Engineering Structures 177 (2018) 489–505

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Review article

Modeling methods for gas quenching, low-pressure carburizing and low- T

pressure nitriding
Emilia Wołowiec-Korecka
Institute of Materials Science and Engineering, Lodz University of Technology, 1/15 Stefanowski street, 90924 Lodz, Poland

A B S T R A C T

Modern thermal and thermochemical treatment of steels is a rapidly growing research area. Thermochemical processes are researched, designed, optimized, and
controlled by computer-aided methods. The goal of the present paper is to provide a comprehensive state-of-the-art review and to analyze the critical mass of
knowledge in the field of modeling methods to describe the interactions between the processing, microstructure and postoperative properties of the treated material.
This article briefly discusses modern low-pressure thermochemical processes to describe the difficulties in modeling their non-equilibrium nature and transient states
that occur in the course of the processes. A synthetic review of modeling methods, in particular, artificial intelligence methods, is presented to investigate low-
pressure thermochemical processes. The opportunities and dangers associated with these methods are also analyzed.

1. Introduction computational methods. The traditional method of achieving optimum

material properties by trial and error is time-consuming and costly,
The basis of an effective thermal and thermochemical treatment is which is an argument for seeking alternative ways to design thermal
to capture the relationships between the treatment processes, micro- and thermochemical treatment processes. Computer modeling and si-
structure and post-process properties of material [1]. The ability to mulations facilitate the design of engineering materials and predictions
identify and control the parameters that affect the intensity of the of the properties of these materials with a significant reduction in time
processes is the key to achieving the desired combination of material and costs. This applies to the processes themselves and to the post-
features required for a specific application. Understanding these re- process properties of the items under treatment [12–21].
lationships is the starting point for further actions, such as the opti- Computational support is of particular importance in low-pressure
mization of a process or initial state of the material, e.g., the compo- thermochemical treatments due to their non-equilibrium nature and
sition of alloys, in terms of the effectiveness, duration, and cost of the transient states during the course of the processes. In this case, the
treatment. It is also important for practitioners to achieve a high level of primary goal of the simulation is to predict the course of the process
conformity for real items under treatment with technological assump- and the final properties of the product, thus ensuring the repeatability
tions [2–5]. Thermochemical treatments of a surface, both carburizing of the process results. In industrial applications, it is a well-known si-
and nitriding, is of great importance in engineering practice [6–9]. tuation, when the range of achieved effective case depth in heat treated
Combined with quenching, thermochemical treatments are widely used elements is too wide. In consequence, some parts not meet the required
in machine parts due to their beneficial effects on fatigue resistance and standard. The problem reason is the batch nature of heat treatment. A
tribological and anti-corrosion properties. However, introducing addi- whole load of elements (details, parts) is situated in one furnace
tional carbon or nitrogen atoms to a workpiece, as well as the changing chamber where the carburizing and quenching take place. In fact, every
the structure by quenching, brings about residual stresses that often element in the chamber is carburized and quenched with different
result in a significant distortion of machine parts and undesired effects parameters: temperature, carburizing gas flow, quenching gas flow, etc.
[10]. To improve such defects, it is necessary to have a thorough un- From a modeling point of view, it is possible to predict that situation.
derstanding of how the process parameters influence the final profiles When characteristics of the furnace are known (temperature distribu-
of carbon/nitrogen, stress/distortion, and microstructures of heat- tion in the chamber, placement of carburizing and hardening nozzles,
treated components [11]. charge density), then the points with the maximum and minimum ef-
As expectations regarding the quality and precision of product ficiency in the chamber can be determined. To reduce the problem
manufacturing continue to grow, further advances in materials en- usually a modification of the way of distributing elements in the batch
gineering are closely associated with the use of computer-aided (based on the simulations calculations) is needed.

E-mail address: [email protected].

URL: https://linkedin.com/in/e-wolowiec.

https://doi.org/10.1016/j.engstruct.2018.10.003
Received 1 March 2018; Received in revised form 9 August 2018; Accepted 1 October 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

Nomenclature Ms initial temperature of martensite transformation

Q heat flux
Symbol notation T temperature
c heat capacity
β coefficient of carbon transport across the gas-metal phase h heat transfer coefficient
boundary t time
ρ density x mole fraction
ε strain
εe elastic strain Abbreviation notation
εp plastic strain
εth thermal strain ADALINE ADAptive LINear Element
εtr phase transformation strain CFD computational fluid dynamics simulation
εtp transformation plasticity strain DM data mining
εdf diffusion induced strain of carbon and/or nitrogen FDM finite differences method
ξ volume fraction of the transformed phase FEM finite element method
ϕk volume fraction of transformed austenite FVM finite volume method
ϕM volume fraction of martensite HPGQ high-pressure gas quenching
σij stress KDD knowledge discovery in databases
A surface area of the solid component k-NN k nearest neighbors method
C concentration of carbon in the material LPC low-pressure carburizing
CMax maximum solubility of carbon within austenite LPN low-pressure nitriding
DC carbon diffusion coefficient MLP multi-layer perceptron neural network
I flux of atoms across the surface of the material NN neural network
K heat conduction coefficient RBF radial basis function neural network
L latent heat due to phase transformation SVM support vector method
Mf final temperature of martensite transformation

Compared to processes that use atmospheric conditions, low-pres- computer modelling of the processes that take place during heat
sure treatments have many advantages. Due to a higher carbon mass- treatment to predict production regimes.
flow rate, the time required for the low-pressure carburizing (LPC)
process can be dramatically reduced, especially for larger case depths. 2. Modeling methods: conventional, numerical and those based
Because of the absence of oxygen (air), the quality and properties of the on knowledge discovery databases
part surfaces are improved [22]. High-pressure gas quenching (HPGQ)
in combination with LPC is commonly used to harden new transmission 2.1. Data analysis methods in model preparation
gears in the automotive industry [20,23–26]. Using gas as the quench
media, as opposed to oil, can reduce deformations within each part due The general procedure of process modeling is shown in Fig. 2.
to the single-phase nature of convection cooling with gas [27]. In the Construction of a model should start with a preliminary analysis, which
case of low-pressure nitriding (LPN), the nitrided layers are less brittle usually consists of three parts. The first part is extrapolation aimed at
than those formed by gas-nitriding, both within the limited zone of preparing the data. Data usually need be cleaned (e.g., removing re-
nitrides and zone of external nitriding [28]. There has thus been in- dundant, contradictory, dubious data) and transformed into a usable
creasing experimental and theoretical interest in mathematical de- form. Preliminary choices of the sub-set and variables (features of the
scriptions of low-pressure carburizing and nitriding processes in recent phenomenon, which are regarded as significant) are made in the next
years [12,20,29,30] (Fig. 1). Some of these processes have been com- step. The aim of these actions is to reduce the number of variables
mercialized [7,22,31–35] (Table 1). under analysis to a rational level to enable an effective analysis. Further
The areas of practical application of computer techniques in mod- actions depend on a specific issue, as mentioned above. Different
elling thermochemical treatment were first outlined by Lakhtin et al. methods can be used, from the simple choice of variables by linear
[36], who identified two basic directions: the use of a computer for regression to examination by graphic and statistical methods to de-
direct control over the production process of heat treatment and termine the most important features as well as the complexity and the

Fig. 1. Number of publications on modeling heat-chemical treatment (till 31 December 2016).

Source: Scopus.com

490
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

Table 1 the final stage, where it will be applied to new data.

A review of the commercially available software for designing and simulating
low-pressure thermochemical problems.
2.3. Conventional modeling methods
No. Provider Product Specification References

1 ALD Vacuum VC-Sim Low-pressure [22]

The literature on this subject provides the criteria for the classifi-
Technologies carburizing cation of modeling techniques into three distinct groups: conventional
2 ECM Technologies Infracarb Low-pressure [23] (physical and mathematical), numerical and ‘statistical,’ each of which
software carburizing has its own advantages and areas of application (Table 2) (Fig. 3). A
3 Ipsen Carb-o-Prof Low-pressure [31]
similar classification was provided by Malinov and Sha in an extensive
carburizing
4 Seco/Warwick SimVac Low-pressure [7,32] review on modeling software and simulation in material-related issues
carburizing [39]. The groups of methods are constantly being improved and applied
5 Lodz University of SimLPN Low-pressure [35] to model different processes and correlations, both in producing pri-
Technology nitriding mary material and in further treatment by material engineering
6 Worcester Polytechnic CarbTool Low-pressure (and [33,34]
Institute (CHTE) gas) carburizing
methods.
Physical methods are usually based on fundamental laws and the-
ories that describe the physical nature of the process. Therefore, the
general nature of the phenomenon, in accordance with the research basic assumption in this type of modeling is that a physical model and
guidelines [37,38]. mathematical equations which describe the phenomenon are perfectly
known as well as the equations can be solved by analytical methods.
2.2. Model construction, assessment of correctness and usability of results
2.4. Numerical modeling methods
The next stage involves the main data analysis. Various models can
be used to reproduce the nature of the phenomenon under considera- Numerical methods are commonly used in applied mathematics.
tion, and the best model is chosen. The quality of the prediction, i.e., They are used when a problem under study does not have an analytical
whether the value of the variable being modeled is correct, and the solution (given by formulae) or when the application of such solutions
stability of results for different samples are the assessment criteria. The is troublesome due to their complexity. The results obtained by such
model that is found to be the best at the data analysis stage passes on to methods are usually approximate, but the accuracy of the calculations

Fig. 2. General procedure of process modeling. 1 – Preliminary analysis (A – Extrapolation, B – Preliminary choices of the variables, C – Determination of the most
important features, the complexity and the nature of the phenomenon); 2 – Model construction and assessment of correctness (main data analysis); 3 – Applying of
the model to new data.

491
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

Table 2
A review of the methods for the modeling of thermal and thermochemical problems.
Group of models Method Application in modeling of thermal and thermochemical References
processes

Conventional (mathematical and physical) Analytical equations Fick’s laws solution [105,109,144,148,149]
models Carbon activity in austenite (function of alloy elements) [147]
calculation
Carbon diffusion coefficient calculation [148,152,153,156,157,163]
Diffusion kinetics for LPC [141,142]
Volume fraction of phases transformed prediction [183,190–192]

Numerical models Computer Fluid Dynamics Pyrolysis of propane and acetylene under vacuum modeling [126,130]
(CFD) methods Heat flow issues [173,178]
Cooling gas flow issues [135,173]
Ammonia decomposition modeling [232]
Finite Element Method Temperature history during cooling [175]
(FEM) Hardness distribution after LPC + HPGQ calculation [19,21,170,184]
Phase fraction after LPC + HPGQ calculation [44]
Phase transformation kinetics during LPC + HPGQ prediction [3]
Strains due to the phase transformation, transformation [42,43]
plasticity, and thermal expansion/ contraction after HPGQ
calculation
Residual stress after HPGQ prediction [43,206,207]
Deformation after LPC + HPGQ prediction [43,181,188,207]
Finite Difference Method Carburizing modeling [140]
(FDM) Fick’s laws solution [151]
Heat transfer coefficient calculation [183]
Finite Volume Method HPGQ process modeling [168,170]
(FVM)

