International Journal of Pressure Vessels and Piping 81 (2004) 499–506 www.elsevier.com/locate/ijpvp
In service embrittlement of cast 20Cr32Ni1Nb components used in steam reformer applications D.M. Knowlesa, C.W. Thomasa,*, D.J. Keenb, Q.Z. Chenc a
Materials Performance Technologies, P.O. Box 31310, Lower Hutt, Ne w Zealand b Plant Reliability Solutions, P.O. Box 263, Carina, Qld 4152, Australia c University of Hong Kong, Pokfulam Road, Hong Kong, China
Abstract
Severe Severe embrittle embrittlement ment has been experienced experienced in a number number of cast manifold manifold components components.. This has manifeste manifested d itself itself as cracking cracking at tee to manifold manifold connections. connections. Attempts Attempts to weld repair proved futile futile leading leading to concern concern about the integrity integrity of the entire system. This experience experience contrasts with similar components that have successfully remained in service for many years. The paper describes the investigations into these failures and laboratory investigations into the properties of cast 20Cr32Ni1Nb alloys. Results indicate that variations in alloy chemistry within the stated allowable range are sufficient to cause embrittlement. q 2004 Elsevier Ltd. All rights reserved. Keywords: Embrittlement; Creep–fatigue
1. Introduction
Steam Steam reform reformer er furnac furnaces es areat thefrontend of a numberof numberof indust industria rially lly import important ant proces processes ses.. These These furnac furnaces es take take a supply feed of methane and steam and ‘reform’ them to hydrogen and carbon monoxide which subsequently subsequently become the basic building blocks in industries such as ammonia, methanol, DR iron production and petroleum refining. A typica typicall reform reformer er furnace furnace consis consists ts of an array array of vertic vertical al tubes tubes in a firebox. The smallest furnaces may have only ten such tubes tubes butthe larges largest, t, with with up to 700tubes, 700tubes, are very very signi significa ficant nt and capital intensive items of plant. These tubes contain a catalyst and the feed gas flows internally from the top to the bottom. Effectively, each tube behaves as a separate reactor. At the bottom bottom of the furnace furnace,, the various various tubes are all connec connected ted to a system system of manifo manifolds lds that that collec collects ts the gas into into a single single stream stream for distri distribut bution ion to furthe furtherr proces processin sing g units. units. The operating temperatures required in these furnaces are high. Skintemperatureofthereformertubesisapproximately850– 950 C and gas outlet temperatures are around 760–850 C. These These temper temperatu atures res and the need need to operat operatee reliab reliably ly for scheduledcampaignsof scheduledcampaignsof possiblyfive possiblyfive yearsput hugedemands on the materials of construction used in these furnaces. The present paper addresses problems encountered in the outlet manifold system. Fig. 1 shows a schematic illustration 8
8
* Corresponding author. Tel.: þ 64-4-569-0027; 64-4-569-0027; fax: þ 64-4-569-0431.
[email protected] (C.W. Thomas). E-mail address:
[email protected] 0308-0161/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijpvp.2003.12.025
of a reformer and an outlet manifold system. Traditionally manifold manifold componen components ts have been manufact manufactured ured from wrought wrought alloy alloy 800H or 800HT. 800HT. As systems have increased increased in size, there has been a shift to the more economic and nominally better performing cast 20Cr32Ni1Nb alloy which has become an industry standard. This alloy is offered by a number of manufacturers using various trade names but in reality, there is little variation between them.
2. The alloy
The material is covered by ASTM Standard A351-94 where it is described as alloy CT15C. Reference to this standard, however, is rarely made and the material is more common commonly ly identi identified fied by its variou variouss trade trade names names such such as CR32W or KHR32C. This standard describes composition and manufacturing requirements but makes no reference to elevated temperature mechanical properties. The material is essentially, a cast version of alloy 800. Alloy 800 contains 20% chromium and 32% nickel with an upper limit of 0.1% carbon. It is a solid solution alloy but also contains small amounts of aluminium aluminium and titanium titanium which lead to the formation formation of carbides carbides and sometimes sometimes,, a small amount of gamma prime g 0 phase. The ‘H’ and ‘HT’ grade gradess invol involve ve mani manipu pulat latio ion n of grai grain n size size and and mino minorr variations in the carbon, aluminium and titanium content.
500
D.M. Knowles et al. / International J ournal of Pressure V essels and Piping 81 (2004) 499–506
Fig. 1. Schematic illustration of a reformer furnace [1].
