Effects of Cooling Rate during Quenching and Tempering Conditions on Microstructures and Mechanical Properties of Carbon Steel Flange

This study investigated the mechanical properties of steel in flanges, with the goal of obtaining high strength and high toughness. Quenching was applied alone or in combination with tempering at one of nine combinations of three temperatures TTEM and durations tTEM. Cooling rates at various flange locations during quenching were first estimated using finite element method simulation, and the three locations were selected for mechanical testing in terms of cooling rate. Microstructures of specimens were observed at each condition. Tensile test and hardness test were performed at room temperature, and a Charpy impact test was performed at −46 °C. All specimens had a multiphase microstructure composed of matrix and secondary phases, which decomposed under the various tempering conditions. Decrease in cooling rate (CR) during quenching caused reduction in hardness and strength but did not affect low-temperature toughness significantly. After tempering, hardness and strength were reduced and low-temperature toughness was increased. Microstructures and mechanical properties under the various tempering conditions and CRs during quenching were discussed. This work was based on the properties directly obtained from flanges under industrial processes and is thus expected to be useful for practical applications.


Introduction
The flange is a component used to interconnect pipelines and must provide a strong and reliable connection. Since pipelines are increasingly being developed for harsh environments such as high pressure, low temperature and corrosive atmospheres, flanges are thus also required to have excellent mechanical properties and reliability [1][2][3][4][5].
The mechanical properties of thick steel plates used for large structures such as flanges are controlled using post-heat treatment [6][7][8][9]. Quenching is used for the purpose of improving strength but it has a detrimental effect on low temperature toughness due to the formation of a hard secondary phase. Thus, tempering has been used to obtain a balance of strength and toughness by optimizing the microstructure [4,10]. In low carbon steels, it is well known that tempering after quenching induces decomposition of as-quenched martensite into tempered martensite and precipitates, static recovery of dislocation and release of residual stress [11][12][13][14]. Martensite after tempering loses strength but gains

Background
Low-carbon steel sheets have various microstructures and therefore various characteristics. Microstructures are distinguished into matrix and secondary phase. According to the morphological characteristics, the matrix is classified as granular bainitic ferrite (GBF), lath bainitic ferrite (LBF), acicular ferrite (AF) and quasi-polygonal ferrite (QPF), and the secondary phases are classified as martensite (M), martensite-austenite constituent (MA), degenerated pearlite/pearlite (DP/P) and carbide (C). GBF and LBF are two different morphologies of bainitic structure, based on descriptions in the literature [23,24]. GBF consists of block ferrite and a granular M/A constituent. GBF is transformed during continuous cooling and can transform over a wide range of CRs. For these reasons, GBF tends to form a major structure in low carbon steels that have cooled at various rates [17,[25][26][27]. After GBF transformation ends, LBF is formed in the region of retained austenite with rich carbon content, but not rich enough to form M/A constituent [11]. LBF has a parallel lath shape and has relatively higher strength than GBF due to the fine laths and rich carbon content [25][26][27].
The transformation temperature ranges of AF and GBF are similar; typically, AF is formed at a faster CR than GBF. AF nucleates within grains in austenite as lath-shaped needles that radiate irregularly in various directions to form basket weaves. These chaotically arranged AF laths impede crack propagation so the AF microstructure produces an excellent combination of strength and toughness [17,25,27].
QPF transforms at slow CRs at high temperatures. It has higher dislocation density than polygonal ferrite and has irregular grain boundaries. Secondary phases often form at QPF grain boundaries [25]. Compared to other matrices, OPF has the lowest strength but excellent ductility.
M is formed by cooling austenite at a greater than critical CR, which prevents carbon from spreading. As a result, carbon forms a solid solution and M is an unstable microstructure with high dislocation density. M has a lath phase in low-carbon steel, so it has very high hardness and strength, but M is brittle, so its presence degrades toughness. MA consists of martensite and retained austenite.
