Multiphysics Modeling of a Welded Furnace Roll for ...€¦ · WELDING RESEARCH NOVEMBER 2016 /...

WELDING RESEARCH

NOVEMBER 2016 / WELDING JOURNAL 431-s

Introduction

Creep and fatigue in combination isthe main reason for the failure ofmany engineering components operat-ing under high temperature and cyclicloading. Many studies have been con-ducted to investigate creep-fatiguecrack initiation and growth (Refs.1–3), creep and fatigue interactions(Refs. 4, 5), damage mechanism (Refs.6, 7), lifetime predictions (Refs. 8–12),and hold time effect on creep and fa-tigue life (Refs. 13–15). Karl (Ref. 7)

tested and modeled smooth andnotched specimens of Type 304 stain-less steel by applying several types ofidealized fatigue loading to obtain aclear picture of the types of damageoccurring in a steam turbine and simi-larly loaded mechanical systems. Zhu(Ref. 11) proposed a low-cycle, fa-tigue-creep life prediction model forgeneral use in isothermal and thermo-mechanical loading based on the theo-ry of ductility exhaustion. With betterunderstanding of the creep-fatiguedamage process, creep-fatigue assess-

ment standards and codes (Refs.16–19) were developed such as theASME Boiler and Pressure Vessel Code(BPVC) Section III, Subsection NH,and the R5 assessment procedure. ASME BPVC Section III, SubsectionNH (Refs. 16–17), considers cyclic fail-ure modes at elevated temperaturesand provides creep-fatigue interactionrules and damage limits. The R5 as-sessment procedure (Refs. 18, 19) wasdeveloped in the UK to address bothcreep-fatigue crack initiation in initial-ly defect -free components and thegrowth of flaws by creep and creep-fatigue mechanisms. Although significant progress hasbeen made to understand the creep-fatigue damage process and design theengineering components according tothe ASME code or the R5 assessmentprocedures for elevated-temperatureservice, creep-fatigue failures are stillobserved in industries because real-world problems are far more complexthan a well-controlled research envi-ronment. A large number of real-worldproblems have to be solved using mul-tiphysics simulation tools to model thephysics involved in a process or serv-ice. However, it is challenging to con-duct a fully coupled numerical analysisamong a separate continuum physicsphenomena. Many successful multi-physics modeling efforts are basedupon loose or one-way coupling be-cause of computational cost and re-sults mapping issues between multi-physics simulation tools (Ref. 20). A catastrophic roll failure was ob-served in a continuous hot-dip coating

Multiphysics Modeling of a Welded Furnace Rollfor Improving CreepFatigue Life

This study proposes methods to improve the creepfatigue lifetime using numerical analyses, including heat transfer analysis, creepfatigue analysis, and CFD analysis

to simulate the multiphysics involved in the roll service

BY Y. P. YANG AND W. C. MOHR

ABSTRACT Heattransfer analysis, creepfatigue analysis, and computational fluid dynamics (CFD)analysis were conducted to understand the failure of a welded roll in a continuous coating line and to propose solutions to improve the creepfatigue lifetime. Analysisresults showed three factors that contributed to the roll failure: difference of materialproperties between filler metal and base metal, furnace temperature variation duringservice, and high operating temperature. Based on these findings, two fatiguelifeimprovement methods, using the electron beam welding (EBW) process without fillermetal to weld the roll and cooling the roll from inside to lower the shell temperature,were proposed. Creepfatigue analyses showed both methods were effective inreducing the creep strain and stress to improve the roll fatigue life. Using EBW, thecreep strain and stress in the interface between the filler metal and the base metalwere completely eliminated. By lowering the shell temperature using the inside rollcooling, the creep strain and stress near the weld were significantly reduced. A coolingsystem was designed and evaluated using a fully coupled CFD and heat transferanalysis. A preliminary welding test showed that EBW could be implemented to weldthe roll, and the CFD analysis results confirmed that effective inside roll cooling wasdifficult to achieve with the current design. In addition, the predicted highly localizedmaximum principal stress at the weld root can be used to explain the observed crack atthe weld root, which indirectly validates the models.

KEYWORDS • Creep • Fatigue • Multiphysics Modeling • Crack • Welding

Y. P. YANG ([email protected]) and W. C. MOHR are with the Edison Welding Institute, Columbus, Ohio.

Yang Nov 16.qxp_Layout 1 10/14/16 11:02 AM Page 431

WELDING RESEARCH

WELDING JOURNAL / NOVEMBER 2016, VOL. 95432-s

line in a high-temperature roll posi-tion. The failure occurred in the weldjoining the end bell to the roll shelland resulted in the complete 360-degseparation of the idle side end bellfrom the roll shell, as shown in Fig. 1.Historically, furnace rolls in this rollposition have shorter service lives(less than one year) in comparison toother lower temperature roll positions.By examining the welded joint, therewere two types of cracks found nearthe weld. The first type of crack initi-ated from the weld root and propagat-ed to the weld’s outer surface, result-ing in the separation between the rollshell and the end bell. The second typeof crack initiated from one of the weldtoes. The service record of the failedroll showed that the roll was routinelyexposed to an operating temperaturebetween 982 and 1066C, with fre-quent temperature fluctuations ofmore than 500 C. Occasionally, theroll chamber temperature dropped toabout 200 C. These temperature vari-ations resulted in a cyclic thermalloading on the roll. In addition, themechanical load used to transport thestrip was applied on a 90-deg sectionin the roll. Since the roll was rotatingcontinuously, the mechanical load wasa cyclic-type loading. Therefore, the