Table 1 Chemical compositions of CT15C and alloy 800H
CT15C 800H Sample
%C
%Cr
%Ni
%Si
%Mn
%Nb
0.05–0.15 0.1 max 0.11
19–21 19–23 19.74
31–34 30–35 31.37
0.5 –1.5 1.0 max 1.03
0.15– 1.5 1.5 max 0.90
0.5–1.5
The adoption of a cast variant saw the use of similar levels of the prime alloying elements, chromium and nickel. Titanium and aluminium were not used in the cast variant. However, in parallel with the development of spun cast reformer tube alloys involving alloying with small amounts of niobium, the manifold alloy similarly was alloyed with approximately 1% of niobium to improve creep properties. Table 1 lists chemical composition requirements of ASTM A351 alloy CT15C and compares them with the wrought alloy 800H. The similarities are clearly shown. There are no standardised creep or stress rupture properties for the cast 20Cr32Ni1Nb alloy. Instead, designers are required to make use of data supplied by manufacturers. Alloy 800 on the other hand, has been in use for many years and stress rupture data is available from
%Ti
%Al
0.15–0.6
0.15 –0.6
0.98
a number of sources. Fig. 2 shows a comparison of published stress rupture data for the cast material (based on a manufacturer’s data) and the wrought alloy 800H [2] (based on API 530 data). According to these data, the expected life of the cast material is an order of magnitude higher than that of the wrought material at typical design stress levels around 10 MPa.
3. The problem
The problem that was encountered with this material was severe in-service embrittlement. After only relatively short periods in service, routine inspection at scheduled plant outages led to the discovery of cracking at the weldments
Fig. 2. Comparison of stress rupture data for cast alloy CT15C and wrought alloy 800H.
D.M. Knowles et al. / International Journal of Pressure Vessels an d Piping 81 (2004) 499–506
501
Fig. 3. Schematic illustration of typical manifold and bull-T arrangements.
connecting the main manifold arms to the bull-T. The bull-T typically connects the arms of the manifold to the main transfer line. Typical schematic manifold/bull-T arrangements are shown in Fig. 3. Fig. 4 shows a bull-T in position below the furnace. This problem was encountered at two independent plants. In service, the manifold undergoes significant thermal expansion. The tube to manifold pigtail connections, despite being deliberately flexible to account for thermal expansion loads are numerous and, in acting together, are capable of exerting significant system loads on the manifold. These thermal loads lead to significant bending applied to the manifold arm to bull-T weldment and it is not uncommon for cracking to be found at this location. The most common failure mode in these manifold systems is in fact, creep– fatigue at the welds and this was found to have occurred at two independent plants. Fig. 5 shows the sort of cracking that was observed. Based on the stress rupture data shown in Fig. 2, an effective life of 5 years and a service temperature of 760 C, an approximate mean stress of 45–50 MPa is implied. This is significantly higher than pressure based hoop stresses which are typically below 10 MPa and upon which design is based and indicates the relatively high level of thermal stress.
What was not expected, however, was the extreme brittleness of the parent bull-T material. Attempts to grind out and re-weld the damage simply led to the generation of more cracks as can be seen in Fig. 6. It is not uncommon for in-situ solution anneal heat treatments to be conducted on these materials to improve weldability. The temperature required, however, is in excess of 1100 C and a recent recommendation was that this should be increased to 1200 C. Such heat treatments can be readily done for small areas in the immediate vicinity of the weldment. This did not, however, provide confidence that the remaining parts of the manifold system were sufficiently ductile to ensure their integrity when returned to service.
Fig. 4. Typical bull-T.
Fig. 5. Creep fatigue crack.
8
8
8
502
D.M. Knowles et al. / International J ournal of Pressure V essels and Piping 81 (2004) 499–506
Fig. 6. Cracking after and during attempted weld repair.
This, led to the decision in both cases investigated, that temporary repairs should be made until such time that replacement components could be installed. One of the bullTs was made available for metallurgical investigation. It had been in service for approximately 5 years at service temperatures of approximately 760 C. The analysis of this material is included in Table 1. 8
Fig. 7. Charpy impact data.
