Validity of Heat Transfer Coefficient based on Cooling Time, Cooling Rate, and Heat Flux on Jominy End Quench Test A. Sugianto Graduate School of Engineering, Utsunomiya University, Utsunomiya, Japan M. Narazaki, M. Kogawara, and A. Shirayori Dept of Mechanical Systems Engineering, Utsunomiya University, Utsunomiya, Japan Keywords: Heat Transfer Coefficient, Cooling Curve, Cooling Rate, Heat Flux, Simulation Abstract Jominy end-quenched test is one of tools to know the hardenability character of a steel being hardened by heat treatment. As a coupon test, successful prediction of mechanical property and microstructure on Jominy test piece will increase the accuracy prediction of actual hardened steel parts. In this study, two-dimensional finite-element analysis was performed to predict thermal history and radial distortion of end-quenched steel bar using FEM-based software DEFORM-HT. It incorporates transformation-induced volume change, latent heat, transformation plasticity, and temperature-dependent material properties of individual phase of austenite, pearlite, bainite, and martensite. The following assumption was made; simplified diffusion at heating, diffusional transformation from autenite to bainite and pearlite at slow cooling, and diffusionless transformation from austenite to martensite at fast cooling. There are two thermal boundary condition used in this study i.e. water cooled area and still-air cooled area of the Jominy model. To study the validity of Heat Transfer Coefficient (HTC) employed on water-cooled area, the HTC was taken from several works of Narazaki, Homberg, Shorin, Inoue, and Hunkel. Measured and predicted cooling time and cooling rate curve inside Jominy stainless steel 34 then is compared to validate the accuracy of the HTC. Time-dependent heat flux is also reported to analyze the heat transfer mechanism. The most accurate employed HTC, Narazaki and Shorin method, then is applied on Jominy carbon steel S45C and alloy steel SCr42H. Predicted cooling curve, Jominy hardness and Jominy cooling time, and radial distortion then is compared to the measured one with relatively good agreement. Introduction Computer simulation in heat treating industry is widely applied to design the process parameter and to predict mechanical property microstructure relationship. Due to complexity of cooling characteristic given by real fabricated parts, the simplest way to test the reliability of the computer simulation on predicting the property of hardenable steel is through Jominy endquenched hardenability test. The simulation of heat-treated Jominy bar will be useful for validating the accuracy of the input data. It is because of simple shape and simple cooling method. However, successful prediction greatly depends upon material property data, boundary condition, and heat transfer coefficient (HTC). This study limits the influence of HTC on prediction accuracy of temperature-cooling time curve and temperature-cooling rate curve on 171
end-quenched austenitic stainless steel 34. Many works on prediction of HTC had been studied by numerous researcher based on mathematical modeling. However, the validity of HTC employed on accurate prediction of thermal history is rarely studied. Then, the most accurate HTC will be determined for predicting thermal history and property relationship as well as radial distortion of steel S45C and SCr42H. The most interest area is thermal boundary condition of water-cooled area at endquenched position. Narazaki used combination of lumped-heat capacity method and inverse method with the least residual method on water-cooled area, whereas the former method on aircooled area not separating between air convection and radiation [1]. Homberg assumed that HTC was less temperature dependent [2]. Hypotheses of Shorin based on global analysis of the lower surface results higher variation of the HTC [3]. Hunkel et al considered constant 32 W/(m 2 C) HTC at water-cooled area and zone-based HTC on peripheral surface (non contact with water) cooled by air convection and radiation [4]. Inoue simply predicted the HTC by inverse calculation for infinite half space [5]. The current work is to obtain