The distribution of martensite in MA depends on the stability of retained austenite and affects the mechanical properties of MA [4,17,[28][29][30].
P is a eutectoid structure composed of ferrite and cementite with a lamellar structure and forms at high temperatures with slow CR. The mechanical properties of P are between those of martensite and ferrite. DP has almost the same mechanical properties as P. DP is transformed by low carbon concentration or increased CR compared to that of P [31].
C is usually observed on grain boundaries and impurities form and aggregate at the end of phase transformation. C is also observed inside GBF and may be finely precipitated by tempering. The influence of C on the mechanical properties depends on the shape and size [32].

Materials and Methods
In this study, low-alloy steel ( Table 1) was used to make 10-inch-diameter weld neck flanges. The raw steel was heated at 1250 • C for 12 h in a furnace. Hot-forging and punching were performed at 850 to 1250 • C using a hydraulic press. The completed flanges were then subjected to post-heat treatment of the entire flange in a furnace ( Figure 1). All flanges were heated at a heating rate of 80 • C/h to 940 • C and held at that temperature for 4 h and quenched in a water bath that had a temperature of 25 to 50 • C. Afterward, flanges were reheated at 80 • C/h to T TEM = 450, 530 or 600 • C, held at that temperature for t TEM = 1, 3 or 5 h, then air-cooled. Specimens were named according to their tempering conditions (Table 2).
CRs during quenching by location of flange were estimated using FEM simulation (Forge NxT 3.0, Enginsoft, Rome, Italy). Thermal parameters for the heat transfer simulation during annealing and quenching were set as those of a low-alloy steel (C45 steel), provided by the commercial FEM software. Metal to metal contact and metal to fluid quenching were applied to the heat transfer simulation. Three rectangular zones (20 mm × 19 mm) were then selected (   [4,17,[28][29][30]. P is a eutectoid structure composed of ferrite and cementite with a lamellar structure and forms at high temperatures with slow CR. The mechanical properties of P are between those of martensite and ferrite. DP has almost the same mechanical properties as P. DP is transformed by low carbon concentration or increased CR compared to that of P [31].
C is usually observed on grain boundaries and impurities form and aggregate at the end of phase transformation. C is also observed inside GBF and may be finely precipitated by tempering. The influence of C on the mechanical properties depends on the shape and size [32].

Materials and Methods
In this study, low-alloy steel (Table 1) was used to make 10-inch-diameter weld neck flanges. The raw steel was heated at 1250 °C for 12 h in a furnace. Hot-forging and punching were performed at 850 to 1250 °C using a hydraulic press. The completed flanges were then subjected to post-heat treatment of the entire flange in a furnace ( Figure 1). All flanges were heated at a heating rate of 80 °C/h to 940 °C and held at that temperature for 4 h and quenched in a water bath that had a temperature of 25 to 50 °C. Afterward, flanges were reheated at 80 °C/h to TTEM = 450, 530 or 600 °C, held at that temperature for tTEM = 1, 3 or 5 h, then air-cooled. Specimens were named according to their tempering conditions (Table 2).     CRs during quenching by location of flange were estimated using FEM simulation (Forge NxT 3.0, Enginsoft, Rome, Italy). Thermal parameters for the heat transfer simulation during annealing and quenching were set as those of a low-alloy steel (C45 steel), provided by the commercial FEM software. Metal to metal contact and metal to fluid quenching were applied to the heat transfer simulation. Three rectangular zones (20 mm × 19 mm) were then selected ( Figure 2a  Microstructure observation, hardness test, tensile test and Charpy impact test were conducted using specimens taken at three locations (Figure 3a,b). To observe microstructure, 5 × 10 × 3 mm specimens were obtained along the vertical direction of the flange and then polished physically using SiC paper and diamond paste. Then, they were etched using 4% Nital solution (4% NHO3 + ethanol), and then the matrix was observed using an optical microscope (Olympus corporation, GX51F, Okayama, Japan) and the secondary phase was identified using a scanning electron microscope (JEOL, 7200F, Tyoko, Japan) and energy dispersive spectrometer (EDS) (Oxford Instruments, Nanoanalysis, Oxford, England). Using these microphotographs, microstructure types were classified according to their morphological features and their area fractions were calculated. To confirm the presence of MA constituents in the microstructure, the specimens were etched using Klemm's I solution (50 mL saturated aqueous Na2S2O3 + 1 g K2S2O5) for 40 to 50 s and then observed using an optical microscope. To understand phase evolution during tempering, equilibrium phases were predicted based on themodynamics calculation using Thermocalc with TCFE 9 database (Thermocalc, ver. 2020b).
