Verification of the Simulated Carburizing Process in Different Bore Sizes

Abstract

Carburizing is one of the leading surface treatments in the industry. For this study, 20MnCr5 steel was gas carburized and quenched in real circumstances and simulated with Simufact software. The research investigated the dimensions and types of bores. A through and blind bore was used in this study to analyze how the geometry affects the created layer and, additionally, it takes into account the placement in the heat treatment furnace. An optical microscope and Vickers hardness tester were used to detect the changes in microstructure and measure the layer thickness. After the experiments, a simulation calculated the same variables to compare and validate the results to each other. It can be stated that the placement in the chamber did not influence the form of the high carbon content layer. The simulation and the measured results were in good agreement. The maximum hardness difference was 17%, but the calculated layer thicknesses were always between the measured data. For example, in the case of a small blind bore, the calculated layer thickness was 1.68 ± 0.18 mm, while the measured value was 1.54 ± 0.37 mm. Additionally, the hardness change in depth was similar in both cases. After this validation process, the residual stresses and plastic strains were determined. The maximum residual stresses were similar for every case, namely around 1900 MPa, while the maximum plastic strain was measured in a small blind bore with a value of 0.18. The minimum plastic strain was 0.04 in the through bore.

1. Introduction

Carburizing is a well-proven process in the industry if a hard and wear-resistant layer on the surface is needed while the core remains tough. It is usually used in the case of low-carbon steels to increase the carbon content in the surface layer. The carbon diffuses onto the surface at generally high temperatures (870–1000 °C), and because it is a diffusion-controlled process, the heat treatment time is long, typically from a few hours to a day. This process can be performed in different media such as gas, liquid, solid or plasma [1,2,3,4]. Carburizing is a surface treatment that can be followed by quenching and tempering. During the heat treatment, a phase transformation is created in the microstructure that can serve the desired properties of the part. After carburizing, a carbon concentration gradient appears in the surface layer while the core remains unchanged [5,6,7]. Due to the quenching, the layer contains a martensitic microstructure with high hardness, providing wear resistance.

The diffusion of carbon in iron can be described by Fick’s II law which uses a constant diffusion coefficient [8,9,10]. The base concept of this equation can be expressed as a temperature and carbon concentration-dependent Arrhenius function [11,12]. Otherwise, many researchers have measured the carbon diffusivity in austenite with empirical equations. Lee et al. summarized these data in their paper [13]. Carbon, as a strengthening element, can appear not only on the surface but also in the alloy. Increased carbon content is the typical usage for tool steels. It can create carbide with other elements, which improves the hardness and wear resistance of the material [14,15]. On the other hand, it could be present as a form of nano-particle or fiber as reinforcing agents for the light metal matrix composites [16,17].

Based on the numerical theories, simulation software can calculate the treated material’s hardness, residual stress, and microstructure. In most cases, gears are investigated because of their common usage. For instance, 20MnCr5 was used in the following papers, which show the properties of this material after carburizing in real circumstances. In most cases, researchers focused on residual stress, which is significant for distortion behavior during heat treatment [18,19]. The variation of the carbon content and the retained austenite influenced the residual stress distribution, but despite the different values (25–900 MPa), all of it was compressive [20,21,22].

Many different types of software have been used for the finite element analysis and simulation of carburizing. In this paper, Simufact was used, which is rarely known for heat treatments, as it is typically used for welding and forming [23,24].

Not only the software but also the material is not common for simulations. Similar compositions can be found, which are summarized in the following. Liang et al. [25] chose a 20CrMnMo steel spur gear part for simulation and experiments. Carburizing and quenching were used where the key process parameters were the temperature, the carbon potential, and the time. The simulated and the real martensite content was the same, but the predicted hardening depth was 11.4% less than the measured depth. They also used an artificial neural network that predicted the hardening depth to be 1.58 mm, which was almost a 9% difference between the measurement and the simulation. Wang et al. [26] focused on 20CrMnTiH bevel gear. The carbon potential was changed during the carburizing process which was followed by the simulation. The heat transfer coefficient was also changed during the cooling process because a steam film was created around the specimen. The deformation of the gear was measured after the heat treatment, and it was only a few micrometers between the simulated and the real one. The simulation has given a good estimation of the experimental results. Rangaswamy et al. [27] found higher differences in maximum compressive stress between the simulation and the experimental measurements. The range and the value showed disparity, but the other parameters (hardness, retained austenite, carbon content) agreed well. Rudnizki et al. [28] used the pulse carburizing method of high-temperature vacuum carburizing, which was composed of several enrichment and diffusion steps on 20MnCr5B + Nb. A diffusion coefficient was chosen from a MOB2 mobility database but also calculated based on the previously mentioned equations. It was stated that the two values would not fit well with each other. In contrast, the calculations perfectly fit the experimental data.

