heinzboehmer's 2002 Topaz 6MT Coupe

[heinzboehmer]

Sat down with a friend (who works for an OEM) and he walked me through doing FEA on my design. Learned a ton, so documenting here for my own sake.

We didn't really know what the typical loads that these braces would see were, so instead we approached the analysis by using the max load they'll ever see. Assumptions follow:

1. BMW designed the E86 braces to bend in a crash, so the max load my design will ever see is just below the load that causes the braces to fail.
2. The braces must fail below the shear strength of the fasteners that hold them to the chassis, otherwise the feature would be useless.
3. BMW fastens these braces with class 10.9 M10 bolts on the E85/E86 chassis. Shear strength for the bolts is 23.2 kN, which is the absolute theoretical load limit for the braces (real limit is below that due to the bending).
4. Braces are at a 46.4 deg angle outward from the centerline of the car towards the strut towers.
5. Materials for all my pieces are 6061 T6 aluminum. Figured we would do the analysis on the worst case (in terms of 6061 vs 7075) just to see.
6. 6061 tensile yield strength is 265 MPa (bit of extra padding cause why not).

With all this in mind and a bit of math, we get that the max x component (front to back) of the load is 16 kN and the max y component (side to side) is 16.8 kN (for each brace).

Plugged all this into the FEA tool, set up all the connections and constraints and done. Analysis came back saying what every mechanical engineer I've shown this to has said: my design is way, WAY overkill.

I forget the actual numbers, but the piece that bonds to the windshield has an insane safety margin (think 10 or so). Also, as expected, the piece that the braces bolt to sees the most load by far, but was still only seeing around half the tensile yield strength for peak load. Keep in mind that all these numbers come from loads that the brace mounting points will never see, since the braces are designed to bend before the fasteners fail.

What this all means is that I can make this design significantly lighter, so lots of iterations coming up!

The following is the initial list of TODOs:

• Remove that aluminum extension on the passenger side. I thought this would help with distributing the load to the sheet, but simulation said it was basically useless.
• Decrease wall and rib thickness. It will be interesting to find a good tradeoff between overkill and too little bonding area.
• Play around with the thickness of the flange that the studs press into.

I'll likely target a safety factor of 1 for future simulations, since the normal loads the assembly will see are way below the peak loads I'll be plugging in to the SW.

And just for fun, a couple images I found online to substantiate the claim that the braces will bend before the fasteners fail:

[0-60motorsports]

Sat down with a friend (who works for an OEM) and he walked me through doing FEA on my design. Learned a ton, so documenting here for my own sake.

We didn't really know what the typical loads that these braces would see were, so instead we approached the analysis by using the max load they'll ever see. Assumptions follow:

1. BMW designed the E86 braces to bend in a crash, so the max load my design will ever see is just below the load that causes the braces to fail.
2. The braces must fail below the shear strength of the fasteners that hold them to the chassis, otherwise the feature would be useless.
3. BMW fastens these braces with class 10.9 M10 bolts on the E85/E86 chassis. Shear strength for the bolts is 23.2 kN, which is the absolute theoretical load limit for the braces (real limit is below that due to the bending).
4. Braces are at a 46.4 deg angle outward from the centerline of the car towards the strut towers.
5. Materials for all my pieces are 6061 T6 aluminum. Figured we would do the analysis on the worst case (in terms of 6061 vs 7075) just to see.
6. 6061 tensile yield strength is 265 MPa (bit of extra padding cause why not).

With all this in mind and a bit of math, we get that the max x component (front to back) of the load is 16 kN and the max y component (side to side) is 16.8 kN (for each brace).

Plugged all this into the FEA tool, set up all the connections and constraints and done. Analysis came back saying what every mechanical engineer I've shown this to has said: my design is way, WAY overkill.

I forget the actual numbers, but the piece that bonds to the windshield has an insane safety margin (think 10 or so). Also, as expected, the piece that the braces bolt to sees the most load by far, but was still only seeing around half the tensile yield strength for peak load. Keep in mind that all these numbers come from loads that the brace mounting points will never see, since the braces are designed to bend before the fasteners fail.

What this all means is that I can make this design significantly lighter, so lots of iterations coming up!

The following is the initial list of TODOs:

• Remove that aluminum extension on the passenger side. I thought this would help with distributing the load to the sheet, but simulation said it was basically useless.
• Decrease wall and rib thickness. It will be interesting to find a good tradeoff between overkill and too little bonding area.
• Play around with the thickness of the flange that the studs press into.

I'll likely target a safety factor of 1 for future simulations, since the normal loads the assembly will see are way below the peak loads I'll be plugging in to the SW.

And just for fun, a couple images I found online to substantiate the claim that the braces will bend before the fasteners fail:

Thats some great news! Really looking forward to this especially if and im sure it will work with the OE cabin filter housing (i have the OE CSL One). Also I love the cleaner looking engine bay this will give and the OE CSL intake will look so much nicer on Full display. Thanks for working on this.

