喜歡就充值下載吧，，資源目錄下展示的全都有，，下載后全都有，dwg格式的為CAD圖紙，有疑問(wèn)咨詢QQ：414951605 或1304139763 Thermal constraint considerations in design of a heatrecovery boilerR. Caligiuri, J. Foulds*, R. Sire, S. AndrewExponent Failure Analysis Associates, 149 Commonwealth Drive, Menlo Park, CA 94025, USAReceived 23 September 2005; accepted 8 October 2005Available online 9 February 2006AbstractHeat recovery boilers are typically designed to conform to the rules of the ASME Boiler and Pressure Vessel Code, Sec-tion I on Power Boilers. While the unfired pressure vessel design rules of Section VIII of the ASME Code have evolved toaccommodate a wide range of vessels and loading conditions (including system thermal loads), Section I of the Coderemains characteristically simple, specifying explicit rules only for primary loading (pressure mainly). By and large, thecombination of boilermaker experience and Section I rules makes for robust boilers. However, departure from conven-tional design, although in conformance to the explicit rules of Section I, can result in premature failures. This paperdescribes one such design of a heat recovery boiler in a hardwood Kraft pulp mill, wherein system thermal stresses inducedon superheater tube-to-tube tie welds were high enough to result in weld toe cracking and early tube leaks. A finite elementstress analysis of the design showed that operational thermal stresses due to constraint imposed by the tie welds were inexcess of the tube material yield strength and above a level that would be considered acceptable for design. The analysishighlights the need for considering effects of thermal constraint in designs that otherwise meet the basic Code requirementsfor boilers.? 2005 Elsevier Ltd. All rights reserved.Keywords: Heat recovery boiler; Tie weld; Superheater; Thermal constraint; Design1. Introduction and backgroundHeat recovery boilers are commonly used in a variety of processes involving combustion and production ofexhaust heat. While gas turbine-based combined cycle power plants represent a relatively recent, popularapplication of heat recovery boilers (commonly referred to as heat recovery steam generators or HRSGs),other chemical and industrial processes having heat recovery boilers to produce steam and/or electric powerhave been in existence for well over 50 years. Kraft pulp mills comprise one such industrial process utilizingheat recovery boilers. A Kraft pulp mill heat recovery boiler is the subject of this paper, although the lessonslearned apply to boilers in any application.1350-6307/$ - see front matter ? 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.engfailanal.2005.10.003*Corresponding author. Tel.: +1 650 326 9400/688 7226; fax: +1 650 326 8072/328 2990.E-mail address: (J. Foulds) Failure Analysis 13 (2006) 13881396Fig. 1 is a schematic of the heat recovery boiler used to recover heat generated in the burning of blackliquor, a part of the hardwood Kraft pulping process. Of particular interest to this study is the pendant super-heater, consisting of three banks as identified the low, intermediate, and high temperature sections. The boi-ler comprises many such repeat units or elements, all connected to common headers. For size perspective, theintermediate section is about 43 feet (13.1 m) tall.Heat transfer is counter-flow for the low and intermediate sections and parallel-flow for the high temper-ature section. Each section consists of four tubes arranged in serpentine loop fashion from an inlet to an outletheader. Fig. 2 is a schematic of the tube arrangement for the intermediate section. Also seen in the schematicare horizontal lines to indicate locations of tube-to-tube tie welds.Within 18 months of the start of operation, leaks developed in the superheater tubes. Inspections followingboiler shutdown revealed that tube leaks had occurred from cracks that had initiated at the toes of tube-to-tube tie welds and propagated into the tube base material, breaching the tube wall. The intermediate and hightemperature sections exhibited cracking. Fig. 3 shows macrophotographs of typical tie welds after cleaningand a metallographic cross-section showing tube cracking at a tie weld.2. Design and performance specificationsThe pendant superheater was designed per ASME Boiler and Pressure Vessel Code, Section I: PowerBoilers 1 (ASME I). Nominal performance specifications on the superheater were as follows: inlet steam(to the low temperature section) at 712 psig (4.9 MPa), 505 ?F (263 ?C)1and outlet steam (from the hightemperature section) at 630 psig (4.3 MPa), 752 ?F (400 ?C). Steam temperature gradients were nominallyFig. 