Hardness tests were performed using a Rockwell hardness (HRC) tester (Hanbando commerce, HBD-DBE SW2259, Seoul, Korea) on specimens with dimensions of 13 × 13 × 5 mm taken along the vertical direction of the flange. Hardness was measured using a C-type scale. Nine measurements Microstructure observation, hardness test, tensile test and Charpy impact test were conducted using specimens taken at three locations (Figure 3a,b). To observe microstructure, 5 × 10 × 3 mm specimens were obtained along the vertical direction of the flange and then polished physically using SiC paper and diamond paste. Then, they were etched using 4% Nital solution (4% NHO 3 + ethanol), and then the matrix was observed using an optical microscope (Olympus corporation, GX51F, Okayama, Japan) and the secondary phase was identified using a scanning electron microscope (JEOL, 7200F, Tyoko, Japan) and energy dispersive spectrometer (EDS) (Oxford Instruments, Nanoanalysis, Oxford, England). Using these microphotographs, microstructure types were classified according to their morphological features and their area fractions were calculated. To confirm the presence of MA constituents in the microstructure, the specimens were etched using Klemm's I solution (50 mL saturated aqueous Na 2 S 2 O 3 + 1 g K 2 S 2 O 5 ) for 40 to 50 s and then observed using an optical microscope. To understand phase evolution during tempering, equilibrium phases were predicted based on themodynamics calculation using Thermocalc with TCFE 9 database (Thermocalc, ver. 2020b).
Hardness tests were performed using a Rockwell hardness (HRC) tester (Hanbando commerce, HBD-DBE SW2259, Seoul, Korea) on specimens with dimensions of 13 × 13 × 5 mm taken along the vertical direction of the flange. Hardness was measured using a C-type scale. Nine measurements were taken from each parts and then the maximum and minimum values were discarded and the average was calculated for the rest. machine (MTS criterio, model 45, Eden Prairie, MN, USA) and were gauged once at a crosshead speed of 1.5 mm/min at room temperature.
The Charpy impact test was performed on specimens that had been machined according to ASTM-E23 standard; these samples were taken along the horizontal direction of the flange. Charpy impact tests were conducted once at −46 °C using an impact tester (ZwickRoell, PSW750+TZE, Ulm, Germany).

Results and Discussion
Q-condition specimens had a matrix that included LBF, AF, GBF and QPF and secondary phases M, MA, DP/P and C ( Figure 4). These microstructure types were classified by morphological features (Table 3) [23][24][25]33,34]. The area fraction of phases ( Figure 5) was measured three times in each and averaged using an image analyzer [35]. C microstructures were smaller than other microstructures and very rare, so C was excluded due to difficulties in identification and measurement of area fraction.
GBF was a major microstructure at all CRs ( Table 3). As CR increased, the fractions of LBF, AF and M increased but the fractions of QPF and DP/P decreased. Tensile test specimens were round according to sub-size standard of ASTM-E8 and were prepared along the horizontal direction of the flange. The tests were conducted using a tensile testing machine (MTS criterio, model 45, Eden Prairie, MN, USA) and were gauged once at a crosshead speed of 1.5 mm/min at room temperature.
The Charpy impact test was performed on specimens that had been machined according to ASTM-E23 standard; these samples were taken along the horizontal direction of the flange. Charpy impact tests were conducted once at −46 • C using an impact tester (ZwickRoell, PSW750+TZE, Ulm, Germany).

