Study of Cooling Medium Variables during Quenching in SAE 4340 Steel Using Statistical and Modeling Tools

Abstract

Although quenching is one of the most widely used heat treatments in the metal-mechanical industry to improve the mechanical properties of steels, it is also responsible for the generation of residual stress, distortion, and fractures in the treated parts. The high-temperature gradients present during quenching and martensitic transformation are the main failure mechanisms. Cooling is the critical quenching stage where several variables that need to be controlled are involved in reducing these problems. The objective of this research was to evaluate the main variables in the quenching process in SAE 4340 steels, which promote distortion, residual stress accumulation, and fracture failures. A 2ᴷ factorial experiment was designed, samples with C-ring geometry susceptible to changes in quenching variables were used, and the variables studied were the agitation and temperature of the quenching medium. Experimental measurements, statistical tools and modeling were used to evaluate and predict the distortion generated in quenched samples. Such tools include Minitab 21^® software and its statistics utilities. Furthermore, a finite element method model was carried out using STFC Deform^®. The results suggest that there are optimal conditions in the quenching process to minimize distortion and residual stresses and to improve mechanical properties of quenched parts; therefore, the methods used in this work could be useful to detect and control the appearance of defects in an industrial environment.

1. Introduction

Heat treatment is one of the most used processes to improve the mechanical properties of components such as hardness, strength, and toughness [1,2,3,4,5]. This process consists of heating the parts to the austenitizing temperature, keeping them for a certain time for microstructural homogenization, and cooling them at different cooling rates, which allows modifying the initial properties of the parts. Cooling is the most critical stage of heat treatment because it controls both the microstructure and the final mechanical properties of the quenched pieces. At low cooling rates, microstructures such as ferrite and perlite predominate, favoring a low level of residual stress and distortion, but with lower strength and hardness. In contrast, at high cooling rates, a complete transformation of austenite into martensite is obtained, increasing the mechanical properties. However, it is also known that this transformation can lead to a higher magnitude of residual stresses and increase the probability of distortion and fracture failures in the pieces [6,7,8,9]. Jami et al. [10] analyzed the behavior of the torsional strength and hardness of an AISI 4340 steel after quenching and tempering, using an austenitizing temperature of 860 °C and tempering temperatures of 300 °C, 350 °C, and 400 °C for a soaking time of 120 min. The results indicated that the maximum torsional strength and hardness were achieved with the tempering temperature of 300 °C. Clark et al. [11] studied the effect of austempering, quenching–tempering heat treatments on distortion, residual stress, and percentage of retained austenite in samples of Navy C-rings of SAE 4320, 8822, and 8620 steels, using X-ray Diffraction and CMM techniques. It was observed that the austempering heat treatment for the SAE 4340 steel decreases distortion; however, it increases the magnitude of the residual stresses, while in the heat treatments of quenching–tempering for all the steels, distortion is acceptable. García-Navas et al. [12] studied the gear manufacturing process that includes forging and heat treatments (normalizing, quenching, and tempering) on the final residual stress state, microstructure, and hardness of AISI 4140 steel through experimentation and FEM simulation, using X-ray Diffraction and Vickers microhardness. They found a good relationship between the experimental and simulated processes when the hardness and residual stresses were analyzed.