[heinzboehmer]

Thats some great news! Really looking forward to this especially if and im sure it will work with the OE cabin filter housing (i have the OE CSL One). Also I love the cleaner looking engine bay this will give and the OE CSL intake will look so much nicer on Full display. Thanks for working on this.

Thanks! Looking forward to having this done and installed on the car as well.

I don't think this will work with the stock cabin air filter housing, unfortunately. Gonna need to chop it up to make clearance. Ultimately, I want to print a mold and make it out of prepreg CF, but that's a project for the future.

Also, just to be clear, I plan on running this in addition to the stock strut bar, but I guess you could run it in place of if you wanted to.

[Bry5on]

I would leave the ribs for bonding area, and leave the extended surface to stiffen up the windshield sheet to avoid peeling adhesive as much as possible. That’s the place that makes the most sense to go overkill to me. You could also tab and slot the machined part and sheet then just weld it up. That would allow you to ignore bonding area internal to the part.

For the load case, you can just calculate the bucking load for the beam. All you need is diameter, length, wall thickness and material. Since it’s notched, it will fail below this number because it has an initiation point. That way you can use a real number and get a real safety factor.

[heinzboehmer]

For the load case, you can just calculate the bucking load for the beam. All you need is diameter, length, wall thickness and material. Since it’s notched, it will fail below this number because it has an initiation point. That way you can use a real number and get a real safety factor.

Added to TODO list!

I would leave the ribs for bonding area, and leave the extended surface to stiffen up the windshield sheet to avoid peeling adhesive as much as possible. That’s the place that makes the most sense to go overkill to me. You could also tab and slot the machined part and sheet then just weld it up. That would allow you to ignore bonding area internal to the part.

You bring up an excellent point that I did not consider yesterday. We ran the analysis assuming the top surface of the aluminum piece was fixed in place, not bonded. I'll mess around to see if I can produce something useful that takes into account the adhesive area.

Also, ran some super quick tests in CAD just to get an idea of the theoretical min weight of the assembly. Baseline is the overkill design, which weighs 579 g total.

First up, extended surface fully removed, but everything else the same. This saves 111 g (19%):



Next up, extended piece removed, ribs removed and wall thickness decreased to 2mm. This saves 235 g (41%):



This means that a realistic minimum weight I can get this assembly to is 400-450 g, which raises an interesting question: do I really care about saving max ~200 g? The weight is being added to a terrible spot in the chassis, but the potential tradeoff of saving the 200 g is ending up with a subpar piece that will fail and essentially require a chassis replacement. It might make more sense to focus my attention on saving weight in other pieces (CF cabin air filter housing and firewall plug for instance).

I still want to run the FEA on everything (because new toy to play around with), but I'm not sure I'll actually end up making any changes to the design.

[heinzboehmer]

Alright, signed up for the 14 day SimSolid trial. Integrates very nicely with Onshape, but it's $675 per month once the trial ends, so I've got a deadline for finishing this project now!

[heinzboehmer]

Drove the car a good amount today and broke a couple things.

First up, the zip tie that holds the right side brake duct upright broke and allowed it to pivot down. Was scraping nicely on the wheel when turning:




Will need to print another one and reattach it with higher quality zip ties, that Bryson recommended.

I also snapped the top piece of my phone holder:



Of course it landed inside the dash, so retrieving it was fun:




This is a tricky piece to print since it needs to be strong along multiple orthogonal axes. I considered experimenting with printing it at an angle, but instead decided to try this approach:



Blue piece will be printed upright (as seen in the screenshot) and yellow will be printed laying down. Yellow is designed to press fit into blue and should help to keep it from breaking again in the same way it did today.

[heinzboehmer]

Wait what am I doing. I just need to print the entire tab laying down...

[bigjae46]

Not sure what exact part is the issue but GSR Motorsports makes a sturdy aluminum part that holds the lower fender liner in place.

https://gsrmotorsports.net/products/spec-e46-coupe-splash-shield-brackets

[heinzboehmer]

Not sure what exact part is the issue but GSR Motorsports makes a sturdy aluminum part that holds the lower fender liner in place.

https://gsrmotorsports.net/products/spec-e46-coupe-splash-shield-brackets

Ah this brake scoop attaches to the control arm, not the stock brake duct bracket. But thanks for the link, good to know that exists,.

[ethan]

First up, the zip tie that holds the right side brake duct upright broke and allowed it to pivot down. Was scraping nicely on the wheel when turning:

I've had good luck with these ties so far with my set of Bry5on ducts. Only street and one track day so far though - fingers crossed.

https://www.mcmaster.com/6614K57/

[heinzboehmer]

Back to chassis mods. Be warned, there are LOTS of pictures in this one, but as previously mentioned, my FEA SW is on a free trial, so documenting all the results here

Discussed the design of the piece that bonds under the windshield with Bryson and we (mostly him, if I'm being honest) came up with some alternative designs/construction methods.

