Heat Exchanger tube Failure mechanism

[yamoffathoo]

I am very interested in a failure mechanism that is associated with low and intermittent tube flow that causes tube failure and tube movement. This has been observed in verticaly oriented, u-tube heat exchangers and requires a hot shell (secondary) side, cold tube (primary) side, liquid medium, a thick tubesheet and double rolled tubes - each roll is 1.25 x tubesheet thickness and located adjacent to the secondary and primary tube sheet faces.

The tube flow is comprised of a constant tempering flow in parallel with a control valve that responds to shell vapour pressure. The cooling flow requirement is slightly more than tempering, therefore, the valve modulates cyclically, alternating between a minute of low tempering flow to 10 seconds of flows 20x higher.

My theory is that, as a result of this flow transient, the following mechanism occurs:

- during periods of low tempering flow, the radial thermal gradient through inlet tubes and tube sheet is small due to the insulating laminar heat transfer coefficient
- when flow quickly increases, the tube gradient increases sufficiently to relax the upper rolled joint and permit axial tube contraction reaction forces to develop between the lower rolled joint and the seal weld.
-by the time tube contraction reaches the lower rolled joint, the upper tube sheet has cooled and locks in some residual tensile tube strain.
- tensile stress in the tube section between upper and lower rolled joints is released when the lower rolled joint relaxes.
- after this 10 second cooling cycle, there follows 60 seconds of tube sheet heating
- with 500,000 cycles occurring each year, cracks are initiated at stress risers in the root of the fillet/seal weld and propagate at 45 degrees from tube OD to ID along the heat affected zone adjacent to the fillet weld.
- when the tube has cracked circumferentially through wall, it is free to inch-worm down the tube sheet in response to the 'thermal peristaltic' action.

Does this sound plausible, and would eliminating the double roll by rolling the section between, prevent tube cracking and movement (if the flow transient could not be changed)?

 

[GenB]

Not an easy task you are proposing.
you should look at old technology: locomotives,
they had the same problem, most adopted a certain tube rolling technique called... you have to look at Sect. I, they still mention it there. I got very interested and I called the retrofit shop at the Museum in Sacramento CA.

 

[yamoffathoo]

Section I PFT-12.2.2 mentions expanding tubes using the Prosser method. Could you point me to a good source of historical information on this method? Would you have a contact at the Museum in Sacramento CA? Is it the re-rolling 'task' that you have reservations about?

 

[GenB]

I will see if I still have their phone #
you can try calling the shop and talk to a senior boiler maker.
they will treat you well.

 

[davefitz]

sounds like a classic ledineg static flow instability problem.

Vertical , heated u-tube configuration will have flow instability unless inlet orifice ( or capillary tube) frictional pressure drop is added to each tube.

research older ( 1950's) topic of heated parallel channel instability or ledineg instability. Or add inlet capillary tubes or an efective inlet orifice.

 

[yamoffathoo]

Ledinegg flow instability is observed in two phase flow regimes - this application is single phase. There was some interest in single phase flow mal-distribution and instability during thermal transients the late 70's early 80's ref Argonne National Laboratory ANL/NE-07/21 paper on this subject, however, no subsequent work.

 

[davefitz]

to the contrary, ledineg instability can also occur in single phase systems with low friction as well. Also, there are several published cases oF supercritical fluids undergoing the ledinegg instability - I believe the First BW supercritical unit ( Philo) and at least one russian supercritical unit had such instabilities in their superheater sections during startup operations ( when the fluid is yet cool, and may be entering the u-tube platen or penedant superheater elements at below the psuedo-critical point.)

 

[yamoffathoo]

In any case, the Richardson number (R) during tempering flow is very low, approximately 3 per degF imposed temperature difference, so I believe recirculation is not the cause of tube thermal fatigue.

 

[muld0020]

The first thing you should do is pull a tube sample and have a lab analyze the circ. cracks. There are a lot of plausable failure mechanisms for circ. cracking. You should note exactly where the leaks are with regard to the bundle and tubefield, and make sure there are no cracks down the length of the tube. With low flows you could be seeing hot spots on the shell side, but I doubt the differential leg expansion is generating significant additional axial loads. I would not expect differential leg expansion to be the culprit if the inner bend dia. is 4 * tube OD or greater. Axial loads on a U-tube bundle are very small (the U's act similarly to a thermal expansion loop). If thermal axial loads were the culprit it would be on the inner leg(s) only, and especially when there is a tight bend dia.

500,000 cycles a year sounds like a lot of cycles. If the equipment is older it probably wasn't designed for that kind of cycling. What kind of ramp rates does your exchanger see? If you keep heating and cooling cycles in the 100-200 F/hr your cyclic fatigue limits will greatly increase. Also, the highest stresses due to pressure loadings are in the outer most edge of the outer most tube (generally).

The 1-1/4" roller expansion is a function of the roller expansding mandrel, any longer an the mandrels tend to break. Thicker tubesheets are step expanded several times to about 1/8" from the shell side of the tubesheet. If the tubesheet is thick (6"+), it may be better to use explosive expansion (or hydro-expansion) for a uniform and gradual transition from expanded area to un-expnaded tube. Either of these expansion methods could be performed again on-site to make sure all the joints are tight, and repair any tubejoint welds that may not be holding.