air-cooled HTC from 93 o C (normalizing process of SUS34), whereas the previous work of Waki [6] performed on measured data of cooling curve, hardness, and radial distortion used for basis comparison. Materials and Methods Jominy test bar with shape, dimension, and position of 5 point measuring of thermal history is shown in the Fig. 1. Jominy end-quenched test was conducted using standardized height of free jet water flow, volumetric flow rate, and space distance between end-quenched specimen and discharge nozzle according to JIS G561 using equipment of Fig. 2 [7]. Five sheathed thermocouples with 1.6mm diameter were inserted into the cylindrical center of both specimen in the arranged order of 1.8, 5, 1, 25, and 5mm from end-quenched position. Material Selection There was three materials used i.e. austenitic stainless steel SUS34 (AISI34), carbon steel S45C (AISI 145), and alloy steel SCr42H (AISI512). The chosen material for studying heat treatment simulation is not only conducting joint research project between JSHT and JSMS but also for predicting distortion due to the fact that both steel are widely applied on throughhardened structural and case-hardened structural component [8]. JIS SCr42H is low alloy steel having chemical composition as follows.21%c,.26si,.82%mn,.16p,.17s,.2ni, 1.15Cr,.1Mo, and Fe as remainder. Whereas, S45C contains.47%c,.17si,.62%mn,.19p,.2s,.3ni,.13cr,.1mo, and Fe as remainder Heat Treating Process The steel bar was heat-treated under 1-hour holding time at heating temperature 85 o C (for SUS34 and S45C steel), and temperature 93 o C (for SUS34 and SCr42H). Rapid cooling by jet water on end-quenched area was conducted based on Jominy test method. However, for taking HTC on air-cooled area, SUS34 was heated to 93 o C followed by air-cooling into ambient temperature. Hence, there are 2 regions of cooling as a thermal boundary condition i.e. end-quenched region by 3 o C jet water flowing and side-and-top-quenched region by free convection and radiation of 3 o C ambient air. 172
The current work, SUS34 Jominy bar heated to 93 o C, was intended to take the thermal history on air-cooled area because this kind of stainless steel has no transformation-induced latent heat. The data acquisition of thermal history then was inputted into LUMPPROB computer program based on lumped heat capacity method, to calculate the HTC of air-cooled area [9]. In spite of different material and dimension of specimen used, the precision of obtained HTC on peripheral surface is relatively good agreement compared with other literatures as shown in left hand of Fig. 3 [1,11]. The previous work, SUS34 Jominy bar heated to 85 o C, was intended to take thermal history under Jominy testing. h air 1 1 1.8 25 5 5 12 M5x.5 h air Enlarged Bottom Area 848 Nodes 78 Elements Bore1.7 Dia25 h water Figure 1 Jominy Specimen with 5 wedged thermocouples and 2D FEM mesh symmetrical model Figure 2 Experimental equipment comprising of Pump (1), Heater (2), Valve (3), Thermocouple (4), Water tank (5), Jominy specimen (6), and Water quenchant (7) Simulation Process Heating and cooling simulation was undertaken by DEFORM-HT v9. FEM-based tool. Due to simple shape, two-dimensional calculation is fair enough with symmetrical geometry assumption. 4-noded quadrilateral meshed model can be seen on Fig. 1 having 848 nodes and 78 elements. Discrete model varies.3,.6,.8, 1, and 2mm mesh thickness from end quench to the top position. All node and element are treated as a rigid condition for SUS34 Jominy bar, whereas elasto-plastic condition for S45C and SCr42H Jominy bar. Restraint velocity is node 1 on the top position. Except validation by Hunkel method, there are two thermal boundary condition containing water-cooled area and air-cooled area. The air-cooled area at area except end-quenched region with temperature-dependent HTC is given in the Fig. 3 (left hand), whereas various HTC employed on water-cooled area is given in the Fig. 3 (right hand). The results of predicted temperature-cooling time and temperature-cooling rate curve inside SUS34 Jominy bar will be compared by measured one to choose the most precise HTC. Furthermore, the chosen HTC will be employed to S45C and SCr42H Jominy bar for predicting thermal history, hardness, and radial distortion. High linearity of physical, mechanical, and thermal property is assumed for the material model. During quenching of SUS34 Jominy bar, thermal property data includes conductivity, 173
