Convection Section Failure Analysis and Fitness-for-Service Assessment for an Ammonia Plant
By: James R. Widrig, Quest Integrity Group and Neil Schroetlin, Dakota Gasification Company
As published by the 2014 AIChE Annual Safety in Ammonia Plants and Related Facilities Symposium - Download the PDF version.
In-service equipment failures present a considerable challenge to reliability engineers. Since these events are unplanned, the spare parts and resources to quickly and efficiently make repairs may not be readily available. Facility management is faced with making decisions to perform short-term repairs to equipment to safely operate until repair parts, equipment, manpower and product supply permit an extended outage and long-term repair. In these instances, the reliability engineer will need to answer the question, "We had a failure of convection tubes in our furnace. Can we restart and continue to operate until our next planned shutdown?”
Dakota Gasification Company owns and operates the Great Plains Synfuels Plant located near Beulah, North Dakota. The Synfuels Plant is the only commercial-scale coal gasification plant in the United States that manufactures natural gas. Each day the Synfuels Plant converts approximately 18,000 tons of lignite coal into an average 145 million cubic feet of synthetic natural gas for home heating and electricity generation. In addition to synthetic natural gas, the Great Plains Synfuels Plant produces numerous products from the coal gasification process that have added great diversity to the plant's output. One of the products produced is anhydrous ammonia; Dakota Gasification Company has the ability to produce about 400,000 tons per year.
The processing steps used to produce anhydrous ammonia are (see Figure 1):
(1) Rectisol synthesis gas, air and steam are heated by a fired heater using synthetic natural gas as a fuel. These heated gases are introduced into (2) the reformer where the methane is converted to carbon monoxide, carbon dioxide and hydrogen. The exit gases from the reformer enter (3) shift conversion where the carbon monoxide and water react to form hydrogen and carbon dioxide. In the (4) carbon dioxide removal vessel, an absorption process is used to remove the carbon dioxide. The stream from the carbon dioxide removal system still contains small amounts of carbon monoxide and carbon dioxide so it is sent through (5) the methanator where they react with hydrogen to form methane. Before the stream can be introduced to the ammonia synthesis loop, the (6) dryers and cold box remove water and methane along with excess nitrogen so that the hydrogen and nitrogen ratio equals three to one. (7) The synthetic gas compressor takes the fresh feed from the cold box plus the ammonia loop recycle stream and compresses it to 2,500 psig. As the stream travels through (8) the radial flow ammonia converter, the hydrogen and nitrogen react to form ammonia. (9) The ammonia recovery unit then cools the stream and the liquid ammonia is sent to the 30,000 ton ammonia tank where it is stored until it is sold. The recycle gases are then returned to the synthetic gas compressor for reintroduction into the ammonia converter.
Figure 1 - Dakota Gasification Anhydrous Ammonia Process Diagram
The Ammonia Plant Feed Gas Heater, BA-4403, is a multiple service, balanced draft, single cell, vertical cylindrical heater. The convection section preheats Rectisol synthesis gas (syngas), air and the syngas/steam mixture in separate convection section coils prior to conversion in the reformer reactor (similar in design to a secondary reformer). The heater has a radiant section with two coils. One coil is used for steam generation the second for final pre-heating of the syngas /steam mixture before reforming.
Tube Failure and Analysis
On April 15, 2012, the Ammonia Plant Feed Gas Heater sustained a tube failure (Tube 15V) in Lower Convection Section 1, the Lower Rectisol/Steam Coil. The failure occurred in one of the eight passes in the second row of finned tubes (from coil inlet) near the top of the lower convection section (see Figure 2, Figure 3, and Figure 4).
Figure 2 - Tube Failure at Center Support
The convection section tubes were 6.625 in. (168 mm) OD with 0.550 in (14 mm), MSW, 23.6 ft. (7.2 meters) in length, and constructed of centrifugally cast HK-40 material with solid continuous welded 304 stainless steel fins (see Table 1) .
