Boss Oring failure

[Piyush582]

We recently conducted a high pressure impulse cycle (50 to 4500 to 50 psig) test on a manifold assembly with boss fittings on it. The port is per AS5202-12 with MS21902J12 fitting and boss O-ring (NAS1612-12) installed in it. After about 9000 cycles at 140F, the O-ring extruded and broke down resulting in leakage through the port. The other ports in the manifolds were checked and even though we did not have leakage through the other ports, we did find O-ring nibbling and O-ring particles in the port implying the O-ring could have failed after a few more cycles. We inspected the port dimensions and the port was iaw AS5202 spec, including all tolerances and roughness. We are in process of inspecting Fitting. We already check the compatibility of the O-ring with respect to Skydrol and found to be good at 140F. What could be the cause for this failure? The unit has to go through 200,000 cycles of impulse pressure cycles with 150,000 cycles at 140F, 20,000 cycles at 212F, 20,000 cycles at 14F and 10,000 cycles at 248F.
Any kind of help, advice is highly appreciated.

 

[israelkk]

To my opinion the o-ring extruded because the metal to metal contact is lost or too low. You need to calculate the force due to inside pressure applying on the MS21902J12 fitting and then clamp the fitting using a controlled clamping torque that will produce a metal to metal contact force larger than the force that the inside pressure applies every cycle.

 

[Piyush582]

The manifold material is 7075-T73 Aluminum alloy and the fitting is 304 CRES. We did follow the recommended torque for the 304 CRES fitting. The yield for the 7075 Aluminum is more than 304 CRES so I assume the recommended torque for the fitting should be sufficient to avoid any loosening under that pressure and should not damage the Al manifold threads.

 

[israelkk]

I do not know where did you take the recommended torque for the 304 CRES fitting. Does it take into account the 4500psi pressure applying on it?

To my calculations the force due to the pressure on the fitting can go up to 5523 lbf based on the maximum sealing diameter "A" in AS5202-12. Therefore, the minimum clamping torque (without any margin of safety) is 87.3 lbf-feet based on the mean diameter between the diameter "A" in AS5202-12 to the diameter "E" in MS21902J12.

 

[Piyush582]

IsraelK, thanks for the quick response. The recommended torque values we get are from the AS18280 procurement spec for that fitting. For the -12 size, the recommended torque is 855-945 lb-in or 78ft-lb max. I agree the force on the fitting under 4500 psi is 5523 lbf but when the pressure is applied, based on the stiffness of the fitting and manifold material, the force will distribute between the fitting and manifold in the ratio of their stiffness. In this particular case, the stiffness, based on our calculation, is coming out to be approx. .54 for fitting and .46 for manifold. which implies when the pressure is applied (i.e. 5523 lbf), approx. half of that force will be absorbed by the manifold and half by the fitting. Under preload torque, the fitting is under tension while the manifold port is under compression. So, only half of 5523lbf or 87.3lbf-feet is the required min torque without any safety factor to sustain pressure of 4500 psi.

 

[israelkk]

To my beat judgement I think you mix stiffness which is how much the fitting is extended due to the clamping torque that create the tension in the fitting and the stiffness of the manifold that is compressed in the contact face of the manifold with the hex head of the fitting. The stiffness of the manifold is how much compressive "strain" the manifold sees and the stiffness of the fitting is determined by how much is the tensile "strain" the fitting sees. However, both the fitting and the manifold sees same force the 5523 lbs. You can not split the force. When the pressure is applied it releases the compressive strain in the manifold and creates a gap in the contact surface. Causing the o-ring to extrude through the gap. This is the reason why you need to clamp the fitting to the manifold with a force greater than the force the pressure pushes the fitting thereby, avoid creating a gap when the internal pressure is applied.

 

[Piyush582]

ISraleK, it seems like you are making a lot of sense. To my understanding, we are applying a preload force of 4235 lbf with 900 in-lb torque and assuming k factor of 0.2. If the pressure load is less than the torque preload, then the force will distribute between the fitting and manifold in the ratio of stiffness. But in this particular case, the pressure force is more than the preload torque, so it will overcome compressive strain (preload torque) of the manifold. The open question I have here is - if the pressure force is more than the torque preload, then will it not distribute the pressure force in the ratio of stiffness? I am currently going through the "Preloaded bolts under static loading" section of Machine design by Robert Norton. They have some examples but in all the cases, their external load is less than the preload torque. They did specify the external load will get distributed in the ratio of stiffness between the bolt and the clamped material but all their examples have external load less than preload torque. I just need to confirm that this phenomena of force distribution is not valid when the external load is more than preload torque value.

 

[Tmoose]

Have you shown the damage to your o-ring supplier?
I'd confirm the o-ring material is proper for the temperature and pressure.

First I'd determine empirically if the flanges are separating at 4500 psi.
If they are, then the tendency will be greater for the o-ring to extrude into the gap and be damaged.

With ~rectangular groove o-rings, "backer rings" are used to prevent o-ring extrusion and nibbling damage in high pressure service.