roll was working under cyclic thermaland mechanical loading and high-temperature conditions. Creep-fatiguedamage could occur during the rollservice. Finite element analyses, including a heat transfer analysis and a creep-fatigue analysis, were conducted tomodel the cyclic thermal and mechani-cal process of the furnace roll in a con-tinuous hot-dip coating line. The heattransfer analysis was conducted to pre-dict the temperature history of the rollby modeling heat convection from hotair inside the furnace. The creep-fatigue analysis was performed to pre-dict creep strain, stress, and deforma-tion by inputting the predicted tem-perature history and applying mechan-ical loads. The deformation was vali-dated by experimental measurement.Analysis results showed that the fail-ure resulted from a creep-fatiguemechanism rather than a creep mecha-nism. The difference of material prop-erties between the filler metal and thebase metal is the root cause for the rollfailure, which induces higher creep

strain and stress in the interface be-tween the weld and the HAZ (Ref. 21). The research in this paper is a con-tinual study based on the conclusionsof Ref. 21. The main objective of thisstudy is to propose methods to im-prove the creep-fatigue lifetime usingnumerical analyses, including heattransfer analysis, creep-fatigue analy-sis, and CFD analysis, to simulate themultiphysics involved in the roll serv-ice. Based on the analytical results,two improvement methods using elec-tric beam welding (EBW) and applyinginside cooling were proposed and stud-ied using numerical analysis. Analyti-cal results confirmed that EBW was ef-fective to improve the creep-fatiguelife, and the inside roll cooling methodwas effective but difficult to imple-ment in practice. The steel manufac-turer confirmed that EBW withoutfiller metal could be used to produce adefect-free weld of MO-RE®1 from apreliminary welding test. This pro-posed EBW solution could be an effec-tive method to improve the creep-fatigue life of the furnace roll in thecontinuous hot-dip coating line, whichcould solve this long-term problemand provide huge cost savings for thesteel manufacturer. However, weld residual stresseswere ignored in the present analysesdue to two reasons. The first reasonwas that postweld heat treatment(PWHT) was applied to relieve resid-ual stress after welding. The secondreason was that the roll worked at ahigh temperature and residual stresscould further be relieved during serv-ice. To confirm that EBW is still effec-tive if weld residual stresses are in-cluded in the analysis, a follow-upstudy was performed to include weldresidual stress as initial conditionsduring creep-fatigue analysis for acomprehensive assessment of the

Fig. 1 — Roll failure in the vertical furnace of a continuous hotdip coating line.

Fig. 2 — Finite element mesh for a newdesign: A — New design; B — finite element mesh.

Table 1 — Creep Properties

SS310 MORE1 N117 A n T A n T A n T (MPa)–nh–1 (C) (MPa)–nh–1 (C) (MPa)–nh–1 (C)

1.44E15 3.8998 537.8 8.67E13 3.8998 871.1 3.00E51 17.373 593.3 9.35E16 4.2936 593.9 9.12E13 4.2936 926.7 1.93E39 13.515 648.9 2.29E15 4.4792 649.2 1.74E12 4.4792 982.2 4.35E30 12.043 760.0 8.43E16 5.2800 704.5 9.62E13 5.2800 1037.8 1.86E16 6.8728 871.1 4.40E12 3.5001 759.8 8.32E10 3.5001 1093.3 1.90E13 6.3135 982.2 1.22E11 3.9389 813.6 1.06E08 3.9389 1148.9 1.55E11 6.5634 1093.3

A

B


WELDING RESEARCH


creep-performance of the welded furnace

roll. The results of the followingstudy were reported in Ref. 22 inwhich both flux-cored arc welding(FCAW) and EBW were modeled topredict residual stress by inputtingthe detailed welding information,such as welding parameters and num-ber of weld passes. The same conclu-sion that EBW without filler metal isan effective method to improvecreep-fatigue life was obtained whenincluding weld residual stress.

Analysis ProcedureFinite Element Model

A new design (Fig. 2A) of the fur-nace roll in the continuous hot-dipcoating line was developed and con-sisted of two journals made of stain-less steel 310 (SS310), two end bellsmade of high-temperature alloysMO-RE®1, and a roll shell made ofMO-RE®1. There were two weld typesin the roll: the shell weld joining theend bell to the shell and the journalweld joining the journal to the end

bell. Because ofloading and geomet-ric symmetry, a halfthree-dimensional(3D) finite elementmodel was used inthe analysis andsymmetric bound-ary conditions wereapplied in the sym-metric plane, asshown in Fig. 2B.