1. Schematic of heat recovery boiler showing the pendant superheater that is the focus of this study.1US customary units were specified in the design and were used in all analyses; SI units are reported in parentheses.R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 13881396138983 ?F (46 ?C), 115 ?F (64 ?C), and 117 ?F (65 ?C) across the low, intermediate, and high temperature sec-tions, respectively, with a steam attemperation stage in the crossover between the intermediate and hightemperature sections.Most of the pendant superheater tubes had a nominal outside diameter of 2 in. (50.8 mm). Tube wall thick-ness varied, being 0.165 in. (4.2 mm), 0.180 in. (4.6 mm), or 0.220 in. (5.6 mm). Tube materials were, perASME I, SA-192 (CMn steel), SA-209 (MnMo steel), and SA213 (T11) (CrMo steel). The 3 in.(76 mm) long tube-to-tube tie welds were shop-produced gas metal arc welds (GMAW). Reported design tem-perature in all cases was 800 ?F (427 ?C). For SA-192, the lowest-strength material at design temperature, andthe one representing the majority of tubes, the minimum yield strength ranged from about 26 ksi (179 MPa) atroom temperature to about 14.5 ksi (100 MPa) at 800 ?F (427 ?C). The applicable ASME I-allowable stress at800 ?F (427 ?C) was 9 ksi (62 MPa).3. Finite element stress analysesPrior to performing a detailed three-dimensional (3D) finite element (FE) stress analysis, a relatively simpletwo-dimensional (2D) elastic analysis was conducted to quantify the effect of pressure and thermal gradientacross the tube wall. The 2D analysis naturally could not provide any indication of the effect of thermal gra-dients along the length of individual tubes. The stress analyses results and conclusions made here are for theintermediate and high temperature sections.Fig. 2. Schematic of intermediate superheater section showing tube arrangement; horizontal lines indicate tie-weld locations.1390R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 138813963.1. Local 2D analysisFig. 4 illustrates the FE model used for both heat transfer and stress analysis. The nominal tube dimensionsof 2 in. (50.8 mm) outside diameter and 0.165 in. (4.2 mm) wall thickness were used in the analysis. A 2D planestrain analysis was performed using the ANSYS program 2. The assumed conditions were: internal steamFig. 3. (a) Photograph of tie welds (after cleaning) and (b) metallographic cross-section showing tube cracking at tie weld (arrows).Fig. 4. 2D FE model of tube-tie weld.R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 138813961391pressure of 630 psig (4.3 MPa), steam temperature of 700 ?F (371 ?C), gas temperature of 1425 ?F (774 ?C),convective heat transfer coefficients of 1.93E-05 BTU/s/in2/?F (0.057 kJ/s/m2/?C), and 4.07E-04 BTU/s/in2/?F (1.2 kJ/s/m2/?C) for the gas and steam side, respectively. The coefficients were computed using the methodsin Ref. 3, with inputs for fluid velocities from plant performance specifications.The analysis showed metal temperature varying between 721 and 739 ?F (383 and 393 ?C) with a max-imum through-wall gradient of 5 ?F (3 ?C). The stress analysis results showed stress due to pressure atthe tubetie-weld interface was only about 4 ksi (27.6 MPa), well below the ASME I allowable of 9 ksi(62 MPa) at 800 ?F (427 ?C). The through-wall gradient was such that resulting stresses at the tube-weldinterface were compressive (due to higher tube outside surface temperature). In the worst case, where cooldown could produce tensile stresses at the outer tube surface, the contribution from the thermal gradientwas estimated at no more than 0.5 ksi (3.4 MPa). In effect, the pressure and tube wall thermal gradient stres-ses were determined to be low and of no design concern. This finding is not surprising since boiler tubedesigns conforming to the ASME I requirements are expected to have acceptable pressure stresses (bythe very intent of ASME I), and thermal gradients across tube walls even in high heat flux regions of boilershave not generally been a source of integrity problems. What is of particular interest here is the effect ofthermal gradients along individual tubes that are tied together and configured in such a manner that thelongitudinal differential expansion between adjacent segments causes high stresses at the tie welds. The3D analysis sought to quantify this effect.3.2. Global 3D analysisA global 3D analysis was used to obtain relative stress levels throughout the superheater to identify tie-weldlocations exhibiting the highest global stresses. Again, the ANSYS program 2 was used. Refer to Fig. 2 for aschematic of the global model. The model is composed primarily of pipe and beam elements. The tubes aremodeled using ANSYS pipe element Type 16, a specialized beam element adapted to model tubular geome-tries. The U-bends were modeled using ANSYS curved pipe element, Type 18, and the tie welds using beamelement, Type 4. Typical weld dimensions were used in computing sectional properties (area and moment ofinertia) of the beams. At the top, truss bar elements, Type 8, were used to simulate the boiler ceiling penetra-tion, and were assigned temperatures to permit free expansion of the model. While the elements used here arecapable of describing out-of-plane loading, this form of loading was considered minor and