Results and Discussion
Q-condition specimens had a matrix that included LBF, AF, GBF and QPF and secondary phases M, MA, DP/P and C ( Figure 4). These microstructure types were classified by morphological features (Table 3) [23][24][25]33,34]. The area fraction of phases ( Figure 5) was measured three times in each and averaged using an image analyzer [35]. C microstructures were smaller than other microstructures and very rare, so C was excluded due to difficulties in identification and measurement of area fraction.
GBF was a major microstructure at all CRs ( Table 3). As CR increased, the fractions of LBF, AF and M increased but the fractions of QPF and DP/P decreased.
In MC-Q (Figure 4c,d), the matrix was mainly composed of GBF (51 ± 4.7% area fraction) but included QPF (6 ± 3.4%), and the secondary phases were M (11 ± 3.1%) and DP/P (10 ± 0.7%); compared to FC-Q, this was a lower area fraction of Am and a higher area fraction of DP/P. The area fraction of MA was still low (2 ± 0.1%), and the grains were coarser than in FC-Q.
In SC-Q (Figure 4e,f), the matrix was composed of similar area fractions of OPF (39 ± 7.9% area fraction) and GBF (37 ± 5.9%); AF and LBF were not observed. The main secondary phase was DP/P (15 ± 1.4%); M (7 ± 1.3%) was relatively uncommon. The area fraction of MA was the same as in FC-Q and MC-Q. Grains in SC-Q were coarser than in FC-Q and MC-Q, because in SC-Q, the CR was slow so the specimen stayed at a high temperature for a relatively long time.
In MC-Q (Figure 4c,d), the matrix was mainly composed of GBF (51 ± 4.7% area fraction) but included QPF (6 ± 3.4%), and the secondary phases were M (11 ± 3.1%) and DP/P (10 ± 0.7%); compared to FC-Q, this was a lower area fraction of Am and a higher area fraction of DP/P. The area fraction of MA was still low (2 ± 0.1%), and the grains were coarser than in FC-Q.
In SC-Q (Figure 4e,f), the matrix was composed of similar area fractions of OPF (39 ± 7.9% area fraction) and GBF (37 ± 5.9%); AF and LBF were not observed. The main secondary phase was DP/P (15 ± 1.4%); M (7 ± 1.3%) was relatively uncommon. The area fraction of MA was the same as in FC-Q and MC-Q. Grains in SC-Q were coarser than in FC-Q and MC-Q, because in SC-Q, the CR was slow so the specimen stayed at a high temperature for a relatively long time.
The area fractions of MA were not affected by CR, unlike the other microstructure types. The effect of CR on the formation of MA has not been established. Previous studies have reported that the area fraction of MA increased [26,35] or decreased [36,37] as CR increased.      The area fractions of MA were not affected by CR, unlike the other microstructure types. The effect of CR on the formation of MA has not been established. Previous studies have reported that the area fraction of MA increased [26,35] or decreased [36,37] as CR increased.
The area fraction of MA was significantly reduced after tempering [38,39]. Q specimens (Figure 5a,b,c) had numerous small MA inclusions. In 450_1 specimens, some of the relatively coarse MA remained, but the rest disappeared (Figure 5d,e,f). In 450_5 specimens (Figure 5g,h,i) with increased tempering time and in 600_1 specimens (Figure 5j,k,l) with increased T TEM , almost all of the MA disappeared. In 600_5 (Figure 5m,n,o) with increases in both T TEM and t TEM , MA was not observed. From the literature [4,10,29,30], the disappearance could be explained by decomposition of MA to ferrite matrix and cementite. Decomposition of MA could be expected to reduce hardness and strength and improve toughness. However, the area fraction of MA was 2% in all specimens before tempering (Table 3) and so should not affect their mechanical properties.