Heat Transfer Analysis

The heat transfer analysis was gov-erned by the following equation:

The heat transfer analysis was gov-erned by the following equation:

where (T) is the density of material,cp(T) is the specific heat, Kx(T), Ky(T),Kz(T) are the thermal conductivity co-efficients for three space directions, T

is the temperature, and t is the timeelapsed. In the analysis, thermal con-ductivity was assumed to be the samein all three space directions. Temperature-dependent materialproperties were included in the finiteelement model for material stainlesssteel 310 (Refs. 23, 24), MO-RE1(Ref. 25), and N117 (Ref. 26). Figure3A shows that the filler metal has thehighest elastic modulus and MO-RE®1has the lowest elastic modulus amongthe three materials. Figure 3B showsthat N117 has the lowest coefficient ofthermal expansion (CTE) among thethree materials. Figures 3C–E showcomparisons of yield strength, tensilestrength, and elongation between thethree materials at different tempera-

� T( )cp T( )�T�t

= ��x

Kx T( )�T�x

��

��

+ ��y

Ky T( )�T�y

�

��

�

�+

��z

Kz T( )�T�z

��

��

1( )

Fig. 3 — Comparison of material properties between three materials: A — Elastic modulus; B — coefficient of thermal expansion; C —yield strength; D — tensile strength; E — elongation.

A

C D

E

B


WELDING RESEARCH


tures. When the temperature is lowerthan 600C, MO-RE®1 has the lowestyield strength, tensile strength, andelongation among the three materials. There is an increase in the yieldstrength of materials MO-RE®1 andN117 from 600 to 800C. This phe-nomenon is known as yield strengthanomaly (YSA) in which the yieldstrength actually increases at elevatedtemperatures. The mechanism of YSAcould be attributed to the phase trans-formation and the precipitates at 800C.

The roll was heated by heat convec-

tion from hot air insidethe furnace. Surface heatconvection (qs) was calcu-lated using Equation 2:

qs = h(T–Ts) (2)

where h was heat convec-tion coefficient (h), and Ts

was the air temperatureinside the furnace. T was the tempera-ture on the outer and inner surfaces ofthe furnace roll. Since Ts and T werechanged with the time and the loca-tion on the roll surface, a user subrou-tine was developed to automaticallycalculate the heat flux (qs). The usersubroutine was developed on the plat-form of ABAQUS (Ref. 27). For a giventime and location on the roll surface,ABAQUS provided a temperature (T)to the subroutine. The subroutine de-termined Ts based on the heating zoneshown in Fig. 4 and the heatingprocess diagram shown in Fig. 5 and

then calculated a heat flux (q). By in-putting the heat flux to ABAQUS, anew temperature (T) was calculatedand inputted to the subroutine. Thislooping process was done for all timeincrements and locations on the roll inthe thermal analyses. As the roll tem-perature reached the furnace air tem-perature, the heat flux approachedzero. By using the subroutine, the heatflux distributions on the roll and theheat flux transient changes duringheating and cooling were modeled dur-ing the thermal analyses. The typical heat convection coeffi-cient (h) for air is between 10 and1000 W/(m2K). Several heat transferanalyses were conducted to determinethe heat convection coefficient to heatthe roll based on the input furnacetemperature, as shown in Fig. 5. It wasfound that the roll can be heated tothe designated temperature accordingto the time specified in Fig. 5 in 24 hby inputting the heat convection coef-ficient, 20 W/m2C. After heating for

Fig. 4 — Heating sections along the roll axial direction. Fig. 5 — Furnace roll chamber temperature as a function of time.

Fig. 7 — Mesh for the cooling system design. A — Overall design; B —holes at the end; C — cooling channels near the journal.

Fig. 6 — Mechanical boundary and loading conditions. A— Forces on the roll; B — pressure on the roll; C — modeling contacts.

A AB

B

C

C

Yang Nov 16.qxp_Layout 1 10/14/16 11:02 AM 34

WELDING RESEARCH


16 h, the roll was heated to about200C. After heating for 24 h, the rollwas heated to about 1066C. Then, theroll temperature was kept constant for100 h and then cooled to about 200Cwithin 24 h. During creep-fatigueanalysis, the heating cycle was repeated 10 times.

CreepFatigue Analysis

A creep-fatigue analysis procedurewas developed using the ABAQUS com-mercial finite element code in whichisotropic creep- and plasticity-coupledbehavior was modeled by solving a cou-pled system of constitutive equations.This analysis procedure has been suc-cessfully used in modeling the creep be-havior of a nuclear pipe system (Ref.28). For simplicity, the power-law creepmodel was used in this study. The me-chanical load and gravity were appliedas a constant load and the thermal loadwas applied as low cyclic loads. Figure 5shows a one-cycle thermal load in whichthe holding time is about 100 h. Thelong hold time made the power-lawcreep model suitable for the analysis ofthe furnace roll. During heating andcooling, a dynamic analysis was con-ducted and creep process was not mod-eled because the temperature quicklydropped to a low value. At a given tem-perature, the rate of steady-state creepcan be calculated by the power law(Refs. 29, 30):

where is creep strain rate, is theapplied stress, A is a constant depend-ent on the material, and n is an expo-nent dependent on the creep mecha-nism. The creep constants A and n at

different temperatures, listed in Table1, were obtained by fitting the datafrom Refs. 23 to 26 and the data froman industrial customer. Steel sheet tension of 31136 N was

transferred to the roll through contacton one-fourth of the roll circumfer-ence over a width of 1270 mm in thecenter of the roll length, as shown inFig. 6A. The combined force in the 45-

�� = A�n 3( )Fig. 9 — Predicted effective creep strain after 10 loading cycles for both cool to 204Cand cool to 427C.