specific heat, and density [12]. During quenching of S45C and SCr42H Jominy bar, the steel is assumed as a mixture of individual phase of austenite, pearlite, bainite, and martensite having particular input data. It therefore incorporates temperature-dependent property such as plastic flow stress property (yield strength, modulus of strain hardening), elastic property (Modulus of Young, Poisson ratio, and linear thermal expansion coefficient). Volume change due to transformation dilatation is also incorporated as well [13]. The coefficient of transformation plasticity of SCr42H and S45C came from different literature. Heat Transfer Coeff. (KW/m 2 C).15.125.1.75.5.25 Scott HTC Yamaguchi HTC SUS34 HTC 2 4 6 8 1 Temperature ( ) Heat Transfer Coeff (KW/m 2 C) 6 5 4 3 2 1 Narazaki-HTC Homberg-HTC Shorin-HTC Hunkel-HTC Inoue-HTC 2 4 6 8 1 Temperature ( ) Figure 3 Temperature-dependent HTC at air-cooled area (LH) and water-cooled area (RH) Results and Discussions Actually, only predicted and measured results of temperature cooling time curve, temperature cooling rate curve, heat flux curve of SUS34 and temperature cooling time curve, Jominy hardness-cooling time curve, and radial distortion distribution of S45C and SCr42H will be presented due to limitation of space in this report. Predicted temperature cooling time curve, cooling rate curve, and heat flux curve is required for validating the employed HTC. The most accurate of predicted curve to the measured curve will be chosen for further application of thermal history, hardness, and distortion simulation onto carbon and alloy steel. Temperature Cooling Time and Cooling Rate Curve on Stainless Steel Jominy Bar To evaluate the accuracy of prediction of thermal history, temperature cooling time curve is needed. Dennis stated that the use of TTT diagrams is more general but better agreement can be obtained by using continues cooling diagram with a cooling function close to the actual [14]. As shown in Fig. 4, there are 5 curves illustrating measured and predicted temperature cooling time curve that was obtained from 5 positions of thermocouple. The temperature cooling time graph was derived from embedded thermocouple measuring the temperature change as function of time. Measured time-dependent temperature change is amplified and linearized by a thermoelectric converter. With known frequency, the converter is connected to the PC for data acquisition. Via high-speed analog to digital conversion machine (Kyowa Electronic Instruments K.K. Sei ADC-116A), the analog data is converted into digital data. 174
Measured Predicted by NarazakiHTC Predicted by HombergHTC Predicted by ShorinHTC Predicted by HunkelHTC Predicted by InoueHTC Cooling Rate (C/s) 1 2 3 4 5 9 Cooling Rate (C/s) 9 5 1 15 2 75 6 45 3 Thermocouple No. 1 75 6 45 3 Thermocouple No. 2 15 15 1 2 3 4 5 6 Cooling Rate (C/s) 9 25 5 75 1 1 2 3 4 5 6 Cooling Rate (C/s) 1 2 3 4 5 9 75 6 45 3 Thermocouple No. 3 75 6 45 3 Thermocouple No. 4 15 15 1 2 3 4 5 6 1 2 3 4 5 6 Cooling Rate (C/s) 9 5 1 15 2 75 6 45 3 15 Thermocouple No. 5 1 2 3 4 5 6 Figure 4 Temperature cooling-time and cooling-rate curve on stainless steel SUS34 Jominy test bar 175
Concerning the prediction accuracy of thermal history inside the specimen during quenching, only temperature cooling time curve data is not enough for basis reference of prediction. The second important data is temperature cooling rate curve. The cooling rate should also be included when evaluating the thermal history [15]. As shown also in Fig. 4, there are 5 positions of thermocouple showing measured and predicted cooling rate. The temperature cooling rate curve was calculated from measured data of temperature cooling time curve using LUMPPROB Cooling Rate program [16]. The program is based on lumped heat capacity method assuming that uniform temperature inside and surface of the probe specimen and neglecting