Abstract
In the present study, the tensile strength of ferritic and austenitic functionally graded steels produced by electroslag remelting has been modeled. To produce functionally graded steels, two slices of plain carbon steel and austenitic stainless steels were spot welded and used as electroslag remelting electrode. Functionally graded steel containing graded layers of ferrite and austenite may be fabricated via diffusion of alloying elements during remelting stage. Vickers microhardness profile of the specimen has been obtained experimentally and modeled with adaptive networkbased fuzzy inference systems (ANFIS). To build the model for graded ferritic and austenitic steels, training, testing and validation using respectively 174 and 120 experimental data were conducted. According to the input parameters, in the ANFIS model, the Vickers microhardness of each layer was predicted. A good fit equation which correlates the Vickers microhardness of each layer to its corresponding chemical composition was achieved by the optimized network for both ferritic and austenitic graded steels. Afterwards; the Vickers microhardness of each layer in functionally graded steels was related to the yield stress of the corresponding layer and by assuming Holloman relation for stressstrain curve of each layer, they were acquired. Finally, by applying the rule of mixtures, tensile strength of functionally graded steels configuration was found through a numerical method. The obtained results from the proposed model are in good agreement with those acquired from the experiments.
chemical concentration profile; microhardness; tensile strength; ANFIS; ESR; ferritic FGS; austenitic FGS
Application of ANFIS for analytical modeling of tensile strength of functionally graded steels
Ali Nazari^{*} * email: alinazari84@aut.ac.ir
Department of Materials Engineering, Savhe Branch, Islamic Azad University, Saveh, Iran
ABSTRACT
In the present study, the tensile strength of ferritic and austenitic functionally graded steels produced by electroslag remelting has been modeled. To produce functionally graded steels, two slices of plain carbon steel and austenitic stainless steels were spot welded and used as electroslag remelting electrode. Functionally graded steel containing graded layers of ferrite and austenite may be fabricated via diffusion of alloying elements during remelting stage. Vickers microhardness profile of the specimen has been obtained experimentally and modeled with adaptive networkbased fuzzy inference systems (ANFIS). To build the model for graded ferritic and austenitic steels, training, testing and validation using respectively 174 and 120 experimental data were conducted. According to the input parameters, in the ANFIS model, the Vickers microhardness of each layer was predicted. A good fit equation which correlates the Vickers microhardness of each layer to its corresponding chemical composition was achieved by the optimized network for both ferritic and austenitic graded steels. Afterwards; the Vickers microhardness of each layer in functionally graded steels was related to the yield stress of the corresponding layer and by assuming Holloman relation for stressstrain curve of each layer, they were acquired. Finally, by applying the rule of mixtures, tensile strength of functionally graded steels configuration was found through a numerical method. The obtained results from the proposed model are in good agreement with those acquired from the experiments.
Keywords: chemical concentration profile, microhardness, tensile strength, ANFIS, ESR, ferritic FGS, austenitic FGS
1. Introduction
Functionally graded materials (FGM) possess properties that vary gradually with location within the material^{1}. An FGM comprises a multiphase material with volume fractions of the constituents varying gradually in a predetermined (designed) profile, thus yielding a nonuniform microstructure in the material with continuously graded properties^{2}. There are not enough studies on the plastic behavior of FGMs. Among these few works, most of the researchers have been modeled their work with the aid of conventional flow theories which are the one of the best tools that has ever proposed. For example, some of them have tried to use J_{2} flow theory^{35} but the empirical investigations haven't been linked to the obtained results because of the difficulty of FGMs fabrication. Okolednik^{6} although has used J integral concept to model several materials with yield stress gradient, but his studies was not confirmed by the experimental results. One of the FGMs with elasticplastic behavior is functionally graded steel (FGS) which have recently been produced from austenitic stainless steel and carbon steel using electro slag refining (ESR) method^{7,8}. In these composites, by selecting the appropriate arrangement and thickness of the primary ferritic and austenitic steels as electrodes, it is possible to obtain composites with several layers consist of ferrite, austenite, bainite and martensite. The resultant composites using two slices of original ferrite (α_{0}) and original austenite (γ_{0}) is as below:
where α, β and γ are ferrite, bainite and austenite phase in the final composite respectively; el is electrode; com is composite; and R is remelting.
Diffusion of chromium, nickel and carbon atoms which taking place at the remelting stage in the liquid phase controls the chemical distribution of chromium, nickel and carbon atoms in the produced composites. The transformation characteristics of FGSs have previously been investigated, in that the diffusion coefficients of chromium, nickel, and carbon atoms at temperatures just above the melting point of iron were estimated. Also, the thicknesses of the emerging bainite and martensite phases were determined^{7}.