the service temperature. The intention was to determine if the apparent loss of ductility was a low temperature phenomenon or if toughness was also compromised at service temperatures. Tensile testing similarly was undertaken at room temperature and at 800 C to simulate service conditions. The solution annealing heat treatment involved holding at 1100 C for 3 h followed by air cooling. The small laboratory specimens were removed from the furnace and cooled relatively rapidly in air. The Charpy impact test results are listed in Table 2 and illustrated in Fig. 7. The tensile test results are contained in Table 3. The Charpy data exhibited some scatter. However, while there was a modest increase in toughness with increasing temperature, the toughness as revealed by impact testing remained low even at temperatures close to the operating temperature. The tensile data also revealed extreme brittleness at room temperature for the as-received exservice material with no measurable elongation on the test piece itself. The stress–strain curve for this sample showed a plastic strain of less than 0.4%. Only one test was undertaken for this condition because the material was so brittle, the duplicate specimen failed during machining. The as-received material however, had significant ductility at service temperatures. In fact, the elongation and reduction 8
8
4. Mechanical testing
The issue of concern was apparent extremely low ductility that led to an inability to weld the bull-T without cracking. Consequently, the mechanical testing undertaken to date has concentrated on toughness using Charpy impact testing on standard 10 £ 10 mm samples. In addition, a series of tensile tests have also been undertaken. The testing has been conducted on the material removed from service (as received) and after a solution annealing heat treatment. For the as-received material, a series of Charpy impact tests were undertaken at a series of temperatures approaching Table 2 Charpy Impact Results 8
Condition
Temperature ( C)
Impact energy J
Annealed Annealed Annealed Ex service Ex service Ex service Ex service Ex service Ex service Ex service Ex service Ex service Ex service Ex service
23 23 23 23 23 23 105 200 295 320 390 590 710 760
32 38 26 6 9 7 14 11 13 20 10 12 25 10
Table 3 Tensile test results Condition
Temperature ( C)
UTS (MPa)
0.2%Proof (MPa)
% Elong
% ROA
20 20 800 800 20 800 800
340
158
180 171 200 171 174
91 80 177 80 82
15 16 39 39 0 51 34
13 15 45 54 0 55 33
8
Sol annealed Sol annealed Sol annealed Sol annealed Ex service Ex service Ex service
D.M. Knowles et al. / International Journal of Pressure Vessels an d Piping 81 (2004) 499–506
503
Fig. 8. As-cast 20Cr32Ni1Nb.
Fig. 9. 20Cr32Ni1Nb alloy after 1 day at service temperatures.
of area results were as high or higher for the asreceived material than the solution annealed test pieces at 800 C. 8
The density of intra-dendritic carbides typically increases dramatically when the material first enters service. Fig. 9 shows a similar sample to that shown in Fig. 8 except that it has been held at 800 C (approximate service temperature) for 24 h. The number and density of intra-dendritic secondary carbides is extremely high. It is these carbides and the complexity of the interdendritic eutectic carbides that have been attributed with generating the good creep properties of similar alloys [3]. In the ex-service material, the relatively long term aging at approximately 800 C had led to the agglomeration and dissolution of many of the intra-dendritic carbides. A significant number however, remain as can be seen in Fig. 10. 8
5. Metallography
The microstructure of 20Cr32Ni1Nb consists of an interdendritic network of primary carbides in an austenitic matrix. The microstructure is illustrated in Fig. 8. Despite the interdendritic eutectic carbides and the relatively low carbon content of this material, the austenite matrix is typically supersaturated with carbon and some fine intradendritic carbides can be seen in the matrix.
Fig. 10. 20Cr32Ni1Nb alloy ex-service.
8
504
D.M. Knowles et al. / International J ournal of Pressure V essels and Piping 81 (2004) 499–506
Fig. 11. 20Cr32Ni1Nb alloy ex-service and solution annealed.
Thein-serviceexposurehasnotledtowholesalechangesinthe microstructure. It is considered significant, however, that the networkof primarycarbides appearscontinuousin manyareas and has adopted a two phase appearance. After solution annealing (3 h at 1100 C), the microstructure has again not changed in any major way. The density of intra-granular secondary carbides remained similar to the exservice material except a precipitate free zone has developed adjacent to the primary carbides. In addition, the dual phase natureof the eutectic carbides wasremoved and thestrings of interdendritic precipitate that were present in the ex-service material tended to break up into discreet particles (Fig. 11). As part of an on-going program to understand the microstructure of this alloy, the ex-service material has been examined using a scanning electron microscope. The dual phase nature of the primary carbides was evident as illustrated in Fig. 12 which shows an example of primary eutectic precipitation viewed using the back scatter detector to highlight atomic weight differences. Analysis of these precipitates using an energy dispersive X-ray analysis system in conjunction with the SEM showed the two phases within the precipitates to be strongly segregated (Fig. 13). One was chromium rich while the other was essentially free of chromium and contained niobium and silicon. This silicon/niobium rich phase was not found in the as-cast or solution annealed materials. 8
of toughness. These mechanical test results are consistent with the observation that the original bull-T was extremely difficult to weld. However, the relatively high tensile ductility at operating temperatures suggest that integrity at operating temperature is not an issue. Thermal loads at start up or shut down however may lead to cracking while the bull-T is relatively cold. The metallographic examination revealed no immediately obvious reason for the embrittlement. More in depth examination however, revealed the presence of a silicon and niobium rich phase in the interdendritic precipitates. These materials are part of an on-going investigation to characterise the microstructure and establish the influence of these phases on material properties. Similar phases in 20Cr32Ni1Nb have however been identified by other workers[4,5] who have identified the silicon rich phase as ‘G-phase’ reported to be Ni16Nb6Si7. In similar but higher carbon alloys such as HP50Nb reformer tube materials [6], the silicon rich phase was identified as a silicide having an h-carbide (M6C) structure.. In all these cases, extreme brittleness at ambient temperatures have resulted. It is concluded therefore that the problem of brittleness in the outlet manifold components examined was caused by
6. Discussion
The Charpy impact testing confirmed the brittle nature of the ex-service material. There was a modest improvement in toughness with increasing temperature but the impact energy of the ex-service material was always below that of the solution annealed material. The low ambient temperature tensile tests also showed the exservice material to be extremely brittle with no measurable elongation being recorded on the test pieces. It was of interest however, that in tensile tests at 800 C, tensile ductility of ex-service material was high. The solution annealing heat treatment re-established a significant level 8
Fig. 12. SEM backscattered electron image of primary eutectic interdendritic precipitates in ex-service material.
