Assuring the Integrity of Air-cooled Heat Exchanger Tube

to Tubesheet Welds

David Spencer
Dr. Jeremy Leggoe & Pr. Xiaozhi Hu
School of Mechanical Engineering
Gordon Fuller & Darryl Godfrey
CEED Client: BP Refinery Kwinana

Abstract

Air-cooled heat exchangers can be used for several applications on a crude oil refinery,
one of the more common uses is for cooling process fluids. These heat exchangers are
generally of a single pass design which consists of two large rectangular ‘header’ boxes,
an inlet and outlet connected by a number of finned tubes with air blown past them by a
fan. Due to the simultaneous failure of several welded tube to tubesheet joints in a set of
high pressure heat exchangers at BP Refinery Kwinana, a better method of assuring the
integrity of these joints is being explored. Investigations on samples of the failed header
boxes concluded the damage mechanisms causing failure were a combination of
corrosion and an environmentally induced cracking mechanism (Verbruggen, 2007). This
created a desire for methods of obtaining two sets of joint quality information; a
determination of the extent of the external corrosion of the joint, and a method of
detecting any cracks originating from the inaccessible weld root. This project includes an
investigation into the stress distribution in the joints by means of finite element analysis
and testing on a sample header box manufactured to investigate the information that can
be acquired utilising different Non-Destructive Testing (NDT) methods and devices.

1. Introduction
1.1 Air-Cooled Heat Exchangers

Air-Cooled heat exchangers use finned tubes to transfer heat from a fluid contained in the
tubes to air which is forced past by a fan. The most common applications for fin fan heat
exchangers are process fluid coolers, condensers and air pre-heaters. This project is focussed
on the connection between the finned tubes and the long rectangular fluid inlet/outlet boxes
known as header boxes. These header boxes consist of a tubesheet, a plugsheet with top and
bottom plates and end plates completing the box construction. There are also stay plates
parallel to the top and bottom plates for structural integrity. The tubesheet is a usually steel
plate with holes into which tubes are inserted then expanded and/or welded to create a sealed
joint, the plugsheet has the same number of holes as the tubesheet but they are threaded and
during operation are sealed using gaskets and removable plugs. These plugholes are the only
inspection access after fabrication of the heat exchanger and are similarly the only access
when joining the tubes to the tubesheet during fabrication as the header box is completely
constructed and heat treated before the tubes are joined.

7
CEED Seminar Proceedings 2009 David Spencer: Heat Exchanger Joint Integrity Assurance

Figure 1: This diagram shows a single pass air-cooled heat exchanger tube
bundle and inlet/outlet header boxes, (American Petroleum
Institute Standards, 2006).

These heat exchangers have highly varied operating conditions, particularly internal pressures
and temperatures but also differing process fluids. Industry standards use these factors to
determine the minimum specifications of the header box materials, thicknesses and tube to
tubesheet joint types.

1.2 The Failure

In June 2005 a set of air-cooled heat exchangers on the BP Kwinana Refinery suffered
failures in several of the welds joining the tubes to the tubesheet. These heat exchangers were
part of the Hydrofiner 2 processing unit, a ‘hydrotreating’ process unit which primarily
removes sulphur from the stream by adding heat in the presence of hydrogen and a catalyst.
This causes a reaction in which the sulphur from the stream is converted to gaseous hydrogen
sulphide and is removed as part of the reactor effluent. These heat exchangers are used to cool
a mixture of wash water and reactor effluent, this fluid is comprised of a variety of harmful
and corrosive gases approximately 3% of which is Hydrogen Sulphide. In these heat
exchangers the fluid is under high pressure, greater than 6300 kPa gauge pressure and up to
136 oC at the inlet side of the exchanger.

In this and many other air-cooled heat exchangers on site the tubes have a 25.4 mm outer
diameter, while the minimum diameter of the plugholes ranges from 24mm to 34mm
depending on the type of tube to tubesheet joint. The heat exchangers with welded joints have
the larger plugholes at minimum 34mm as these plugholes are not only used for inspection
but also as access when welding the tube ends during manufacture, where smaller plugholes
exist the tubes are usually only expanded into the tube holes. The magnitude of the internal
pressure contained by these specific heat exchangers combined with a hazardous process
fluid, causes the requirement for the thickness of the tubesheet and plugsheet to be 40mm, the
tube end joint required is a seal weld. This creates problems during welding as a tubesheet of
this thickness acts as a heat sink causing the welded steel to cool very quickly, dramatically
increasing the final hardness of the weld.