Your boss o-ring appears to live in a triangular groove, and would be forced by pressure into a 30 degree corner.
I have no idea if high pressure applications require that geometry be modified to prevent damage or not.

 

[Piyush582]

We have checked that O-ring for compatibility and the results came out good. The O-ring is per NAS1612, material controlled by NAS1613 with allowed temperature range upto 250F.
I read the Machine design section again, and I confirm that the portion of the total load which is taken by the manifold should be less than the torque preload to avoid any loose contacts. Its not that the Preload force needs to be more than the total load to avoid loosening of contacts.

 

[israelkk]

Piyush582clamped

Your confusion is that the stiffness ratio is relevant only for the "added" force to the initial constant clamping force. For example: if the manifold stiffness was 10 times the fitting stiffness the application of the added force due to the pressure will add to the fitting only 10% of the "initial clamped" force because the fitting will stretch only 10% (because of the manifold high stiffness) compared to initial stretch during the torque clamping. Thereby, add only 10% to the initial clamping force already in the fitting. This is the reason for initial clamping of bolted joints that makes them safe against fatigue during cyclic loading added to the clamped joint. If the bolts were initially tighten lightly instead of full clamping tight then with every load cycle the bolt/fitting would see the full cyclic load instead of only 10% cyclic addition to the initial clamped force.

 

[israelkk]

Just wanted to add that if the added load overcomes the initial clamped force and a gap is created the bolt/fitting will see 200% of the initial clamped force due to the loss of the joint/manifold stiffness positive contribution.

 

[Piyush582]

Let me share with you all the calculation data and then you can point where exactly I misunderstood.
The torque of 900 in-lb is applied to the fitting which created a preload force of 4235 lbf (tensile in fitting, compression in manifold). The force here will be the same for both manifold and fitting, but the deflection would be different and will depend upon the stiffness.
Now when a pressure load of 4500 psi is applied, it will generate a force of 5523 lbf based on the fitting cross section area. Since the manifold/fitting was under preload, this additional force will create the same deflection for both manifold and fitting, and the force will get distributed in the ratio of stiffness. According to our calculation, the stiffness % for manifold (aluminum alloy) is coming out to be 46% and fitting (steel) is 54%.So, 46% of 5523 lbf i.e.2540 lbf is added to the fitting tensile preload force and 54% of 5523 lbf, i.e.2982 lbf is reduced from the manifold compression preload force. The total tensile force on fitting under pressure load would be 4235 + 2540 = 6775 lbf and the total compression force in the manifold under pressure load would be 4235 - 2982 = 1253 lbf. As long as this net compression force (1253 lbf) in the manifold does not go to zero, we are good.

Are we talking the same thing or are you trying to convey something different? can you point out where am going wrong?

 

[israelkk]

You can use the stiffness to know how much the manifold will compress or decompress under the clamping and the inside forces respectfully. This decompression will manifest in a stretch in the fitting. For example: if the manifold was compressed 0.001" during clamping (4235 lbs) and the fitting stretched a little bit less (more stiff according to your calculations), then when the pressure inside system is applied the fitting continue to stretch and the manifold compression is reduced. Whem the pressure just reach the point where the load on the fitting is equall to 4325 lbs the contact force between the fitting and the manifold becomes zero. The manifold expanded 0.001" and the bolt additionally extended by same 0.001" because due to previous clamping the fitting will "feel" the decompression of the manifold first. Only after the clamping force is overcomes then and only then the fitting will be stretched based only on its stiffness. Until then when some clamping force exists the fitting will be stretched by the same amount the manifold is decompressed. This is the reason that a bolted joint is favorable only when the joint stiffness is much larger than the bolts stiffness. You need to feel the formulas you see in the engineering books. The matter is poorly explained. It shows the two lines of the bolt and the joint stiffness where the joint is much stiffer then the bolts but not really explain what happens to the joint and the bolt when an external load is applied to disengage the two clamped pieces of the joint. This is what I tried to explain in my posts. Finally, you "can not" split the force between the manifold and the fitting. Both feel same load. The stiffness ratio only dictates how much the fitting will be stretched when the pressure is applied as long as the force due to the pressure is "less" than the clamping force. When the force due to the pressure exceeds the clamping force a gap is created, the o-ring extrudes and the fitting starts to resist/carry the load due to the pressure by itself and stretched according to its stiffness only.