 

[yamoffathoo]

Tube samples have been milled out and beach marks on tube crack surfaces indicate high cycle fatigue and the branching morphology indicates OD to ID cracking direction.

The 100 degF water in the tubes condenses 300 psig saturated steam on the shell side. Temperature gradients in the tube walls are due to heat conducted from the condensing steam, through the tubesheet into the tubes at the two rolled joints and convected into the cooling water. Thermal fatigue from changes in these temperature gradients are only due to cooling water flow fluctuations in the tubes - tempering to full flow. Shell thermal cycles are very few, limited to vessel warm-up and cool down.

Circumferential tube cracking has been observed on both inner and outer tube locations. Tube bend radii are all greater than 4 x tube OD (5/8"dia). Tube material is Inconel 800. In approximately two years of operation, one tube cracked through-wall and inch-wormed 2" down the tubesheet. More than half of the other tubes were cracked circumferentially and had moved various amounts up to 1/4".

Continuously rolling the 6" thick tubesheet with explosive or hydro-expansion is being proposed, as you recommend. The effectiveness of this repair to stop tube cracking and inch-worming would be masked by re-instating tube seal welds. The plausibility of this failure mechanism is the question...

 

[rmw]

What does "each roll is 1.25 x tubesheet thickness" mean?

rmw

 

[yamoffathoo]

I apologize, each roll is 1.25" long. The upper roll begins 1" below the primary tube sheet face and the lower roll begins 1/16" above the secondary tube sheet face.

The upper tube sheet face is cladded with 1/4" thick Inconel 800 (to which the tubes are seal welded with 1/4" fillets), while the lower face is not cladded (and not seal welded).

A few more construction details that might bear on the thermal fatigue mechanism:

- this is a four pass, ten foot long tube bundle with tube cracking and movement only occurring in the first quadrant
- each pass is 39 tubes and the first pass is encapsulated on all sides by a two foot long, pie shaped off-gas cover
- off-gases are not vented at any time during two year periods of continuous operation

 

[GenB]

i am curious, are the gasses added on purpose or are gasses produced by the fluids?
if the gasses need not be there, can you store the gasses in an added vessel?
does the problem exist at the section with accumulated gasses?
is the first pass the coolest of the four or th ehotest?
Seems to me that the movement, expand and contract of the tubes create the cracking, all will agree to that.
finding the cure:
incoloy tends to corrode and gas in the system could affect the tube at the most weak part.
design: we mke small shell steam generators and we never pay attention to expansion and contraction of the tubes because they are for lower pressure 150 psi sat.steam.
If these were for higher pressure and temp, we wold have the same problem you are having, some tubes would crack at the hotest ends due to expansion and contraction; we would have to address the problem and find a solution to avoid loosing tubes.
Since the solution can not be done by working the tube and yes, you can corrugate (bellows) each tube but then how would you insert them.
but why not corrugate(bellow) the shell, that way the expansion and contraction of the tubes will push the shell and not the tubes.
now the test would be how many cycles the shell will widstand.
I do not know that asnwer because I do not have the equipment, software and expertise. I think you do!

 

[davefitz]

yamoffathoo:

are you certain the fluid pressure on the tube side (cold water) is above 300 psig at all times? If not , the low flow / low friction cold fluid may be boiling and easily cause flow stagnation and flow instability.

 

[yamoffathoo]

Davefitz - pressure in the tube side is 1200 psig guaranteed at all times. In one vessel, tube movement resulted in tube fretting, subsequent blow-out and severe thermal shock to the tube sheet during periods of tempering flow, which permitted reverse flow in every tube except the one with the hole.

Genb - gases are produced whenever water condenses - there is no deaerator in the process. Tube cracking and movement occurs only adjacent to seal welds in the cold (inlet) quadrant. There is no possibility of thermal interaction between shell and tube bundle - the u-tubes are free to expand vertically down (axially) from the tube sheet, which is clamped between shell flanges.

 

[rmw]

I take it by your description that your tubes are not rolled full length (of the tubesheet?) That gives me some pause.

If you are not venting non condensables, then you may also have pockets in the Hx where the tubing is blanketed by the gasses and not being heated as are other parts of the tubing.

rmw

 

[yamoffathoo]

Yes, I would imagine that most off-gassing occurs where condensation rates are highest on the coldest tubes - the first pass. The two foot long shroud around the inlet quadrant tubes is designed to funnel non-condensibles to the off-gas collection system. Because this system is not in service, a gradually thickening gas barrier to condensation causes the lower tubesheet face to gradually cool.

A cooler lower tubesheet face means a tighter lower rolled joint which means higher tensile stresses generated in the tubes during the full flow thermal transients...

 

[yamoffathoo]

RMW, please explain what it is about 5/8" dia x .0445" thick SB-163 UNS N08800 tubes not being rolled full length in a 6" thick SA 508 Gr3, Class 2 tube sheet that "gives you pause".

I have been searching high and low for any evidence of this 'inch worm', 'slinky motion', 'catch and release' or 'thermal peristalsis' occurring between autofrettaged tubes and their tube sheets. My Schmidt plots indicate that the through-wall thermal gradients are steep enough and the tube contraction loads are high enough to cause slip, but how do I prove that cyclic flow transients can coordinate this interaction between rolled joints?

Perhaps there is empirical evidence in the field of reciprocating compressors?