Table 1 - Convection Tube Details
tube bundle is supported by three tube sheets (tube supports), two on the ends
and one in the middle. Subsequent
inspection identified that both end tube supports had dropped down due to
failure of the four support bolts (two for each end support). The loss of tube
support at the end tube sheets resulted in the tube coil having a high tensile
stress, with a bending and possibly a torsional component, on the top of the
tubes at the failure site(s) along the middle support. It is unlikely that the failure of Tube 15V
caused the supporting bolts on the end tube sheets to fail. This is due to the
location of the observed failure on the top of the tube. (The bolt failure is
discussed in a later section of this paper.)
Figure 5 - Cross Section of Fracture Surface
The tube coil had been in-service since 2003. The operating temperature near the outlet of the tube is ~1,255°F (680°C) and the maximum design tube wall temperature is 1,482°F (806°C). This is sufficient time at temperature to form sigma phase in HK-40 castings. Also, welds typically have high ferrite content and are therefore susceptible to sigma phase formation. The base metal microstructure showed no evidence of prolonged overheating above the design temperature.
Repairs and Testing
Dakota Gasification Company (DGC) had an inventory of replacement tubes for the convection section to allow for immediate replacement of several tubes. During the removal of the damaged tube and an adjacent tube (found to be cracked on the outside), no additional tubes were found to be cracked by visual inspection. However, the full extent of the tube cracking was unknown from the inspection due to the configuration of the convection section and the use of finned tubes. Visual or NDT testing of the tubes would be difficult or impossible without the removal of a substantial number of tubes. Visual inspection, dye penetrant inspection or a proof hydrotest would only locate tubes which had through wall cracks or tubes where cracking could be located on the external surface. Existing cracks which were not through wall or initiated from the ID could not be found with confidence using these conventional methods.
The Failure Assessment Diagram (FAD) Method
Rather than proceeding with a full repair, a hydrotesting procedure was developed using fracture mechanics and a failure assessment diagram (FAD) analysis to proof test the remaining tubes (see Figure 6). FAD is a fracture mechanics model for assessing crack-like flaws. A series of calculations are performed using the component geometry, crack dimensions and material properties. These calculations result in an assessment point on the FAD. Failure is predicted if the point falls on or outside of the curve. The crack is considered to be stable if the point falls inside of the curve. The FAD assessment method is implemented in API 579, Part 9, Level 2.
Figure 6 - FAD Diagram
The proof hydrotest was to be performed at a much higher pressure than required by API 530 for testing new tubes during construction. The desired pressure would be high enough that existing crack-like flaws of a critical size or greater in the tubes would propagate to failure under the hydrotest conditions. The maximum pressure would also be limited by the API 560 limitation of 90% of yield strength at ambient temperature for testing (see Table 2). In theory, the tubes that survive the hydrotest pressure could then be assumed to have existing crack-like flaws that would not fail in operation.
Table2 - Comparison of Test Pressures
The following are example calculations and assumptions used to define the test conditions:
Case 1: Hydrostatic Test at Ambient Temperature.
Given the assumed fracture toughness, the 5 in. (127 mm) x 0.3 in. (7.6 mm) crack is borderline for failing or passing the proof hydrotest (see Figure 7). Note that this result does not mean that the proof hydrotest will definitely identify cracks larger than 5 in.(127 mm) x 0.3 in. (7.6 mm) The actual toughness of the HK-40 weld is unknown at this point, and it may be significantly higher than assumed. Therefore, it is possible that larger cracks may survive the test.
Figure 7 - FAD Diagram, 5 in. x 0.3 in. Crack
Under Hydrotest Conditions
Case 2: Operating Conditions
In this case, the assessment point falls well inside of the FAD (see Figure 8). Therefore, if the convection tubes survive a 3,500 psig (24.1 MPa) proof hydrotest at ambient temperature, any cracks that remain will not lead to failure (in the near term) under a reasonable operating scenario.