Knowledge Discovery Basic statistical Analysis of variance Evaluation of the significance of factors that affect the [37,38]
Databases (KDD) models phenomenon
Multiple regression Determination of the carbon diffusion coefficient [148,152,153,156,157,163]
Data mining Regression neural networks Multiphase diffusion during LPN prediction [29]
models Forecasting of nitrided layer hardness after LPN [29]
Prediction of the formation and dissolution of carbide [248]
precipitations during LPC
Hardness distribution after HPGQ [195]
Precipitation phenomena prediction during LPC [67,249]
Heat treatment process parameters optimization [15,16,63,67]
Steel hardenability and other properties modeling [55,60,62]
TTT transformation diagrams prediction [58]
Classifying neural networks Categorization of materials by their features [29]
Heuristic method Surface phenomena during LPC prediction [67,249]
Genetic algorithm Heat treatment parameter optimization [15,16]
Alloy materials design [90]
Evolutionary algorithms Heat treatment process parameter optimization [89]
Database models Relational databases Gathering material data for analyses and calculations [19]
Tabulation Acceleration of calculations [19]

can be pre-defined as needed. Numerical methods consist of closing in methods of data analysis and an analytical approach are insufficient.
on the final result by using a sequence of approximations. In effect, the This applies to situations when analysis of a large amount of data is
result will always have some error, although the error can be as small as required, and a number of parameters need to be taken into account,
one chooses. In the past two decades, numerical simulation by the finite which can have a significant effect on the process features or item that
element method (FEM) has been widely used in engineering problems is being processed. Also, in non-equilibrium processes, the relationships
due to its ability to reduce time-consuming and trial-and-error experi- between the process parameters and features of the item under treat-
ments that plague engineers. According to Kim et al. [40], finite ele- ment can be unknown, and it is risky to make assumptions regarding
ment models are currently convenient to predict the deformation, the type of relationship that they have (e.g., that it is linear). Such cases
temperature history, phase fraction, and hardness during carburizing require a flexible approach.
heat treatment as well as to calculate the strains due to phase trans- The use of data mining (DM) methods has been a common trend in
formation, transformation plasticity, and thermal expansion/contrac- modeling and simulations for several years. To quote Berry and Linoff,
tion after high-pressure gas quenching. Therefore, numerous re- data mining is a process of studying and analyzing large amounts of
searchers who study workpiece properties following thermochemical data by automatic and semi-automatic methods aimed at discovering
treatment [41–43] have stopped using analytical calculations and use significant patterns and rules [48,49]. Model quality assessment is
FEM software instead, such as HEARTS [44], ABAQUS [45], SYSWELD based on a test sample that is usually taken from an experimental da-
[46] and DEFORM-HT [47]. However, in these commercialized pro- tabase. DM is part of a broader set of methods of knowledge analysis,
grams, some problems remain imperfectly modeled, such as the effect called knowledge discovery in databases (KDD), which involve the use
of the diffusion of nitrogen or simultaneous diffusion of carbon and of data mining methods as well as database-related techniques. Migut
nitrogen upon subsequent thermo-mechanical simulation [11]. and Demski mention four key features that distinguish data mining
from traditional statistical data analysis [50]:
2.5. Modeling methods based on data mining
1. Ability to analyze large bodies of data.
When knowledge of a phenomenon is incomplete, the traditional 2. Orientation towards practical results and applications rather than on

492
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

compression and classification [77].

• The RBF (Radial Basis Function) networks [78,79] have similar
applications for perceptron networks (MLP). They are the most
useful for the classification of objects when class elements are
grouped around points in multidimensional space (or oscillate
around a common center). Networks of this type cope very well in
such cases, even when the number of neurons is very small.

3. Problem of the proper choice of a model

To summarize the modeling methods, the choice of a model depends

on the task. Conventional methods can be successful in modeling phe-
nomena that are well-known and described by analytical equations.
When the boundary conditions are properly selected, conventional
models cover the whole spectrum of parameters. However, the right
choice of coefficients is of key importance for model correctness [39]. If
the relationships are more complex, such as for phenomena described
by differential equations, then numerical methods have to be applied,
which may not produce an ideal solution, but they are sufficiently
precise to make the result of the calculations sufficiently close to the
expected value. Phenomena that cannot be described by mathematical
equations or for which no mathematical apparatus is available can be
processed using statistical methods, including data mining (DM)
methods [48,49]. It is recommended that the search for solutions
should begin with simple models and methods that can be made in-
creasingly complex as needed. The author of this review also suggests
not discriminating against the simplest methods as long as they prove
sufficiently effective.
Fig. 3. Modeling methods classification for modeling of gas quenching, low- According to Berry and Linoff [48], two principal groups of mod-
pressure carburizing and low-pressure nitriding. elling tasks include predictive issues and discovering new knowledge,
with predictive issues including classification, regression and time
developing and testing theories. series, as well as discovering new knowledge including detecting de-
3. Applying methods that are capable of reproducing virtually any viations, segmentation, cluster analysis, analysis of associations and
relationship between variables. sequences, summary listings, visualization and text mining. The studies
4. Using existing data first. conducted for this review primarily concerned predictive issues,
whereas an extensive description of the whole subject matter can be
Artificial neural networks (NN) are a type of DM method. They take found in papers by Weiss and Indurkhya, Ratner, Hassti et al., Braha
inspiration from the human nervous system and are widely used in and others [37,38,48,49,80–85].
nonlinear modeling [51,52]. The use of ANNs is not very common in
thermal and thermochemical treatments due to the need to gather vast
training databases, but they are of increasing interest to researchers 3.1. Predictive application of modeling methods
because of their capability to go beyond conventional mathematical
methods. Neural models cannot be interpreted because they are a ‘black Predictive issues involve situations in which there is a need to de-
box’ type of model; however, this inconvenience is accepted by many velop a model that can be used to predict a value of a feature (depen-
researchers, and the number of studies in which such methods are dent variable) from a set of other features (independent variables, such
employed has been steadily growing [39,53–67]. The interest in neural as process parameters). A model is trained on an example set of objects
networks means that we will likely see a rapid increase in their number (cases) for which both independent and dependent variables are
in the near future. known. Predictive issues include classification and regression. Apart
Proper selection of the architecture of neural networks is of parti- from conventional and numerical methods, the following DM methods
cular importance to the teaching process and subsequent operation are applied to solve regression problems [38,48,49,80,86]: linear re-
[68]. There are no set guidelines that assign a type of neural network to gression, regression trees, linear models and generalized linear models,
a specific issue, but statistical analysis of an issue indicates that certain MAR splines, neural networks, boosted trees, the k nearest neighbours
types are applied more frequently for specific issues than others: (k-NN) method (which can be useful in supplying missing data), and the
Support Vector Method (SVM). The following methods are applied to

• The Madaline network, consisting of the linear elements of solve classification problems: neural networks, classification trees,
discriminative analysis, logistic regression, generalized linear models,
ADALINE (ADAptive LINear Element), is useful in recognizing ob-
jects, filtering signals, grouping, and modeling [69,70]. general additive models, boosted classification trees, the k-NN method,

• The Kohonen network has been used successfully for classifying and naive Bayes classifier, SVM, and MAR splines. These methods are quite
different, and it is difficult to select the best method. The effectiveness
recognizing objects [71–73].
• The Hopfield network (with feedback) has also been used success- of a given method (or a group of methods) depends on the objectives of
the analysis as well as the structure and quality of the data; therefore,
fully in recognizing and segmentation [74,75].
• The MLP (Multi-Layer Perceptron) network [71–73], i.e., a multi- Migut and Demski analyzed the literature on the subject and suggested
[87] that the assessment should be aspect-related, as summarized in
layer perceptron (sigmoidal neuron) network, is a universal ap-
proximator; it has been used successfully in modeling [76], Table 3.

493
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

Table 3 Calculations are made rapidly in the first case; in the other, there is a
A review of the selected DM methods that are helpful for solving regression and time-consuming network training stage, but it is a one-off action
classification problems [87]. [68,89].
Methods of modeling: To summarize the data from the literature and the author’s personal
1. NN, 2. SVM, 3. Trees, 4. MARS, 5. K-NN, 6. Linear models experience, it is recommended to take into account the following cri-
(GLM, GRM, GDA), 7. Boosted trees, 8. Naive Bayes method teria in methods selection for modeling individual problems:
1 2 3 4 5 6 7 8
1. Data quality (error, noise, outliers, dummy variables, and redundant
Handling data of – – + + – – + + variables, i.e., those that provide repeated information),
various types 2. Data quantity,
naturally
3. Implementation speed (expected date of model launching),
Missing data – – + + + – +/− –
Robustness to – – + – + – + – 4. Expected relationship complexity,
outliers 5. The need for an easy interpretation of the model or the possibility to
Scalability – – + + – + – +/− accept a black box-type model,
(N 6. Specificity of the issue (the practical experience of the implementing
observations)
researcher, ‘suggesting’ a method for a specific type of problem),
Scalability – – + +/− +/− + +/− +/−
(N variables) 7. Economic and rational factors (applying a method that is the best
Robustness to – – + + – +/− – – but is time-consuming and that requires financial outlays that may
insignificant not be justified, e.g., in preliminary or estimation studies).
variables
Interpretability – – + +/− – +/− – +
Predictive power + + – +/− + – + – 4. Single-phase diffusion in low-pressure carburizing

Low-pressure carburizing (LPC) is a heat treatment process that is

3.2. Seeking the optimum process parameters
used for surface hardening. It is a nonequilibrium process in which
small amounts of carburizing gas are continuously pumped in and out
Reversal of the direction of the links between the input parameters
of a furnace at a rate that allows the carbon on the surface of the steel to
of processes and their outcome values with the aim of determining the
reach the austenite solubility limit [94]. It was invented in late 1968,
best process parameters is an optimization issue. Evolutionary algo-
patented and fully commercialized [7,95–103]. LPC is operated by a
rithms, in particular, genetic algorithms, are the most precise tools that
pulse technique in which cycles of saturation and annealing are re-
can be used to solve these types of problems [88–90]. However, for
peated at a constant temperature [94,104,105]. During the saturation
many cases, this solution is not acceptable due to the prolonged cal-
stage (boost stage), the steel surface is saturated by carbon from a low-
culations necessary to yield a result. Therefore, there are two better
pressure gas atmosphere and diffuses into the material. Then, the gas
models: a mathematical-heuristic model and a model that uses neural
supply is stopped, and only carbon diffusion occurs (Fig. 4). Currently,
networks. A heuristic model is usually developed by a trial-and-error
the process has been elucidated quite well, both in terms of its kinetics
method. It is a combination of physical equations and engineering
[106–111], the effects of individual factors [112,113], surface phe-
simplifications. The characteristic feature of a heuristic model is that it
nomena [114] and the final properties of carburized material
‘defends itself’ by the conformity of its output data with the results of
[115–120]. It was proven that LPC provides a more uniform carburized
real-life experiments; however, there is no mathematical proof that it is
oxide-free case depth, cleaner parts and less part distortion
the best existing solution to the problem [91–93]. An algorithm based
[94,121,122].
on a neural network is an alternative method to optimize a process.
Mass transfer during carburizing proceeds in three stages: carbon

Fig. 4. The change of surface concentration of carbon during low-pressure carburizing.

494
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

transport from the atmosphere to the steel surface, chemical reaction at long boost stage during the process forms not only carbides but also a
the gas-steel interface, and carbon diffusion between the carbon-rich carbon deposit on the surface of the parts [105]. Gorockiewicz, La-
atmosphere and steel surface, which provides a driving force for the pinski, and others confirmed these reports [106,121,136,137].
mass transfer phenomena and establishes a carbon concentration gra-
dient at the surface of the carburized part [123]. Zajusz et al. [124] as
4.3. Carbon diffusion in austenite
well as Wolowiec et al. [110] report that the carburizing layer thickness
and carbon distribution in the diffusion zone depend on the process
Unfortunately, the sizes of carbon and nitrogen atoms are very small
parameters as well as mass transport (Fig. 5a). The latter is described by
and it is a very hard thing to obtain the profiles of those atoms.
the two main parameters: the mass transfer coefficient through the gas-
Probably the most popular method is a planing thin chips of steel,
solid interphase β(t) and diffusivities of the carbon and steel compo-
burning them and then analyzing the existing elements. However, it is a
nents.
time-consuming method that does not guarantee the precision of the
result. Another method is determining carbon concentration profiles in
4.1. Carbon transport from the atmosphere to the steel surface the surface layer by glow discharge optical emission spectrometry
(GDOES). The surface of the sample is fired and analyzed simulta-
Carbon used in diffusion saturation of austenite is obtained in the neously. It can be analyzed up to a depth of 0.1 mm at a time. Then the
process of thermal decomposition of aliphatic hydrocarbons, as de- sample need be polished about 0.1 mm deeper and the analysis may be
scribed by Grafen and Edenhofer as well as Kula et al. [105,125]. Ki- repeated until the base concentration is obtained. It is the very precise
netic models of the decomposition of light hydrocarbons, such as measurement of carbon profile (with accuracy ± 0.002%C), but still, it
ethylene, acetylene, and propane, have been developed by Khan et al., is an expensive and time-consuming method, mostly not acceptable in
Matsugi and Miyoshi, Yada and Watanabe and other researchers industrial conditions. In laboratory conditions that technique may be
[126–135]. The impact of the acetylene flow on the carburized layer also used for nitrogen profile determining. However, many researchers
characteristics was described by Rokicki and Dychton [94]. report it is some difficult to obtain a nitrogen standard for calibration of
the device in that investigation.
4.2. Chemical reaction at the gas-steel interface To avoid the time-consuming measurements of real carbon profiles,
researchers studying low-pressure carburizing have to develop their
The phenomena occurring at the gas-metal phase boundary that own models beginning with the basics: developing a physical model and
determine the rate at which the carbon concentration changes at the the analytical equations that describe it. First attempts to develop
surface, including the most important catalytic interactions between the carburizing models were based on the simplest, one-dimensional form
carburizing atmosphere, are reflected in the mathematical model by the of the mass conservation law (the second Fick’s law) as well as con-
atomic carbon flux IC crossing the boundary: centration-dependent carbon diffusivity [138–140]. A general analy-
IC = −β [CA −(C(t))S] tical low-pressure carburizing model based on the Devian-Osterend
(1)
equation was presented by Tsepov [104]. More complex analytical
where IC – flux of carbon atoms across the surface of the carburized models to calculate the effect of carburizing on the steel properties were
material, cA – concentration of carbon in the material at equilibrium described by Morral et al. [141,142], Cavaliere et al. [143], and Se-
with the carburizing atmosphere, (c(t))s – current concentration of menov et al. [144,145]. The solution of the problem in all studies
carbon at the material surface, and β – coefficient of carbon transport consists of calculating the C(x, t) function, which is the concentration of
across the gas-metal phase boundary. carbon varying with time and distance from the surface.
The boundary condition for the low-pressure carburizing process is Kolodziejczyk [146] described a model in which the carbon diffu-
different from that of the gas carburizing process. In the gas carburizing sion coefficient DC in austenite depends on the temperature, carbon
process, the carbon potential can be monitored and well controlled. concentration and presence of alloy additives. The carbon activity in
During a low-pressure carburizing process, the surface carbon increases austenite in the presence of alloy additives affects the carbon diffusion
dramatically and can reach the maximum solubility. Saturation of the coefficient and allows for the computation of the maximum solubility of
carbon decreases with the diffusion of carbon. Kula et al. found that a carbon within austenite CMax at temperature T for a particular material.