 

[Piyush582]

Israelkk
I agree with most of your theory and am on the same page with you, except for the "you "can not" split the force between the manifold and the fitting. Both feel same load. The stiffness ratio only dictates how much the fitting will be stretched when the pressure is applied as long as the force due to the pressure is "less" than the clamping force". The reason I would like to discuss this more is because if the pressure distribution theory is wrong, then basically I am challenging the methodology being used in our company. Secondly, I will have to apply a torque which would be more than the recommended torque value provided in the AS18280 spec. So, I hope you understand my situation and will not get pissed off by my continuous examples, evidences etc.
I have attached 2 pages of that book where they have shown with figures and also force distribution. Page 918 pretty much explain all the discussion we are having. fig 14-24 (a) shows when only preload torque is applied and the corresponding force Fi and the different deflection in both manifold and fitting, and the figure 14-24 (b) is when external load P is applied. According to this book, P will have two component Pb and Pm, where Pm goes to manifold and Pb goes to bolt. and then the last sentence of page 918 where the book has explained that if the applied force P is large enough to cause the component Pm to exceed the preload force Fi, then the joint will separate and the bolt will feel the full load of the applied load P.
Now if you refer back to my calculations, the P force is the pressure load of 5523 lbf, Pb is 2540 lbf for fitting, Pm is 2982 lbf for manifold and as long as the component Pm which is 2982 lbf less than the preload Fi which is 4235 lbf, there should not be any separation. I humbly request you to please spend some time and page 918 (attached)thoroughly. If after reading also you still think that to avoid separation the total pressure load P should be less then the Torque preload Fi, and am wrong in saying that to avoid separation Pm needs to be less than Fi, then let me know. I will consider this as the potential root cause of leakage and will initiate a serious discussion with our internal team and customers.

 

[Piyush582]

I don't see the attachment in my previous post. Let me try it again through this post. if you don't see it attached, try this link.

http://files.engineering.com/getfil...10-9f66bfcdc8f1&file=Bolt_torque-_preload.jpg

 

[israelkk]

Which version of Robert Norton are you using?

I understand your point and where it comes from but, I still have a difficult to point exactly what is bugging me. I will try to go deeper at my own pace but not now. I may find that you are correct as it may seam and it is always good to learn new things. Anyway, if the o-ring extruded that means joint separation or low contact resulting force as a result of the applied pressure which could not resist the extrusion of the seal. This can be a result of a difference between your calculated stiffness of the fitting and the manifold and the actual stiffness of the fitting and the manifold. This is why it is a good practice to have a robust design and to be on the side and clamp the joint to at least 5523 lbf. I understand that this raises a problem that the current fitting which is made of a weak material 304 CRES. You can manufacture a replacement fitting from much stronger material such a 15-5PH-H1025 or in worst case Custom 455 H1000 and even stronger metals. This may raise another problem, is the manifold material 7075-T73 Aluminum alloy can withstand the high clamp, so you may need to use longer thread or replace to stronger material. I recall from my 35 year experience working with R&D of much higher pressure manifolds, fittings, valves, etc., that there where times where we had to make the manifold from high strength stainless steel in the cost of larger weight manifold. One more thought to take into account is that because the stiffness ratio is low, with every cycle the fitting sees large force addition, the statement in page 918 "if the bolt doesn't fail when preloaded it probably won't fail at service" no longer apply because it assumes high joint_to_bolt stiffness ratio (close to 10). Therefore, maybe if you solve the o-ring extrusion problem you may find a fitting failure as a result of fatigue.

 

[Piyush582]

I think as long as the torque force does not damage the manifold, it should be fine as during pressure load the stress levels in the manifold will go down. Regarding the fitting, if I go with the worst case scenario of 5523 lbf torque force (1150 in-lb) and 5523 lbf pressure force, then the fitting tensile stress margin of safety (Y.S./Tensile stress - 1) is coming out around .74 and fitting thread shear margin of safety is coming around 2.95 considering static load (fatigue would be worse). The threads will not shear but the fitting may yield based on the lower tensile yield margin of safety.
In that case, you are right, we might have to go with 15-5 CRES steel if we decide to go with higher torque. The margin on the manifold as a result of higher torque is greater than 5 for both compression and shear, we should not have any problem with manifold.
I will check the version of the book tomorrow, its in my office.

 

[saplanti]

1. Check o-ring cavity dimensions for the o-ring diameter against the other suggested o-ring cavity dimensions, perhaps from the Parker O-Ring Handbook just to compare/verify.
2. You may need to use back-up rings for applications that may extrude o-ring from the gap which normally should not be available. Check handbooks/internet for back-up rings for o-ring applications.
3. Check the o-ring material and hardness suitability for the application and conditions.
4. Sometimes the installation can be problem. You may be damaging/scratching o-ring without even noticing, check your procedure for installation.

 

[Piyush582]

The boss O-ring gland is pretty much standard and Parker refers to the same spec for O-ring cavity dimensions.
The backup rings are not required in this case as it is backed up by metal. The O-ring gland is created by metals (fitting and manifold) on all side. As long as the metals stay in contact, o-ring cannot squeeze out.
Already checked the O-ring material. There is only one hardness for this NAS1612 O-ring which is compatible to skydrol. If we want to go to higher hardness, then we are deviating from the spec. Then entire Aerospace industry use this O-ring for Skydrol high pr application.
Already building tools for installation to avoid O-ring contact with fitting threads.

I am more and more inclined towards 304 cres Fitting yielding as the root cause of leakage. Spending a lot of time in fine tuning calculations for fatigue analysis

 

[saplanti]

This may be helpful (I guess this is the area that you work):

https://en.wikipedia.org/wiki/O-ring_boss_seal