Figure8 - FAD Diagram of 3 in. x 0.3 in. Crack Under Operating Conditions
The proof hydrotesting of the furnace tubes was accomplished by testing each of the eight passes in the convection coil at 3,500 psig (24.1 MPa) and holding the pressure for one hour. During the testing, one additional tube failed at 900 psig (6.2 MPa) and was replaced. The tube was removed and the failure location was preserved for failure analysis.
Upon completion of the hydrotesting, the furnace was successfully returned to operation on May 20, 2012. However, these repairs were considered temporary.
When Should We Plan for Replacement of the Tubes?
The initial tube failure event likely resulted in additional crack-like flaws in surrounding tubes. Further failure analysis and engineering assessment were planned to fully understand the root cause of the failures and to determine the remaining life of the convection coil tubes.
The desired outcome of the failure analysis and engineering assessment was to determine if the convection tubes were fit to operate until either of two future shutdown opportunities in November 2012 or June 2013.
In order to accurately calculate critical flaw sizes, material properties such as toughness, yield strength and ultimate tensile strength must be identified. These material properties were obtained by destructive tests from tube samples removed from the heater.
Maximum Possible Existing Flaw Sizes
Maximum possible existing flaw sizes in the tubes were calculated using the material properties data obtained from the ex-service tube. The analysis used Quest Integrity Group’s commercial software Signal™ Fitness-For-Service, a Windows®based program that uses the calculation guidelines contained in the British Standards BS-7910 and API 579/ASME FFS-1 .
Critical Crack Size Curves
The result of this analysis is a plot of flaw depth (a) versus flaw length (2c) for an assumed elliptical surface flaw. Additionally, it is possible for a 360 ºsurface flaw to exist in the tubes. This analysis was also done in order to identify a critical depth for a 360ºsurface flaw. Figure 9 shows the resulting critical crack size curve. The critical 360ºflaw is plotted with a flaw length equal to the outer circumference of the tube. These flaw sizes represent the maximum possible flaw sizes that could withstand hydrotest conditions of 3,500 psig (24.1 MPa) and a temperature of 70 °F (21°C).
Figure 9 - Critical Existing Flaw Sizes, Calculated from Hydrotest Conditions
Any combination of flaw depth and flaw length that falls on or below the curve can exist in the tubes. The actual possible crack configurations could not be narrowed down beyond this due to the limitations of "back-calculating” an existing flaw size from hydrotest conditions.
Extent of Existing Flaw Size
In theory, it is possible that a long and shallow crack or a 360ºsurface crack would exist in the tubes that passed the proof test. However, a qualitative conclusion regarding which flaw length/depth combinations are more likely to occur can be reached based upon the axial stress profile of the tubes.
The axial stress due to bending at the top of the pipe cross-section is tensile, while the axial stress due to bending at the bottom of the pipe cross section is compressive. This results in the overall axial stress that tends to grow a crack being higher at the top of the pipe cross-section than at the bottom leading to the identification of the likely and unlikely crack growth scenarios shown in Figure 10.
Figure 10 - Possible Crack Growth Scenarios for Existing Flaws in Convection Tubes
Another factor that supports the idea that the existing flaws in the tubes are more likely to be shorter and deeper can be drawn from the probable cause of the crack formation. The existing flaws were probably initiated and propagated significantly when a tube sheet support failed and overloaded the welds in the tubes. Alternatively, the crack may have grown after the tube sheet failure due to tube vibration. Such a scenario would cause the crack to break through at the 12 o’clock position rather than grow as a circumferential surface crack. The cracks observed in the sampled tubes that failed the hydrotest at 900 psig supports the hypothesis that cracks will be shorter and deeper rather than longer and shallower.
Remaining Life of Maximum Possible Existing Flaw Sizes
Metals subjected to high temperatures and stresses will creep over time. Many methods exist for characterizing creep behavior. The work described in this paper utilize the MPC Omega method which allows calculation of the creep life, strains and strain rates and incorporates the acceleration of creep toward the end of life.