The control of the variables involved in a manufacturing process allows us to ensure the quality of the components, according to their mechanical properties and geometric dimensions. The dimensional instability of the parts can be minimized through analysis of the process capability and statistical quality control, allowing a 95% reduction in the losses of the finished product. This analysis requires the study of components that allows the identification of phenomena such as distortion and residual stresses, which have been widely investigated [10,13,14,15,16,17]. Navy C-rings are widely used as a specimen type for this purpose, which are very sensitive to geometrical changes and mechanical properties during heat treatments. Manivannan et al. [18] investigated the dimensional changes of Navy C-rings of SEA 1010 steel with three different thicknesses, subjected to the heat treatment of ferritic nitriding with gas and gaseous carbonitriding, using a CMM gauge to determine the distortion and FEM modeling to predict the residual stresses. The results indicated dimensional changes in the opening (C-ring gap), which was more evident in carbonitriding. Yu et al. [19] studied the distortion of a C-ring piece of an AISI 4140 steel at the austenitizing temperatures of 850 °C, 900 °C, and 950 °C and different agitation rates, evaluating the mechanical properties, microstructure, and distortion after quenching. They found that the most effective heat transfer and the highest microhardness were obtained with the maximum agitation speed at a temperature of 850 °C; however, for this condition, the distortion was higher. Hernandez et al. [20] analyzed the distortion and internal stresses after quenching of AISI 304 stainless steel Navy C-rings cooled in water at 40 °C; they measured the temperature changes in the core of the most robust zone of the part and compared them with simulation software results. The experimental and the simulation results coincided satisfactorily; the residual stresses obtained were tensile, and for this reason, a notable distortion was generated in the C-ring gap. Da Silva et al. [21] studied the behavior of geometric processes in a C-ring of AISI 4140 steel using both experimental and simulated tests. They found that the experimental results are in good agreement with the simulated ones and the ring distortion is mainly related to the martensitic transformation. Boyle et al. [22] investigated the magnitude of residual stress, retained austenite, and distortion after carburizing heat treatment of SAE 8620 and PS 28 steels in Navy C-ring specimens, using X-ray Diffraction techniques, a CMM gauge, and optical microscopy. The results obtained indicate greater distortion in the ring gap separation; the retained austenite was 15 to 20% in the most robust zone of the piece and the residual stresses on the surface were compressive, while in the core, they were tensile.

The low quality and noncompliance with the specifications of industrially manufactured parts are mainly caused by high concentrations of residual stresses, distortion, and fracture failures. Recent studies have shown that quenching heat treatment has caused losses of around 20 to 40% in the metal-mechanical industry [12,23,24]. During the quenching heat treatment, some critical variables are involved, which must be controlled and predicted, such as the austenitizing temperature and quenching medium, due to their impact on mechanical properties and quenching-related problems. Zordão et al. [25] evaluated the hardness and cooling curves of an SAE 1045 steel cylinder after quenching heat treatment by studying variables such as agitation, liquid temperature, and different cooling environments. They found that the greatest hardness is achieved with agitation at a temperature of 45 °C in a liquid environment using Na₂SO₄ and NaNO₂ as the cooling medium. Lopez-Garcia et al. [26] analyzed the dimensional behavior and residual stresses of a complex piece of AISI 4340 steel, modifying some variables of the quenching heat treatment, such as the immersion position and quenching medium. They used finite element modeling to compare experimental and simulated results, in addition to optical microscopy and distortion measurement. The data show that the greatest hardness and distortion were obtained when the parts were quenched in a water medium and in a horizontal position, which was in good agreement with the FEM model.

Another important factor during quenching is the heat transfer coefficient (HTC), which is defined quantitatively using the cooling curves, establishing the cooling conditions on the surface and core of the components [27,28]. Taraba et al. [29] calculated the heat transfer of an ISOPARID 277 HM oil with different agitation using ANSYS software modeling and the inverse heat transfer coefficient (IHTC) method. They concluded that there is a clear difference between the HTC of agitated and unstirred oil, with agitation favoring a higher HTC. Sedighi and McMahon [30] studied the effect of HTC and residual stress development in a BS970 steel cylinder using part position and oil agitation after the quench heat treatment as variables. They used X-ray Diffraction, the inverse method to calculate HTC, and compared the results with finite element simulations. The results showed that with agitation and in the transverse position, a higher heat extraction is produced, while in the vertical position, the highest magnitude of tensile residual stresses was generated, with both results being very close to the simulation.

In recent years, many investigations have been carried out on the phenomena of distortion and residual stress in metallic components, mainly in processes that require high-dimensional precision, such as welding, machining, and quenching heat treatment, using finite element methods (FEM). The results obtained have contributed significantly to reducing productivity losses in manufacturing industries and increasing production processes [26,31,32,33,34,35,36]. Cho et al. [37] employed FEM to evaluate the effect of quenching heat treatment on distortion for bevel gears in machined and forged SAE 4118H steel. They used a three-dimensional scanner to determine the distortion and DEFORM^® software was used to compare the simulation results with the experimental results. Slováček [38] developed computational simulation that contemplates the thermal and mechanical metallurgical analysis of the quenching heat treatment using air, water, and oil as the cooling medium in components such as shafts and plates used in ships. Using SYSWELD^® software, they compared the magnitude of residual stresses and hardness under different quenching environments. They showed that the finite element simulation successfully predicted the maximum magnitude of residual stresses, hardness and yield strength when cooled in water.