Fig. 5. Selection of low-pressure (vacuum) carburizing and gas quenching modeling aspects.

495
 E. Wołowiec-Korecka

Table 4
Empirical equations for the diffusivity of carbon in austenite. (D – the diffusivity of carbon in cm2/s, C – the carbon content in mass percent, xC – the mole fraction of carbon, RC – the gas constant
as 1.986 cal/mol/K, RJ – the gas constant in 8.314 J/mol/K, and T – the temperature [K]).
Carbon diffusivity in austenite Ref.

153593 − 15370C − 10727C2 +2432.9C3 [152]

D (T, C ) = (0.47263−0.55169C + 0.22848C2−0.02965C3)·exp(− ) (2)
RJ T
2
38300 − 1.9·105xC + 5.5·105xC [153]
D (T, C ) = 0.5·exp(−30x C )·exp(− ) (3)
RC T
37460 [154]
D (T) = 0.668·exp −
( )(4) RC T

36000 [155]
RC T
D (T) = 0.49·exp (− )(5)
32000 [155]
RC T
D (T, C) = (0.07 + 0.06C)·exp (− )(6)
31350 [155]
RC T
D (T, C) = (0.04 + 0.08C)·exp (− )(7)
37000 − 6600C [157]

496
RC T
D (T, C) = 0.47·exp(−1.6C)·exp (− )(8)
180590 [158]
RJ T
D (T) = 0.4718·exp (− )(by liquid carburization) (9)
127950 [158]
RJ T
D (T) = 0.0052·exp (− )(by solid carburization) (10)
107400 [159]
RJ T
D (T) = 0.0077·exp (− )(11)
112000 [160]
RJ T
D (T) = 0.02·exp (− )(12)
144300 − 15000C + 370C2 − 4366.3Mn + 4050.7Si − 1240.7Ni + 7726Cr + 12126.6Mo − 6788.6Al [162]
D (T, C) = (0.146−0.036C(1−1.075Cr )−0.0315Mn + 0.0509Si−0.0085Ni + 0.3031Mo−0.052Al)·exp ⎛− ⎞ (13)
⎝ RJ T ⎠
−32000 [148]
DC = (0, 07 + 0, 06Cγ )·exp − ( kT )(14)
−37000 − 6600·Cγ [157]
DC = 0, 38·exp − γ
( 1, 986·T )−1, 6·C (15)
− 18900 + 0,38 + 4300 − 2,63 Cγ1,5
( ) ( ) [156]
DC = 0, 68·e T T (16)
DC = (1, 876·10−11·T2−4, 468·10−8·T + 2, 669·10−5)·exp [(−0, 0034·T + 5, 06)·Cγ ](17) [161]
19700 [124]
1.43exp(− + 0.002242exp(6790 / T ) cC )
T
DC (cC ) = (18)
1 − 0.232cC
Engineering Structures 177 (2018) 489–505
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

It can be calculated using an analytical method, for instance, a method cooling [164].
based on equations provided in work [147].
Q = hAT(2) (20)
The development of carbon profiles in austenite depends on both
the phenomena occurring at the gas-metal phase boundary and on the where Q is the heat flux, A is the surface area of the solid component, T
diffusion of carbon in austenite. Carbon transport in austenite has been is the temperature difference between the surface of the component and
described by Fick's diffusion laws. The first law describes the flux of the the fluid gas, and h is the heat transfer coefficient, which combines the
diffusing component. The second law describes the change in the influences of the convection and radiation.
carbon concentration as a function of time. Fick’s laws are solved by an A number of studies exist that simulate the quenching processes
analytical method, as described in papers by Goldstein and Moren using the finite element method (FEM), including work by Inoue and
[148], Jimenez et al. [149], Ochsner et al. [109], Demenov et al. [144] Arimoto [44], Watt et al. [165] and Wallis et al. [166,167], Wang et al.
and Karabelchtchikova et al. [150]. It is also possible to use the Crank- [168], Kopun et al. [169], Elkatatny et al. [170]. This method is
Nicolson finite differences method (FDM) [33,151]. Finite element available ‘ad hoc’ in the ANSYS-CFX tool and has the advantage that
calculations were used in work by Cavaliere at al. [143]. The model of when performing numerical computations, one can solve a system of
Sisson, Karabelchtchikova, and Wang that predicts the carbon con- three equations: conservation of mass, conservation of momentum and
centration profile in steel in both gas and low-pressure carburizing conservation of total energy [171]. A reliable review of the application
processes is also based on analytical equations [4,33,150]. of the FEM in analyzing and simulating the quenching process in over
Functions describing the relationship between the carbon diffusion 290 cases was completed by Mackerle [172]. Gas quenching flow in a
coefficient DC and temperature or time have been presented in nu- quenching chamber is as frequent an issue as heat flow in a workpiece.
merous articles [148,152–162]. All researchers agree that the correct Practically all researchers [173–179] have noted the usefulness of the
relationship is of key importance for model consistency with real ex- ANSYS-CFX software [171] based on the finite volume method (FVM)
periments, although they provide varying relationships. The first stu- and of the ANSYS-FLUENT software [180] based on the finite element
dies of the model were based on the carbon diffusion coefficient for- method (FEM).
mula proposed by Wells et al. [152]. This relationship turned out to be The importance of the heat transfer coefficient h is similar to that of
insufficient, and therefore, formulas were proposed by Leyens [156], the diffusion coefficient DC in carburizing. It has a significant influence
Goldstein [148], and Tibbets [157]. Relationships concerning carbon on the quenching results, including distortion, residual stresses, and
diffusivity in austenite are presented in Table 4. More papers related to hardness distribution, and therefore, a number of reports have been
these issues are discussed by Ochsner et al. in [109]. published on the determination of its optimum value
[17,164,181–184]. According to its physical meaning, the combined
5. Heat flow during high-pressure gas quenching heat transfer coefficient is influenced by the surface temperature of the
component. The heat transfer coefficient schedule can be calculated
Carburizing is inseparably linked with quenching. Low-pressure according to the cooling history by the trial and error method [181],
carburizing with high-pressure gas quenching is one of the fastest inverse techniques [185] or finite difference method [183].
growing manufacturing processes in both the automotive and aerospace Various mathematical models of quenching, with different ac-
industries. The critical parameters that determine the carburized ele- curacies and complexities, have been developed in the past two dec-
ment properties after quenching are the cooling efficiency (dependent ades. According to Davis et al. [186], there are three general types of
on the type of cooling media, pressure and velocity of gas flux), che- models used in property prediction and computer simulation: static
mical composition and thermal properties of the cooled element models, dynamic models, and programs with both static and dynamic
(Fig. 5b); transformations during cooling; and resulting internal stresses models. Static models are based on simple empirical formulas derived
[163]. Dowling observed, and other researchers confirmed, that it is from experimental data and regression analysis. Dynamic models are
important for the problem at hand to take the following issues into based on the solution of a series of differential equations. An important
account [13]: category of dynamic models is the use of finite element analysis to
predict the distortion, residual stresses, and hardness distribution
(1) The effects of the chemical composition and physical properties of [44,165–167,187]. The finite element simulation of the quenching
the material being cooled on its post-cooling features. process includes heat transfer, phase transformations, and deformation.
(1) The effects of the geometry and mass of the cooled items on the These three parts are influenced by each other and should be coupled
cooling rate. together during the simulation process [181]. In this case, finite ele-
(2) The relationship between the thermal conductivity coefficients, ment analysis of the gas quenching process can be implemented by
material density, heat capacity and temperature. using DEFORM-HT [188].
(3) The item cooling rate for a point in the material taken into con-
sideration and its distance from the surface. 5.2. Phase transformation kinetics

5.1. Heat transfer Phase transformations during the quenching process lead to an ex-
tensive rearrangement of atoms. Depending on whether diffusion oc-
The relationships above are interlinked with the equation of non- curs, phase transformations during quenching can be classified as dif-
steady-state heat transmission (Fourier’s formula): fusional and martensitic phase transformations. The diffusional phase
∂T ∂ ⎛ ∂T ⎞ transformation involves the diffusion of carbon and other solute atoms,
ρc = K −σij εijp + LI ξI and it is time-dependent. Martensitic phase transformation is diffu-
∂t ∂X ⎝ ∂X ⎠ (19)
sionless and occurs very rapidly [181].
where ρ is the density, c is the heat capacity, K is the heat conduction Diffusional phase transformation includes transformations from
coefficient, LI is the latent heat due to phase transformation, σij is the austenite to ferrite and cementite. The isothermal transformation ki-
stress, εijp is the plastic strain rate, and ξI is the volume fraction of the I- netics of ferrite/pearlite and bainite are modeled according to the law
th phase transformed. The second and third terms of the right side of developed by Johnson-Mehl-(JMA) [189] and confirmed by others
the equation represent heat due to plastic deformation and latent heat [190–192]. It is used to predict the volume fraction of the transformed
due to phase transformations, respectively. The rate of heat transmis- phases:
sion by convection and radiation between the surface of the solid
component and fluid gas is calculated using Newton’s equation of ϕk = 1−exp(−bt n ) (21)