the small change of surface area. Studying the characteristic of cooling curve from thermocouple No. 1 to 5, the predicted cooling curve varies with the employed HTC. The variation of cooling curve becomes closer as the position of thermocouple is away from the end quench. There is a big variation among predicted cooling curve on thermocouple No. 1, whereas there is no variation of cooling curve on thermocouple No. 5. At 1.8mm distance from end quench has been strongly influenced by heat dissipation due to forced convection. At 5mm distance from end quench, the cooling mode is due to heat conduction and free convection. Thereby water-cooled HTC has no effect to the variation of predicted cooling curve on thermocouple No. 5. Consequently, predicted and measured cooling curve is perfectly matched. It indicates that SUS34 thermal property input data used is precise. Comparing between predicted and measured cooling time and cooling rate curve shown on thermocouple 1 to 4, none of the employed HTC gives the highest accuracy of prediction. The most accurate prediction of thermal history is given by employing the Narazaki HTC, Homberg HTC, and Shorin HTC showing that between predicted and measured cooling time curve and cooling rate curve almost matched. However, Narazaki HTC gives better agreement because predicted cooling curves are almost corresponding to the measured values. The worse agreement is achieved by employing both Hunkel HTC and Inoue HTC, in spite of incorporating the radiation heat transfer on peripheral surface. Both of the two are low enough to produce the cooling characteristic except thermocouple No. 5. Time-dependent Heat Flux on Stainless Steel Jominy Bar To evaluate the accuracy of HTC prediction on end quenched area, analysis of the timedependent Heat Flux (HF) is also required to support the analysis of cooling curve and cooling rate curve. During quenching into water either by free convection e.g. still quenching or by forced convection e.g. spray quenching or impinging water jet on Jominy test, it mostly includes three stages of heat transfer mechanism i.e. film boiling (vapor blanket), nucleate boiling (vapor collapse), and convective cooling stage. It is therefore HF and HTC should include the three stage of the heat transfer mechanism in which both are nonlinear temperature and time dependent. The location of maximum heat flux at the hot surface is very important for understanding the behavior of quenching phenomena. The location is neither in the film boiling stage, nor in the transition boiling stage, but in the fully wetted region where the convective cooling starts [17]. The heat flux density Q depends on HTC h, surface area A, end quench surface temperature T s, and ambient temperature T or linearly corresponds with heat capacity C and cooling rate dt s / dt according to equation 1. Narazaki HTC and Shorin HTC produce not only the higher variation of temperature dependency producing rapid cooling on thermocouple No.1, but also the three stage of heat transfer mechanism shown at Fig. 5. 176
Q dts ha( Ts T ) = C (1) dt = Heat Flux (KW/m 2 ) 2 175 15 125 1 75 5 25 Predicted by NarazakiHTC HombergHTC ShorinHTC HunkelHTC InoueHTC 2 4 6 8 1 Figure 5 Predicted time-dependent heat flux during cooling Prediction of Temperature-Cooling Time Curve on Carbon and Alloy Steel Jominy Bar Due to page limitation pages, Fig. 6 shows measured and predicted temperature cooling time curves using only Narazaki HTC plotted on respective TTT diagram of alloy steel SCr42H and carbon steel S45C. It can be seen that both measured and predicted curve nearly resemble each other. Even though 1.8mm point tracking does not give accurate prediction below 5 o C, however temperature cooling time between 8-5 o C, where this interval temperature is responsible to the transformed phase during cooling [18,19], is good agreement with the measured one. Thus, the employed HTC will be relatively accurate to predict the cooling curve of both SCR42H and S45C. Measured and predicted cooling curve on the other thermocouples (No. 2, No. 3, No. 4, and No. 5) agree well. It is therefore thermal property input data is sufficiently accurate for predicting volume fraction of bainitic and pearlitic transformation. 