High Cycle Thermal Fatigue 1

[fastphantom]

Hi
I have an application that in normal operation would cycle from ambient to operating temperature - approx 450-500 deg C, and should sit at that temperature for a month or two, come off for maintenance and be turned back on. It could have between 10 and 20 thermal cycles per year. This under normal circumstances is fine. The components over time suffer some cracking which slowly propagates and would normally need replacing every 10-15 years.

This is all okay, however the operating regime of the equipment over the last 18 months has had to change and now the units go through thermal cycling up to two times a day (300 to 600 cycles per year). The result is the component cracks beyond repair in under a year and requires complete replacement. During that year, cracks are welded, and rewelded and patches put in place where chunks of material have been lost.

The original material is 304 Stainless Steel.

Can anyone provide other material options that would be better suited to high thermal cycling fatigue.

Thank you

 

[metengr]

Material options is not necessarily the correct approach. I would evaluate design options to determine if I could redistribute the stresses or determine if the peak stresses are more severe. By the way this is still low cycle fatigue damage.

 

[EdStainless]

1. Use thinner material and allow distortion. The metal will expand, you can either try to restrain it and create very high stresses or simply allow a bit of deformation
2. Eliminate sharp corners and places where materials meet at 90degs, you want weaker joints that will have some flex and not be highly braced.
3. If you want to change materials be careful. While a ferritic stainless (439) has lower thermal expansion it will also loose ductility with high temp exposure. Its properties at temp will be fine, but when cooled it will be brittle, so heat up can be a real issue. There are no good alternatives if you need corrosion resistance (adding Cr to steel increases the thermal expansion).

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[weldstan]

While other alloys like Nickel based alloys might prove more effective, as both metengr and Ed have stated, a change in design is the best way to achieve longer life regardless of the alloy selected. Avoid sharp material thickness changes, allow for even heating and cooling rates, etc.

 

[davefitz]

Both ed stainless and meteng are correct, but to pinpoint the best improvement you would need to define the component and the failure location.

If it is a thick walled component, then one can either reduce the rate of change of process temperature during startup and shutdown, or one could use a material with a higher thermal diffusivity , such as a ferritic material. Redesigning the component for lower peak stresses is also a valid approach , usually implies a FEA analysis.

If it is cracking near a weld, then the component geometry or weld geometry may need to change. Using a full penetration weld in lieu of a fillet weld may provide as much as a 20 fold improvement in fatigue life. Refer to the weld fatigue details in En 12952-3 tables B1,B2,B3. The gradient in wall thickness near welds can also be made more shallow at an angle less than 18 degrees.

"...when logic, and proportion, have fallen, sloppy dead..." Grace Slick

 

[btrueblood]

The temperature of operation is squarely in the sensitization range for stainless steel, which would make me worry, though you say these are cracking in a year vs. 15 years (is that data or just a design life prediction?). You could switch to a stabilized alloy like 321 stainless to see if it helps. But stress concentration reductions are the bigger bang for the buck, as others have mentioned.

 

[EdStainless]

Honestly I have seen better performance from using very low C alloys (304L with 0.015%C) than stabilized grades.
Uniform heating and cooling is the first thing to address, it may mean making it a slower and more controlled process, but investing 5min per cycle will prevent days of downtime for failures.
I have worked with heat treat fixture for years, and the more rigid it is the worse the cracking is.
I recall a T fitting where hot and cold gas streams mixed, the turbulence from the mixing resulted in hundreds of temp cycles (100C-1000C) per minute. The ended up designing a T with well radiused joints (not 90deg) and made it thinner to allow some flexing.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[Tmoose]

http://farm3.staticflickr.com/2872/9541265461_af1ecf4377_o.png

 

[fastphantom]

Wow, thanks for the quick replies. And I do appreciate all the information offered.

Should have been a little more specific in my initial question and given the application - it may change some opinions.

The application is the exhaust diffuser on a gas turbine. It is subject to enormous turbulence and like I said in its normal mode of operation it lasts 15 years before replacing. This is data - not prediction. The turbines are 25 years old, the original diffusers lasted 15 years before serious issues were found, they were replaced around the 20 year mark. The replacements were perfect right up until the change of operating them(1-2 starts per day).

So the original design is capable of withstanding the turbulence in the jet stream - proven by past performance. The firing rate of the turbine hasn't changed - only the number of starts.

Unfortunately the design is full of 90 degree plate intersections - however welds are full penetration and weld radius has been increased to reduce stress raisers.

Can not make the item out of thinner plate sections - it will shake itself to bits.

Grateful for your help.

 

[fastphantom]

another photo

 

[EdStainless]

If this is what your exhaust look like I can imagine how bad the burner cans look.
So they last for 1000-1500 starts? That may be the life that you are stuck with.
Higher O&M is part of the price of higher cycling rates.