497
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

where ϕk is the volume fraction of austenite transformed, b and n can be traditional liquid quenching, the thermal boundary conditions during
determined by the experimental data of isothermal phase transforma- gas quenching are relatively simple to control. By adjusting the gas-
tion, and t is the phase transformation time. K = 2 ferrite/pearlite, quenching furnace pressure, flow speed, or spray nozzle configuration,
k = 3 bainite. The volume fraction of the constituent in continuous heat-transfer coefficients can be designed in terms of both the compo-
cooling can be obtained by mathematical transformation [183]. nent geometry and quenching time [198].
The volume fraction of the constituent in martensite is calculated For deformation, the incremental strain is assumed to consist of
using the following formula [183]: several terms:
2.5 dε = dε e + dε p + dε th + dε tr + dε tp (24)
ϕM = (1−ϕ1−ϕ2) ⎡ ⎛ T −Mf ⎞ ⎤
⎢1− Ms−T
⎜ ⎟
⎥
⎣ ⎝ ⎠ ⎦ (22) where the superscripts e, p, th, tr, and tp represent elastic, plastic,
thermal, phase transformations, and transformation plasticity, respec-
where Ms and Mf are the initial temperatures and the final temperature
tively. Phase transformations and transformation plasticity make the
of martensite transformation, ϕM is the volume fraction of martensite.
quenching process a highly non-linear problem [181]. The formulas for
It also can be modeled by Magee’s equation [193]:
ε e, ε pε th are well presented in literature [199,200].
ϕM = 1−exp(ψ1 T + ψ2) (23) The transformation plasticity is often described as ε tp = K ·σ·f (w ) ,
where σ is the applied stress, K is the transformation plasticity para-
where ϕM is the volume fraction of martensite, T is temperature. ψ1 and meter and f(w) is a formula describing the evaluation of the transfor-
ψ2 are constants, determined by using the martensitic transformation’s mation plasticity depending on the amount w of the formed phase
starting and finishing temperatures. [200]. According to Ferro et al. [199] phase transformation plasticity
During the quenching process of the steel component, both diffu- can be defined as an anomalous plastic strain which can be observed. In
sional and martensitic phase transformations can occur simultaneously the case where the phases are ideally-plastic, an explicit expression for
at different locations of the component. The accuracy of the phase the transformation plastic strain was given in [201]:
transformation models is very important for the quenching simulation
results. Phase transitions can be conveniently modeled with tools using ⎧ 0
if p ≤ 0.03
finite element models, as mentioned above. This principle has been ε tp = 3Δεith σ eq

confirmed by Sugianto et al. [3], who used the available tool DEFORM
⎨−
⎩ σiy
→j
Sh ( ) (lnp) ṗif
σy
p > 0.03
(25)
2D and 3D to study the prediction phase transition kinetics during
where p is the volume proportion of the phase j (formed by the trans-
heating, carburizing, diffusing and quenching processes.
formation), σ eq is the macroscopic Von Mises stress, σ y is the macro-
Cheng et al. [194] claimed that heat conduction during quenching is
scopic yield stress evaluated either by a simple linear mixture rule, or
a nonlinear problem and that the calculation of temperature is closely
by more accurate nonlinear mixture role [202], S is the macroscopic
related to the surface heat-transfer coefficient and phase transforma- stress deviator, Δεith→ j is the difference of thermal strain between the two
tion. The authors indicated the advantages of the finite element method y
phases (ε th th
j −εi ), σi is the yield stress of phase i and the function
for mathematical modeling of martensite in a continuous cooling and
h (σ eq/ σ y ) is defined [199] as follows:
calculating temperature field in part during gas quenching at high
pressure. σ eq 1
σ eq ⎧ 1 if ≤
σy 2
Phase transitions are inherently linked to steel properties changes h⎛ y ⎞ = σ eq 1
after quenching. Accurate prediction of the mechanical properties in ⎝σ ⎠ ⎨ ⎩
1 + 3.5 σ y − 2 if ( ) σ eq
σy
>
1
2 (26)
quenched steel parts has been considered by many researchers
[172,173,184]. Currently, a new trend is observed. It is using data The setting of p = 3% accounts for the fact that transformation
mining methods (especial neural network models) to predict the plasticity analytical formulation (due to the assumption that all the
properties of a quenched element. Kianezhad et al. [195] have studied austenitic phase undergoes plasticity from the beginning of ideal plas-
ANN approaches to predict the hardness of quenched steel parts in ticity) overestimates the phenomenon for very low values of p where
several different quench media. They indicated that ANNs are a pro- the strain is essentially elastic [201]; while function (26) accounts for
minent method for accurately predicting the hardness of steel parts and the nonlinearity with respect to the stress applied, which becomes
can be used to predict accurate hardness for a whole range of hard- important only when the stress is of the order of σ y [201]. The phase
nesses and different phases formed in quenched steel parts. transformations strain was given by Mittemeijer and Somers [203] as:
Podgornik et al. [196] claimed that neural modeling could be suc- ε tr = ∑i yϕi εϕtri (27)
cessfully used for detailed and parallel analyses of different influential
parameters as well as to optimize the vacuum heat treatment process of where yϕi is the variation with time of the volume fraction of phase ϕi
hot work tool steels depending on their composition. Their feedforward and εϕtri is phase transformation strain rate due to the precipitation/
neural network with an error backpropagation training scheme as well dissolution of phase ϕi.
as the four layers of neurons scheme was based on experimentally ob- Ju et al. [9] report that the total strain rate ε after heat treatment,
tained tempering diagrams of ten different hot work tool steel compo- including carburizing or nitriding, is assumed to be the sum of the rates
sitions. To our knowledge, the network architecture seems to be too of the elastic strain εe, plastic strain εp, thermal strain εth, transforma-
complex because according to Cybenko, one hidden layer is sufficient to tion strain εtr and diffusion-induced strain εdf of carbon and nitrogen as
transform each continuous function [197]. well. The authors also provide the mathematical relationships that can
be used to calculate the component.
5.3. Stress and deformation analysis Most researchers have reported that in their studies they use the
finite elements method or the response surface method to determine the
According to Cheng et al. [183], to obtain perfect thermal stresses, distortion, residual stress, average surface hardness and standard de-
strains during gas quenching and residual stresses after gas quenching viation of surface hardness [17,170,181,204–208]. Finite element
in the determination of surface heat-transfer coefficients, the coupled analyses for predicting residual stress and minimizing distortion during
effect between temperatures and phase transformations must be taken heat treatments of steel have been reported in the literature
into account. In consequence, the distortion that results from the [43,207,209,210]. Several commercial simulators, such as HEARTS
quenching process is predominantly due to the thermal gradient and [211], SYSWELD [212], DEFORM-HT [194], DANTE [213,214], and
phase transformations within the component. Compared with COSMAP [3], have been developed and have been used by industries

498
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

that manufacture steel parts, cars, mechanical equipment, and machine the reaction kinetics and mechanism of ammonia decomposition, but
tools [215]. there are still very few predictive modeling attempts that have been
reported to explain the decomposition chemistry [227,228]. Alhumaizi
6. Multi-phase diffusion in low-pressure nitriding et al. [229] mathematically analyzed the one-dimensional re-
action–diffusion–convection system using standard reduction techni-
Nitriding is a thermochemical treatment of diffusion saturation of ques, such as finite difference, finite element, and orthogonal colloca-
the surface layer of steel with nitrogen in heating in an appropriate tion, based on the method of lines. The numerical solution of the
medium (Fig. 6). Lakhtin's and others' works [216–218] have proven governing equations can be stiff in the temporal direction due to dif-
that steel has a high surface hardness, a low tendency toward formation fusive terms and steep in the spatial direction due to convective terms.
of scratches, high wear resistance, endurance limit, and cavitation re- The authors analyzed both the steady-state and transient behavior of a
sistance, and good corrosion resistance in the atmosphere, fresh water, homogeneous tubular reactor with an autocatalytic reaction and con-
and steam as the result of nitriding. During nitriding, the treated parts cluded that methods including the nonoscillatory method and total
experience small deformations, which is an advantage with important variation diminishing method, which are special versions of the high-
meaning. In nitriding, the alloy elements Cr, Mo, W, V, Ti, and so on the resolution finite difference technique, enhance accuracy and eliminate
increase the solubility of nitrogen in the α-phase, dissolve in the ε- and oscillations, but at the expense of increased computational costs. In
γ′-phase and form special nitrides as described by Yang and Sisson 2005, a reaction kinetics model for ammonia decomposition was pro-
[219]. posed by Waghode et al. [227]. Their flow model (flow equations) was
The composition of the diffusion layer may be controlled in low- based on finite element techniques. The Computational Fluid Dynamics
pressure nitriding (LPN) [220–222]. Unfortunately, the processes are (CFD) simulations were used to study the effect of the flow direction on
characterized by a large number of interrelated physicochemical the ammonia decomposition rates in a nitriding process by Ozdemir and
parameters, the measurement, and control of which are practically Akar [230].
impossible during the production operation. The complexity of the low- The most important parameters that affect the kinetics of the growth
pressure nitriding process causes difficulty in the a priori determination of the nitrided layer and its phase composition in the gas nitriding
process include the nitrogen potential, temperature, composition and
of the production conditions necessary for obtaining certain service
properties of parts [28,218]. The main phases of the treatment are ε, γ′ flow rate of the nitriding atmosphere; composition of the material being
nitrided; and preparation of the surface that is to be nitrided. As a result
and α, so diffusion must be modeled for all the phases with singular
conditions and also for the interfaces between the different phases. of nitrogen mass transfer from the atmosphere to the steel surface, a
nitrogen concentration gradient is built up and represents the driving
force for the diffusion of nitrogen in α-Fe. The equation for N diffusion
6.1. Nitrogen transport from the atmosphere to the steel surface for gas nitriding is presented by Cavaliere et al. [143]. In low-pressure
nitriding, the nitrogen potential no longer matters. However, another
Mostly, nitrogen for the nitriding process is produced during the parameter may appear. Kula et al. [28] described the efficacious pro-
ammonia decomposition process. Decomposition of ammonia is a gas cess of low-pressure nitriding. They divided time of the process into
phase reaction, and the rate expression can usually be expressed in alternate stages of material saturation and annealing at the nitriding
terms of the partial pressures of reacting components. Many attempts temperature without the nitriding gas.
can be found in the literature devoted to the chemical kinetics of the
ammonia decomposition reaction [216,223–225], and mostly the rate
of reaction is considered to be of the first order concerning the am-
monia partial pressure. Yin et al. [226] presented a concise review of

Fig. 6. Microstructure of X38CrMoV5-1 steel (H11, 1.2343) after low-pressure nitriding (temperaturę 540 °C, time 4 h).

499
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

6.2. Nitrogen diffusion in austenite 7. Precipitations phenomena

A general multi-phase diffusion model solution using numerical The alloy elements Cr, W, Mo, V, Ti, Nb and Fe produce carbide
techniques was proposed by Engstrom et al. [231] and Andersson et al. precipitation with simple and/or complex structures during both car-
[232]. However, there is a lack of comprehensive information on the burizing and nitriding process [236]. When being heated for thermal
kinetics of the low-pressure nitriding process in the literature. It is a treatment, carbides dissolve in austenite, changing the hardening ca-
non-steady state process, and therefore, solving Fick’s equations of pacity and kinetics of this phase transformation. Cementite also de-
diffusion–which describe the transport of nitrogen atoms necessary to composes and becomes a source of carbon atoms [237]. In further
create the beneficial structure of the surface layer–by analytical or stages of the treatment process, carbide-forming elements from dis-
numerical methods is a considerable challenge due to the non-linear solved carbides remain in the material. They can enter martensite
relationship between the diffusion coefficient and rate of diffusion. The (perlite, bainite) during the cooling phase or earlier – upon overcooling
problem is additionally complicated by a large variety of steel and cast of austenite – and can form carbides (Fig. 8). Utility consequences of the
iron grades (different chemical alloying) that strongly influence the occurrence of carbides in iron alloys can be both positive and negative.
diffusion rate and phase transformations [29,219]. The system of Fine-grained spherical carbides that precipitate from inside grains
equations that describes the relationships can only be solved when a strengthen the material. On the other hand, the appearance of the ce-
simplified approach to the phenomena that accompany the nitriding mentite lattice entails a range of adverse effects: local decarburizations
process is followed, which results in unacceptable discrepancies be- appear in the material [238]; the percentage of alloy elements increases
tween the calculations and observations. The problem can only be [161]; where carbides occur, the material becomes harder and more
solved by an optimization procedure that takes into account several brittle [239,240] and tends to crack [239,241]; and the material be-
variable parameters, assuming that the others are constant. The best comes more susceptible to corrosion [238].
solution in this case, as practice and theory indicate, is computer-aided Depending on the chemical composition, carbides crystallize in a
design based on neural networks. Assumptions about the neural net- regular, hexagonal or rhomboid lattice [242,243]. Carbide clusters
work model and its applications for the low-pressure nitriding of tools often contain carbides of different types and their chemical composition
have been presented in several papers by Zhecheva et al. [64] and changes during thermal treatment. Moreover, each of these chemical
Wolowiec et al. [29,35] (Fig. 7). compounds is characterized by their own kinetic equations [244–247].
One of the appealing attributes of the nitriding process is that the It is not possible to develop an analytical model for such complex
deformations in parts can be kept to a minimum [233]; however, it is phenomena. In such cases, only statistical (analysis of variance, mul-
also an object of research investigations [234]. Zoteva and Kotsyu- tiple regression) and heuristic methods are applied, as well as neural
binskii [235] proposed an analytical-based method that takes into ac- networks [67,248,249].
count the elastic-viscous-plastic interaction of the nitrided case and the
core.

