Impact of Cooling Rate during High-Pressure Gas Quenching on Fatigue Performance of Low Pressure Carburized Gears

Abstract

The impact of cooling rate during high-pressure gas quenching on the fatigue performance of low-pressure carburized spur gears was studied for steel grades 20MnCr5 and 17NiCrMo6-4. The results show an increased fatigue limit by 10 to 11% when applying a slower cooling rate for both steel grades. Moreover, for 20MnCr5 the slower cooled gears show an increase in compressive residual stresses by 130 MPa compared to the faster cooling, although no significant difference was observed for 17NiCrMo6-4. It is also seen that the cooling rate affects the core hardness for both steel grades, while other properties like surface hardness, case-hardness depth and martensite variant pairing were unaffected. The results for the retained austenite content and average martensite unit size show no clear effect of the cooling rate. The possible influence of different carbon distributions after quenching for the two used cooling rates on the carbide precipitation and fatigue limit is discussed.

1. Introduction

The combination of a hard surface layer and a tough core makes case-carburized and hardened gears an excellent choice for power transmission in, e.g., cars, trucks and wind turbines. The high hardness and compressive residual stresses of the surface layer results in a high wear resistance and fatigue strength while the softer core gives a higher impact toughness [1,2].

In the last 20 years, low-pressure carburizing (LPC) and subsequent high-pressure gas quenching (HPGQ) of gears has become a standard for industrial production of case-hardened gears. LPC can realize the industries ambition to reduce gas emissions and usage of mineral oil quenchants, as well as the need for cleaning parts after heat treatment [3,4].

HPGQ allows control of the cooling rate by changing the gas pressure and fan speed, which has been exploited to lower distortion and improve fatigue performance [3,4,5,6,7]. Moreover, slower cooling rates below the martensitic start temperature (Ms) in HPGQ can significantly improve fatigue performance of case-hardened gears, when compared to faster quenching rates [5,6]. However, the impact of cooling rate has only been studied for a few alloys and it is not clear yet if the fatigue improvement is a general effect, due to the cooling treatment, or if it is a composition dependent effect. Additionally, the exact mechanism of this improvement is not yet understood. The fatigue performance of case-hardened gears has been reported to be influenced by the hardness [8,9], residual stress profile [10,11,12], case-hardness depth [9,12,13], retained austenite content [14,15,16], surface finish [11,17,18,19], defects like cracks and inclusions [11,20,21] as well as the prior austenite grain size (PAGS) and martensite unit size [10,16,22,23]. The fatigue performance dependence on these properties might be complex and interrelated, e.g., a change in type and amount of microcracks could depend on PAGS [22,23].

In this work, the effect of slow and fast cooling through the martensite formation regime during HPGQ is investigated. Hardness, residual stresses, retained austenite and microstructural features are quantified to gain more fundamental understanding of the relation between processing route, microstructure and improved fatigue performance.

2. Materials and Methods

The alloying content of the investigated steel grades, 20MnCr5 and 17NiCrMo6-4, provided by the steel supplier and confirmed by X-ray fluorescence analysis are given in

Table 1. Composition of main alloying elements (wt% > 0.1) as well as S and P before LPC.

Composition [wt %]	C	Cr	Mn	Ni	Si	Al	Mo	Cu	S	P
20MnCr5 (A)	0.20	1.15	1.15	0.14	0.14	0.11	0.06	0.17	0.001	0.004
17NiCrMo6-4 (B)	0.22	1.05	0.94	1.10	0.22	-	0.12	0.19	0.025	0.013

. The 20MnC5 steel grade is a high cleanliness ingot cast material and the 17NiCrMo6-4 is a continuous cast material. 20MnCr5 will be called A and 17NiCrMo6-4 B in the figure legends. Spur gears were machined from both steel grades and hard machining was done after LPC heat treatment to comply with tolerances for fatigue testing. The machining before characterization was focused on the gear flanks to minimize the influence of distortions from the cooling treatment, but no machining was performed in the root area, which is the area with the highest stresses during fatigue testing [12]. The spur gears had 20 teeth, an outer diameter of 83 mm and a gear module of 3.65.