Cracks in individual tubes are reasonably common, and the action taken is to plug the tube
ends to individually prevent flow through cracked tubes. This prevents leaking tubes and a
hydrostatic test is then performed to assure the seal of the plugged tube. However in 2005 the
cracks occurred in welds between the tube and tubesheet, making them undetectable by the
Non-Destructive Testing (NDT) methods used and not treatable by plugging the tube end.
This caused a lack of containment incurring significant financial penalties due to unit
downtime and for the reduced production during the manufacture of new heat exchangers.

8
CEED Seminar Proceedings 2009 David Spencer: Heat Exchanger Joint Integrity Assurance

The failed heat exchangers were replaced by six new heat exchangers consisting of two new
header boxes and 213 new tubes per exchanger. The new tube to tubesheet joints were
upgraded from seal welds to multi-pass strength welds. The multi-pass welds have a lower
hardness as the maximum temperature during welding is lower because the weld is split into
multiple passes between which the joint is allowed to cool. The upgraded joints also required
the tubes to be expanded into the tube hole behind the weld, this was to absorb stress caused
by any tube vibration that would encourage the progression of cracks in the tube end welds.

2. Processes
This project includes the use of several approaches and techniques to develop a successful
solution to the situation including the development of a finite element model in order to
analyse the stresses occurring in and around the tube to tubesheet joints. Also a sample header
box was designed and fabricated as a part of this project, testing on this and a sample of a
failed header box will assess the quality of the information that can be obtained utilising
different inspection methods and devices.

2.1 Development of a Finite Element Model

A finite element model has been created utilising Abaqus/CAE version 6.8. The Current
model makes several broad assumptions including constant material density and stiffness.
This model was created to get an idea of the magnitude of the stresses acting in the joint and
their distribution during operation.

In the model the weld is modelled by merging the element nodes, behind the 2mm deep weld
there is a 0.1mm gap between the tube and tubesheet. The loads included in the model are the
internal pressure which acts on the inner surface of the tube and the front face of the
tubesheet. Gravity is modelled by applying a uniform load acting on the nodes of the tube
which are unsupported by the tubesheet, creating an exaggerated gravity load.

2.2 Sample Header Box

The Sample header box is not a closed box instead it has open ends and a detachable front
face to allow testing on tube to tubesheet joints with and without the access constraints of the
actual situation. The Sample header also provides a sample of the joints as they were before
being commissioned for service, this enables a comparison between the joints at manufacture
and failure. There are four different tube joints included on the sample representing the tube
end joints found on air-cooled heat exchangers on the Kwinana Refinery. Each tube to
tubesheet joint is modelled on two tube ends, the joints from left to right are; seal welded,
strength welded and expanded, seal welded and expanded, expanded only.

Figure 2: This diagram shows a exploded schematic of the Sample header box.

9
CEED Seminar Proceedings 2009 David Spencer: Heat Exchanger Joint Integrity Assurance

2.3 Non-Destructive Testing Investigation

An investigation into NDT methods and equipment was undertaken to determine possible
techniques for indicating the extent of joint corrosion and for detecting cracking originating at
the weld root.

The Potential techniques identified for crack detection were Radiography, Eddy current
testing, Magnetic Particle Inspection, Ultrasonics, Dye Penetrant Testing and Vibration
Sensitivity testing. All of these types of NDT are capable of detecting cracks however major
issues were identified with each of these methods in attempting to detect the cracking
occurring in this situation.

Magnetic Particle inspection, Dye Penetrant and eddy current testing can only reliably detect
surface cracks. This is a fatal flaw for this application as the surface from which the cracks are
originating is completely inaccessible, as it exists approximately 38mm down a 1mm gap
between a steel plate and tube.

Radiographic, Ultrasonic and Vibration Sensitivity techniques are all capable of detecting
non-surface cracking, due to the geometry of this application a meaningful Radiographic
image is unobtainable. Ultrasonics presents several techniques often used for testing inside
long tubes and thus access is not a major issue. Unfortunately the techniques used aren’t
capable of reliably detecting cracks as the readings obtained are averaged over the tested area,
usually of approximately a square centimetre. Vibration sensitivity crack detection requires
previous knowledge of the modal frequencies of the entire heat exchanger, obtaining this
information would require the construction of a very large and expensive model. This
combined with the potential harm and detection reliability terminated investigation into this
crack detection method.

For assessing the extent of the corrosion of the tube end joint several potential techniques and
devices were identified. Prior to the commencement of this project, a bore gauge was
purchased in the hope that there may be detectable increase in tube bore readings. Another
potential method identified was taking a measurable mould of a tube end joint capable of
providing information on the geometry of the entire joint. Two other potential methods
included the use of a pin profile gauge or laser profilometry technology in order to gain an
accurate indication of the geometry of the tube to tubesheet joints.