1 A 9 9 Ac3 A 8 Ac3 8 Ac1 7 1 Ac1 7 No.1 Measured 5. 5mm No.1 Measured 6 1 25 No.2 Measured 6 1.8 No.2 Measured 5. P No.3 Measured 25 5mm 5 No.3 Measured 5 1.8 No.4 Measured P No.4 Measured 4 B No.5 Measured 4 Ms No.5 Measured Ms No.1 Predicted 3 3 M5% No.1 Predicted No.2 Predicted No.2 Predicted 2 No.3 Predicted 2 M9% No.3 Predicted No.4 Predicted No.4 Predicted 1 No.5 Predicted 1 No.5 Predicted.1 1 1 1 1 1 1.1 1 1 1 1 1 1 Time (s) Time (s) Figure 6 Jominy cooling time curve of SCr42H (LH) and S45C (RH) Jominy test bar plotted on its TTT diagram 177
Prediction of Hardness Distribution on Carbon and Alloy Steel Jominy Bar As shown in Fig. 7, there are measured and predicted distribution of Jominy hardness and Jominy cooling time. Narazaki HTC and Shorin HTC were employed to both steel SCr42H (Left Hand) and steel S45C (Right Hand). It can be seen that Jominy cooling time has significant effect to the prediction accuracy of hardness distribution. Jominy hardness and Jominy cooling time of SCr42H coincide well with the measured one, whereas both distributions on S45C agree tendentiously on J 1 (Jominy distance 1mm) to J 1 (Jominy distance 1mm distance). The tendentious agreement on S45C may be caused by the accurate cooling time input data. It is because measured Jominy cooling time that was used for comparison on both steels, was taken from 8-5 o C cooling time from most structural steel [18]. However, Wefer and Rose did not clearly identify the exact steel grades. Therefore, the accuracy of cooling time data will not be same for both steels. In the case of SCr42H, the used cooling time is sufficient, whereas in the case of S45C, it will be insufficient. Hardness (HRC) 6 5 4 3 2 Measured Predicted by NarazakiHTC Predicted by ShorinHTC 3 25 2 15 1 Hardness (HRC) 6 5 4 3 2 Measured Predicted by NarazakiHTC Predicted by ShorinHTC 3 25 2 15 1 1 5 1 5 2 4 6 8 1 Distance from End Quench (mm) 2 4 6 8 1 Distance from End Quench (mm) Figure 7 Measured and predicted Jominy hardness and cooling time on SCr42H (LH) and S45C (RH) Prediction of Radial Distortion on Carbon and Alloy Steel Jominy Bar Fig. 8 shows measured and predicted radial distortion distribution after Jominy test of SCr42H alloy steel (Left hand) and S45C carbon steel (Right Hand) using both HTC. Predicted results are calculated from transformation plasticity coefficient K taken from literatures [2,21]. From the graph, it is clearly seen that there is 3 distinct regions of radial distortion i.e. expanded area, contracted area, and unaffected area. Expanded area consists of fully martensite (BCT structure) having crystallite size greater than bainite or pearlite (BCC structure). Couple between transformation plastic and volumetric strain strongly induce the total strain distribution so that radial expansion and contraction produce the final shape. Carbon content and alloying element has significant influence to the amount of expansion-contraction. In the case of SCr42H, good agreement has been made between J 5 and J 1, whereas J -J 5 agrees relatively. Radial distortion on S45C gives the tendentious agreement on along the distance from end quench to top position (J J 1 ). Further simulation by modifying latent heat, transformation volumetric, and transformation plastic strain may result better accuracy. 178
Radial Distortion (mm).6.5.4 Measured Predicted by Narazaki HTC Predicted by Shorin HTC.3.2.2.1.1 -.1 -.1 -.2 -.2 1 2 3 4 5 6 7 8 9 1 -.3 1 2 3 4 5 6 7 8 9 1 Distance from end-quenched (mm) Distance from end-quenched (mm) Figure 8 Measured and predicted radial distortion on SCr42H (LH) and S45C (RH) Conclusions The two HTC validated on austenitic stainless steel 34 from which the most workers derived from, is sufficiently accurate to predict thermal history and hardness distribution of carbon and alloy steel. However, for predicting distortion distribution on end-quenched steel still needs the phase transformation-related input data. The detail conclusion can be summarized as follows: During normalizing of Jominy bar made of stainless steel 34 from 93 o C, all thermocouples gave almost similar