Table 2 — Air Properties as a Function of Temperature (Ref. 32)

Temperature Specific Heat Viscosity Thermal Conductivity Density (C) (J/kg C) (10–5 kg/m s) (10–2 W/m C) (kg/m3)

26.85 1004.9 1.846 2.624 1.177 101.85 1010.6 2.181 3.186 0.941 226.85 1029.5 2.670 4.041 0.706 326.85 1051.1 3.017 4.661 0.588 426.85 1075.0 3.332 5.236 0.504 526.85 1098.7 3.624 5.774 0.441 626.85 1120.9 3.897 6.276 0.392 726.85 1141.1 4.153 6.754 0.353 826.85 1158.9 4.396 7.209 0.321 926.85 1174.6 4.626 7.640 0.294 1026.85 1188.4 4.846 8.054 0.272

Fig. 8 — A quarter model for solid parts and fluid inside the roll.

•


WELDING RESEARCH


deg direction was 44033 N, as shownin Fig. 6A, which was applied on thearea of 912073 mm2 resulting in apressure of 0.048 MPa. The finite ele-ment grids (see Fig. 6B) show the pres-sure applied to the surface that is a 90-deg section in the designed roll. Analy-ses were conducted to evaluate the in-dividual effect of pressure and temper-ature on stress. It was found that thestress induced by pressure was muchsmaller than the temperature changes.Therefore, a constant pressure was ap-plied in the analyses. Only the thermalload was applied as a cyclic load. Weld residual stresses and weldingprocesses were not modeled in thisstudy because PWHT was used in pro-ducing the roll, and PWHT may reduceresidual stresses from welding. In addi-tion, the roll served in a high-tempera-ture environment was essential for thecreep process to reduce the weld resid-ual stress further. A follow-up study

(Ref. 22) was conductedto include weld residualstress during the creep-fatigue analyses. It wasfound that the sameconclusions can be ob-tained with and with-out weld residual stressduring the creep-fatigue analyses.

A Fully CoupledHeat Transfer andCFD Analysis

A fully coupled heattransfer and CFD analysis was con-ducted to simulate the effect of high-pressure air cooling inside the roll onthe roll temperature, which was one ofthe proposed methods to improve thecreep-fatigue life of the furnace roll inthis study. The CFD analysis was gov-

erned by incompressible Navier-Stokesequations. The equations can be readi-ly found in the ABAQUS manual (Ref.26) and the CFD textbook (Ref. 31)and will not be discussed here. TheCFD model and boundary conditionsare discussed in detail below.

Fig. 10 — Effect of coolto temperature between cycles on maximum principal stress. A— Cool to 204C after 10 loading cycles; B — Cool to 427C after 10 loading cycles.

Fig. 11 — Cracks observed near the weldedjoint (Ref. 21).

Fig. 12 — Evolution of temperature, effective creep strain, and maximum principal stress at location 1 of Fig. 10 for case 1: cool to 204Cbetween cycles.

Fig. 13 — Hoop stress near a weld at 427C and observedcracks. A — Hoop stress; B — failed roll; C — crack 1 near aweld; D — crack 2 near a weld.

A

A

B

C D

B


WELDING RESEARCH


A cooling system was designed bymodifying the new design of the fur-nace roll (Fig. 2A). The roll was en-closed by the journal welds, as shownin Fig. 7A. To allow air to flow into theroll, the journal at the right-hand sidewas redesigned to allow air flow intothe roll through a center tube and outof the roll via the four holes aroundthe center hole, as shown in Fig. 7B. A one-fourth heat transfer modeland a CFD model (Fig. 8) were createdfor computational efficiency to evalu-ate the effect of the cooling design onthe temperature of the roll shell byconsidering the symmetry of geome-try, air flow, and heat convection onthe roll. The heat transfer model hassolid parts that include the roll shell

wall and the journals. The outer sur-face of the roll shell was heated fromthe hot air in the furnace, as shown inFig. 8. The CFD model includes thefluid that is air in this analysis. Thetemperature-dependent properties areshown in Table 2. The CFD boundaryconditions include inlet pressure (0.62MPa), outlet pressure (0.1 MPa), no-slip/no-penetration conditions at awall, and air temperature at the inlet(37C). The inlet is the center hole andthe outlet is the four holes, as shownin Fig. 6B. The analysis was conducted usingthe cosimulation capability of theABAQUS/CFD module and theABAQUS/standard module. ABAQUS/CFD simulated the air flow inside the

roll and ABAQUS/standard modeledthe heat transfer between the air andthe roll and in the roll. For a smalltime increment, a CFD analysis wasconducted to predict temperature,pressure, and velocity of the air andoutput heat flux in the interface be-tween the air and the roll. A heattransfer analysis was conducted by in-putting the heat flux from the CFDanalysis to predict the temperature inthe roll and output the temperature inthe interface between the air and theroll. This analysis process was repeateduntil the end of the simulated time.Since each time increment was small,this coupled heat transfer and CFDanalysis was computationally expen-sive. Therefore, a one-fourth model

Fig. 14 — Effective creep strain and maximum principal stress for EB welding. A — Effective creep strain; B — maximum principal stress.

