Undesirable stresses due to attaching (with adhesive) 2 different CTE parts 3

[elinah34]

hey,

I have ULTEM 1000 (CTE=55e-6 1/c) part that is attached (using adhesive) to a steel part (CTE=10e-6 1/c).
The ULTEM 1000 part is the orange ring in the photos below:

By the way, the parts' materials and the joining method by adhesion are constraints.
We are very worrid about stresses that will develop in the interface (adhesive) when exposing these parts to temperature change.
Someone claimed that splitting the ULTEM 1000 ring to some segments may help to reduce the thermal stresses, but I am not sure about it.

What can you say about it? an explanation would be great.

 

[SWComposites]

Thermal stress is a function of the length of the joint between dissimilar materials. Have you done a simple thermal stress calc?

 

[IRstuff]

What adhesive are you using? What is its CTE and tensile properties?

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[moon161]

A thicker bond line will accomodate a greater strain than a thin one

 

[elinah34]

The adhesive is Axon and its CTE is almost identical to ULTEM 1000.
About the question if I did a thermal stress analysis - so, as a one that isn't deeply familiar with simulations (except for simple parts cases) I wasn't sure how to correctly model the adhesive layer, so I assumed a bonded relation between the parts and tried to see the interface stresses and it seemed quite the same in both cases, something that made me raise the question here...
I originally thought that splitting the ring may help...if I take the simplified case of adhesion a short simple straight part Versus a long one, the answer is clear...but the geometry here and the analysis results made me confused a little bit. Do you think the splitted ting is better?

 

[Compositepro]

Splitting the ring will reduce stresses but can possibly introduce peeling stress. In any case the part looks like a seal ring, and you do not want to split a seal. Reducing the thickness of the ring can also reduce shear stress at the bond line.

 

[elinah34]

Thanks

 

[drawoh]

elinah34,

I regard this analysis as simple. You have a [Δ]T. You have a circumference. You can work out the strain due to the temperature, and you can work out the resulting stresses. Your adhesive and design either works, or it doesn't.

If it doesn't, either you need to select materials with matching CTE, or you need a new way to retain the part. Can you cut a retaining slot?

--
JHG

 

[EdStainless]

I had a similar situation, we cycled the units from 20C to 200C frequently. We fit the parts and bonded them at 100C, effectively cutting the delta T in half. As the parts would age and the bond weaken we would get failures in parts while sitting at room temp, we preferred this over having them fail hot.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed

 

[drawoh]

EdStainless,

I recall glueing moderately large mirrors to metal substrates. Glass has a CTE of 9[×]10^-6K[°]^-1. Our preferred material, aluminium, is around 23[×]10^-6K[°]^-1. 300 grade stainless is about 18[×]10^-6K[°]^-1. 400 grade stainless is around 10[×]10^-6K[°]^-1. I called up 416MX, and I added an inspection note stating that the desired stainless steel grade was magnetic.

--
JHG

 

[elinah34]

Hey again,

My colleagues and I had a discussion today regarding this question, and I came out a little confused. They insisted that splitting the ring doesn't decrease the thermal stresses in the adhesive... they said that in each case the mechanical strain equals to the thermal strain, and as a result a same stress would develop.

About the suggestion of making a simple analysis - I don't know how to correctly model the adhesive, and as a result the interfacial stresses may be wrong.

EdStainless, do you have any example of how to manually calculate the expected stresses in each case mentioned above?

To be honest - I am a little embarrassed how simple problem makes me wonder a lot how to properly analyze this joint.

 

[drawoh]

elinah34,

The length of your split ring closely approximates the circumference of an un[‑]split ring.

For each diameter, you have a circumference C_T0 and C_T1. Your adhesive has a thickness, an elastic modulus and a shear strength.

Can you retain with a thick layer of something flexible, like RTV?

Can you split your ring, and apply a small dab of adhesive opposite to the split? This works better if your ring is a little oversized, and it sits in an orthogonal pocket.

--
JHG

 

[Compositepro]

The confusing issue with this problem is that size matters. Whether or not there is going to be a problem depends on the scale of the parts. You cannot look at an unscaled drawing and say the issue is the same regardless of if the dimensions are in microns or meters.

An epoxy adhesive can easily bond a 1 meter square of aluminum foil to a 1 meter square of 1 cm thick steel plate through a 100C temperature cycle. Yet, there is no adhesive that can work if the aluminum is also 1 cm thick.

A better example might be bonding aluminum foil to steel foil in alternating layers. No problem.

Now try bonding alternating layers of one inch thick plates. It cannot be done. The shear stresses between plates is greater than any adhesive can bear.

You have to consider the exact stresses and strains each material will will experience due to the particular geometry in each case. At the ends of bonded joints tensile and compression forces are zero (by definition) while shear stresses are at maximum. This is the reason why splits in the ring might help.

 

[3DDave]

Strain, as a measure of deformation, is already on a per-area or per-length basis, so it's mostly correct to say that the extent doesn't matter.

However, boundary conditions are a different matter.

The unrestricted/unbonded top surface of that plastic ring is attempting to shrink, just as the plastic is at the adhesive layer. The difference is the directly bonded material is being resisted by the adhesive's grip on the metal. This produces a torque where the plastic at the top is trying to peel the ring loose.

That unrestricted shrinkage is largely resisted at the edge of the ring. Splitting it may help enough with that, but it does add more edges which means more peeling-failure initiation sites.

If the ring had a rectangular section and was fully in the tube, as a bushing might be installed, the unbonded surface would uniformly add to the tension load when it shrinks from cold.

Typically the adhesive layer is too thin to matter. If you have a layer that is 0.001 inch thick and the unrestricted delta in diameter is 0.030 inches, that's above any elastic limit one might find**. IOW, it may be worthwhile to ignore the adhesive. In addition if the adhesive CTE and elastic modulus are similar to the plastic, it won't matter anyway.

I haven't seen the delta-T expected for this application, nor if it is large enough to change the elastic modulus of the plastic.

**I've seen failures in thin adhesive joints between much more closely matched CTE metals using silicone elastomer adhesives. The problem was fixed by bumping the joint thickness to a more reasonable 0.020 inch.

 

[3DDave]

Just to add - another feature that can make things bad is thermal inertia/thermal conduction. If the temp changes rapidly the plastic may not respond as quickly as the metal part, making the strain difference even larger than expected.

 

[elinah34]

The temperature change is from 25 Celsius to 50 Celsius, so it's not so dramatic. But this part should carry loads (and not only itself) along its service, and that's the reason I want to make sure how much pre-stress might exist due to different CTE.
You mentioned good things I wasn't familiar with. Thank you

 

[mfgenggear]

Elinah34
Would it be possible to form a counter bore
At the end of the tube and nest the ring in it.with two surfaces to apply adhesive.
Instead of an angle.

 

[elinah34]

Unfortunately I can't machine the part

 

[mfgenggear]

elinah34
"Unfortunately I can't machine the part"
is this in the developing stage or completed parts?
it would be a simple forming operation

 

[EdStainless]

This is the worst geometry for this application, it should be a square shoulder.

Have you picked an adhesive? do you have detailed properties for it (strength and modulus)?
So take sizes at 25C and then apply your delta of 25C. Now just figure out the force to squeeze the plastic ring to the new size of the metal part. That force will be distributed across you adhesive joint.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, consulting work welcomed