Creep Crack Growth Properties
Using the material specific Omega law properties identified in the previous section, a creep crack growth rate (da/dt and dc/dt) was calculated. The creep crack growth rate is modeled using a power law equation:
The quantity Ct is referred to as the "power release rate” and is derived from the stress intensity factor. This value is calculated from the stress profile at the crack tip. The H and q quantities are constants.
The remaining life was calculated for eight of the critical flaw sizes (2c vs. a pairs) identified from the hydrotest results.
Table3 lists these flaw sizes.
See Figure 11 for the corresponding crack growth versus time plots.
Figure 11 - Crack Growth Versus Time for the Maximum Possible Existing Flaw Sizes
These curves terminate at a point that represents 100% of the remaining creep life. It is recommended that the 80% remaining life be used to determine the remaining life of the flaw. This is due to the fact that towards the end of life the crack growth becomes very rapid. Once the crack begins to grow more rapidly, any change in operating conditions could cause the crack to become unstable and lead to sudden failure. Table 4 provides the remaining life values for the eight critical flaw sizes analyzed.
Table 4 ”" Remaining Life for Critical Flaw Sizes
Any surface flaw with a length, 2c less than or equal to 7.35 in.(186.7 mm) (127°around circumference) would operate safely until June 2013 while any surface flaw with a length, 2c less than or equal to 8.22 in. (208.8 mm) (142° around circumference) would operate safely until November 2012. Any flaw with a size greater than this would become unstable before the scheduled November 2012 shutdown.
Assessment of Failed Bolts
The ultimate cause of the tube sheet failure was a failure of the bolts that supported the tube sheets. See Figure 12for a picture of one of the failed bolts. The bolts were originally designed to withstand a temperature of 680 °F (360°C). Even though there was insulation surrounding the bolts, the fact that the tube sheet was in direct contact with each bolt means that heat will be conducted to the bolt. Subsequently, the bolts were most likely operating at a temperature closer to 1,300-1,400 °F (704-760°C).
The bolts were 1 in. (25.4mm) diameter B8M (SS 316) with the "threads included” configuration.
The bolt assessment consisted of 13 different permutations. These permutations used static strength based criteria (yield strength, ultimate tensile strength), creep strength based criteria (10,000 hour and 100,000 hour creep strength) and an allowable stress criteria (from ASME, BPV Section VIII Div 2) for temperatures of 680 °F, 1,300 °F, and 1,400 °F (360, 704, and 760°C). The design should be based on satisfying both static and creep strength criteria. All but one of the permutations (ultimate tensile strength at 680 °F (360°C)) showed that the bolts were loaded beyond the allowable capacity. This assessment leads to the conclusion that the bolts were not adequate for the temperatures which they were exposed to in operation.
Figure 12 - Failed Tube Support Bolt
A fractured finned convection tube (Tube 15V) from a convection coil failed down the center of a circumferential tube weld which was embrittled by sigma phase (Ïƒ).
Both end tube supports for the convection coil had dropped down due to failure of the four support bolts (two for each end support). The loss of tube support at the end tube sheets resulted in the tube bundle having a high primarily high tensile stress on the top of the tubes at the failure site(s) along the middle support.
In order to safely resume operations and avoid a potentially long repair timeframe, hydrotesting at a proof pressure was used to provide reasonable assurance that tubes in the convection section would not fail upon restart or in the near term operation.
Further failure analysis and engineering assessment were performed to fully understand the root cause of the failures and to determine the remaining life of the convection coil tubes.
The failure analysis and engineering assessment results were successfully used to evaluate if the convection tubes were fit to operate until either of two future shutdown opportunities in November 2012 or June 2013.
The methods discussed in this paper exemplify how failure analysis and fitness-for-service provided the operator a sound basis to evaluate the business and safety risks of continued operation until long-term repairs and system improvements could be implemented.
The American Petroleum Institute and The American Society of Mechanical Engineers, Fitness-for-Service API 579/ASME FFS-1 (API 579 Second Edition). © API Publishing Services June 5, 2007.
ASME (2010), "ASME Boiler and Pressure Vessel Code”, Section VIII, Division 2, the American Society of Mechanical Engineers, New York, NY.