3. Results and Discussion

3.1 Finite Element Analysis

The results of analysis performed on the finite element model thus far have identified a stress
concentration at the weld root, these results are consistent with the findings of examinations
performed on samples of the failed header boxes (Verbruggen, 2006). The model can be
further refined in multiple areas, one of these is modelling a local increase in hardness in and
around the weld. Including this variable hardness or more loading conditions will increase the
accuracy of the maximum stress estimation in the tube end joint.

10
CEED Seminar Proceedings 2009 David Spencer: Heat Exchanger Joint Integrity Assurance

Figure 3: This diagram shows the stress distribution around a tube end joint

3.2 Header Testing

Thus far the testing methods performed on the sample header box include moulding and bore
gauge testing. The bore gauge and a pin profile gauge have both been used to test a sample of
the failed header box from 2005.

Comparing the results of the bore gauge on the sample and failed headers indicates an average
increase in maximum bore diameter of approximately 0.1mm, a value representing the total
increase in bore diameter from fabrication to failure. As this device measures the bore of the
tube, it is therefore incapable of detecting preferential corrosion on a single side of the tube.
These findings indicate this device will be incapable of producing results that are able to be
used to accurately determine the condition and remaining service life of the tube end joints.

A pin profile gauge is capable of quantifying the extent the tube end joint remains protruding
from the front face of the tubesheet. Pin profile gauge testing on a failed header box sample
found values ranging from 1.0mm to 0.0mm protruding joint remaining. As the weld
procedures specify 3 millimetres of tube to be protruding at fabrication, this method of testing
has the potential to produce a valuable indication of tube end joint integrity. A major issue
with this device is that it is only capable of testing a single location at a time. This means
testing with such a device is very labour intensive, as it takes a minimum of eight readings to
achieve an accurate measure of the lowest remaining weld height on each tube end joint.

Figure 4: Design for a pin profile gauge testing a undamaged tube end weld.

11
CEED Seminar Proceedings 2009 David Spencer: Heat Exchanger Joint Integrity Assurance

The cured joint moulds give an accurate assessment of the geometry over the full
circumference of the tube, and preferential corrosion of the joints would also be identifiable.
The quantifiable results of the moulds are less accurate than those of the bore or pin profile
gauges due to the low stiffness of the cured mould material and 0.5% linear shrinkage of the
material. The other major issue with the use of moulding to obtain a geometric representation
of the joint is the curing time, if the mould curing time is 24 hrs and two mould devices are
used, only four tube joints can be examined in 48 hours.

Laser profilometry potentially provides a quick and accurate technique capable of measuring
the bore of the tube, detecting pits and preferential corrosion at the joint. With some slight
modifications it may also be capable of measuring the profile of the tubesheet surface around
the joint, providing accurate digitalised results of the protruding joint material remaining. The
cost of this equipment is one of the deterring factors.

4. Conclusions and Future Work

Laser profilometry and a limited access pin profile gauge will be trialled on the sample and
failed header boxes to assess the quality of the results in estimating the remaining service of
the tube end joints.

By the further refinement of the finite element model the accuracy of the maximum stress
occurring in the joint can be improved. The model can be enhanced by the inclusion of the
variable hardness over the weld and heat affected zone and by creating an interference fit
between the tube and tubesheet to model the expansion of the tube into the joint. The
inclusion of a vibration load acting on the tube would be a more accurate model of the worst
case scenario of the loads acting on the joint.

In February 2010 the replacement heat exchangers are due for inspection, this opportunity
will be taken to assess the condition of the tube end joints utilising effective testing devices
and techniques indentified in this investigation. An Inspection plan will also be developed
outlining the procedure for the use of these devices and the rejection criterion.

5. References
American Petroleum Institute Standards. (2006). Air-Cooled Heat Exchangers for General
Refinery Service (ANSI/API Standard 661 Sixth Edition, February 2006). Retrieved January
15, 2009, from the IHS database.

Verbruggen, H. (2006). Examination of C3904E Fin Fan Tube Sheet Samples.

Report 6B5-MET24, November 21, 2006, Mayfield Engineering Pty. Ltd. Welshpool, W.A.
ABN: 22112829624

Verbruggen, H. (2007). Examination of a Cracked C3904E Fin Fan Tube Sheet Weld Sample.
Report 7B5-MET5, April 2, 2007, Mayfield Engineering Pty. Ltd. Welshpool, W.A.
ABN: 22112829624

Verbruggen, H. (2007). Examination of C3904E Fin Fan Tube Sheet Samples.

Report 7B5-MET12, July 13, 2007, Mayfield Engineering Pty. Ltd. Welshpool, W.A.
ABN: 22112829624

12