Fig. 7. Exemplary scheme of a system for modeling the low-pressure nitriding process [35].

500
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

Fig. 8. Process of carbide clusters formation on the surface and on particle borders (boost/diffusion stage); A. Temp.: 1273 K, segments: 4/0 min., B. Temp.: 1273 K,
segments: 6/0 min., C. Temp.: 1273 K, segments: 12/0 min., D. Temp.: 1273 K, segments: 18/0 min. [248].

8. Summary and future outlook treatment will emerge, with unelucidated and unexplored phenomena
(low-pressure thermal treatment of sinters, low-pressure thermal
Computer-aided modeling of phenomena and material features is treatment of 3D-printed elements, etc.). It seems that methods of arti-
becoming a common tool, and the contemporary progress of material ficial intelligence will be used first, both with regard to predictions
engineering is closely associated with the application and development (artificial neural networks) and optimization (genetic and evolutionary
of computation methods. Computer-aided reproduction of the processes algorithms). However, teaching databases will increasingly help to
and phenomena helps to predict the properties of materials with a improve the precision of their predictions.
considerable reduction of the financial outlays and the time needed for
experiments and implementation. The near future will see an increasing Acknowledgements
demand for effective, precise and easy-to-handle computation tools for
predicting and designing thermal and thermochemical low-pressure This research did not receive any specific grant from funding
treatments. Both conventional and numerical models, as well as those agencies in the public, commercial, or not-for-profit sectors.
of the data mining type, are already widely applied in modeling such
processes. Disclosure statement
Conventional models based on analytical equations undoubtedly
yield the most accurate results. However, the constant growth of the The author declares that she has no conflict of interest.
computation capacity of computers favors the development of numer-
ical methods. It is expected that results will be obtained by numerical Appendix A. Supplementary material
methods more quickly and more accurately. This process is favored by
the development of knowledge of low-pressure treatments, which en- Supplementary data to this article can be found online at https://
ables the establishment of the coefficients describing the process in doi.org/10.1016/j.engstruct.2018.10.003.
individual devices with greater precision.
An increase in the body of knowledge will be accompanied by an References
increase in the amount of data regarding LPC, LPN and HPGQ pro-
cesses, and the need for effective analysis, faster searches and acquiring [1] Mittemeijer EJ. Fundamentals of materials science. The microstructure – property
of new knowledge from data are going to lead to an increase in the relationship using metals as model systems. Berlin: Springer; 2010.
[2] Mukai R, Matsumoto T, Ju D, Suzuki T, Saito H, Ito Y. Modeling of numerical
applications of DM methods. simulation and experimental verification for carburizing-nitriding quenching
Due to their nature, some phenomena are still going to be too process. Trans Nonferrous Met Soc China 2006;16:566–71.
complicated, or their modeling by conventional and numerical methods [3] Sugianto A, Narazaki M, Kogawara M, Shirayori A, Kim S, Kubota S. Numerical
simulation and experimental verification of carburizing-quenching process of
will be time-consuming. Moreover, new applications of thermal

501
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

SCr420H steel helical gear. J Mater Process Technol 2009;209:3597–609. https:// [40] Kim D, Cho H, Lee W, Cho K, Cho Y, Kim S, et al. A finite element simulation for
doi.org/10.1016/j.jmatprotec.2008.08.017. carburizing heat treatment of automotive gear ring incorporating transformation
[4] Zhang L. Modeling and verification of simulation tools for carburizing and car- plasticity. Mater Des 2016;99:243–53. https://doi.org/10.1016/j.matdes.2016.03.
bonitriding PhD Thesis Worcester Polytechnic Institute; 2016 047.
[5] Tsuji S, Ishigami I, Yamanaka K. Vacuum carburizing of low carbon steel with [41] Liu CC, Xu XJ, Liu Z. A FEM modeling of quenching and tempering and its ap-
methane. Trans Jpn Inst Met 1987;28:48–56. plication in industrial engineering. Finite Elem Anal Des 2003;39:1053–70.
[6] Totten GE, editor. Steel heat treatment: equipment and process design. Boca Raton: [42] Li Z, Freborg AM, Hansen BD. Modeling the effect of carburization and quenching
Taylor & Francis; 2007. and quenching on the development of residual stresses and bending fatigue re-
[7] D.H. Herring, Vacuum heat treatment. Principles | Practices | Applications, BNP sistance of steel gears. J Mater Eng Perform 2013;22:664–72.
Media II, Troy; 2012. [43] Lee S-J, Matlock DK, van Tyne CJ. Comparison of two finite element simulation
[8] Gawroński Z, Malasiński A, Sawicki J. Elimination of galvanic copper plating codes used to model the carburizing of steel. Comput Mater Sci 2013;68:47–54.
process used in hardening of conventionally carburized gear wheels. Int J Automot https://doi.org/10.1016/j.commatsci.2012.10.007.
Technol 2010;11:127–31. [44] Inoue T, Arimoto K. Development and implementation of CAE system HEARTS for
[9] Sawicki J, Górecki M, Kaczmarek Ł, Gawroński Z, Dybowski K, Pietrasik R, et al. heat treatment simulation based on metallo-thermomechanics. J Mater Eng
Increasing the durability of pressure dies by modern surface treatment methods. Perform 1997;6:51–60.
Chiang Mai J Sci 2013;40:886–97. [45] Yaakoubi M, Kchaou M, Dammak F. Simulation of heat treatment and materials
[10] Stachurski W, Zgórniak P, Sawicki J, Przybysz M. Hardening-related deformations with the use of the Abaqus software. Met Sci Heat Treat 2013;55:386–92. https://
of gear wheels after vacuum carburising and quenching in gas. Adv Sci Technol – doi.org/10.1007/s11041-013-9641-5.
Res J 2017;11:237. [46] Pont D, Guichard T. Sysweld: welding and heat treatment modelling tools. In:
[11] Ju D, Liu C, Inoue T. Numerical modeling and simulation of carburized and ni- Atluri SN, Yagawa G, Cruse T, editors. Comput. Mech. ’95. Berlin, Heidelberg:
trided quenching process. J Mater Process Technol 2003;143–144:880–5. https:// Springer Berlin Heidelberg; 1995. p. 248–53https://doi.org/10.1007/978-3-642-
doi.org/10.1016/S0924-0136(03)00378-9. 79654-8_41.
[12] Kula P, Korecki M, Pietrasik R, Wołowiec E, Dybowski K, Kołodziejczyk Ł, et al. [47] “Virtual Heat Treating” using DEFORM-HT, Aircr. Eng. Aerosp. Technol. 70; 1998.
FineCarb – the flexible system for low-pressure carburizing. New options and https://doi.org/10.1108/aeat.1998.12770dab.015.
performance. Jpn Soc Heat Treat 2009;49:133–6. [48] Berry MJ, Linoff G. Data mining techniques: for marketing, sales and customer
[13] Dowling WE, Pattok T, Ferguson BL. Development of a carburizing and quenching support. Hoboken: John Willey & Sons; 1997.
simulation tool: program overview. HTM J Heat Treat Mater 1997;1:1–6. [49] Hastie T, Tibshirani R, Friedman J. The elements of statistical learning.
[14] Pan J, Li Y, Li L. The application of computer simulation in the heat-treatment Heidelberg: Springer-Verlag; 2002.
process of a large-scale bearing roller. J Mater Process Technol 2002;122:241–8. [50] Statistica. Data Miner, StatSoft, Cracow; 2002.
[15] Song RG, Zhang QZ. Heat treatment optimization for 7175 aluminum alloy by [51] Gianola D, Okut H, Weigel KA, Rosa GJ. Predicting complex quantitative traits
genetic algorithm. Mater Sci Eng C 2001;17:133–7. with bayesian neural networks: a case study with Jersey cows and wheat. BMC
[16] Song RG, Zhang QZ. Heat treatment technique optimization for 7175 aluminium Genet 2011;12:87. https://doi.org/10.1186/1471-2156-12-87.
alloy by an artificial network and a genetic algorithm. J Mater Process Technol [52] Okut H, Wu X-L, Rosa GJ, Bauck S, Woodward BW, Schnabel RD, et al. Predicting
2001;117:84–8. expected progeny difference for marbling score in Angus cattle using artificial
[17] Huiping L, Guoqun Z, Shanting N, Yiguo L. Optimization of technology parameters neural networks and Bayesian regression models. Genet Sel Evol GSE 2013;45:34.
for the plane-strain component in the process of gas quenching. Appl Math Model https://doi.org/10.1186/1297-9686-45-34.
2008;32:860–72. https://doi.org/10.1016/j.apm.2007.02.018. [53] Aliofkhazraei M, Rouhaghdam AS. Neural networks prediction of different fre-
[18] Jacquet P, Rousse DR, Bernard G, Lambertin M. A novel technique to motor car- quencies effects on corrosion resistance obtained from pulsed nanocrystalline
burizing process. Mater Chem Phys 2002;77:542–51. plasma electrolytic carburizing. Mater Lett 2008;62:2192–5. https://doi.org/10.
[19] Kula P, Atraszkiewicz R, Wołowiec E. Modern gas quenching chambers supported 1016/j.matlet.2007.11.052.
by SimVaC Plus Hardness application. Ind Heat 2008;3:55–8. [54] Dobrzański LA, Sitek W. Comparision of hardenability calculation methods of the
[20] Kim D, Cho Y, Kim S, Lee W, Lee M, Han H. A numerical model for vacuum car- heat-treatable constructional steels. J Mater Process Technol 1997;64:117–26.
burization of an automotive gear ring. Met Mater Int 2011;17:885–90. [55] Dobrzański LA, Sitek W. Application of a neural network in modelling of hard-
[21] Collin R, Gunnarson S, Thulin D. A mathematical model for predicting carbon enability of constructional steels. J Mater Process Technol 1998;78:59–66.
concentration profiles of gas-carburized steel. J Iron Steel Inst 1972;10:785–9. [56] Genel K. Use of artificial neural network for prediction of ion nitrided case depth in
[22] Loeser K, Gornicki B. Low pressure carburizing of large transmission parts. Gear Fe–Cr alloys. Mater Des 2003;24:203–7. https://doi.org/10.1016/S0261-3069(03)
Technol 2009;9–10:67–70. 00002-5.
[23] Otto FJ, Herring DH. Vacuum carburizing of aerospace and automotive materials. [57] Genel K, Ozbek I, Kurt A, Bindal C. Boriding response of AISI W1 steel and use of
Heat Treat Prog 2005;1–2:33–7. artificial neural network for prediction of borided layer properties. Surf Coat
[24] Pawelski Z, Pałczyński T. Influence of steering course of CVT to diesel fuel con- Technol 2002;160:38–43.
sumption an exhaust emission at ECE cycle. J KONES 2007;14:335–42. [58] Malinov S, Sha W. Application of artificial neural networks for modelling corre-
[25] Pawelski Z, Pałczyński T. The properties of powertrain with diesel engine and CVT lations in titanium alloys. Mater Sci Eng A 2004;365:202–11. https://doi.org/10.
transmission steering. J KONES 2008;15:433–41. 1016/j.msea.2003.09.029.
[26] Pałczyński T. Simplified model of city bus dynamics as a tool of an energy con- [59] Malinov S, Sha W, Guo Z. Application of artificial neural network for prediction of
sumption estimation. In: Awrajcewicz J, editor. Dyn Syst. Model. Cham: Springer; time–temperature–transformation diagrams in titanium alloys. Mater Sci Eng A
2016. 2000;283:1–10.
[27] Loeser K, Seemann R. High pressure gas quenching and the influence of material [60] Reddy NS, Krishnaiah J, Hong S, Lee J. Modeling medium carbon steels by using
hardenability of case hardening steels in the gear industry; 2004. artificial neural networks. Mater Sci Eng A 2009;508:93–105. https://doi.org/10.
[28] Kula P, Wolowiec E, Pietrasik R, Dybowski K, Januszewicz B. Non-steady state 1016/j.msea.2008.12.022.
approach to the vacuum nitriding for tools. Vacuum 2013;88:1–7. https://doi.org/ [61] Sha W, Edwards KL. The use of artificial neural networks in materials science
10.1016/j.vacuum.2012.08.001. based research. Mater Des 2007;28:1747–52. https://doi.org/10.1016/j.matdes.
[29] Wolowiec E, Kula P, Januszewicz B, Korecki M. Mathematical modelling the low- 2007.02.009.
pressure nitriding process. Appl Mech Mater 2013;421:377–83. [62] Sitek W, Dobrzański LA, Zacłona J. The modelling of high-speed steels’ properties
[30] Jung M, Oh S, Lee Y. Predictive model for the carbon concentration profile of using neural networks. J Mater Process Technol 2004;157–158:245–9. https://doi.
vacuum carburized steels with acetylene. Met Mater Int 2009;15:971–5. https:// org/10.1016/j.jmatprotec.2004.09.037.
doi.org/10.1007/s12540-009-0971-1. [63] Xu L, Xing J, Wei S, Zhang Y, Long R. Optimization of heat treatment technique of
[31] Alquicer R. Controls software in atmosphere equipment. Gear Solut 2016;1:26. high-vanadium high-speed steel based on back-propagation neural networks.
[32] Jaster M. Why vacuum carburizing? Gear Solut 2010;3–4:31–5. Mater Des 2007;28:1425–32. https://doi.org/10.1016/j.matdes.2006.03.022.
[33] Wei Y, Zhang L, Sisson RD. Modeling of carbon concentration profile development [64] Zhecheva A, Malinov S, Sha W. Simulation of microhardness profiles of titanium
during both atmosphere and low pressure carburizing processes. J Mater Eng alloys after surface nitriding using artificial neural network. Surf Coat Technol
Perform 2013;22:1886–91. 2005;200:2332–42. https://doi.org/10.1016/j.surfcoat.2004.10.018.
[34] Karabelchtchikova O, Wang G, Sisson RD. New carburizing calculation tool for gas [65] Dobrzański LA, Trzaska J. Application of neural networks for the prediction of
and low-pressure carburizing. Heat Treat Prog 2008;3–4:18. continuous cooling transformation diagrams. Comput Mater Sci 2004;30:251–9.
[35] Kula P, Pietrasik R, Wołowiec E, Januszewicz B, Rzepkowski A. Low-pressure ni- https://doi.org/10.1016/j.commatsci.2004.02.011.
triding according to the FineLPN technology in multi-purpose vacuum furnaces. [66] Liujie X, Jiandong X, Shizhong W, Tao P, Yongzhen Z, Rui L. Artificial neural
Adv Mater Res 2012;586:230–4. https://doi.org/10.4028/www.scientific.net/ network prediction of heat-treatment hardness and abrasive wear resistance of
AMR.586.230. High-Vanadium High-Speed Steel (HVHSS). J Mater Sci 2007;42:2565–73. https://
[36] Lakhtin YM, Kogan YD, Bulgach AA. Prospects for computer usage in heat treat- doi.org/10.1007/s10853-006-1278-y.
ment and casehardening. Met Sci Heat Treat 1984;26:3–9. [67] Wolowiec E, Kula P. The application of artificial neural networks in optimization
[37] Weiss SM, Indurkhya N. Predictive data mining. A practical guide. Burlington: of heat treatment processes of steel. J Appl Comput Sci 2011;19:161–9.
Morgan Kaufmann Publisher; 1998. [68] Hagan M, Demuth H, Beale M. Neural networks design. Boston: PWS Publishing
[38] Braha D. Data mining for design and manufacturing. Methods and applications. Company; 1996.
Heidelberg: Kluwer Academic Publisher; 2001. [69] Hornik K, Stinchcombe M, White H. Multi-layer feed forward networks are uni-
[39] Malinov S, Sha W. Software products for modelling and simulation in materials versal approximations. Neural Netw 1989;2:359–66.
science. Comput Mater Sci 2003;28:179–98. https://doi.org/10.1016/S0927- [70] Rumelhart D, Hinton G, Wiliams R. Learning internal representations by error
0256(03)00106-X. propagation. Cambridge: MIT Press; 1986.