Statistical tools such as analysis of variance (ANOVA), analysis of means (ANOM), and others are widely used to compare the variance between the means of groups of response variables to find a relationship between factors even when it is not obvious to the naked eye. These tools are especially useful in large production batches or in large productions over time where the large amount of data make it possible to study and control the process variables practically in real time. Several research studies mention that ANOVA has proved to be a very useful statistical tool to identify the factors of the most important variable in industrial processes [39,40]. Agboola et al. [41] tested samples of medium C steel to experimental quenching by modifying the liquid medium, the soaking time and the austenitizing temperature. They used mechanical tests to measure hardness and tensile strength. In addition, they used statistical tools, such as the Taguchi method, line regression and variance method, to evaluate the variable that has the greatest influence on the properties studied. They concluded that the soaking time was the most influential variable. Nunes et al. [42] analyzed variables such as quenching medium and soaking time at the austenitizing temperature in an AISI 4340 steel, focused on studying the resulting microstructure, hardness, and distortion and generated a statistical correlation by means of ANOVA. The results showed that the hardness is positively influenced when using a temperature of 900 °C and the longest soaking time of 60 min for water quenching.

The objective of this article is to analyze some variables of the heat treatment process, such as temperature and agitation of the quenching medium of SAE 4340 steel in standard Navy C-rings specimens. The results obtained from the heat-treated samples were manipulated using analysis of means (ANOM) to generate a correlation between the studied variables and their effect on distortion and hardness. Minitab^® was used as a statistical tool. In addition, numerical software was used to predict the formation of residual stresses and distortion in the samples.

2. Materials and Methods

The chemical composition of AISI-SAE 4340 steel used in the present research is as follows: 0.41%C, 0.88%Cr, 0.25%Mo, 0.71%Mn, 1.80%Ni, 0.3%Si, 0.021%S, 0.005%P and Fe balance, in weight percent. The quenching samples were taken from a 4 in diameter steel bar, which was machined into the Navy C-ring shape. The specimens were instrumented with a K-type thermocouple tightly inserted in the most robust part of the ring at a depth of 15 mm and connected to a data acquisition module OMEGA model OMB-DAQ 54. The geometry and position of the thermocouple are given in Figure 1a, the initial average dimensions for the C-ring gap, inner and outer diameters were 6.09 mm, 15.87 mm, and 25.40 mm, respectively. The samples were heated at 5 °C/min rate in a Luzeren FL7-N electric resistance furnace to an austenitizing temperature of 860 °C; once reached, it was soaked for 30 min to achieve a homogeneous austenitic microstructure. The quenching was carried out in a liquid medium using industrial quenching oil Equiquench 770^® of Equimsa. The manufacturer’s directions were used, indicating working temperatures up to 60 °C. The Navy C-rings were submerged for 100 s, and the immersion rate used was 53 mm/s in a 35 L rectangular tank (50 cm long, 30 cm high, and 20 cm wide), with agitation supplied by a 1 HP peripheral pump, with four ¾” discharge jets oriented at 90° in the 50 cm wall and 100 L/min of total flow. The conditions of the process of quenching are summarized in

Table 1. Conditions in the quenching heat treatment processes.

Sample	Agitation (L/min)	Austenitizing Temp. (°C)	Soaking Time (min)	Oil Temp. (°C)	Immersion Rate (mm/s)	Immersion Direction
A	100	860	30	60	53	-Y
B	0	860	30	60	53	-Y
C	100	860	30	27	53	-Y
D	0	860	30	27	53	-Y

. It is important to mention that no tempering treatment was implemented in this study, because the greatest occurrence of defects takes place in the “as-quenched” condition.