Fig. 15 — Temperature of the roll with inside cooling and outsideheating.

Fig. 16 — Predicted effective creep strain for ten loading cycleswith inside cooling.

A B


WELDING RESEARCH


was used for this coupled analysis in-stead of a full 3D model.

Analysis Results andDiscussions

Effect of Coolto Temperatureon CreepFatigue Strain andStress

The temperature history of thefailed roll showed that the roll experi-enced an operating temperature be-tween 982 and 1066C, with furnacetemperature variations more than500C. Occasionally, the furnace tem-perature dropped to 204 and 427Cbetween temperature cycles. To identi-fy the effect of the cool-to tempera-ture between cycles on the creep-fatigue strain and stress, sequentiallycoupled heat-transfer and creep-fatigue analyses were conducted fortwo cases. Case 1 cooled the furnace toa temperature of 204C, and Case 2cooled the furnace to a temperature of427C between thermal cycles. Figure 9 shows the predicted distri-butions of effective creep strain forCase 1 and Case 2. The distributions ofeffective creep strain are identical be-tween Case 1 and Case 2 since the onlydifference between these two cases isthe temperature drops between load-ing cycles. Weld toes and weld roothave high effective creep strain, ashighlighted at locations 1–3 in Fig. 9.

Location 1 has the highest creepstrains among the three locations. Thepredicted creep strain is low (0.29%).Ten cycles were modeled, which areequivalent to a two-month time peri-od. With the time increasing, the cu-mulated creep strain will be higher. Al-though the predicted creep strain islow, an elastic-plastic-creep analysishas to be performed to model the plas-tic and creep process during roll serv-ice. A sensitivity study showed that apure elastic-plastic analysis cannotpredict the same stresses as the elas-tic-plastic-creep analysis. Figure 10 shows the predicted dis-tributions of the maximum principalstress for Case 1 and Case 2. Case 1has a much higher maximum principalstress than Case 2. Compressivestresses were shown in the bottom ofthe weld with tensile stresses aroundit. These kinds of stress distributionsresulted from the difference of theCTE expansion between the filler met-al (N117) and the base metal (MO-RE®1). The CTE of the base metal MO-RE®1 is higher than the filler metalN117. At high temperatures, the basemetal may expand more than the fillermetal, while the filler metal restrainsthe base metal expansion to producecompressive plastic strains so the di-mensions of the base metal becomesmall. After cooling to a lower temper-ature, the filler metal will pull the basemetal to the original dimensions.Thus, high tensile stresses show in thebase metal and compressive stresses

show in the bottom of the weld. Figure 11 shows that a crack initiat-ed from the weld root and then propa-gated through the weld. The weldmacrograph was prepared by cuttingone of the failed rolls. The crack ismost likely induced by the high tensilemaximum principal stress shown atthe weld root in Fig. 10A. Therefore,the experimental results indirectlyverified the model predictions. Al-though compressive stresses areshown in the weld before cracking, thestress distributions could be changedafter cracking propagates through theweld. The tension induced from thethermal and mechanical loads wouldovercome the compressive stress andinduce tensile stresses in the crack tipduring the crack propagation. Figure 12 shows the evolution oftemperature, effective creep strain,and maximum principal stress at theweld root for Case 1. Effective creepstrain and maximum principal stressincreases as the time increases. Themaximum principal stress increase re-sults from the plastic strain accumula-tion. This result implies that the weldroot is most likely the failure locationsince the stress range increases as thetime increases. The weld-root crackshown in Fig. 11 confirms this analysisresult. Figure 13A shows a hoop stress dis-tribution in Case 2 where the weld isin compression, and the base metalnear the weld is in tension. Hoopstress was plotted to understand the

Fig 17. — Predicted maximum principal stress near the weld rootafter ten loading cycles with inside roll cooling.

Fig. 18 — Predicted temperature, effective creep strain, and maximum principal stress near the weld root with inside roll cooling.


WELDING RESEARCH


near-weld cracks found in the failedroll, as shown in Fig. 13B. The stressmagnitude is low because the roll is ata temperature of 427C. If the rollcools to room temperature, the stressmagnitude would be higher. Althoughthe modeled weld has a different ori-entation than the failed roll, the hoopstress pattern will be the same be-tween the modeled weld and the failedweld. The tensile stress near the weldcould induce the cracks near the weldas shown in Fig. 13C, D. This study suggests that the cool-totemperature has a significant effect oncreep strain and stress during temper-ature cycles. Therefore, controlling thecool-to temperature between cycleswill reduce the creep-fatigue strainand stress to extend the creep-fatiguelifetime. In addition, the analyses alsoshows that the difference in materialproperties between the weld (N117)and the base metal (MO-RE®1) is theroot cause of roll failure. If the samematerial is used for both the weld andbase metal, the high creep-fatiguestrain and stress could be reduced.