502
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

[71] Haykin S. Neural networks: a comprehensive foundation. New York: Prentice Hall; [115] Rzepkowski A, Rzepkowski A, Kula P. Vacuum carburized layers examinations of
1994. ISO 2639 standard. Acta Metall Slovaca 2004;1:730–3.
[72] Fausett L. Fundamentals of neural networks. New York: Prentice Hall; 1994. [116] Azis SA, Jauhari I, Ahamad NW. Improving surface properties and wear behaviors
[73] Patterson D. Artificial neural networks. Singapore: Prentice Hall; 1996. of duplex stainless steel via pressure carburizing. Surf Coat Technol
[74] Ma J. The object perceptron learning algorithm on generalised Hopfield networks 2012;210:142–50.
for associative memory. Neural Comput Appl 1999;8:25–32. [117] Korecki M, Kula P, Wolowiec E, Bazel M, Sut M. Low pressure carburizing and
[75] Oja E. Unsupervised learning in neural computation. Theor Comput Sci nitriding of fuel injection nozzles. Heat Process 2014;3:59–62.
2002;287:187–207. [118] Carey S, Herring DH. Low-pressure carburizing process development of M50 NiL.
[76] Aitkin M, Foxall R. Statistical modelling of artificial neural networks using the Heat Treat Prog 2007;5–6:43–6.
multi-layer perceptron. Stat Comput 2003;13:227–39. [119] Wei S, Wang G, Zhao X, Zhang X, Rong Y. Experimental study on vacuum car-
[77] Daqi G, Yan J. Classification methodologies of multilayer perceptrons with sig- burizing process for low-carbon alloy steel. J Mater Eng Perform 2014;23:545–50.
moid activation functions. Pattern Recognit 2005;38:1469–82. https://doi.org/10. https://doi.org/10.1007/s11665-013-0762-1.
1016/j.patcog.2005.03.024. [120] Semenov MY. Control of heat-resistant steel carburized layer structure. Part I. Met
[78] Moody J, Darken C. Fast learning in networks of locally-tuned processing units. Sci Heat Treat 2013;55:257–64.
Neural Comput 1989;1:281–94. [121] Kula P, Dybowski K, Wolowiec E, Pietrasik R. Boost-diffusion vacuum carburizing
[79] Lehtokangas M, Saarinen J. Centroid based multilayer perceptrons networks. – process optimisation. Vacuum 2014;99:175–9.
Neural Process Lett 1998;7:101–6. [122] Osterman VM, Jones T. Carburizing for our troops. Ind Heat 2006;5:53–5.
[80] Agresti A. An introduction to categorical data analysis. Hoboken: John Willey & [123] Karabelchtchikova O, Sisson RD. Calculation of gas carburizing kinetics from
Sons; 1996. carbon concentration profiles based on direct flux integration. Defect Diffus
[81] McCullagh P, Nelder JA. Generalized linear model. London: Chapman & Hall; 2007;266:171–80.
1999. [124] Zajusz M, Tkacz-Śmiech K, Dychtoń K, Danielewski M. Pulse carburization of steel
[82] Ratner B. Statistical modeling and analysis for database marketing. London: – model of the process. Defect Diffus Forum 2014;354:145–52. https://doi.org/10.
Chapman & Hall; 2003. 4028/www.scientific.net/DDF.354.145.
[83] Han J, Kamber M. Data mining: concepts and techniques. Waltham: Academic [125] Grafen W, Edenhofer B. Acetylene low-pressure carburizing – a novel and superior
Press; 2001. carburizing technology. HTM J Heat Treat Mater 1999;26:79–83.
[84] Friedman J. Multivariate adaptive regression splines. Ann Stat 1991;19:1–67. [126] Khan RU, Bajohr S, Buchholz D, Reimert R, Minh HD, Norinaga K, et al. Pyrolysis
[85] Pyle D. Data preparation for data mining. Waltham: Academic Press; 1999. of propane under vacuum carburizing conditions: an experimental and modeling
[86] Berry MJ, Linoff G. Mastering data mining. Hoboken: John Willey & Sons; 2000. study. J Anal Appl Pyrolysis 2008;81:148–56. https://doi.org/10.1016/j.jaap.
[87] Migut G, Demski T. Technical information 2012. 2007.09.012.
[88] Goldberg D. Genetic algorithms. Boston: Addison-Wesley Reading; 1989. [127] Matsugi A, Miyoshi A. Modeling of two- and three-ring aromatics formation in the
[89] Velsker T, Eerme M, Majak J. Artificial neural networks and evolutionary algo- pyrolysis of toluene. Proc Combust Inst 2013;34:269–77. https://doi.org/10.
rithms in engineering design. J Achiev Mater Manuf Eng 2011;44:88–95. 1016/j.proci.2012.06.032.
[90] Sitek W, Dobrzański LA. Application of genetic methods in materials’ design. J [128] Slavinskaya NA, Riedel U, Dworkin SB, Thomson MJ. Detailed numerical modeling
Mater Process Technol 2005;164–165:1607–11. https://doi.org/10.1016/j. of PAH formation and growth in non-premixed ethylene and ethane flames.
jmatprotec.2005.01.005. Combust Flame 2012;159:979–95. https://doi.org/10.1016/j.combustflame.2011.
[91] van Wie M. Heuristics, science, and engineering. Mech Eng – CIME 2012;134:8. 10.005.
[92] Kopecek P. Selected heuristic methods used in industrial engineering. Proc Eng [129] Bensabath T, Monnier H, Glaude P-A. Detailed kinetic modeling of the formation
2014;69:622–9. of toxic polycyclic aromatic hydrocarbons (PAHs) coming from pyrolysis in low-
[93] Mistakidis ES, Stavroulakis GE. Nonconvex optimization in mechanics: algorithms, pressure gas carburizing conditions. J Anal Appl Pyrolysis 2016;122:342–54.
heuristics and engineering applications by the F.E.M. Heidelberg: Kluwer https://doi.org/10.1016/j.jaap.2016.09.007.
Academic Publisher; 1998. [130] Buchholz D, Khan RU, Bajohr S, Reimert R. Computational fluid dynamics mod-
[94] Rokicki P, Dychton K. Acetylene flow rate as a crucial parameter of vacuum car- eling of acetylene pyrolysis for vacuum carburizing of steel. Ind Eng Chem Res
burizing process of modern tool steels. Arch Metall Mater 2016;61:2009–12. 2010;49:1130–7. https://doi.org/10.1021/ie900996h.
https://doi.org/10.1515/amm-2016-0324. [131] Saggese C, Sanchez NE, Frassoldati A, Cuoci A, Faravelli T, Alzueta MU, et al.
[95] Gronquist E, Scotto V, Taft D, Westeren H. Method of vacuum carburizing, US 3, Kinetic modeling study of polycyclic aromatic hydrocarbons and soot formation in
796,615; 1974. acetylene pyrolysis. Energy Fuels 2014;28:1489–501. https://doi.org/10.1021/
[96] Krylov VS, Yumatov VA, Kurbatov VV. USSR 668978; 1977. ef402048q.
[97] Kubota K. Vacuum carburizing method and device, and carburized products, US 5, [132] Xu C, Al Shoaibi AS, Wang C, Carstensen H, Dean AM. Kinetic modeling of ethane
702,540; 1997. pyrolysis at high conversion. J Phys Chem A 2011;115:10470–90. https://doi.org/
[98] Poor RP, Barbee GW, Verhoff SH, Brugg JE. Furnace for vacuum carburizing with 10.1021/jp206503d.
unsaturated aromatic hydrocarbons, 7,267,793; 2007. [133] Norinaga K, Deutschmann O, Saegusa N, Hayashi J. Analysis of pyrolysis products
[99] Kula P, Olejnik J, Heilman P. Method for under-pressure carburizing of steel from light hydrocarbons and kinetic modeling for growth of polycyclic aromatic
workpieces, EU 1,558,781; 2006. hydrocarbons with detailed chemistry. J Anal Appl Pyrolysis 2009;86:148–60.
[100] Kula P, Olejnik J, Heilman P. Hydrocarbon gas mixture for the under pressure https://doi.org/10.1016/j.jaap.2009.05.001.
carburizing of steel, EU 1,558,780; 2007. [134] Muller C, Michel V, Scacchi G, Come G. THERGAS: a computer program for the
[101] Jones WR. Versatile high velocity integral vacuum furnace, US 7,514,035; 2009. evaluation of thermochemical data of molecules and free radicals in the gas phase.
[102] Kula P, Olejnik J, Heilman P. Method for under-pressure carburizing of steel J Chim Phys 1995;92:1154–78. https://doi.org/10.1051/jcp/1995921154.
workpieces, US 7,550,049; 2009. [135] Yada K, Watanabe O. Reactive flow simulation of vacuum carburizing by acetylene
[103] Kula P, Olejnik J, Heilman P. Hydrocarbon gas mixture for the under-pressure gas. Comput Fluids 2013;79:65–76. https://doi.org/10.1016/j.compfluid.2013.
carburizing of steel, US 7,513,958; 2009. 03.005.
[104] Tsepov SN. Characteristic features of carburizing of steel during vacuum carbur- [136] Gorockiewicz R, Łapinski A. Structure of the carbon layer deposited on the steel
izing. Met Sci Heat Treat 1979;21:633–8. surface after low pressure carburizing. Vacuum 2010;85:429–33.
[105] Kula P, Pietrasik R, Dybowski K. Vacuum carburizing – process optimization. J [137] Sawicki J, Gutkowski A, Kaczmarek Ł, et al. Modeling of thermal phenomena and
Mater Process Technol 2005;164–165:876–81. economic aspect of configuring furnace graphite insulation. Met Sci Heat Treat
[106] Gorockiewicz R. The kinetics of low-pressure carburizing of alloy steels. Vacuum 2015;56:685–9.
2011;86:448–51. [138] Bongartz K, Lupon DF, Schuster H. A model to predict carburization profiles in
[107] Tanaka K, Ikehata H, Takamiya H, Mizuno H. Calculation of microstructure in high temperature alloys. Metall Trans A 1980;11:1883–93.
vacuum carburizing incorporating kinetics modeling of grain-boundary cementite. [139] Bongartz K, Schulten R, Quadakkers WJ, et al. A finite difference model describing
Iron Steel Inst Jpn 2012;52:134–9. carburization in high-temperature alloys. Corrosion 1986;42:390–7.
[108] Ryzhov NM, Smirnov AE, Fakhurtdinov RS. Control of carbon saturation of the [140] Bongartz K, Quadakkers WJ, Schulten R, et al. A mathematical model describing
diffusion layer in vacuum carburizing of heat-resistant steels. Met Sci Heat Treat carburization in multielement alloy system. Metall Trans A 1989;20:1021–8.
2004;46:340–4. [141] Morral JE, Dupen BM, Law CC. Application of commercial computer codes to the
[109] Ochsner A, Gegner J, Mishuris G. Effect of diffusivity as a function of the method modelling carburizing kinetics of alloy steels. Metall Trans A 1992;23:2069–70.
of computation of carbon concentration profiles in steel. Met Sci Heat Treat [142] Engström A, Höglund L, Ågren J. Computer simulation of carburization in mul-
2004;46:148–51. tiphase systems. Mater Sci Forum 1994;163–165:725–30.
[110] Wolowiec E, Kula P, Kolodziejczyk Ł, Dybowski K, Korecki M. Mathematical [143] Cavaliere P, Zavarise G, Perillo M. Modeling of the carburizing and nitriding
modelling of the vacuum carburizing process. Therm Process Gear Solut processes. Comput Mater Sci 2009;46:26–35. https://doi.org/10.1016/j.
2014;3–4:34–40. commatsci.2009.01.024.
[111] Bhadeshia HK. Diffusion of carbon in austenite. Met Sci 1981;10:477–9. [144] Semenov MY, Smirnov AE, Ryzhova MY. Computation of carbon concentration
[112] Chen F, Liu L. Deep-hole carburization in a vacuum furnace by forced convection curves in vacuum carburizing of steels. Met Sci Heat Treat 2013;55:38–42.
gas flow method. Mater Chem Phys. 2003;82:801–7. [145] Semenov MY, Smirnov AE, Fakhurtdinov RS, Ospennikova OG, Gromov VI.
[113] Widanka K. Effect of phosphorus on vacuum carburising depth of iron compacts. Optimization of modes of vacuum carburizing of gears from heat-resistant steel
Arch Civ Mech Eng 2010;1:85–91. VKS-7 on the basis of computational design. Met Sci Heat Treat 2015;57:28–31.
[114] Kula P, Kaczmarek Ł, Dybowski K, Pietrasik R, Krasowski M. Activation of carbon https://doi.org/10.1007/s11041-015-9829-y.
deposit in the process of vacuum carburizing with preliminary nitriding. Vacuum [146] Kołodziejczyk Ł. Mathematical modeling of the process of vacuum carburizing
2013;87:26–9. PhD Thesis Lodz University of Technology; 2003.