First did the buckling calcs in an effort to get some more realistic loads for the braces. Approximated wall thickness by measuring the thickness of the flat part, dividing by 2 and then rounding up to the nearest value that made sense (flat is wider than OD of bar, so I'm assuming it's squished out a bit). Flat measured in at 2.75 mm, so I used 1.5 mm for the calcs:

E = 2 x 10^11 Pa (these things are made out of some sort of steel)
I = (pi / 4)(r2^2 - r1^2) = (pi / 4)(0.01275^4 - 0.01125^4) m^4 = 8.175 x 10^-9 m^4
L = 0.45 m
K = 1 (you can see in the above Z4 crash images that the brace buckles right after the flat, so counting that as rotation free and translation fixed)

Plugged that all into Euler's formula:

F = (pi^2 * 2 x 10^11 * 8.175 x 10^-9) / (1 * 0.45)^2 = 79.686 kN

Which is, unfortunately, ~3x higher than the shear strength of the bolts used on the Z4. i.e. Useless

Guess that notch weakens the braces much more than I expected, so time for some more involved analysis. Talked to Bryson and he walked me through the statics of this problem. I don't presently feel comfortable enough with the theory to try and explain it, so I'll just use the number he arrived at for the FEA

Statics says it will take ~21 kN to buckle the notched brace, which tells us the max x component (front to back) of the load is ~15.8 kN and the max y component (side to side) is ~13.9 kN (for each brace).

Alright time to test the new designs. First up, a boxed + welded version of the previous design:



In this design, the CF sheet is replaced with an aluminum one, the ribs are entirely removed and wall thickness is decreased to 2 mm. The two pieces are meant to be welded together, which is what gives it its strength. Despite this being constructed entirely out of aluminum, it's lighter than the previous design! Weighs in at 450 g (0.993 lb) for all three pieces. FEA results follow:

Both braces in compression:




One brace in compression, one in tension:




Both braces in tension:




(Ignoring brace attachment point piece for now, as it needs work at this point. More on that later.)

As you can see, the piece that gets bonded to the chassis is a beast. Biggest stress concentrations are at the bosses, but those can likely be mitigated with some design tweaks. However, this is where I stopped iterating on this design and moved on to the other one.

Next up is a fully machinable, one piece design that gets pocketed from the front/sides instead of from the bottom. The result is a V shape that is super strong along the load paths and also cuts down significantly on weight.

This design can also be extended to be a two piece, weldable assembly (machined top part + laser cut sheet metal bottom part). This should help cut the cost due to the sheet metal part, but for now, I'm just gonna go ahead with the fully machined version.


A little heavier, but nothing that makes me want to trade strength for weight. Both pieces weigh in at a total of 499 g (1.099 lb). FEA results follow:

Both braces in compression:




One brace in compression, one in tension:




Both braces in tension:




The part that the braces mount to is the component under the most stress, but I am less concerned about this piece, as it is easily replaceable. I iterated on it a bunch based on the feedback from the previous FEA runs and this is about as good as it's going to get without having to modify the braces. As mentioned before, the braces are not flat up top and this piece has to accommodate for that. I could take a big hammer (really, a press) to them and flatten them, but this seems like it would make it significantly easier to bend the brace at the interface between the flat and cylindrical sections, so I'd rather avoid it. If I run into issues with the current design, then I'll likely have to go this route.

Quick look at safety factor analysis as well, starting with 6061-T6:

Both braces in compression:




One brace in compression, one in tension:




Both braces in tension:




Not bad. Min safety factor is 1.3 for the brace attachment point and 2.5 for the windshield piece.

Also ran this analysis with 7075-T6:

Both braces in compression:




One brace in compression, one in tension:




Both braces in tension:




As expected, it's much stronger. Min safety factor jumps up to 2.4 for the brace attachment point and 4.6 for the windshield piece.

Happy with the V design. Highly likely that this is going to be what's going into my car. I do want to run one final analysis with the full fledged SimSolid just to double check that the panel bond adhesive isn't going to be subject to loads higher than what it's designed for, but I think it'll be alright.

And finally, ran some quotes and the one that made the most sense was right around $950 all in for 6061-T6 (materials, labor, taxes, shipping, etc.) and $1000 for 7075-T6. $50 extra for 7075? Uh, yes please!

[heinzboehmer]

I've had good luck with these ties so far with my set of Bry5on ducts. Only street and one track day so far though - fingers crossed.

https://www.mcmaster.com/6614K57/

Thanks! That's basically exactly what I bought

Should have known better than to throw some cheapo ties at them...

[George Hill]

IF you still have time on your simulation trial it might be fun to see how your first design stacks up against the final design. Would show what you learned and how it affected everything starting with a design that "seemed" good to now.

[heinzboehmer]

IF you still have time on your simulation trial it might be fun to see how your first design stacks up against the final design. Would show what you learned and how it affected everything starting with a design that "seemed" good to now.

That's an excellent idea. I'll run that analysis soon.