Tempering decomposed the secondary phases. As T TEM and t TEM increased, cementite became spheroidized and coarsened. Spheroidization of cementite is thermodynamically driven by a decrease in ferrite/cementite interfacial energy. Small cementite belongs to large cementite by Ostwald ripening. These cementites are coarsened and the total surface area to volume ratio is decreased [40,41].
M content decreased as T TEM and t TEM increased. Comparison to the Q specimen (Figure 6a) indicated that the decomposition of M started in the 450_1 specimen ( Figure 6d); pre-existing M broke up into ferrite and cementite, and then they mixed with fine cementite and precipitated after tempering. In the 450_5 specimen (Figure 6g) and the 600_1 specimen (Figure 6j), M was decomposed completely into fine ferrite and tempered martensite. Cementite became coarsened and nearly spherical in the 600_5 specimen (Figure 6m).  Materials 2020, 13, 4186 9 of 16 DP/P content was affected by T TEM and t TEM . DP/P content did not differ noticeably between Q ( Figure 6b) and 450_1 specimens (Figure 6e). In 450_5 (Figure 6h) and 600_1 specimens (Figure 6k), pre-existing degenerated cementite of DP was spheroidized, and the edges of lamellar cementite of P became rounded. In the 600_5 specimen (Figure 6n), spheroidized cementite was coarsened and non-degenerated lamellar cementite was still present.
C content was not affected by T TEM or t TEM and did not differ noticeably between Q specimens ( Figure 6c) and 450_1 specimens (Figure 6f). This indicates the occurrence of the decomposition and coarsening of C in 450_5 (Figure 6i) and 600_1 specimens (Figure 6l). In the 600_5 specimen (Figure 6o), C was spheroidized in comparison to the specimens of 450_5 and 600_1.
It was attempted to identify carbides in decomposed M during tempering and the results are depicted in Figure 7. As can be seen in Figure 7a,b, EDS analyses of carbides in 450_1 and 600_5 specimens clearly revealed that carbide formation was promoted with increasing tempering time and temperature. The stoichiometry of the carbides was close to M 3 C, cementite carbides. These results were well matched with the stable phase predicted by equilibrium thermodynamics calculation, as shown in Figure 7c. Within the present tempering temperatures, cementite was predicted to be stable. This is also in good agreement with previous reports in which decomposition of M into M 3 C cementite carbides was observed [10,42]. The mechanical properties of the steels were affected by the decomposition of the secondary phase and the decomposition velocity of each type according to tempering conditions. M decomposed faster than DP/P under the same tempering conditions; this phenomenon has been reported previously [43,44]. M is unstable at high temperatures and has a pronounced tendency to decompose to ferrite and cementite, which are thermodynamically stable microstructures. The mechanical properties of the steels were affected by the decomposition of the secondary phase and the decomposition velocity of each type according to tempering conditions. M decomposed faster than DP/P under the same tempering conditions; this phenomenon has been reported previously [43,44]. M is unstable at high temperatures and has a pronounced tendency to decompose to ferrite and cementite, which are thermodynamically stable microstructures. Furthermore, carbon's activation energy for diffusion is low, so carbon has a strong tendency to escape the microstructure. In contrast, DP/P is stable and has markedly lower spheroidizing velocity than that of unstable structures [40]. Moreover, tempering drives various phenomena such as the release of residual stress, decline in dislocation density and appearance of fine precipitates. These phenomena combine with the transformation of MA and decomposition of the secondary phase identified to control the mechanical properties of the steels [4,10,19,30,45].
As the CR decreased, hardness under the same tempering condition decreased (Figure 8). Hardness decreased as a result of the change from LBF and AF to QPF in the matrix and from M to DP/P in the secondary phase. FC-Q, MC-Q and SC-Q had hardness values of 19, 15.7 and 11.7 HRC, respectively. Grain size in each specimen was not quantified, so the contribution of the grain size could not be calculated. However, as CR increased, the grain size decreased and the hardness increased, so we infer that the grain refinement depending on the CR also influenced the increased hardness [18,20,46]. Materials 2020, 13, x FOR PEER REVIEW 11 of 17 As the CR decreased, hardness under the same tempering condition decreased (Figure 8). Hardness decreased as a result of the change from LBF and AF to QPF in the matrix and from M to DP/P in the secondary phase. FC-Q, MC-Q and SC-Q had hardness values of 19, 15.7 and 11.7 HRC, respectively. Grain size in each specimen was not quantified, so the contribution of the grain size could not be calculated. However, as CR increased, the grain size decreased and the hardness increased, so we infer that the grain refinement depending on the CR also influenced the increased hardness [18,20,46].