2.1. Heat Treatment and Quenching

The gears were carburized in an industrial multi-chamber LPC system with modular HPGQ using helium gas. The quenching gas pressure as well as fan speed could be varied to achieve different cooling treatments. However, during cooling only the fan speed could be changed. Fast cooling (18 bar Direct) was implemented as direct quenching using 18 bar gas pressure and 100% max fan speed for 10 min (600 s). The slow cooling (6 bar Stop) was designed as a two-step quench, using 6 bar gas pressure. The first part used 100% max fan speed for 65 s followed by 5 min (300 s) of 20% max fan speed. Quenching speed is not solely governed by fan speed since it also depends on quenching chamber layout, load configuration and turbine blade geometry. The relative fan speeds used here can thus be motivated by the need to actually measure temperature paths experienced by the components if the results would be reproduced in another furnace system.

The temperature evolution during the two studied HPGQ treatments was measured for a test setup of gears with thermocouples, see Figure 1a, using a data logger. The data logger was placed in a thermally insulating box to resist heat and traveled with the load. The data logger was connected with 16 type N thermocouples (TC) that were mounted in holes drilled at different gear positions as well as additional Inconel probes, see Figure 1a,b. The positions were the same for 20MnCr5 and 17NiCrMo6-4 with 2 TCs each at the center of the gear tooth (CT) and center of the gear ring (CR) as well as 1 each in holes close to the surface at the gear root and teeth tip. The holes at CT and CR end at the center of the gear thickness to allow central placement of the TCs. The holes at the tip and root were drilled radially from the center of half gears to their outer diameter, see Figure 1a. Additionally, TCs in 3 Inconel probes were used as reference for the temperature curve due to the absence of phase transformations in the material at the used temperatures. The fixture was filled with dummy material for additional weight and for comparable loading conditions to the industrial process. The gears were fastened to the fixture to avoid position shifts during cooling, see Figure 1a.

Based on the above described cooling treatments, the gears used for characterization were low pressure carburized in two batches with each having gears from both steel grades with subsequent HPGQ with 18 bar Direct or 6 bar Stop. After cooling, the gears were tempered at 180 ${}^{\circ }$C for 2 h. A total of 7 gears of each steel grade were heat treated per batch and the gears were characterized with the methods described in Section 2.2.

2.2. Characterization

The characterization of the gears was focused on the fatigue performance, hardness, residual stresses, retained austenite content as well as the microstructure. The 2 cooling treatments, 18 bar Direct and 6 bar Stop, were studied for each steel grade resulting in a total of 4 fatigue series. The hardness HV1 profile was measured at the gear flank and root with varying step sizes between 100 and 300 $\mathsf{\mu }$m steps down to a depth of 3 mm, see Figure 1b. A total of 2–4 gear teeth were investigated for each condition. Additionally, the core hardness HV30 was determined at the CT, ITR and CR positions. ITR is the interface of gear tooth and gear ring as shown in Figure 1b.

The residual stress and retained austenite content was measured to a depth of 160 $\mathsf{\mu }$m by X-ray diffraction and stepwise electrolytic etching on the machined gear flank since the root was not accessible. The residual stress was determined along the gear profile in the direction from gear tooth tip to root according to the standard EN 15305 [24] and the retained austenite content according to ASTM E975 [25]. The measurements were carried out for 2–4 gear teeth for each condition.

The microstructure characterization was focused on the surface close to the gear root shown in Figure 1b since the stresses are highest at this position during fatigue testing [12]. The cross-section of the area between two gear teeth was hot mounted and mechanically prepared with diamond suspensions, 9, 3 and 1 µm, followed by a final polishing with colloidal silica. For the characterization, a JEOL JSM-7800F field emission gun scanning electron microscope (JEOL Ltd., Tokyo, Japan) with a Bruker e${}^{-}$Flash${}^{\mathrm{HR}}$ electron backscatter diffraction (EBSD) detector (Bruker Corporation, Billerica, MA, USA) was used. The EBSD results were analyzed focusing on the martensite unit size and PAGS. The MTEX toolbox v.5.2 [26] and the prior austenite orientation reconstruction algorithm from Nyyssönen et al. [27] were used for the EBSD analysis. The PAGs were determined from the reconstructed austenite orientation and the amount of PAGs as well as the PAGS distributions were determined in MTEX [26]. Additionally, the variant pairing, which describes the misorientation between neighbouring martensite units, was calculated by Nyyssönen’s algorithm [27] after the definition of Morito et al. [28]. The most relevant variant pairings are the packet boundaries, V1–V2/V3/V4/V5/V6 with V1–V2 being a twin boundary and V1–V4 a low angle block boundary, as well as the V1–V16 pairing typically observed in plate martensite [28,29]. The martensite unit size and variant pairing were summed up over 3 scans for each gear with each scan having a size of 81 µm × 61 $\mathsf{\mu }$m and a step size of 150 nm. The presence of microcracks close to the root surface was investigated by SEM imaging.