Improvement Study 1: ElectronBeam Welding

The study of the effect of cool-totemperature between cycles on thecreep-fatigue process shows that thedifferent material properties betweenthe base material and the filler metalare the root cause of high creep-fatigue strain and stress at the weldtoes and the weld root. One option toreduce the high creep-fatigue strainand stress is to use the same materialfor both the filler metal and the basemetal during the welding of the rollusing FCAW. Another option is to usea newer welding process such as EBWwithout any filler metals. Since usingthe base metal as filler metal could re-sult in weldability problems duringFCAW of the roll, EBW was studied toexplore the possibility of improvingthe creep-fatigue life. After weldingusing EBW, PWHT will be conductedto eliminate the weld residual stressesand restore material properties. There-fore, weld residual stress does notneed to be included in the creep-fa-tigue analysis. The weld material prop-erties could be assumed to be the sameas the base material properties. A ten-cycle creep-fatigue analysis

was conducted by replacing the mate-rial properties in the weld with thebase material properties while keepingthe other conditions the same as Case1 to study the effect of an electron-beam welded joint on creep-fatiguestrain and stress. Figure 14 shows thepredicted effective plastic strain andmaximum principal stress with EBWwithout the use of filler metal. The ef-fective creep strain plot uses the samescale as in Fig. 9. The striking differ-ence in Figs. 14A and 9 illustrates theelimination of the localized concentra-tion of an effective creep strain in Fig.14A. Similarly, maximum principalstress is plotted as the same scale usedin Fig. 10. It is found that high stressat the weld root disappears, as shownin Fig. 14B. Analysis results show that EBWcould be an effective method to im-prove the creep-fatigue life of the fur-nace roll in the continuous hot-dipcoating line. A preliminary welding

test shows that EBW could be used toproduce a defect-free weld of MO-RE®1. A further study was conductedto predict the creep strain and stressin the electric beam welded roll by in-cluding the weld size, shape, and resid-ual stress resulting from EBW (Ref.30). Analysis results confirmed thatEBW is an effective method to signifi-cantly reduce creep strain and stress.

Improvement Study 2:Inside Cooling

A material’s resistance to creepgreatly depends upon the tempera-ture to which the material is exposed.If the roll operating temperature canbe lowered, the creep life of the rollcould be extended. A method was nu-merically examined by applying coolair inside of the roll to lower the rollshell wall temperature. The tempera-ture on the roll shell is the result ofcompeting effects of inside surface

Fig. 19 — Cooling system starting temperature.

Fig. 20 — Fluid temperature evolution during inside cooling.


WELDING RESEARCH


cooling and outside surface heatingby hot combustion flue gas originat-ing from the direct fire furnace im-mediately below. A heat transfer analysis was con-ducted using the analysis proceduresdiscussed in the Finite Element Mod-el section, where it was assumed thatthe cool air flows quickly inside theroll and could keep a temperature of204C (400F) without being heatedby the roll inner surface. The furnacetemperature was cyclic from 1066 to204C. Figure 15 shows the predictedtemperature at three points on theroll shell for ten cycles. The predictedtemperatures on the roll outer sur-face (N147), on the roll inner surface(N148), and on the inner surface ofthe end bell (N3284) are 668, 651,and 646C, respectively. The temper-ature on the roll outer surfaces re-duces from 1066 to 668C. Creep-fatigue analysis was conduct-ed using the analysis procedures dis-cussed in the Finite Element Modelsection by inputting the predictedtemperature history from the heattransfer analysis and applying me-chanical loads. Figure 16 shows thepredicted distribution of effectivecreep strain with inside cooling. Effec-tive creep strain was significantly re-duced by inner surface cooling of theroll by air, as illustrated by compar-isons of Figs. 16 and 9. Figure 17 shows the distributionof the predicted maximum principalstress with inside roll cooling. Stressat the weld root was reduced signifi-cantly with inside cooling, as illus-trated by comparisons of Fig. 10A(without inside cooling) to Fig. 17(with inside cooling). Figure 18shows the evolution of temperatureand maximum principal stress withinside cooling. The stress range withinside cooling is about 105 MPa,while the stress range without insidecooling is about 140 MPa, resultingin a 25% reduction. The numerical evaluation of theinside cooling concept shows that theinside cooling method can lower theroll shell temperature, reduce thecreep strain at the weld toe, and re-duce the stress range at the weldroot. Therefore, this method could bean effective way to improve thecreep-fatigue life of the furnace rollin a production line.