The degree of hardness decreased after tempering and the range of the decrease increased as TTEM and tTEM increased. The range of reduction was decreased as the CR during quenching slowed. The hardness of FC specimens decreased by 5.2 HRC from 19 (FC-Q) to 14.8 HRC (FC-650_5); the hardness of MC specimens decreased by 4.9 HRC from 15.7 (MC-Q) to 10.8 HRC (MC-650_5), and the hardness of SC specimens decreased by 3.8 HRC from 11.7 (SC-Q) to 7.9 HRC (SC-650_5). Previous studies [2,[47][48][49] have reported that TTEM changes the hardness of M, LBF, P and ferrite and that the decreasing range of hardness is in the order of M, LBF and P and is highly dependent on TTEM. In contrast, the hardness of ferrite was not much affected by tempering.
Decomposition of the secondary phase increased as TTEM and tTEM increased (Figures 5 and 6). Hardness was reduced by decomposition and softening of hard secondary phases that were enriched in carbon. Tempering of LBF decreased the hardness due to the diffusion of carbon in LBF and the reduction in the dislocation density. The major reasons for the change in the hardness of ferrite were reduced residual stress and reduced dislocation density [47,49]. Tempering reduced hardness by releasing residual stress, decreasing the dislocation density and causing precipitation of solute carbon in the microstructure.  Figure 4).
Strengths were decreased after tempering (Figure 9), but not as quickly as hardness (Figure 8). Decomposition of the secondary phase was caused by precipitation of solute carbon during tempering ( Figure 6). Carbon is in solid solution and has a strong strengthening effect; therefore the strength decreased significantly as the carbon content in the secondary phase decreased due to the diffusion of carbon atoms into M3C cementite carbides during tempering [10]. The formation of the carbides was indeed observed in tempered specimens, as shown in Figure 7.
The reduction ranges of strength after tempering under CRs were defined as the difference between the Q specimen and the tempered specimen that had the lowest strength. In FC specimens, the reduction ranges were 80 MPa for tensile strength and 55 MPa for yield strength; in both MC and SC specimens, the reduction ranges were 60 MPa for tensile strength and 40 MPa for yield strength. The degree of hardness decreased after tempering and the range of the decrease increased as T TEM and t TEM increased. The range of reduction was decreased as the CR during quenching slowed. The hardness of FC specimens decreased by 5.2 HRC from 19 (FC-Q) to 14.8 HRC (FC-650_5); the hardness of MC specimens decreased by 4.9 HRC from 15.7 (MC-Q) to 10.8 HRC (MC-650_5), and the hardness of SC specimens decreased by 3.8 HRC from 11.7 (SC-Q) to 7.9 HRC (SC-650_5). Previous studies [2,[47][48][49] have reported that T TEM changes the hardness of M, LBF, P and ferrite and that the decreasing range of hardness is in the order of M, LBF and P and is highly dependent on T TEM . In contrast, the hardness of ferrite was not much affected by tempering.
Decomposition of the secondary phase increased as T TEM and t TEM increased (Figures 5 and 6). Hardness was reduced by decomposition and softening of hard secondary phases that were enriched in carbon. Tempering of LBF decreased the hardness due to the diffusion of carbon in LBF and the reduction in the dislocation density. The major reasons for the change in the hardness of ferrite were reduced residual stress and reduced dislocation density [47,49]. Tempering reduced hardness by releasing residual stress, decreasing the dislocation density and causing precipitation of solute carbon in the microstructure.