The high-cycle fatigue evaluation was performed using pulsator testing, which tests the gear tooth root load carrying capacity. The test was done in an MTS servo hydraulic test rig with a R value of 0.1. A total of 9–14 tests were carried out per series with a maximum of 3 tests per gear. From the fatigue data, the fatigue limit was determined using an interpolated Wöhler curve. The calculations were performed with a custom software based on the stress-life method (S-N) [30] using a fixed exponent due to small sample size and a limit of $2\times {10}^6$ cycles as infinite life, i.e., runout.

3. Results

3.1. Quenching

The temperature curves from the central gear tooth (CT) position TCs during HPGQ with 18 bar Direct and 6 bar Stop are shown in Figure 2a. The plots for the TCs from the other positions, CR, Root, and Inconel, can be found in Appendix A for a better overview. The TCs positioned at the same gear positions show strong differences of up to 87 ${}^{\circ }$C at the same time for CT 18 bar Direct and slightly lower maximum differences at other positions. The TCs at the CR and Root position show a clear deviation from the overall temperature behavior around 350 ${}^{\circ }$C for both cooling treatments, see Figure A1a,b.

The differences in temperature are most likely due to inhomogeneous gas flow due to the insulation box geometry as well as the different size of dummy gears resulting in an additional heating source, see Figure 1a. The gears used for characterization should not be impacted by any of these effects since there was no insulation box as well as no dummy gears in the top layer during their cooling treatment.

The comparison of the two cooling treatments is focused on the TCs at the tip of the gear teeth in Figure 2b since they are positioned closest to the surface. The temperature curves show a clear difference between the 18 bar Direct and interrupted 6 bar Stop quenching for the complete temperature range. The 18 bar Direct is significantly faster in the higher temperature region and the TCs at the tip of the gear teeth reach 200 ${}^{\circ }$C in approximately 30 s while it takes around 65 s for 6 bar Stop. Additionally, the 6 bar Stop shows a decrease in its cooling rate at 65 s due to the reduction in fan speed, and thereafter shows a slower cooling rate to room temperature, which is reached after about 200 s.

3.2. Characterization

3.2.1. Hardness

The average hardness HV1 at different depths of the gear root is shown in Figure 3. The hardness profiles show the typical behavior of case-hardened gears with a high hardness at the surface with values around 720–770 HV1 followed by a decrease until reaching a near constant level around a depth of 1.2 mm with values between 390 and 460 HV1 for 20MnCr5 and 450–500 HV1 for 17NiCrMo6-4. The case-hardness depth, CHD 550 HV1, is similar for all test series with values between 0.70 and 0.78 mm. The observed hardness profiles for the flank are nearly identical to the root and are not shown here. Additionally, the core hardness HV30 for all positions, CT, ITR and CR, have a trend to lower values for the 6 bar Stop gears compared to the 18 bar Direct, see

Table 2. Summation of characterization results for 20MnCr5 (A) and 17NiCrMo6-4 (B).