Feasibility Study of InsideCooling

A fully coupled heat transfer andCFD analysis was conducted using theone-fourth model (Fig. 8) to evaluatethe effect of the cooling system designinside the roll on the temperature ofthe roll shell. In the design, cool airflowed into the roll through a centertube and out the roll via the four holesaround the center hole. The outer sur-faces of the roll were heated from thehot air in the furnace. Before turning on the cool air, theroll was heated to the operating tem-perature according to the furnace-heating process — Fig. 5. Figure 19shows the roll temperature at the startof cooling utilizing the cool air. To ap-ply the cooling process smoothly, thecool-air pressure was linearly rampedup from atmospheric pressure to 0.62MPa at the inlet in one minute. Figure 20 shows the predicted airtemperature inside the roll at 3, 15, and27 s. At 3 s, the inlet pressure is 0.130MPa. The air temperature near the endof the center tube is heated up by thehot roll shell. The air temperature isabout 525 K. As time increases, the airtemperature increases. At 27 s, the airtemperature inside the roll reaches 800K. Therefore, instead of cooling the rollshell using the flowing air, the air isheated by the roll shell. The design isnot efficient to cool the roll shell.

Conclusion Multiphysics analyses, including aheat transfer analysis, a creep-fatigueanalysis, and a computational fluid dy-namics analysis, were conducted tounderstand the failure of a welded rollin the vertical furnace of a continuoushot-dip coating line and propose solu-tions to improve the creep-fatiguelifetime. The heat transfer analysiswas conducted to predict the temper-ature history of the welded roll bymodeling heat convection heatingfrom hot air inside the furnace usingABAQUS and a user-developed sub-routine. The creep-fatigue analysiswas performed by inputting the pre-dicted temperature history, applyinga mechanical load, and definingboundary conditions. The CFD analy-sis was used to evaluate a cooling sys-

tem design that intends to cool theroll by flowing high-pressure cool airinside the roll. Based on the analysisresults, the following conclusionscould be drawn: • Three factors: difference of mate-rial properties between the filler metaland base metal, furnace temperaturevariation during service, and high op-erating temperature contributed tothe roll failure. • Reducing the furnace temperaturevariation during roll service can lowerthe stress in the welded joint to im-prove the creep-fatigue life of the roll. • Using EBW without filler metal toreplace FCAW to weld the roll, thecreep strain and stress in the interfacebetween the filler metal and the basemetal can be eliminated so the creep-fatigue life of the roll can be improved. • Applying cool air inside the rollcan lower the roll temperature to re-duce the creep strain and stress to in-crease the roll lifetime, but it requiresan effective cooling system design. • The predicted high tensile hoopstress can be used to explain the ob-served cracks near the weld along thedirection perpendicular to the roll hoopdirection, and the predicted highly local-ized maximum principal stress at theweld root can be used to explain the ob-served crack at the weld root, which val-idates the models indirectly.

1. Ewald, J., Sheng, S., Klenk, A., andSchellenberg, G. 2001. Engineering guideto assessment of creep crack initiation oncomponents by two criteria diagram. Inter-national Journal of Pressure Vessels and Pip-ing 78: 937–949. 2. Nikbin, K. 2013. Creep/fatigue crackgrowth testing, modeling and componentlife assessment of welds. Procedia Engineer-ing 55: 380-393. 3. Marie, S., and Delaval, C. 2001. Fa-tigue and creep-fatigue growth in 316stainless steel cracked plates at 650°C. In-ternational Journal of Pressure Vessels andPiping 78: 847-857. 4. Vacchieri, E., Holdsworth, S. R., Cost,A., Poggio, E., Riva, A., and Villari, P. 2014.Creep-fatigue interaction in two gas tur-bine Ni based superalloys subjected toservice-like conditions. Materials at HighTemperatures 31(4): 348–356. 5. Ji, D. M., Zhang, L. C., Ren, J., andWang, D. 2015. Creep-fatigue interactionand cyclic strain analysis in P92 steel based

References


WELDING RESEARCH


on test. Journal of Materials Engineeringand Performance 24(4): 1441–1451. 6. Goswami, T., and Hanninen, H. 2001.Dwell effects on high temperature fatiguedamage mechanisms: Part II. Materials andDesign 22(3): 217–236. 7. Karl, J. O. 2013. Thermomechanicalfatigue life prediction of notched 304stainless steel. PhD dissertation, Universi-ty of Central Florida. 8. Shi, D., Dong, C., Yang, X., Sun, Y.,Wang, J., and Liu, J. 2013. Creep and fa-tigue lifetime analysis of directionally so-lidified superalloy and its brazed jointsbased on continuum damage mechanics atelevated temperature. Materials and Design45: 643–652. 9. Goswami, T. 2004. Development ofgeneric creep-fatigue life prediction mod-els. Materials and Design 25(4): 277–288. 10. Cui, L., and Wang, P. 2014. Two life-time estimation models for steam turbinecomponents under thermomechanicalcreep-fatigue loading. International Journalof Fatigue 59: 129–136. 11. Zhu, S. P., Huang, H. Z., Liu, Y.,Yuan, R., and He, L. 2012. An efficient lifeprediction methodology for low cycle fa-tigue-creep based on ductility exhaustiontheory. International Journal of Damage Me-chanics 22(4): 556–571. 12. Shi, D., Dong, C., Yang, X., Sun, Y.,Wang, J., and Liu, J. 2013. Creep and fa-tigue lifetime analysis of directionally so-lidified superalloy and its brazed jointsbased on continuum damage mechanics atelevated temperature. Materials and Design45: 643–652. 13. Chakherlou, T. N., Aghdam, A. B.,Akbari, A., and Saeedi, K. 2010. Analysis ofcold expanded fastener holes subjected toshort time creep: finite element modellingand fatigue tests. Materials and Design31(6): 2858–2866. 14. Peng, J., Zhou, C. Y., Dai, Q., and He,X. H. 2015. Dwell fatigue and cycle deforma-