Furthermore it has been shown that the tensile strength of the FGS composites depends on the composition and number of layers and those has been modeled based on the tensile behavior of individual phases^{8}; to do so the yield stress of each element in the composites was related to the microhardness value of that element.
In the previous studies, Chary impact energy of functionally graded steels in both crack divider^{912} and crack arrester^{12,13} configurations was experimentally examined and modeled by different methods. In addition, the ductile to brittle transition of the specimens was studied in a series of works^{1418}. Fracture toughness of these specimens in terms of J_{IC} in both crack divider^{1921} and crack arrester^{21,22} configurations was also investigated. The tensile behavior of oblique layer functionally graded steels was the other property which studied in the previous studies^{23,24}. Prediction Vickers hardness^{25} and tensile strength^{26} of functionally graded steels by the mechanismbased strain gradient plasticity theory was the other works done in this area. In a series of works, Charpy impact energy^{2734} and fracture toughness^{35,36} of functionally graded steels was modeled based on strain gradient plasticity theory.
Several works have addressed utilizing of computeraided prediction of engineering properties including those done by the authors^{14,18,3740}. Adaptive networkbased fuzzy inference systems (ANFIS) is the famous hybrid neurofuzzy network for modeling the complex systems^{41}. ANFIS incorporates the humanlike reasoning style of fuzzy systems through the use of fuzzy sets and a linguistic model consisting of a set of IFTHEN fuzzy rules. The main strength of ANFIS models is that they are universal approximators^{41} with the ability to solicit interpretable IFTHEN rules. Nowadays, the artificial intelligencebased techniques like ANFIS^{42} have been successfully applied in the engineering applications. However, there is a lack of investigations on metallurgical aspects of materials.
In the present work, microhardness profile of functionally graded ferritic and austenitic steels has been modeled by ANFIS and then tensile strength of FGSs has been modeled analytically by means the ANFIS results. To build the model for graded ferritic and austenitic steels, training, testing and validation using respectively 174 and 120 experimental data were conducted. The obtained results have been compared by experimental ones to evaluate the software power for predicting the microhardness profile of functionally graded ferritic and austenitic steels. Two equations were presented by the ANFIS results which correlate the Vickers microhardness profile of both ferritic and austenitic steels to their corresponding chemical composition profile. Afterwards; by supposing suitable relationship between Vickers microhardness and the yield stress of the corresponding layer and by assuming Holloman relation for stressstrain curve of each layer, they were obtained. Finally, by applying the rule of mixtures, tensile strength of FGSs was found. There was a good agreement between the predicted results and those obtained from the experiments.
2. Experimental Procedure
To make FGSs, a miniature ESR apparatus was used. The consumed slag was a mixture of 20% CaO, 20% Al_{2}O_{3} and 60% CaF_{2}. The original ferritic and austenitic steels (α_{0} and γ_{0}) which used as electrodes were commercial type AISI 1020 and AISI 316 steels respectively. The chemical composition of the asreceived ferritic and austenitic steels is given in Table 1.
Ferritic and austenitic steel slices were spot welded in form of 2piece electrode for remelting. The thickness of each slice in the primary electrode was 150 mm.
Remelting processes were carried out under a constant power supply of 16 KVA. After remelting, the composite ingots were hotpressed down to the thickness of 30 mm. Forging and rolling operations were carried out at 980 °C and then specimens were aircooled.
To investigate the variation of hardness in composites, Vickers microhardness test was employed using 100 gf weight.
The concentration of chromium, nickel and carbon in functionally graded steels was determined by dataequipped linear analyzer.
As the previous work^{8} indicate, a bainite layer is produced during remelting stage approximately in the middle of the forged specimen. Therefore, two series of tensile specimens were produced (one from the ferritic and the other from the austenitic graded region) in which bainite layer was not placed in the produced specimens as shown in Figure 1. Tensile specimens from the FGS specimens were made. Tensile tests were carried out under extension rate of 0.1 mm/s. Specimens dimension was in accordance to the ASTM E8 standard and it is shown in Figure 2. The as received rod was annealed at 980 °C and then aircooled.
Tensile strength of asreceived ferritic and austenitic steels which were annealed at 980 °C and then aircooled was also measured.
For metallographic examinations, the plates were sliced, ground, polished, and etched in a "Kalling" solution and 1 pct Nital.