The aim of this paper is to investigate the effect of gas carburization on a bore dimension and depth. Two different bore sizes with two different depths were manufactured. The samples were placed in the furnace next to the gas inlet and far from that to observe the effect of the gas flow in the holes for the carburization. After that, simulations were made with Simufact software with the same parameters to calculate and validate the diffusion layer thickness, the hardness, and the residual stresses.

2. Materials and Methods

Four different bore dimension was used for the investigation, which can be seen in Figure 1. A small (d = 4 mm) and a larger (d = 20 mm) bore were manufactured as a blind and through version; 4 mm is a critical bore size, especially plasma nitriding, where the loss of electrons is low due to the special geometry [29,30]. To find out the differences between a critical and a higher diameter, 4 and 20 mm were chosen.

Table 1. Dimensions of the samples.

Abbreviation	Name	Dimensions (mm)
LBT	Large bore trough	ø20 × 30
LBB	Large bore blind	ø20 × 15
SBT	Small bore trough	ø4 × 30
SBB	Small bore blind	ø4 × 15

shows the dimensions and the names of the samples. One of every type was placed in two different positions in the furnace to examine how the gas flow influences the carburizing. The Degussa Durferrit type of furnace controls the temperature with electrical heating elements to the max. of 1050 °C. The dimensions and the placement of the samples can be seen in Figure 2.

20MnCr5 steel was used in this study with the following chemical composition in wt%: 0.21 C, 0.31 Si, 1.21 Mn, 1.15 Cr, 0.025 P, 0.035 S, 0.40 Cu, Fe balanced. It is a commonly used carburizable material whose properties were already imported into the MSC Simufact Forming 16.0 software from the JMatPro database. The carburizing and heating were made by an industrial partner based on the diagram in Figure 3. In the first hour, the temperature reached the carburizing temperature (920 °C), and the carbon potential was 0.8%, which increased to 1.08% for the remaining 12 h of the treatment in methanol. After the diffusion, cooling was started in N₂ gas to 200 °C. A quenching in oil and a tempering in a salt bath were then followed. Dashed lines showed the steps of the treatments in the diagram.

The cross-section of the sample was prepared with mechanical grinding and polishing, then cleaned in ethanol and dried in hot air [31]. Next, 2% Nital was used to etch the surface to show the microstructure which was observed by Olympus PMG3 optical inverted research metallurgical microscope. The hardness was measured with Buehler IndentaMet 1105 microhardness tester with a 100 g load.

Two-dimensional axisymmetric simulations were created for the investigation in Simufact software; 10 points of quad elements were used to mesh the samples with the quadtree option. Four different models were generated according to Table 1. The initial mesh size was 1 mm, but the mesh was refined close to the surface to properly simulate carbon diffusion. For the refinement, the Cartesian process was used with level 4 and after with level 2 to achieve this mesh size. The smallest elements around the edges have a 60 µm edge length. After the refinement, the mesh contained ~55,000 elements in every model. It was needed because the diffusion is dominant along the edges and must be calculated precisely. The heat treatment process was the same as the real one described in Figure 3. The carbon potential on the surface was changed as previously described. Namely, in the first hour, the carbon potential was 0.8%, which increased to 1.08% for the remaining 12 h. After that, it was changed to zero because the heat treatment process continued in the N₂ atmosphere. The flow of the gas was not simulated because we found that the position inside the furnace did not change the carburized layer thickness inside the samples. The material properties were imported from the software’s material library, which also contains the data for chemical, mechanical, thermal, and phase transformation.

3. Results and Discussions

Based on the experimental microstructure and hardness, there were no differences between the two places in the furnace, so the simulation focused only on one position, which did not consider the gas flow. The results of the similarity are presented in the following sections.

3.1. Microstructure

The inner surface of the bore (named: in) and the top of the specimen (named: top) were measured to observe the effect of the carburizing in a closed and open system. Chen and Liu [21] used a continuous gas in their specimen in a vacuum furnace. A uniform layer was created in the deep hole (ø3 × 110 mm) with this method. In contrast, our samples were placed in a heat treatment furnace with a continuous gas inlet into the chamber, but not directly to the samples.

After the surface and heat treatment, a fine martensite was created next to the edges and a low-carbon martensite, bainite, with a small amount of ferrite was created in the core, which can be seen in Figure 4. All of the samples had the same microstructure after the heat treatment. The samples were etched in 2% Nital, which can distinguish these microstructures, but determining the layer thickness was not obvious because of the blear transitions.