tion of CP-Ti at ambient temperature. Mate-rials and Design 71: 1-1. 15. Chakherlou, T. N., and Aghdam, A.B. 2008. An experimental investigation onthe effect of short time exposure to elevat-ed temperature on fatigue life of cold ex-panded fastener holes. Materials and De-sign 29(8): 1504–1511. 16. Gurumurthy, K., Srinivasan, B., Kr-ishna, P. S., Achary, G. G. S., and Subra-manyam, S. V. R. 2014. Creep-fatigue de-sign studies for process reactor compo-nents subjected to elevated temperatureservice as per ASME-NH. Procedia Engi-neering 86: 327–334. 17. Sheridan, M., Knowles, D., Mont-gomery, O. 2013. Comparison of R5 andASME NH Creep-fatigue damage assess-ment methodologies. ASME 2013 PressureVessels and Piping Conference Vol. 1A(PVP2013-97625), doi:10.1115/PVP2013-97625. 18. Ainsworth, R. A. 2006. R5 proce-dures for assessing structural integrity ofcomponents under creep and creep-fatigueconditions. International Materials Reviews51(2): 107–126. 19. Knowles, D. M. 2014. R5 high tem-perature creep-fatigue life assessment foraustenitic weldments. Procedia Engineering86: 315–326. 20. Cross, M., Croft, T. N., McBride, D.,Slone, A. K., and Williams, A. J. 2007. Mul-tiphysics modeling and simulation:progress and challenges. NAFEMS WorldCongress, Vancouver, Canada. Retrievedfromwww.nafems.org/downloads/NWC07/keynotes/cross.pdf. 21. Yang, Y. P., and Mohr, W. C. 2015.Finite element creep-fatigue analysis of awelded furnace roll for identifying failureroot cause. Journal of Materials Engineeringand Performance 24(11): 4388–4399. 22. Yang, Y. P., and Mohr, W. C. July17–21, 2016. Weld residual stress and

creep-fatigue analysis of a furnace roll in acontinuous hot-dip coating line. Proceed-ings of 2016 ASME Pressure Vessels and Pip-ing Conference (PVP2016), Paper 63733,Vancouver, BC, Canada. 23. British Stainless Steel Association.Elevated temperature physical propertiesof stainless steels. Retrieved fromwww.bssa.org.uk/ topics.php?article=139. 24. Allegheny Technologies Incorporat-ed, ATI 309TM/ATI 309STM/ATI310TM/ATI310STM, Technical Data Sheet.Retrieved fromatimetals.com/Documents/ati_309_309s_310_310s_tds_en3_1_v1.pdf. 25. Duraloy Technologies, Materialspecification MO-RE 1 Alloy, duraloy.com. 26. Harald Pihl AB, Ni-Cr-Co-Mo weldingelectrode, INCONEL Welding Electrode 117.Retrieved from www.haraldpihl.com/ attach-ments/119_Inconel%20WE%20117.pdf. 27. Simulia. 2014. ABAQUS 6.14 usersguide. 28. Yang, Y. P., Brust, F. W., and Oh, J.July 20–24, 2003. Creep behavior of abimetallic welded joint in a nuclear pipingsystem. ASME 2003 Pressure Vessels and Pip-ing Conference, pp. 35–42, doi:10.1115/PVP2003-2045, Cleveland, Ohio. 29. Nabarro, F. R. N. Do we have an ac-ceptable model of power-law creep? 2004.Material Science Engineering A 387–389:659–664. 30. Brust, F. W., Scott, P. M., and Yang, Y.P. 2003. Weld residual stresses and crackgrowth in bimetallic pipe welds. Proc. the17th Int. Conf. on Structural Mechanics in Re-actor Technology. Prague, Czech Republic. Re-trieved fromiasmirt.org/transactions/17/G08-1.pdf. 31. Versteeg, H. K., and Malalasekera,W. 2007. An Introduction to ComputationalFluid Dynamics (Second Edition). PearsonEducation Limited. 32. Retrieved from www.engineering-toolbox.com/dry-air-properties-d_973.html.


Multiphysics Modeling of a Welded Furnace Roll for ...€¦ · WELDING RESEARCH NOVEMBER 2016 /...

Documents

Transcript of Multiphysics Modeling of a Welded Furnace Roll for ...€¦ · WELDING RESEARCH NOVEMBER 2016 /...