3. Experimental Results
Vickers microhardness profile of the ferritic and austenitic regions of the functionally graded steel is illustrated in Figure 1. The concentration profile of chromium, nickel and carbon atoms has been illustrated in Figure 3. All of the results show a good compatibility with the initial work done by Aghazadeh and Shahosseinie^{7}. The mechanism of diffusion of chromium, nickel and carbon atoms has been discussed in that work^{7}. In addition, metallographic studies from the cross section of the produced FGS show that the new stabilized phase (bainite layer) in the FGS produced is similar to those acquired in the previous works^{736} as shown in Figure 4; the thickness of bainite layer is 0.6 mm which was verified by Vickers microhardness examination as it is shown in Figure 1. Finally the stressstrain curves of original ferrite and original austenite specimens have been illustrated in Figure 5.
Tensile strength of FGSs has been illustrated in Table 2. For comparison, tensile strength of the specimens' edge (γ_{0}, γ_{m}, α_{0} and α_{m} layers) has been shown in Figure 5. Electronprobe microanalysis studies illustrate that the chemical composition of the α edge layer is pct C = 0.2, pct Cr = trace, and pct Ni = trace, and that of the γ edge layer is pct C = 0.07, pct Cr = 18.1, and pct Ni = 9.1, which is similar to those of original alpha and gamma steels; this is in accordance to the previous results^{7,8}. Thus, the first boundary condition may be determined using the predicted Vickers hardness value and tensile strength of edge layers (i.e. original austenite, γ_{0} for γ region and original ferrite, α_{0} for α region). To achieve the tensile strength of γ_{m} and α_{m} layers, tensile specimens of the same composition and same mechanical properties to γ_{m} and α_{m} layers were prepared similar to the previous studies^{7,8}. Initially, the average chemical composition of γ_{m} and α_{m} layers was obtained (Table 3). Afterwards, samples with composition in accordance to the average chemical composition of γ_{m} and α_{m} layers were produced by means of a vacuum induction furnace. Similar to the primary composites, the hotpressing process was carried out at 980 °C, followed by air cooling. Through trial and error (i.e., conforming the chemical composition and changing the cooling rate), the sample with the nearest hardness to γ_{m} and α_{m} layers was selected to make tensile test specimens. Tensile test results of (γ_{0}, γ_{m}, α_{0} and α_{m} layers are shown in Figure 5.
4. Architecture of ANFIS
The architecture of an ANFIS model with two input variables is shown in Figure 6. Suppose that the rule base of ANFIS contains two fuzzy IFTHEN rules of Takagi and Sugeno's type as follows:
The basic learning rule of ANFIS is the backpropagation gradient descent, which calculates error signals recursively from the output layer backward to the input nodes. This learning rule is exactly the same as the backpropagation learning rule used in the common feedforward neural networks^{43,44}. Recently, ANFIS adopted a rapid learning method named as hybridlearning method which utilizes the gradient descent and the leastsquares method to find a feasible set of antecedent and consequent parameters^{43,44}. Thus in this paper, the later method is used for constructing the proposed models.
4.1. ANFIS structure and parameters
The structure of proposed ANFIS networks was consisted of the chromium concentration at the first of each layer (fCr), the chromium concentration at the end of each layer (eCr), the nickel concentration at the first of each layer (fNi), the nickel concentration at the end of each layer (eNi), the carbon concentration at the first of each layer (fC), the carbon concentration at the end of each layer (eC) and the distance of the middle of each layer from the specimen edge (D). To achieve a more accurate model, the concentration of chromium, nickel and carbon atoms in the first and the end sides of each layer were considered as inputs of the network to increase the total inputs to seven. It should be noted that the specimen edge was that side of the specimens with similar chemical composition to the original ferritic and austenitic steels respectively for graded ferritic and graded austenitic steels. The value for output layer was the average Vickers microhardness of each layer which was obtained from Equation 4:
where: Hv(x) is the Vickers microhardness profile of functionally graded steels, and x_{i+1}x_{i} represents the thickness of each layer which was considered equal to 100 µm in this study.
The input space is decomposed by three fuzzy labels. In this paper, for comparison purposes, two types of membership functions (MFs) including the triangular (ANFISI) and Gaussian (ANFISII) were utilized to construct the suggested models. To build the models, the thickness of the ferritic and austenitic regions was divided into 100 µm thick layers. Therefore, 174 and 120 layers were achieved for ferritic and austenitic regions, respectively. For graded ferritic region, from 174 collected data, 122 data (70%) were randomly chosen for training set, 26 (15%) data for testing set and the other 26 (15%) data for validation set (ANFISI model). For graded austenitic region, from 120 collected data, 84 data (70%) were randomly chosen for training set, 18 (15%) data for testing set and the other 18 (15%) data for validation set (ANFISII model). Moreover, up to 1000 epochs were specified for training process to assure the gaining of the minimum error tolerance.