503
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

[147] AWT-Fachausschuss 5, Arbeitskreis 4, editors. Die prozessregelung beim [182] Herring DH. A review of gas quenching from the perspective of the heat transfer
Gasaufkohlen und Einsatzharten, Expert Verlag, Renningen-Malmsheim; 1997. coefficient. Ind Heat 2006;2:67–72.
[148] Goldstein JL, Moren AE. Diffusion modeling of the carburization process. Trans [183] Cheng H, Xie J, Li J. Determination of surface heat-transfer coefficients of steel
Metall Soc AIME 1978;9A:1515–25. cylinder with phase transformation during gas quenching with high pressures.
[149] Jimenez H, Staia MH, Puchi ES. Mathematical modeling of a carburizing process of Comput Mater Sci 2004;29:453–8. https://doi.org/10.1016/j.commatsci.2003.11.
a SAE 8620H steel. Surf Coat Technol 1999;120:358–65. 003.
[150] Karabelchtchikova O, Sisson RD. Carbon diffusion in steels: a numerical analysis [184] Pritchard JE, Nurnberg G, Shoukri M. Computer modeling of pressure gas
based on direct integration of the flux. J Phase Equilib Diffus 2006;27:598–604. quenching in vacuum furnaces. Heat Treat Met 1996;4:79–83.
https://doi.org/10.1361/154770306X153611. [185] Morales BH, Brimacombe JK, Hawbolt EB. Determination of quench heat transfer
[151] Crank J. The mathematics of diffusion. New York: Oxford University Press; 1979. coefficients using inverse techniques. Proc. 1st Int. Conf. Quenching Control
[152] Wells A, Batz W, Mehl RF. Diffusion coefficient of carbon in austenite. Trans Distortion 1996:155–64.
Metall Soc AIME 1950;188:553–60. [186] Davis JR, editor. ASM handbook, Vol. 4. Metals Park: American Society for Metals;
[153] Kaufman LR, Radcliffe SV, Cohen M. Decomposition of austenite by diffusional 1991.
processes. In: Interscience Publishers, Chichester; 1962. [187] Guan YH, Chen TL, Wang GH, Zhang JT. The prediction of the mechanical
[154] Smith RP. The diffusivity of carbon in Fe-Co alloys. Trans Metall Soc AIME properties of metal during laser quenching. J Mater Process Technol
1964;230:470–80. 1997;63:614–7. https://doi.org/10.1016/S0924-0136(96)02694-5.
[155] Krishtal MA. Diffusion processes in iron alloys. Israel program for scientific [188] DEFORM Users Manual, Scientific Forming Technologies Corporation, Columbus;
translations. Jerusalem 1963;1970:90–133. 1999.
[156] Leyens G, Woelk G, Wuenning J. Berechnung der Aufkohlung nach dem [189] Johnson WA, Mehl FR. Reaction kinetics in processes of nucleation and growth.
Saettigungs-Ausgleichs-Verfahren. Arch Für Eisenhüttenwes 1976;47:385–90. Trans Metall Soc AIME 1939;135:416–25.
[157] Tibbets GG. Diffusity of carbon in iron and steels at high temperatures. J Appl Phys [190] Price RF, Fletcher AJ. Determination of surface heat-transfer coefficients during
1980;51:4813–6. quenching of steel plates. Met Technol 1980;7:203–11. https://doi.org/10.1179/
[158] Ismail MI, Iskander SS, Saleh EB. Carburizing of steels. Surf Technol 030716980803286739.
1981;12:341–9. https://doi.org/10.1016/0376-4583(81)90028-5. [191] Denis S, Farias D, Simon A. Mathematical-model coupling phase-transformations
[159] Zhang X, Takahashi R, Akiyama T, Yagi J. Carburization rate into solid iron at CO- and temperature evolutions in steels. ISIJ Int 1992;32:316–25.
CO2 atmosphere. Tetsu-Hagane 1997;83:299–304. https://doi.org/10.2355/ [192] Hou L, Cheng H, Li J, Li Z, Shao B, Hou J. Phase composition of T10 steel during
tetsutohagane1955.83.5_299. gas guenching. Proc Eng 2012;31:682–7. https://doi.org/10.1016/j.proeng.2012.
[160] Murakami T, Fukuyama H, Nagata K. Mechanisms of carburization and melting of 01.1086.
iron by CO gas. ISIJ Int 2001;41:416–21. [193] Magee CL. Phase transformations. In: Metals Park; 1970. p. 115–154.
[161] Dybowski K. The computing of effective carbon diffusion coefficient in steels to a [194] Cheng H, Wang H, Xie J. Calculation of coupled problem between temperature and
process of vacuum carburizing control PhD Thesis Lodz University of Technology; phase transformation during gas quenching in high pressure. Appl Math Mech
2005. 2006;27:305–11.
[162] Lee SJ, Matlock DK, van Tyne CJ. An empirical model for carbon diffusion in [195] Kianezhad M, Sajjadi SA, Vafaeenezhad H. A numerical approach to the prediction
austenite incorporating alloying element effect. ISIJ Int 2011;51:1903–11. of hardness at different points of a heat-treated steel. J Mater Eng Perform
[163] Dybowski K, Sawicki J, Kula P, Januszewicz B, Atraszkiewicz R, Lipa S. The effect 2015;24:1516–21. https://doi.org/10.1007/s11665-015-1433-1.
of the quenching method on the deformations size of gear wheels after vacuum [196] Podgornik B, Belič I, Leskovšek V, Godec M. Tool steel heat treatment optimization
carburizing. Arch Metall Mater 2016;61:1057–62. https://doi.org/10.1515/amm- using neural network modeling. Metall Mater Trans A 2016;47:5650–9. https://
2016-0178. doi.org/10.1007/s11661-016-3723-0.
[164] Li Z, Grandhi RV, Shivpuri R. Optimum design of the heat-transfer coefficient [197] Cybenko G. Continuous valued neural networks with two hidden layers are suf-
during gas quenching using the response surface method. Int J Mach Tools Manuf ficient. Medford: Tufts University; 1988.
2002;42:549–58. [198] Li Z, Grandhi RV. Multidisciplinary optimization of gas-quenching process. J
[165] Watt DF, Coon L, Bibby M, Goldak J, Henwood C. An algorithm for modeling of Mater Eng Perform 2005;14:136–43. https://doi.org/10.1361/10599490522257.
micro structural development in weld heat affected zones (Part A) reaction ki- [199] Ferro P, Bobollo F, Tiziani A. The influence of phase transformations on residual
netics. Acta Metall 1988;36:3029–35. stress induced by welding process – 3D and 2D numerical models. Model Simul
[166] Wallis RA, Bhathena NM, Bhowal PR, Raymond EL. Application of process mod- Mater Sci Eng 2006;14:117–36.
eling to heat treatment of superalloys. Ind Heat 1988;55:525–34. [200] Dalgic M, Löwisch G. Transformation plasticity at different phase transformations
[167] Wallis RA, Bhowal PR, Bhathena NM, Raymond EL. Modeling the heat treatment of bearing steel. Mater Werkst 2006;37:122–7. https://doi.org/10.1002/mawe.
of superalloy forgings. JOM 1989;41:35–7. https://doi.org/10.1007/BF03220824. 200500971.
[168] Wang J, Gu J, Shan X, Hao X, Chen N, Zhang W. Numerical simulation of high [201] Leblond JB, Devaux J, Devaux JC. Mathematical modelling of transformation
pressure gas quenching of H13 steel. J Mater Process Technol 2008;202:188–94. plasticity in steels I: case of ideal-plastic phases. Int J Plast 1989;5:551–72.
https://doi.org/10.1016/j.jmatprotec.2007.08.059. [202] Leblond JB, Mottet G, Devaux JC. A theoretical and numerical approach to the
[169] Kopun R, Škerget L, Hriberšek M, Zhang D, Stauder B, Greif D. Numerical simu- plastic behaviour of steels during phase transformations – II. Study of calssical
lation of immersion quenching process for cast aluminium part at different pool plasticity for ideal-plastic phases. J Mech Phys Solids 1986;34:411–32.
temperatures. Appl Therm Eng 2014;65:74–84. https://doi.org/10.1016/j. [203] Mittemeijer EJ, Somers MAJ. Thermochemical surface engineering of steels.
applthermaleng.2013.12.058. Cambridge: Woodhead Publishing; 2015.
[170] Elkatatny I, Morsi Y, Blicblau AS, Das S, Doyle ED. Numerical analysis and ex- [204] Atraszkiewicz R, Januszewicz B, Kaczmarek Ł, Stachurski W, Dybowski K,
perimental validation of high pressure gas quenching. Int J Therm Sci Rzepkowski A. High pressure gas quenching: distortion analysis in gears after heat
2003;42:417–23. https://doi.org/10.1016/S1290-0729(02)00042-X. treatment. Mater Sci Eng A 2012;558:550–7. https://doi.org/10.1016/j.msea.
[171] CFX-5 Solver and Solver Manager, AEA Technology, CFDS Department, Harwell; 2012.08.047.
2002. [205] Li H, Zhao G, Huang C, Niu S. Technological parameters evaluation of gas
[172] Mackerle J. Finite element analysis and simulation of quenching and other heat quenching based on the finite element method. Comput Mater Sci
treatment processes. Comput Mater Sci 2003;27:313–32. https://doi.org/10. 2007;40:282–91.
1016/S0927-0256(03)00038-7. [206] Gawroński Z, Sawicki J. Technological surface layer selection for small module
[173] Lior N. The cooling process in gas quenching. J Mater Process Technol pitches of gear wheels working under cyclic contact loads. Mater Sci Forum
2004;155–156:1881–8. https://doi.org/10.1016/j.jmatprotec.2004.04.279. 2006;513:69–74. https://doi.org/10.4028/www.scientific.net/MSF.513.69.
[174] Krause F, Schüttenberg S, Fritsching U. Modelling and simulation of flow boiling [207] Karabelchtchikova O, Rivero IV, Hsiang SM. Modeling of residual stress distribu-
heat transfer. Int J Numer Methods Heat Fluid Flow 2010;20:312–31. tion in D2 steel via grinding dynamics using a second-order damping system. J
[175] Babu K, Kumar TS. Mathematical modelling of surface heat flux during quenching. Mater Process Technol 2008;198:313–22.
Metall Mater Trans B 2010;41:214–24. [208] Sawicki J, Siedlaczek P, Staszczyk A. Fatigue life predicting for nitrided steel –
[176] Meduri PK, Warrier GR, Dhir VK. Wall heat flux partitioning during subcooled finite element analysis. Arch Metall Mater 2018;63:917–23. https://doi.org/10.
forced flow film boiling of water on a vertical surface. Int J Heat Mass Transf 24425/122423.
2009;52:3534–46. [209] Sawicki J, Kruszyński B, Wójcik R. The influence of grinding conditions on the
[177] Srinivasan V, Wang DM. Modelling and simulation of the heat and mass transfer distribution of residual stress in the surface layer of 17CrNi6-6 steel after car-
characteristics of binary mixtures for boiling flow applications. J ASTM Int burizing. Adv Sci Technol Res J 2017;11:17–22.
2011;8:1–22. [210] Sawicki J, Siedlaczek P, Staszczyk A. Finite-element analysis of residual stresses
[178] Wang T, Yu X, Chen T, Zhang Q. CFD simulations of quenching process for partial generated under nitriding process: a three-dimensional model. Met Sci Heat Treat
oxidation of methane: comparison of jet-in-cross-flow and impinging flow con- 2018;59:799–804. https://doi.org/10.1007/s11041-018-0229-y.
figurations. Chin J Chem Eng 2016. https://doi.org/10.1016/j.cjche.2016.04.039. [211] Inoue T, Ju DY, Arimoto K. Metallo-thermo-mechanical simulation of quenching
[179] Gao BJ, Li HJ, Chen YF, Dong JH. Numerical analysis and structure improvement process – theory and implementation of computer code HEARTS. In: Proc. 1st Int.
for the corrosion and cracking of the mixing tee for the quenching gas of hydrogen Conf. Quenching Control Distortion, ASM International, Chicago; 1992. p. 205–12.
from coal. Proc Eng 2015;130:1246–57. https://doi.org/10.1016/j.proeng.2015. [212] SYSWELD. A predictive model for heat treat distortion, Southwest Research
12.206. Institute, San Antonio; 1992.
[180] Fluent, Fluent 6.3 User’s Guide, Fluent Inc., Lebanon; 2006. [213] Ferguson BL. Effective technical collaboration in heat treatment process model-
[181] Li Z, Grandhi RV, Srinivasan R. Distortion minimization during gas quenching ling: a case study. Int Heat Treat Surf Eng 2012;6:61–6. https://doi.org/10.1179/
process. J Mater Process Technol 2006;172:249–57. https://doi.org/10.1016/j. 1749514812Z.00000000017.
jmatprotec.2005.10.018. [214] Li Z, Ferguson BL, Nemkov V, Goldstein R, Jackowski J, Fett G. Effect of quenching