The simulated carburization can predict the carbon content and the amount of martensite after the quenching. The distribution of the microstructure can be seen in Figure 5 and Figure 6 shows the carbon content after the carburization. Only the blind bore is presented because the results at the beginning of the hole were the same as the through bore. The carbon enrichment was found in the corners because the diffusion was more powerful in nearby places. Finally, 100% martensite can be found until the carbon content reached 0.5%.

3.2. Hardness

The carburized layer thickness was determined by hardness testing, which showed the end of the diffusion zone, where the hardness reached the base plus 50 HV [30]. The hardness was measured on the inner surface of the hole and the top of the sample in the cross-section. The measurement was repeated three times in every section, and the mean values with standard deviation were calculated. These hardness values were determined from the simulation also. However, when the integrated calculation method was used (based on EN ISO 18265:2003 standard [32]), the results were completely wrong, as seen in Figure 7a.

In this case, the maximum hardness was inside the material, not on the surface, which was treated with no significant hardening. After recognizing the error, the calculation method was changed based on the Maynier equations, which results can be seen in Figure 7b [33,34]. The difference was the standard only converted the tensile stress of the material to hardness. However, the Maynier equations first calculated the hardness of the phases which depended on the material composition. After that, the hardness of the given part was calculated from the weighted ratio of the phases. In the software, a user-defined result can be determined, which was calculated after the end of the simulation. The hardness calculation can be optimized because, in the case of the Maynier equations, the hardness of the bainite could change a lot, depending on the material’s composition. The simulation, in some cases, overshot the measured values, which can be caused by more carbon diffusing out from the surface during the end of the heat treatment process. The results from the measurements and the simulations were depicted in the same diagrams, which can be seen in Figure 8, and the maximum hardnesses are summarized in

Table 2. Maximum hardnesses (HV0.1) on different places on the samples after the heat treatments and differences in % according to the simulation’s data.

LBT					LBB					SBT					SBB
Real		Simulated			Real		Simulated			Real		Simulated			Real		Simulated
In	Top	In	Top		In	Top	In	Top		In	Top	In	Top		In	Top	In	Top
785	871	890	871		827	849	1002	1012		834	925	853	886		768	853	854	886
				diff.					diff.					diff.					diff.
			top	0%				top	16%				top	−5%				top	4%
			in	11%				in	17%				in	2.3%				in	10%

.

It can be observed that the maximum hardness was not exactly on the surface but also almost 0.1 mm far from the edge. A retained austenite remained on the surface during the quenching and tempering, which caused a hardness decrease [34,35].

The characteristic of the hardness profile was the same in every case. The maximum hardness (1000 HV0.1) was determined from the simulation, which was an extremely high value compared to the others. However, the decreasing part followed the experimental results well. The average layer thicknesses calculated from the simulations and the measurements are shown in

Table 3. The layer thicknesses with the standard deviation of the samples after the heat treatments.

Average layer thickness (mm)	LBT		LBB		SBT		SBB
	Real	Simulated	Real	Simulated	Real	Simulated	Real	Simulated
	1.86 ± 0.20	1.75 ± 0.04	1.76 ± 0.27	1.86 ± 0.06	1.42 ± 0.50	1.68 ± 0.13	1.54 ± 0.37	1.68 ± 0.18

. In both cases, the layer thickness was bigger in the case of the larger bore. Additionally, the layer thicknesses obtained from the simulation were, in all cases, within the standard deviation of the values determined from the measurements, which validates the simulations.

After this validation, the residual stresses and plastic strains were determined from the simulations, which can be seen in Figure 9. The maximum residual stresses were almost the same in every sample, around 1900 MPa in the outer diameter. However, the plastic strains were different in shape. As one could have expected, there was a peak value at the edge of the blind bores both in the small and the large type, but in the case of the small bore, even if the bore was through, the plastic strains were higher than in the case of the larger. The maximum plastic strains were 0.04 (LBT), 0.09 (LBB), 0.08 (SBT), and 0.18 (SBB), respectively. It can be seen that the plastic strains were doubled in the case of the smaller bore, and it has to be taken into account if the samples are used in the industry.

4. Conclusions

Four different types of samples were carburized in a heat treatment furnace. The carburized layer thickness was determined by hardness testing. The heat treatment process was modeled with finite element simulation with Simufact software. The microstructure, hardness, layer thickness, residual stress, and plastic strain were calculated from the simulation. A user-defined result was described inside the Simufact software with the Maynier equations for the hardness calculation. The results of the simulation were validated with the layer thickness determined from the measurements. It was found that the simulation predicted the layer thickness between the measured data. After this validation process, the residual stresses and plastic strains were determined. The maximum residual stresses were similar for every case, namely, around 1900 MPa, while the maximum plastic strain was measured in the case of a small blind bore with a value of 0.18. The minimum plastic strain was 0.04 in a large through bore.