One of the most difficult tasks in ANFIS studies is to find this optimal network architecture, which is based on the determination of numbers of optimal results. The assignment of initial weights and other related parameters may also influence the performance of the ANFIS to a great extent. However, there is no well defined rule or procedure to have an optimal network architecture and parameter settings where the trial and error method still remains valid. This process is very time consuming^{4548}.
In this study the Matlab NN toolbox is used for NN applications. To overcome optimization difficulty, a program has been developed in Matlab which handles the trial and error process automatically^{4548}. The program tries various functions and when the highest RMSE (Root Mean Squared Error) of the testing set, as the training of the testing set is achieved, it was reported^{4548}.
The IFTHEN rules in this study were achieved as follows. Suppose that the rule base of ANFIS contains two fuzzy IFTHEN rules of Takagi and Sugeno's type:
The corresponding equivalent ANFIS architecture is shown in Figure 7. The functions of each layer are described as follows^{41,42,49,50}:
Layer 1  Every node i in this layer is a square node with a node function:
where fCr, eCr, fNi, eNi, fC, eC and D are inputs to node i, and A_{i}, B_{i}, C_{i}, D_{i}, E_{i,} F_{i} and G_{i} are the linguistic label (fuzzy sets: small, large, ...) associated with this node function.
Layer 2  Every node in this layer is a circle node labeled П which multiplies the incoming signals and sends the product out. For instance,
Each node output represents the firing weight of a rule.
Layer 3  Every node in this layer is a circle node labeled N. The ith node calculates the ratio of the ith rule's firing weight to the sum of all rule's firing weights:
Layer 4  Every node in this layer is a square node with a node function:
where is the output of layer 3, and {p_{i}, q_{i}, r_{i}, s_{i}, t_{i}, u_{i}, v_{i}, w_{i}} is the parameter set.
Layer 5  The signal node in this layer is a circle node labeled R that computes the overall output as the summation of all incoming signals, i.e.,
4.2. ANFIS results and discussion
In this study, the error arose during the training and testing in ANFISI and ANFISII models can be expressed as absolute fraction of variance (R^{2}) which is calculated by Equation 18^{51}:
where t is the target value and o is the output value.
All of the results obtained from experimental studies and predicted by using the training, testing and validation results of ANFISI and ANFISII models are given in Figures 8a and ^{c}; and Figures 9a and ^{c}, respectively. The linear least square fit line, its equation and R^{2} values were shown in these figures for the training, testing and validation data. Also, inputs values and experimental results with testing and validation results obtained from ANFISI and ANFISII models were given in Tables 4 and 5, respectively. As it is visible in Figures 8 and 9, the values obtained from the training, testing and validation sets in ANFISI and ANFISII models are very close to the experimental results. The results of testing and validation phases in Figures 8 and 9 show that the ANFISI and ANFISII models are capable of generalizing between input and output variables with reasonably good predictions.
The performance of the ANFISI and ANFISII models is shown in Figures 8 and 9. The best value of R^{2} is 99.75% for training set in the ANFISI model. The minimum values of R^{2} are 98.34% for testing set in the ANFISII model. All of R^{2} values show that the proposed ANFISI and ANFISII models are suitable and can predict microhardness profile of FGSs values very close to the experimental values.
From the optimized network, the best fit equation to predict Vickers microhardness values by the specific inputs was obtained. These relationships for ferritic and austenitic regions are expressed by Equations 19 and 20, respectively:
where: HV(α) and HV(γ) are the Vickers microhardness of each layer in ferritic and austenitic regions, respectively. The R^{2} values are between experimental results and those obtained by Equations 19 and 20 for ferritic and austenitic regions, respectively.
5. Modeling Tensile Strength
To model tensile strength of functionally graded steels it has been assumed that the austenitic functionally graded steel consists of mγ layers and ferritic functionally graded steel consists of mα layers. According to the previous work^{10}, it has been assumed that tensile strength of each layer is related to its corresponding stressstrain curve.
According to the previous works^{912} it is assumed that the yield stress of each element is proportional to the Vickers microhardness of that element. Therefore, the yield stress of each layer in α and γ regions should also obey the hardness pattern. The yield stress of each layer may be related to the Vickers microhardness of that layer as:
where:
σ_{y}(γ_{0}), σ_{y}(γ_{m}), σ_{y}(α_{0}) and σ_{y}(α_{m}) are the yield stress of γ_{0}, γ_{m}, α_{0} and α_{m} layers, respectively; and HV(γ_{0}), HV(γ_{m}), HV(α_{0}) and HV(α_{m}) are the yield stress of γ_{0}, γ_{m}, α_{0} and α_{m} layers, respectively.