504
E. Wołowiec-Korecka Engineering Structures 177 (2018) 489–505

rate on distortion and residual stresses during induction hardening of a full-float computational tools for materials science. Calphad 2002;26:273–312. https://doi.
truck axle shaft. J Mater Eng Perform 2014;23:4170–80. https://doi.org/10.1007/ org/10.1016/S0364-5916(02)00037-8.
s11665-014-1196-0. [233] Özdemir IB, Akar F. The response of nitriding chemistry to different initial gas
[215] Nart E, Genel K. Investigation of carburising treatment of injector flange in diesel compositions. J Mater Eng Perform 2015;24:3002–7. https://doi.org/10.1007/
engine. Mater Des 2010;31:2110–7. https://doi.org/10.1016/j.matdes.2009.10. s11665-015-1585-z.
030. [234] Murav’ev VI. Effect of residual stresses on distortion of nitrided parts. Met Sci Heat
[216] Smirnov AV, Kuleshov JS. Calculation of diluted ammonia nitriding reactions. Treat 1975;17:796–8. https://doi.org/10.1007/BF00703072.
Metalloved Termicheskaya Obrab Met 1966;5:45–9. [235] Zot’eva AS, Kotsyubinskii OY. Predicting distortion during nitriding. Met Sci Heat
[217] Lakhtin YM, Kogan YD, Struve N. Entwicklungstendenzen des Gasnitrierens. Neue Treat 1977;19:986–9. https://doi.org/10.1007/BF00670173.
Hutte 1977;6:320–4. [236] Ikehata H, Tanaka K, Takamiya H, Mizuno H, Shimada T. Modeling growth and
[218] Lakhtin YM. The present state of the nitriding process. Met Sci Heat Treat dissolution kinetics of grain-boundary cementite in cyclic carburizing. Metall
1993;35:376–81. Mater Trans A 2013;44:3484–93. https://doi.org/10.1007/s11661-013-1722-y.
[219] Yang M, Sisson RD. Alloy effects on the gas nitriding process. J Mater Eng Perform [237] Schneider A. Iron layer formation during cementite decomposition in carburising
2014;23:4181–6. https://doi.org/10.1007/s11665-014-1187-1. atmosperes. Corros Sci 2002;44:2353–65.
[220] Haś Z, Kula P. Nitrovac’79 – ein neues Verfahren für thermo-chemische [238] Saltykov P, Putilina M. Three effects of the cementite phase of iron-carbon alloys
Behandlung. Pol Eng 1983;6:4–5. on their electrochemical behaviour in an axalic acid medium. Prot Met
[221] Lakhtin YM, Kogan YD, Soshkin SM. Nitriding of steels in vacuum. Met Sci Heat 2006;42:372–99.
Treat 1980;22:635–8. [239] Stokes B, Gao N, Reed PAS, Lee KK. Effects of carbides on fatigue characteristics of
[222] Soshkin SM, Lakhtin YM, Kogan YD. Structure of the diffusion layer with vacuum austempered ductile iron. Metall Mater Trans A 2005;36:977–88. https://doi.org/
nitriding. Met Sci Heat Treat 1984;26:521–3. 10.1007/s11661-005-0291-0.
[223] Löffler DG, Schmidt LD. Kinetics of ammonia decomposition on polycrystalline Pt. [240] Yusa S, Hara T, Tsuzaki K. Refinement of grain boundary cementite in medium-
J Catal 1976;41:440–54. carbon tempered martensite by thermomechanical processing. Mater Sci Eng A
[224] Ertl G, Huber M. Mechanism and kinetics of ammonia decomposition on iron. J 1999;273–275:462–5.
Catal 1980;61:537–47. [241] Todaka Y, Umemoto M, Ohno A. Dissolution of cementite on carbon steel by ball
[225] Mariadassou G, Shin C-H, Bugli G. Tamaru’s model for ammonia decomposition drop deformation and laser heating. J Alloys Compd 2007;434–435:497–500.
over titanium oxynitride. J Mol Catal A 1999;141:263–7. [242] Jack D, Jack K. Invited review: carbides and nitrides in steel. Mater Sci Eng A
[226] Yin SF, Xu BQ, Zhou XP, Au CT. A mini-review on ammonia decomposition cat- 1972;11:1–27.
alysts for on-site generation of hydrogen for fuel cell applications. Appl Catal A [243] Shatov A, Ponomarev S, Firstov S. The contiguity of carbide crystals of different
2004;277:1–9. shapes in cemented carbides. Int J Refract Met Hard Mater 2004;24:61–74.
[227] Waghode AN, Hanspal NS, Shigidi IM, Nassehi V, Hellgardt K. Computer model- [244] Lee E, Park W, Baik K, Ahn S. Different carbide types and their effect on bend
ling and numerical analysis of hydrodynamics and heat transfer in non-porous properties of a spray-formed high speed steel. Scr Mater 1998;39:1131–8.
catalytic reactor for the decomposition of ammonia. Chem Eng Sci [245] Stavrev D. Formation and changes in carbides in iron-carbon alloys under con-
2005;60:5862–77. centrated energy fluxes. Met Sci Heat Treat 2003;45:407–10.
[228] Deshmukh SR, Mhadeshwar AB, Vlachos DG. Microreactor modelling for hydrogen [246] Wiengmoon A, Chairuangsri T, Bown A. Microstructural and crystallographical
production from ammonia decomposition on rubidium. Ind Eng Chem Res study of carbides in 30wt.%Cr cast irons. Acta Mater 2005;53:4143–54.
2004;43:2986–99. [247] Zhang M, Kelly P. Crystallography and morphology of widmanstatten cementite in
[229] Alhumaizi K, Henda R, Soliman M. Numerical analysis of a re- austenite. Acta Mater 1998;46:4617–28.
action–diffusion–convection system. Comput Chem Eng 2003;27:579–94. [248] Kula P, Wołowiec E, Pietrasik R, Dybowski K, Klimek L. The precipitation and
[230] Özdemir IB, Akar F. The effect of flow orientation on nitriding process. Vacuum dissolution of alloy iron carbides in vacuum carburization processes for auto-
2015;116:104–9. https://doi.org/10.1016/j.vacuum.2015.03.004. motive and aircraft applications – Part I. Adv Mater Res 2012;486:297–302.
[231] Engström A, Höglund L, Ågren J. Computer simulation of diffusion in multiphase [249] Kula P, Wołowiec E. The application of artificial intelligence to modelling and
systems. Metall Mater Trans A 1994;25:1127–34. https://doi.org/10.1007/ evaluation of machines parts. In: Grzech A, Świątek P, Brzostowski K, editors.
BF02652288. Appl. Syst. Sci., EXIT, Warszawa. 2010. p. 315–20.
[232] Andersson J, Helander T, Höglund L, Shi P, Sundman B. Thermo-Calc & DICTRA,

505