To evaluate the distortion, the measurements of the dimensions of the Navy C-rings before and after the quenching heat treatment were carried out using Mitutoyo Absolut Digimatic 500-197-30 equipment with a 0.01 mm resolution. The dimensional changes in the A-B distance (GAP), inner diameter (ID), and outer diameter (OD) for each sample were analyzed, according to Figure 1a. Measurement of residual stresses (RS) after quenching was carried out by means of the X-ray diffraction method using a STRESSTECH G2R-E unit. To evaluate the distribution of residual stresses, they were measured at 3 measurement points with 1.5 mm between each point as shown in Figure 1b.

Both the distortion and the residual stresses were analyzed by numerical simulation and compared with the experimental results. The finite element method (FEM) was used through a three-dimensional model with 30,000 tetrahedral elements and 435,000 nodes. The computed domain of geometry is shown in Figure 2 and the conditions of the simulation processes are shown in

Table 2. Simulations parameters.

Simulation Parameters	Value(s)
Number of simulation steps	400
Number of elements	30,000
Number of nodes	43,500
Initial temperature of the nodes (°C)	860
Environment temperature (°C)	27
Quenching oil temperature (°C)	27 and 60
Iteration method	Newton–Raphson
Heat transfer coefficient (kW/m2K)	Temperature dependent, see Figure 3

. A precise description of the heat transfer and phase transformation models used in this study are published in the former work of Lopez-Garcia et al. [26].

The simulation variants to reproduce the experimental quenching conditions were two different quenching media temperature (27 and 60 °C), and two different HTC as follow: the 100 L/min oil flux case was calculated by the inverse method using the experimental cooling curves, the details of which will be addressed in an unpublished work in progress by the authors. The second case for nonagitated oil convection HTC was retrieved from the bibliography (Sedighi and McMahon [30]). Prior to the oil contact, 20 s of natural convection in air [43] were simulated to reproduce experimental quenching delay, which is the transportation and immerse time. All HTCs used in simulations can be found in Figure 3.

After the heat treatment, the microstructures were observed using an optical metallographic microscope. The specimens were ground using 80 to 2400 grit SiC abrasive papers and polished with a 1 μm diamond compound then etched in a 3% Nital solution for 30 s. Vickers microhardness (HV) was measured in the cross-section of the specimen using a diamond pyramidal indenter and a LECO LM300AT microhardness tester. All microhardness measurements were evaluated using a standard procedure of loading for 15 s and 1000 g; 12 measurements were made at the core of interest zones defined as C1 and C2, which are the heaviest and thinnest zones, respectively (see Figure 4). An average value was reported.

In the present study, ANOM was used to analyze the effect of the variables (temperature and agitation of quenching oil) on the distortion, residual stresses, and hardness. The Minitab 21^® statistical software was used for the analysis, whose results are shown in the next section. This analysis reveals if there is a relationship between the quenching conditions and both experimental distortion and hardness; it also shows graphically how strong the effects of the factors are with respect to each other and whether there are interactions or not [44].

3. Results

Heat treatment of quenching in oil was carried out in C-ring samples of SAE 4340 steel; the effect of the quenching conditions and simulation results are shown below.

3.1. Microstructure

Figure 4 shows photomicrographs using optical microscopy; the images were obtained from the C2 zone of the Navy C-rings at the surface and core of the piece, Figure 5a,b, respectively.

3.2. Residual Stresses

The results of the residual stress measurements for samples A, B, C, D and an additional sample in nontreated condition identified as the initial condition (IC) were determined by X-ray diffraction, and three measurements per sample were made on the outer face of the zone defined as C1, according to Figure 1b. Figure 6 shows the magnitude distribution of the residual stresses obtained before heat treatment (IC) and after the quenching process in all studied samples (A–D). In addition to this simulation of residual stress performed using Deform^®, the results of this simulation can be seen in Figure 7, which shows the behavior of the same experimental points for residual stress and four quenching conditions.

3.3. Distortion of the C-Rings

The distortion of the Navy C-rings was measured as the dimensional change in the GAP, OD, and ID after the quenching heat treatment; four repetitions were evaluated per quenching condition, and the results are summarized through its box plots in Figure 8. The ANOM test is shown in Figure 9.

3.4. Analysis of Hardness

The hardness in C-rings is presented in Figure 10; for C1 and C2, respectively, average values taken from the core of each zone were reported to evaluate the effect of thickness. The experimental data are presented in box charts and ANOM test in Figure 11 and Figure 12, respectively.