D.M. Knowles et al. / International Journal of Pressure Vessels an d Piping 81 (2004) 499–506
505
Fig. 13. Element map showing distribution of silicon, chromium and niobium in interdendritic precipitates in ex-service material.
the formation of intermetallic niobium rich silicide phases. The observation that improvement in weldability can be obtained by solution annealing is consistent with the observation that these silicide phases were not present in the solution annealed samples. Based on the observations made, the formation of these deleterious phases can be considered a normal consequence of in-service aging. This is clearly an undesirable situation. It has been suggested [4] that the formation of the silicides can be controlled if not prevented, by ensuring that the niobium level is maintained below that necessary to stoichiometrically accommodate all the carbon as NbC, i.e. the wt% ratio of niobium to carbon should be held below 7.7. The observation of silicides in HP50Nb materials where the carbon content is higher at 0.4–0.5% would suggest that this is not sufficient to prevent the formation of silicides. It would appear that niobium is a prerequisite for the formation of silicides. It is therefore questioned why niobium needs to be used in this application at all. Alloying of cast alternatives to alloy 800 with niobium appears to have been adopted by manufacturers because of experience with tube alloys and an apparent improvement in creep properties that this produces. Experience with HP50 alloys [7] has shown that in aged materials, the stress rupture strength of tube materials are not significantly better than in the niobium free versions, i.e. any strength advantage gained by alloying with niobium is soon lost once the tubes have entered service, probably because of the formation of intermetallic silicides. In addition, in manifold components, wall thickness is not important and strength can therefore be obtained by design rather than material strength.
However, more work is clearly required before the use of niobium alloyed materials should be condemned in this application. For example, a benefit that has been observed in the performance of aged niobium alloyed materials is their relatively good creep ductility [7]. This property is important in maximising resistance to creep fatigue, which after all, was the problem that initiated this investigation in the first place. The good high temperature ductility of the 20Cr32Ni1Nb material is suggested by the high tensile ductility of the ex-service material which was superior to the solution annealed material.
7. Conclusions
20Cr32Ni1Nb cast material has become an industry standard for reformer furnace outlet manifold components In-service aging of this material, however, leads to serious embrittlement that yields the material unweldable. This problem can be at least mitigated by solution annealing at temperatures in excess of 1100 C, but this is not practical for complete outlet manifold systems which remain embrittled. Furthermore, solution annealed materials are likely to re-embrittle on further exposure. The cause of the embrittlement is the formation of niobium rich silicide intermetallics which have also been observed in closely related HP50Nb reformer tube alloys. The presence of niobium appears to be a prerequisite for the formation of these intermetallics. It is therefore suggested that niobium free grades are developed to avoid the problem of in service embrittlement. 8
506
D.M. Knowles et al. / International J ournal of Pressure V essels and Piping 81 (2004) 499–506
References [1] Rostrup-Nielsen JR. Catalytic Steam Reforming. Berlin: Springer; 1984. [2] API 530 Calculation of heater tube thickness in petroleum refineries, American Petroleum Institute, 1220 L Street, Northwest Washington, DC 20005. [3] Hou W-T. Mater Sci Technol 1985;1(5):385– 7.
[4] Hoffman JJ, Gapinsky GE. Ammonia Plant Safety. AIChE 2002;42: 10–21. [5] Shibaski T, Mohri T, Takemura K. Ammonia Plant Safety, AIChE 1994;34:166–76. [6] Thomas CW, Stevens KJ, Ryan MJ. Mater Sci Technol 1996;12: 469–75. [7] Thomas CW, Tack AJ. Proc Int Symp. Case Histories on Integrity and Failures in Industry, Milan Sep 1999;28:1.