Tensile strength and yield strength decreased with decreasing CR under the same tempering conditions. The tensile strength and yield strength were highest in FC-Q (730 and 578 MPa, respectively) (Figure 9a All specimens had high strength due the high content of GBF in the microstructure. Microstructure types and their fractions were affected by CR and affected the samples' strengths (Table 3, Figure 4). Tempering reduced strength by decomposing hard phases, and the reduction ranges increased with the increase in the area fraction of the hard phase. The similarity of the reduction ranges of strength in MC and SC may be a result of complex phenomena that include not only the fraction of the hard phase but also grain size and fine precipitates. Strain-stress curves of the entire specimen showed dome-shaped, continuous yielding behavior ( Figure 10). This phenomenon could be related to high mobile dislocation density without pinning effect of carbons or precipitates [50][51][52][53]. On the contrary, some previous studies have shown discontinuous yielding after tempering or ageing [50][51][52][53][54][55][56]. It is well accepted that discontinuous yielding is caused by unpinning of dislocation from the Cottrell atmosphere, which formed by the interaction between dislocation and interstitial atoms such as carbon and nitrogen [52,54,55]. Heat treatment could induce diffusion of interstitial atoms into the dislocation core so dislocation is pinned during deformation and able to escape from the atmosphere by reaching the upper yield point. Moreover, this discontinuous yielding can be caused by the pinning effect of fine precipitates of critical size on dislocation [53]. For breaking away from pinning precipitates, higher stress will be required to drive dislocation. Therefore, it can be postulated that both chemical composition and the tempering conditions of this study may have been insufficient to form the Cottrell atmosphere or to grow precipitates to the critical size that is necessary for the development of discontinuous yield behavior. Chary V-notch (CVN) impact energy was not significantly affected by CR ( Figure 11). The experimental temperature (−46 °C) was assumed to be in the ductile-to-brittle transition temperature Strengths were decreased after tempering (Figure 9), but not as quickly as hardness ( Figure 8). Decomposition of the secondary phase was caused by precipitation of solute carbon during tempering ( Figure 6). Carbon is in solid solution and has a strong strengthening effect; therefore the strength decreased significantly as the carbon content in the secondary phase decreased due to the diffusion of carbon atoms into M 3 C cementite carbides during tempering [10]. The formation of the carbides was indeed observed in tempered specimens, as shown in Figure 7.
The reduction ranges of strength after tempering under CRs were defined as the difference between the Q specimen and the tempered specimen that had the lowest strength. In FC specimens, the reduction ranges were 80 MPa for tensile strength and 55 MPa for yield strength; in both MC and SC specimens, the reduction ranges were 60 MPa for tensile strength and 40 MPa for yield strength. Tempering reduced strength by decomposing hard phases, and the reduction ranges increased with the increase in the area fraction of the hard phase. The similarity of the reduction ranges of strength in MC and SC may be a result of complex phenomena that include not only the fraction of the hard phase but also grain size and fine precipitates.
Strain-stress curves of the entire specimen showed dome-shaped, continuous yielding behavior ( Figure 10). This phenomenon could be related to high mobile dislocation density without pinning effect of carbons or precipitates [50][51][52][53]. On the contrary, some previous studies have shown discontinuous yielding after tempering or ageing [50][51][52][53][54][55][56]. It is well accepted that discontinuous yielding is caused by unpinning of dislocation from the Cottrell atmosphere, which formed by the interaction between dislocation and interstitial atoms such as carbon and nitrogen [52,54,55]. Heat treatment could induce diffusion of interstitial atoms into the dislocation core so dislocation is pinned during deformation and able to escape from the atmosphere by reaching the upper yield point. Moreover, this discontinuous yielding can be caused by the pinning effect of fine precipitates of critical size on dislocation [53]. For breaking away from pinning precipitates, higher stress will be required to drive dislocation. Therefore, it can be postulated that both chemical composition and the tempering conditions of this study may have been insufficient to form the Cottrell atmosphere or to grow precipitates to the critical size that is necessary for the development of discontinuous yield behavior.