The hardness in C-rings is directly related to the volume fraction of martensitic transformation as well as the chemical composition of the material. It can be seen that except for sample D, the highest hardness values correspond to the section marked as C2 (see Figure 10 and Figure 11). This is related to a higher area/volume ratio at the edges (C2 zone) with respect to the volume of the whole piece, so the heat extraction is higher and the martensitic transformation occurs first in a shorter time. According to the experimental data, the sample B C2 zone presented the highest hardness of 731 HV (61.5 HRC), which corresponds to the variables of oil without agitation and a temperature of 60 °C. Before the quenching heat treatment, the SAE 4340 steel samples had an average hardness of 265 HV (25 HRC); however, as expected after quenching, the hardness increased in a range between 561 HV to 731 HV (53 HRC at 61.5 HRC), which agrees with the type of steel and the microstructure found in Figure 5.

The ANOM test for the C1 zone in Figure 12 shows that the agitation has a dominant effect over the hardness of the samples and no interaction with the temperature factor. The variations oscillate between 585 HV and 660 HV (see Figure 10 and Figure 11) and show that when there is a good agitation in the liquid bath, the hardness increases, because the more the agitation, the more effective the heat transfer. On the other hand, the temperature of the oil did not show a significant effect on the hardness, generating an average magnitude of 620 HV. In contrast, for the C2 zone, the result is that the oil temperature has the greatest effect on hardness, which is greater when the oil temperature is 60 °C, when hardness levels reach above 710 HV. However, the agitation of the liquid did not show such a severe effect but interacted strongly with the temperature in this zone.

4. Discussion of Results

In this section, the most relevant findings are discussed as follow:

4.1. Residual Stress

It can be observed in Figure 6 that the initial values for the C-rings without heat treatment are in a range of −10 MPa to −190 MPa, after the quenching process, and under a different set of quenching variables, the magnitude of the residual stresses changed to positive values in all cases. The lowest magnitude of residual stresses was obtained in samples A and B with 78 MPa and 351 MPa, respectively, which represent the cooling conditions with agitation at 60 °C and without agitation at 60 °C, respectively. In contrast, the highest values of residual stresses were presented in samples C and D with average values of 535 MPa and 386 MPa, respectively, which correspond to the cooling condition with agitation at 27 °C and without agitation at 27 °C of the quenching oil, respectively. It can be observed that the effect of the temperature of the oil in the liquid bath plays a more important role than the agitation of the oil in the formation of a high level of residual stress in the C-rings; in fact, a residual stress measurement in sample C is close to the yield stress of the material (≈700 MPa). In this sense, sample A presented a lower level of residual stress in the studied areas, which corresponds to a good agitation of oil at 100 L/min and a temperature of 60 °C. On the other hand, the residual stresses predicted by the FEM model (Figure 7) show greater agreement with the case of sample A, and it underestimates the stress values obtained at the end of the modeled quenching process in all the other cases (after 100 s of simulation). This may be related to the use of an experimentally calculated HTC, which contains the effect of many other factors that are not considered by a simpler model, such as the shape of the pieces, the size of the tank and its fluid dynamics effects, and the viscosity of the oil that were correctly contemplated in the calculation of the mentioned experimental HTC using real cooling curves.

It is important to mention that the difference between the stress values measured by X-ray diffraction and those modeled by FEM may also be related to the type of detectable stress. The X-ray diffraction technique have the highest definition, since it is sensitive to the microstructure effects (type II and type III stresses), while the model only summarizes the macroscopic effects (type I stresses).

A high magnitude of residual stresses associated with a heterogeneous distribution on the quenched pieces indicates a deficient and unreliable process in quality control, causing distortion in the pieces, lack of dimensional accuracy, or fractures that can occur more frequently and prematurely before the necessary tempering process or late in their service life.

4.2. ANOM for Distortion

Because the C-ring samples were specially designed to induce distortion in the GAP, it can be inferred from Figure 8 that the GAP distortion is the one with less dispersion in the measurements. Also, in the ANOM test for all distortions (see Figure 9), the following is observed: The ID distortion does not show interaction between the factors and the temperature is the dominant factor on the measured response, while agitation does not show significant effects. In other words, this effect would be mainly related to thermal contraction rather than quenching conditions. Meanwhile, OD distortion seems to have a combined consequence of temperature and agitation as well as the GAP distortion but with a stronger relationship with agitation as shown by the main effects of agitation. Although the ANOM test is easy to implement and interpret, the use of more complete statistical tools such as the ANOVA test is recommended when a large amount of experimental data are available, such as in industrial applications as a powerful process control tool.