cooling curve, it is confirmed that all air-cooled surfaces have almost same HTC. Therefore HTC of air-cooled area will be treated as uniform cooling. Based on the corresponding value between predicted and measured 8-5 o C cooling time curve and cooling rate curve on stainless steel 34, the most accurate of water-cooled HTC was given by Narazaki, Homberg, and Shorin method. The three most accurate HTC is then validated by evaluating the time-dependent heat flux. Only by consideration of three stages of heat transfer mechanism will Narazaki HTC and Shorin HTC be considered as the most accurate HTC. Plotting the measured and predicted cooling time curve onto TTT diagram of both steels, the cooling curve of No.1 thermocouples is less accurate on low temperature. However, 8-5 o C cooling time shows reasonable agreement. Predicted hardness distribution has good agreement with the measured one. Jominy cooling time has strong influence to the accurate prediction of hardness distribution. Predicted radial distortion has relative agreement on radially expanded area around J -J 1. The input data related to the phase transformation i.e. latent heat, transformation volumetric, and transformation plastic strain may affect the total strain distribution. Acknowledgments The authors would like to express their gratitude to Scientific Forming Technologies Co., Columbus, USA and Yamanaka Engineering Co. Ltd., Chiba, Japan for the cooperation with the numerical simulation using FEM-based tool of DEFORM-HT. Radial Distortion (mm).6.5.4.3 Measured Predicted by NarazakiHTC Predicted by Shorin HTC 179
References [1] M. Narazaki, M. Kogawara, A. Shirayori, and S. Fuchizawa, Proc. of 4 th Intl Conf. on Quenching and Control of Distortion, ASM Intl, Material Park, OH, 23, p 97-14 [2] D. Homberg, A Numerical Simulation of the Jominy End-Quench Test, Acta Mater., Vol.44(No. 11), 1996, p 4375-4385 [3] A. Shorin, Rapport de stage - Centre d Etude des Structures et des Matériaux Navals (CESMAN), 1997 [4] M. Hunkel, T. Lubben, F. Hoffmann, and P. Mayr, Using the Jominy End-quench Test for Validation of Thermo-metallurgical Model Parameters, J. Phys. IV France, Vol. 12, 24, p 571-579 [5] T. Inoue and Y. Morimoto, Proc. of 1 st Intl Surface Engineering Congress and 13 th IFHTSE Congress, ASM Intl, Material Park, OH, 22, p 395-41 [6] M. Waki, Master Thesis, Dept of Mechanical System Eng, Utsunomiya Univ., 24, p 14 [7] JIS G561, Method of Hardenability Test for Steel (End Quenching Method), Japanese Standards Association, Tokyo, 1994, p 315-319 [8] Doc of Joint Research Results Symposium, JSHT and JSMS, Kyoto, Japan, 24, p 5-62 [9] M. Narazaki, M. Kogawara, A. Shirayori, and S. Fuchizawa, Proc. of 6 th Intl IFHTSE Congress, ASM Intl, Material Park, OH, 1997, p 428-435 [1] H. Scott, The Problem of Quenching Media for the Hardening of Steel, Trans. ASM, Vol. 22, 1934, p 577-64 [11] T. Yamaguchi et al., Mitsubishi Heavy Industries Technical Review, 6(1), 1969, p 3-39 [12] Heat Transfer Eng Doc, Japan Society of Mechanical Engineers (JSME), 4 th ed, Tokyo, 1986 [13] J. Fluhrer, DEFORM-HT (2D ver9.) Users Manual, Scientific Forming Technology Co Ltd, Columbus, 27 [14] S. Dennis, Prévision des Contraintes Résiduelles Induites par Traitement Thermique et Thermochimique, Revue de Métallurgie Vol. 77, (1997), p 158-176 [15] M. Narazaki, Private lectureship, Utsunomiya, 25 [16] M. Narazaki and G.E.Totten, Steel Heat Treatment Handbook, 2 nd Edition, Edited by G.E. Totten, CRC Press, Boca Raton, FL, 27, p 67-65 [17] J. Hammad, Y. Mitsutake, and M. Monde, Movement of Maximum Heat Flux and Wetting Front during Quenching of Hot Cylinder Block, Int. J. Thermal Sci., Vol. 43, 24, p 743-752 [18] F. Wefer and A. Rose, Atlas zur Warmebehandlung der Stahle, Verlag Stahleisen M.B.H., Dusseldorf, 1958, p 28 [19] B. Smoljan, Prediction of Mechanical Properties and Microstructure Distribution of Quenched and Tempered Steel Shaft, J. Mater. Proc. Technol., Vol. 175 (No. 1-3), 26, p 393-397 [2] Japan Society of Heat Treatment Quenching and simulation Research Group, JSHT, Tokyo, 22 [21] Japan Society of Heat Treatment Quenching and simulation Research Group, JSHT, Tokyo, 23 18