excluded in theanalysis, for practical reasons.Global model loading included pressure, dead weight, and a linear thermal gradient along each tube withmetal temperature at inlet and outlet equal to the steam temperature of the performance specification. Theserpentine tube arrangement (Fig. 2) results in varying temperature differentials between adjacent tubes,depending on location. The outer loop locations expectedly experienced the most severe tube-to-tube gradi-ents, and the analysis captured this effect.The outputs of primary interest were the forces and moments on the tie welds. These forces and moments atthe tie welds were then used as boundary conditions on a local 3D model to obtain detailed stresses around thetie welds.3.3. Local 3D analysisA local 3D analysis (also using the ANSYS program 2) was used to obtain the through-wall stress distri-bution in the tubes at the tie-weld locations. Hoop stress was used as a severity metric since the predominantcracking orientation was normal to the hoop direction.A generalized method was used to make the local analysis results useful in computing stresses at any of thetie-weld locations without having to perform repeated FE analyses for each location. The forces and momentsfrom the global 3D analysis were applied to the three-tube model using 24 generalized degrees of freedom asshown schematically in Fig. 5.The generalized method required analysis for each of the degrees of freedom independently, performed byapplication of a unit displacement at the degree of freedom while constraining displacements in all other gen-eralized degrees of freedom.1392R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 13881396This suitably accounts for reaction forces, so that the result of 24 analyses provides the interdependencebetween forces and displacements per:F K ? U1where F is the 24 1 matrix of generalized force; K is the 24 24 stiffness matrix constructed from the reactionforces of each degree of freedom analysis; and U is the 24 1 matrix of generalized displacements.Eq. (1) may be rewritten as:U K?1? F2defining the individual contributions of each degree of freedom to the applied loading.Fig. 6 shows the local FE model used to analyze the 24 generalized degrees of freedom results. The FEmodel is seen to physically represent only one-eighth of the three-tube geometry shown in Fig. 5. To reducemesh size and enhance computational efficiency, initially a quarter-symmetry segment of the three-tube modelwas considered. A further reduction in model size was made by slicing the quarter-symmetry geometry in half,front to back.Note the relative mesh refinement in the tie-weld region, more so near the weld crown. Symmetric or anti-symmetric displacement boundary conditions were applied to the planes of geometric symmetry to appropri-ately simulate response to the various unit load cases. The internal tube pressure load case was conductedseparately.3.4. Analysis resultsThe local 3D analysis gave detailed through-wall hoop stress distribution estimates at the various tie-weldlocations in the intermediate and high temperature superheater sections. The maximum hoop stress varied as aFig. 5. Three-tube model showing the generalized degrees of freedom in applied forces and moments derived from the global model.R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 138813961393function of location, with the peaks in the range of 5560 ksi (379414 MPa) generally at the uppermost tiewelds (Level A in Fig. 2) and at the outer-loop tubes, where the tube-to-tube temperature differences werethe highest. Additionally, the peak stresses occurred at the weld toe in the crown region. Fig. 7 shows typicalhoop stress contours at the surface in the vicinity of the tie weld. Both the global location of predicted max-imum hoop stress (Level A, outer loop) and the local weldment location of stress (weld crown), coincide withareas of observed cracking.Fig. 8 is a graphical representation of the elastically computed through-wall stress distribution at one of themost highly loaded tie-weld locations in the high temperature superheater section.The hoop stress distribution was linearized in order to eliminate the local effect of the weld toe geometry,and evaluate the overall hoop stress distribution against the material yield strength and the ASME Codeallowable stress for the material. The linearized hoop stress distribution produces the same net axial forceand moment on the tube wall as does the actual predicted stress distribution. The membrane stress is11.3 ksi (81.4 MPa), and the linearized maximum bending stress is 30.8 ksi (212 MPa) at the OD surface, mak-ing the near-surface stress, exclusive of the weld toe geometry effect, 42.1 ksi (290 MPa).Fig. 6. Outside and inside views of one-eighth symmetry FE model of three-tube configuration.1394R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 138813964. Discussion and conclusionsA number of interesting facts are apparent