Moreover, this discontinuous yielding can be caused by the pinning effect of fine precipitates of critical size on dislocation [53]. For breaking away from pinning precipitates, higher stress will be required to drive dislocation. Therefore, it can be postulated that both chemical composition and the tempering conditions of this study may have been insufficient to form the Cottrell atmosphere or to grow precipitates to the critical size that is necessary for the development of discontinuous yield behavior. Chary V-notch (CVN) impact energy was not significantly affected by CR ( Figure 11). The experimental temperature (−46 °C) was assumed to be in the ductile-to-brittle transition temperature Chary V-notch (CVN) impact energy was not significantly affected by CR ( Figure 11). The experimental temperature (−46 • C) was assumed to be in the ductile-to-brittle transition temperature (DBTT) range. Steels that had a similar composition to that of the steels in this experiment had DBTTs in the range of −25 • C to −100 • C, as in earlier investigations [9,16,30,[57][58][59].
After tempering, all specimens under various CRs had increased CVN impact energy ( Figure  10). It increased with both TTEM and tTEM. By observing the fracture surface (Figure 12a,b), it was found that tempering increased the area fraction of the dimpled surface (red lines in Figure 12a), indicating that tempering promoted ductile fracture. The size distribution of the dimple did not show any feasible difference after tempering at 450 °C but became irregular after tempering at 600 °C. This increase in CVN impact energy could be possibly explained with the tempering effect which induced martensite to lose its brittleness by rejecting carbons and annihilating dislocation. Moreover, as evidenced in Figure 6, the rejected carbons formed cementite carbides in tempered martensite and those carbides were spheroidized with increasing tempering time and temperature. It can be postulated that the carbides with high aspect ratio could cause local concentrated stress, which can lead to fracture, but spheroidized cementite relieves local concentrated stress, and thereby inhibit initiation and propagation of cracks and increases toughness (Figure 12c) [13]. Tempering also reduces dislocation density and residual stress, and these changes also contribute to the increase in toughness [19,[60][61][62].   After tempering, all specimens under various CRs had increased CVN impact energy ( Figure 10). It increased with both T TEM and t TEM . By observing the fracture surface (Figure 12a,b), it was found that tempering increased the area fraction of the dimpled surface (red lines in Figure 12a), indicating that tempering promoted ductile fracture. The size distribution of the dimple did not show any feasible difference after tempering at 450 • C but became irregular after tempering at 600 • C. This increase in CVN impact energy could be possibly explained with the tempering effect which induced martensite to lose its brittleness by rejecting carbons and annihilating dislocation. Moreover, as evidenced in Figure 6, the rejected carbons formed cementite carbides in tempered martensite and those carbides were spheroidized with increasing tempering time and temperature. It can be postulated that the carbides with high aspect ratio could cause local concentrated stress, which can lead to fracture, but spheroidized cementite relieves local concentrated stress, and thereby inhibit initiation and propagation of cracks and increases toughness (Figure 12c) [13]. Tempering also reduces dislocation density and residual stress, and these changes also contribute to the increase in toughness [19,[60][61][62].

Conclusions
The present work investigated the effect of heat treatment on microstructure and mechanical properties of flanges by directly obtaining local properties of interest from the flanges under the heat treatment process. Conclusions are as follows: (1) During quenching, cooling rates varied among locations due to non-uniform cooling in flanges.
Microstructure was strongly affected by cooling rates in a way that area fraction of either hard or soft constituent phases was determined by cooling rate. FC-Q has the highest area fraction of hard phase such as LBF, AF and M while SC-Q showed the area volume fraction of softer phases like QPF and DP/P. (2) Both strength and hardness were dependent on cooling rates; faster cooling rates induced hard phases so that hardness and strength resultantly increased. CVN impact energy at −46 • C, however, did not show clear dependence on cooling rates.