	20MnCr5 (A)		17NiCrMo6-4 (B)
	18 bar Direct	6 bar Stop	18 bar Direct	6 bar Stop
Fatigue limit [MPa]	$503\pm 24$	$553\pm 27$	$428\pm 29$	$476\pm 16$
Core hardness CT [HV30]	$445\pm 7$	$419\pm 5$	$475\pm 3$	$462\pm 7$
Core hardness ITR [HV30]	$439\pm 11$	$404\pm 8$	$472\pm 5$	$463\pm 3$
Core hardness CR [HV30]	$429\pm 9$	$390\pm 14$	$479\pm 13$	$463\pm 2$
Surface hardness root [HV1]	$767\pm 9$	$760\pm 13$	$736\pm 13$	$723\pm 17$
CHD 550 HV1 [mm]	$0.75\pm 0.03$	$0.70\pm 0.03$	$0.78\pm 0.02$	$0.75\pm 0.02$
Avg. residual stress between 75 and 150 $\mathsf{\mu }$m [MPa]	$-58\pm 13$	$-193\pm 12$	$-94\pm 16$	$-123\pm 9$
Avg. martensite unit size [$\mathsf{\mu }$m${}^2$]	0.65	0.60	0.74	0.65
Avg. martensite unit size $>5\mathsf{\mu }$m${}^2$ [$\mathsf{\mu }$m${}^2$]	8.75	8.41	9.08	9.28
Amount PAGs $>50\mathsf{\mu }$m${}^2$	61	56	65	58
Avg. PAGS $>50\mathsf{\mu }$m${}^2$ [$\mathsf{\mu }$m${}^2$]	125	157	112	124

. The core hardness for 17NiCrMo6-4 is also significantly higher than for 20MnCr5. The values for surface hardness HV1, core hardness HV30 and case-hardness depth are summarized in Table 2.

3.2.2. Residual Stress and Retained Austenite

The average residual stress and retained austenite content at different distances from the ground flank surface down to 150 $\mathsf{\mu }$m is shown in Figure 4. The residual stress in the direction from gear tooth tip to root in Figure 4a shows compressive stresses over the complete measured range. The 17NiCrMo6-4 results are similar for both cooling treatments, while the 6 bar Stop has a significantly higher compressive stress than 18 bar Direct for 20MnCr5. The scatter between measurements for 6 bar Stop for 20MnCr5 is larger than for all other conditions, but every 6 bar Stop measurement has higher compressive stresses compared to the 18 bar Direct. For comparison the average of the residual stresses between 75–150 $\mathsf{\mu }$m is calculated with resulting values between −58 to −193 MPa, see Table 2.

The average retained austenite content has the highest values, 20.6 to 22.6%, at the surface and decreases continuously moving away from the surface. Small differences in the trends of average retained austenite content are observed between cooling treatments, but these are within the error margins, which are up to 3.5%, due to the strong variation between measurements.

3.2.3. Microstructure

The EBSD orientation maps of martensite in 20MnCr5 for 18 bar Direct and 6 bar Stop are shown in Figure 5 and the 17NiCrMo6-4 martensite orientation maps are shown in Figure A2. The microstructure of both steel grades consists of a complex mixture of large martensite units with a length of around 10 µm and smaller martensite units. There are no clear qualitative differences between the microstructure due to the two cooling treatments based on EBSD results. The corresponding PAGs show small differences in their size distribution as well as in the amount of PAGs larger than 50 $\mathsf{\mu }$m${}^2$, see Table 2 and Figure A3. The analysis is focused on larger PAGs due to uncertainties in determination of smaller PAGs. The SEM microstructure analysis shows no presence of microcracks in the root area before fatigue testing.

Statistical Analysis

The martensite unit size distribution is summed up over 3 different scans per gear and the average unit area of all units is slightly larger for 18 bar Direct compared to 6 bar Stop for both steel grades. For marteniste units larger than 5 $\mathsf{\mu }$m${}^2$, the same trend is observed for 20MnCr5, while 17NiCrMo-64 shows the opposite trend with 6 bar Stop having a larger average unit area, see Table 2.

The relative martensite unit area for a binning size of 2 $\mathsf{\mu }$m${}^2$ is shown in Figure 6 with the exclusion of units below 2 $\mathsf{\mu }$m${}^2$ in the inset. The value for the smallest units between 0 and 2 $\mathsf{\mu }$m${}^2$ is higher for 6 bar Stop in both steel grades. The number of large units is similar for both cooling treatments, but some variations in the exact distribution are observed. 18 bar Direct shows a trend towards more units with an area between 4 and 12 $\mathsf{\mu }$m${}^2$ compared to 6 bar Stop for both steel grades.