from the hoop stress distribution of Fig. 8.? The predicted peak stress at the outside tube surface, even without the geometric effect of the weld toe, isabout 42 ksi (290 MPa), nearly three times the yield strength of SA-192 tube material.? The peak linearized stress exceeds the ASME I allowable for SA-192 at 800 ?F (427 ?C) by more than afactor of 4.5 and even the room temperature allowable by nearly a factor of 4.? The linearized hoop stress exceeds the material ASME I 800 ?F (427 ?C) allowable over more than half thetube wall, and exceeds the corresponding material yield strength over more than one-third of the tube wall.? While not explicitly shown in Fig. 8, one can easily deduce that the major contribution to the hoop stress isfrom thermal constraint, given the primary stress due to pressure is roughly 4 ksi (27.6 MPa), only about10% of the peak stress. Further, in the absence of the thermal constraint, stresses are well within designallowables.-20-1001020304050600.0000.0150.0300.0450.0600.0750.0900.1050.1200.1350.1500.165Distance Through Wall from OD (in)Stress (ksi)HoopAxialRadialHoop (linearized)Fig. 8. Through-wall elastically computed stress distribution for one of the most highly loaded high temperature superheater section tubesat a tie-weld location; the hoop stress linearization was performed to analytically remove the geometric effect of the weld toe.Fig. 7. Typical local analysis results showing hoop stress contours in vicinity of the tie weld. Units in ksi with maximum of 56 ksi(386 MPa) and minimum of ?20 ksi (?138 MPa).R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 138813961395It is evident from the preceding discussion, that tube stresses arising from thermal constraint imposed bythe tube-to-tube tie welds are very high, well in excess of pressure stresses and the material ASME Code-allow-able levels, and remain so over a significant portion of the tube wall. These conditions strongly suggest zero ornegative margins against failure, and are indicative of a significant design deficiency. This underscores the factthat when thermal constraint is ignored in application of ASME I, a deficient design may appear to be Code-compliant. This is because ASME I, while it does have rules (PW-43) for limiting loads on rigid attachments2(used generally to accommodate primary loads), it has no explicit rules for designing to accommodate signif-icant thermal or system loads. This is also a prime example of where the following, from the foreword of theASME Code, applies: The Code is not a handbook and cannot replace education, experience, and the use of engi-neering judgment.Nevertheless, established thermal design methods for accommodating system loads are available, and onesuch method is part of the ASME Boiler and Pressure Vessel Code, Section VIII for pressure vessels 4(ASME VIII).Section VIII, Division 2 of the Code, Alternative Rules Rules for Construction of Pressure Vessels, pro-vides means by which a manufacturer can demonstrate that a design detail is as safe as otherwise providedby the rules in the Code. Although fired process tubular boilers are outside the scope of Section VIII, the anal-ysis methodology and allowable stresses in Section VIII may be utilized to provide a quantitative indication ofthe design suitability or lack thereof.In this case, the predicted maximum stress intensity (defined as the difference between the maximum andminimum principal stresses) of 40 ksi (276 MPa), is compared against the Code-specified allowable stressintensity given by three times the average of the allowable stress at the extremes of the operating cycle tem-perature. The latter translates to roughly 3X (average of 11.8 ksi at room temperature and 9 ksi at 800 ?F)or 31.2 ksi (215 MPa). In effect, the peak stresses at the superheater tie welds exceed the ASME VIII-specifiedlimit by about 30%. This exercise confirms the deficiency in superheater design related to the constraintimposed by the tube-to-tube tie welds. The actual cracking experience was entirely consistent with the resultsof this analysis, and not at all surprising in light of this work.AcknowledgmentsDr. T. Kim Parnell, formerly with Exponent Failure Analysis Associates, is acknowledged for contribu-tions to the FE analysis effort. Domtar Inc. is thanked for supporting this research.References1 ASME boiler and pressure vessel code, Section I: Power boilers. New York (NY), American Society of Mechanical Engineers, 1983.2 ANSYS engineering analysis system users manual, Revision 4.4. Houston (PA), Swanson Analysis Systems Inc., 1990.3 Chapman AJ. Heat transfer. 3rd ed. New York: MacMillan; 1974.4 ASME boiler and pressure vessel code, Section VIII: Pressure vessels. New York (NY), American Society of Mechanical Engineers,1983.2The effective (thermal) load here is more than an order of magnitude greater than the PW-43 allowed attachment load.1396R. Caligiuri et al. / Engineering Failure Analysis 13 (2006) 13881396