= = = = = = = = = = = = = = = = = = = =
P.E. Metallurgy, Plymouth Tube

 

[davefitz]

OK, the cyclic thermal stresses are related to the end reactions as the exhaust ( and the engine) expand from ambient to 600 C ,with perhaps a horizontal growth of 0.3 m. The end reaction is related to the design of the exhaust seal and expansion joint. Other reinforcements used to stiffen the exhaust ( to protect from vibration damage) also contribute to high local thermal stresses, as the stiffer the structure becomes the higher the resulting thermal stresses . The mfr likely has addressed these issues on later models, so the lowest cost best fix is to ask the OEM for details on the upgrade to the exhaust structure.

If a detailed FEA model is made the structure and correctly modeled to simulate startup and shutdown evetns then one could postulate changes to the reinforcing sections to include "thermal stress reliefs" that limit the thermal stresses while retaining their reaction to vibratory forces. But it is more likely that the OEM has the deep pockets for such a analysis.

"...when logic, and proportion, have fallen, sloppy dead..." Grace Slick

 

[fastphantom]

Thanks again
Edstainless - surprisingly the cans show no difference - we are running on starts based maintenance - so based on starts the HGP and CI periods just come around quicker - we are not really seeing anything different here than we have before
davefitz - OEM have offered a new and improved diffuser, but they say based on our operating regime - which leads the world in starts based maintenance btw - they will not give any warranty - not even a month. So they are no help to us.
Played with FEA model but have not really come up with anything ground breaking - hence starting this thread here. Hoping to get something bold - or new on the material front that is just so different that it might work.
Thanks again

 

[davefitz]

I could not tell from the photo, but it looks as if there is not provided any insulation between the liner and the exhaust casing. The failure of the casing could be postponed if there is installed a flexible liner and sufficient annular insulation to reduce the rate of increase in casing temperature to 20% of the current rate. That would reduce the thernmal stress by a factor of 5 and perhaps increase the fatigue life ( due to thermal stress) by a factor as high as 40.

"...when logic, and proportion, have fallen, sloppy dead..." Grace Slick

 

[Kwan]

I would suggest having some of these cracks investigated with a SEM and test the material (strength, etc). This would tell you if you have cracking mostly due to your temp delta or are these cracks due to vibration. It will reveal crack origin, can give insight if the crack started due to damage (nick), and can determine if the exhaust gas is accelerating the cracking. Find out all of this and then the path forward may become more evident. Hope this helps.

 

[WKTaylor]

BE VERY CAUTIOUS diagnosing thermal cycles as the sole culprit. Suggest instrumentation at known fatigue damage points to look for resonance-band(s) [flutter] as engine power [air-flow, RPM, thermal rise/fall] is advanced/retreated.

I have witnessed that odd combinations of thermal strain + high airflow rates + sonic impingement + RPM can create transient resonance on low stiffness assys, when engine power rises/falls. Hence engines that stay on-line at non resonant power settings/conditions could function indefinitely with only minor wear-tear damage. However OPERATION within the known [tested/verified] resonance operational parameters for these engines is generally forbidden; and these engines must transit-thru the resonance range within seconds to avoid resonance-induced damage.

In this case, material-changes can help only to limited degrees... radical redesign and/or changes to operational procedures will pay much higher rewards relating to fatigue life improvements.

Experience: military jet-engine test facility was subject to continuous deterioration [cracking/etc] of the turning vane system ~100-Ft aft of F-100-100 engines during test runs. Poor CRES material choices [affecting weld durability] + obvious potential for all sorts of supersonic airflow + localized sonic/heating-impingement issues + crazy/arbitrary design choices for attachments by the [foreign] contractor made this ETF an nightmare to maintain... much less analyze for improvements.

Regards, Wil Taylor

o Trust - But Verify!
o We believe to be true what we prefer to be true. [Unknown]
o For those who believe, no proof is required; for those who cannot believe, no proof is possible. [variation,Stuart Chase]
o Unfortunately, in science what You 'believe' is irrelevant. ["Orion", Homebuiltairplanes.com forum]

 

[davefitz]

You should be able to easily demonstrate that the failure of the weld at the casing is due to thermal stress by attaching thermocouples to the casing, the middle of the reinforcing rings , and the OD of the reinforcing rings. Monitor the DT ( Temp of casing - temp of middle of ring) during startup and shutdown, and estimate the stress at the weld interface to be about 3* E*a*DT / ( 1-n) ( E= youngs mod, a=linear coef of thermal expansion, n= poisons ratio). The "alternating stress range" may be approximately the difference betweenthat peak stress and the peak reverse stress during shutdown. If that alternating stress range is several times yiled stress at 600 C , then the failure is due to cyclic thermal stress.

If this proves to be the case, then the DT can be reduced by adding annular insulation, as previously discussed. If that is not considered feasible, there are several major changes to the reinforcement weld that could be considered, but it is better handled by the OEM

"...when logic, and proportion, have fallen, sloppy dead..." Grace Slick