The variant pairing of martensite was summed up over 3 scans per gear and is shown in Figure 7. The variant pairing is nearly identical for both cooling treatments with only small maximal differences around 1%. The strongest variant pairing is the V1–V2 with a value of around 25%. The other packet variant pairings [28], V1–V3/V5/V6, have a significant value of around 10% while the block variant pairing V1–V4 is around 5%. V1–V16 has a value around 6–8%.

3.2.4. High-Cycle Fatigue

The measured stress vs. fatigue life (S-N) data for a total of 9–14 measurements for each of the 4 combinations of steel grade and cooling treatments with the corresponding calculated fatigue limit is shown in Figure 8. Both steel grades show a clear improvement of fatigue performance for the 6 bar Stop compared to 18 bar Direct. The fatigue limit estimated from the Wöhler curve is increased by 10% from 503 MPa to 553 MPa for 20MnCr5 and by 11% from 428 MPa to 476 MPa for 17NiCrMo6-4, see Table 2.

4. Discussion

The influence of two different cooling rates during HPGQ on the fatigue performance of low pressure carburized gears shows an improvement in fatigue limit by 10–11% for the slower cooling, 6 bar Stop, compared to the faster cooling, 18 bar Direct, as shown in Figure 8 and Table 2, for both materials. This improvement agrees well with previously reported positive effects of slower cooling rates, especially close to Ms at the surface, for case-hardened gears [5,6]. The observation in two additional steel grades indicates a general positive effect of the slower cooling rate and should thus be independent of composition.

The compressive stresses at the gear flank in 20MnCr5 are increased by about 130 MPa for slower cooling, 6 bar Stop, compared to faster cooling, 18 bar Direct, while there is no significant difference for 17NiCrMo6-4, see Figure 4a and Table 2. The exact estimation of the influence is made difficult by the large scatter between the 6 bar Stop 20MnCr5 measurements. This difference could be an explanation for the increased fatigue limit for 20MnCr5 due to the positive contribution of compressive stresses in the case-hardened layer to the fatigue performance of gears [10,11,12], but it cannot explain the increase for 17NiCrMo6-4. However, the correlation between increased compressive residual stresses and improved fatigue performance has also been observed by Fahlkrans et al. [6]. The lower impact of the cooling rates on the compressive residual stresses in 17NiCrMo6-4 could be related to the higher Ni content, see Table 1, resulting in a higher hardenability of the core [31], which is in good agreement with the observed higher core hardness compared to 20MnCr5, see Figure 3. Additionally, the correlation of preferential stress states measured at the gear flank to the root area, where the highest load is applied, is unclear since simulation results have shown a significant difference between residual stresses in both areas [12].

The core hardness for both steel grades show a trend towards higher hardness values for 18 bar Direct compared to 6 bar Stop, see Table 2. This increase is between 6 and 10% for 20MnCr5 and 2–4% for 17NiCrMo6-4. The impact of the core hardness on the fatigue properties of gears is not well understood due to correlation with other parameters like residual stresses and surface hardness [32]. An effect on the fatigue performance has been reported from theoretical calculations, but the core hardness is expected to have a significantly weaker impact compared to the surface hardness [33,34]. Overall, it is not expected to be the main contributing factor to the improved fatigue performance due to the small differences between cooling treatments as well as the used experimental setup that applies the highest stress near the surface of the root [12].

The surface hardness, case-hardness depth and retained austenite content are known to influence the fatigue performance [8,9,10,11,12,13,14,15,16], but no significant differences between the two cooling treatments are observed, see Table 2. Therefore, the surface hardness and case-hardness depth can be excluded as possible explanations for the improved fatigue performance in this case. The scatter between the retained austenite measurements for the same conditions could mask an underlying difference due to the cooling rate, especially since an impact of the cooling rate on the retained austenite content has been reported [35,36,37].

