what I’m doing, hopefully both of them worked. Thanks for any info.