Figure 13 shows the simulated distortion GAP for all quenching conditions. It is clear that there is no agreement with the experimental results. Simulated results suggest that martensite fraction is overestimated, despite that the use of different HTC and the effect of temperature show results with the expected effect, that is, the HTC with greater agitation will generate greater distortion. As discussed in the residual stress section, it seems that the employment of an accurate HTC cloud generates better results. In addition to the HTC for no agitation conditions, the model for transformation kinetics must be improved. For this research, an algorithm based on classic Time-Temperature diagram was used.

4.3. ANOM for Hardness

The hardness in C-rings is directly related to the volume fraction of martensitic transformation, as well as the chemical composition of the material. It can be seen that, except for sample D, the highest hardness values correspond to the section marked as C2 (see Figure 10 and Figure 11), which is related to a higher area/volume ratio at the edges (C2 zone) with respect to the volume of the whole piece, so the heat extraction is higher, and the martensitic transformation occurs first in a shorter time. According to the experimental data, sample B C2 zone presented the highest hardness 730 HV (61.5 HRC), which corresponds to the variables of oil without agitation and a temperature of 60 °C. Before the quenching heat treatment, the SAE 4340 steel samples had an average hardness of 265 HV (25 HRC); however, as expected after quenching, the hardness increased in a range between 561 HV to 730 HV (53 HRC at 61.5 HRC), which agrees with the type of steel and the microstructure found in Figure 5.

The ANOM test for C1 zone in Figure 12 shows that the agitation has a dominant effect over the hardness of the samples and no interaction with the temperature factor. The variations oscillate between 585 HV and 660 HV (see Figure 10 and Figure 11) and show that when there is a good agitation in the liquid bath, the hardness increases, because the more agitation, the more effective the heat transfer; on the other hand, the temperature of the oil did not show a significant effect on the hardness, generating an average magnitude of 620 HV. For the zone C2, although there is an interaction between the effects of temperature and agitation, the effect of oil temperature appears to be the most dominant according to the ANOM analysis (Figure 12); despite this, the hardness values of samples A and B are very similar, of 710 HV and 730 HV (61–62 HRC), respectively. In both cases, this similarity can be explained by the small volume of the C2 zone, where a complete martensitic transformation is expected, regardless of the level of agitation, as shown in Figure 11.

5. Conclusions

The results show that there is a relationship between the magnitudes of the distortion, hardness, and residual stresses reached in the C-ring specimens caused by the conditions of the quenching oil bath (temperature and agitation).

It was found that the highest residual stress values were related to the quenching conditions where the oil temperature was lower; however, it is the lack of agitation that will cause greater temperature gradients, generating greater dispersion in the experimental values of residual stresses. These values are very close to the yield stress of the material, creating a high probability of irreversible failures, such as excessive distortion and fractures in quenched parts under the as-quenched condition prior to temper treatment. Such conditions must be avoided in industrial applications using this type of steel.

A simple two-factor ANOM test was performed to compare the effects caused by the studied variables on distortion and hardness values. This analysis shows the great effect that agitation has with respect to temperature in applications prone to permanently deform during the quenching heat treatment; thus, agitation systems should be strictly controlled in industrial quenching facilities. Similarly, the hardness values have better performance in areas with higher thermal inertia (higher area/volume ratio) when agitation is the main controlled variable; however, in applications with lower thermal inertia (thin pieces), a better control of the hardness is found by manipulating the temperature variable of the quenching medium.

The use of a numerical model used to estimate the behavior of both distortion and residual stress values after was not very precise in predicting the amount of stress and distortion. Although this model must be improved, results show the importance of using an experimentally obtained heat transfer coefficient.

Using precise FEM modeling in industrial environments can be very time-consuming activity; however, it can be used to identify risk points during the design of heat treatments in new number parts, such as stress concentrators and prone-to-distortion zones.