The root surface microstructure of both steel grades for the two cooling rates, Figure 5 and Figure A2, is qualitatively very similar and the statistical analysis of the variant pairing shows nearly identical results, see Figure 7. Based on these results, a possible explanation of the improved fatigue performance due to different martensite boundaries can be quantitatively excluded. The average martensite unit size for all units has a smaller value for the slower cooling rate in both steel grades. This trend can be partially explained by the higher amount of martensite units smaller than 2 $\mathsf{\mu }$m${}^2$ that have the highest uncertainty due to possible preparation influences. For the units larger than 5 $\mathsf{\mu }$m${}^2$, no consistent trend depending on the cooling rate is observed, see Table 2. The relative martensite unit size distribution shows small differences between the two cooling treatments in both steel grades, see Figure 6. These differences are relatively small with maximum deviations around 1.5% for units larger than 2 $\mathsf{\mu }$m${}^2$ and are most likely due to the limited statistics since the size distribution and amount of PAGs in Table 2 and Figure A3e,f show small differences as well. The overall PAG distribution is expected to be similar since the gears of each steel grade were heat treated in the same way except for the HPGQ and should be a result of limited statistics due to the size of the PAGs.

The effect of microcrack formation during cooling, which is detrimental for fatigue performance [11,17,18,19], can also be excluded for both steel grades since no microcracks are observed in the microstructure close to the root surface. Another explanation for the improved fatigue performance could be difference in the surface finish [11,17,18,19], but it is unlikely due to the identical machining of the gears.

The presented results show differences between the cooling rates, but a consistent mechanism for the fatigue performance improvement in both steel grades has not been found. Therefore, the possible influence of a stronger autotempering/partitioning in the high carbon region of the surface layer for the slower cooling rate is discussed based on literature. It was shown that the cooling rate around Ms impacts the martensite tetragonality in high carbon steels with a stronger heterogeneous distribution for slower cooling rates [35]. Local differences in tetragonality in the same microstructure have been connected to a changed carbon distribution with a stronger local heterogeneous distribution for low tetragonality compared to a more homogeneous distribution for high tetragonality regions [38]. This could cause a difference in carbide formation during tempering of martensite, which is split into multiple stages of carbon segregation and carbide precipitation for Fe-C martensite [39,40,41]. The tempering kinetics can be influenced by different cooling rates and heat treatment [39,42], but in-depth studies focused on the influence of different initial carbon distribution on the final carbide distribution for a high carbon steel have not been carried out yet to the authors knowledge. It has been shown by calculations that the final carbide size and fraction depends on the initial martensite structure and especially the nucleation site distribution for high temperature tempering [43]. Additionally, a heterogeneous carbon distribution with a resulting influence on the carbide precipitation kinetics and final distribution upon tempering has been reported for low carbon steels depending on the transformation order due to different diffusion length [44], which is similar to the expected effect from different cooling rates. The degree of autotempering/partitioning is expected to be similar in both steel grades for the same cooling rate since there are no large differences in the composition of strong carbide forming elements like Cr and Mo and it mainly depends on Ms [35]. Thermo-Calc calculations [45,46,47,48] show similar Ms values for both steel grades, 206 ${}^{\circ }$C for 20MnCr5 and 197 ${}^{\circ }$C for 17NiCrMo6-4, for a carbon content of 0.7 wt% and the Cr, Mn, Ni and Si composition from Table 1. Overall, the reported difference in carbon distribution after the initial cooling [35,38] could lead to changes in carbide precipitation during tempering and possibly explain the improved fatigue performance.

5. Conclusions

The effect of two different cooling rates during high pressure gas quenching on the fatigue performance of case-hardened gears is studied in two steel grades, 20MnCr5 and 17NiCrMo6-4. The results show a clear improvement of 10–11% for the fatigue limit for both steel grades for slower cooling. The slower cooling results in higher compressive residual stress by around 130 MPa compared to the faster cooling for 20MnCr5, while 17NiCrMo6-4 shows no significant difference. The core hardness shows small differences around 2–10% between the two cooling treatments in both steel grades. The average martensite unit size and retained austenite content do not show a clear change depending on the cooling rate.

The surface hardness, case-hardness depth as well microstructure features like variant pairing and crack formation during cooling are very similar for the two cooling rates and can be excluded as mechanism for the fatigue performance improvement. Based on literature results, an in-depth characterization of the effect of autotempering/partitioning on the sub-micron level on the fatigue performance would be a natural extension to the current work to further explore plausible explanations.