If it is assumed that the stressstrain curve of each layer obeys the Holloman relation, the imposed stress to each layer at yield strain of α_{m} and γ_{m} layers may be given as;
where:
σ' and σ" are the imposed stress to each layer in α and γ regions at yield strain of α_{m} and γ_{m} layer, respectively; Ε_{α}m and Ε_{γ}m are the yield strain of α_{m} and γ_{m} layers, respectively; and n(α) and n(γ) are the strainhardening coefficient of each element in α and γ regions, respectively. It is assumed that the strain hardening coefficient of each element in the studied
where n(γ_{0}), n(γ_{m}), n(α_{0}) and n(α_{m}) are the workhardening exponent of γ_{0}, γ_{m}, α_{0} and and α_{m} layers, respectively and x_{γ0}, x_{γm}, x_{α0} and x_{αm} are the positions of γ_{0}, γ_{m}, α_{0} and α_{m} layers, respectively.
Ε(α) and Ε(γ) in Equations 23 and 24 are defined as the yield strain of each layer in α and γ regions, respectively. By considering the suitable boundary conditions:
where: E is the Young modulus.
Therefore the stressstrain curve of each element could be determined. By applying tension loading, all of layers will fail at the ultimate tensile strain of the strongest layer (i.e. γ_{m} in austenitic FGS and α_{m} in ferritic FGS). Therefore, tensile strength of each composite could be obtained as:
where: σ_{ts}(α  FGS) and σ_{ts} (γ  FGS) are the tensile strength of ferritic FGS and austenitic FGS, respectively.
The obtained results from the mathematical model are given in Table 2; there is a good agreement between the experimental and theoretical results.
6. Conclusions
In this work, tensile strength of FGSs produced by ESR process was examined experimentally and predicted by means of ANFIS. Variation of tensile strength in FGSs is due to variation of alloying elements such as chromium, nickel and carbon during the remelting stage. As alloying elements diffuses, FGSs with ferritic and austenitic graded layers together with bainite intermediate layer are created. A model based on ANFIS was introduced to predict the Vickers microhardness of ferritic and austenitic regions. The performance of the acquired optimized network was examined by R^{2} values. All of the obtained R^{2} values showed that ANFIS are capable to predict the Vickers microhardness of FGSs very close to the experimental data. It has been found that ANFIS models will be valid within the ranges of variables. Finally, two empirical equations obtained from optimized ANFISI and ANFISII networks was presented with an excellent accuracy to predict the Vickers microhardness of each layer by means of its input data. The Vickers microhardness of each layer in functionally graded steels was related to the yield stress of the corresponding layer and by assuming Holloman relation for stressstrain curve of each layer, the area under each stressstrain curve was acquired. Tensile strength of FGSs was obtained by means of the rule of mixtures. A good agreement was obtained between the predicted results and those obtained from the experiments.
Received: October 15, 2011
Revised: February 7, 2012
 1. Sankar BV. An elasticity solution for functionally graded beams. Composites Science and Technology 2001; 61:689696. http://dx.doi.org/10.1016/S02663538(01)000070
 2. Jin ZH, Paulino GH and Dodds Junior RH. Cohesive fracture modeling of elasticplastic crack growth in functionally graded materials. Engineering Fracture Mechanics 2003; 70:18851912. http://dx.doi.org/10.1016/S00137944(03)001309
 3. Carpenter RD, Liang WW, Paulino GH, Gibeling JC and Munir ZA. Fracture Testing and Analysis of a Layered Functionally Graded Ti/TiB Beam in 3Point Bending. Materials Science Forum. 1999; 308311:837842. http://dx.doi.org/10.4028/www.scientific.net/MSF.308311.837
 4. Williamson RL, Rabin BH and Drake JT. Finite element analysis of thermal residual stresses at graded ceramicmetal interfaces. Part I. Model description and geometrical effects. Journal of Applied Physics. 1993; 74:13101320. http://dx.doi.org/10.1063/1.354910
 5. Giannakopoulos AE, Suresh S, Finot M and Olsson M. Elastoplastic analysis of thermal cycling: layered materials with compositional gradientes. Acta Metallurgica et Materialia. 1995; 43:13351354. http://dx.doi.org/10.1016/09567151(94)00360T
 6. Kolednik O. The yield stress gradient effect in inhomogeneous materials. International Journal of Solids and Structures. 2000; 37(5):781808. http://dx.doi.org/10.1016/S00207683(99)000608
 7. Aghazadeh Mohandesi J and Shahosseinie MH. Transformation Characteristics of Functionally Graded Steels Produced by Electroslag Remelting. Metallurgical and Materials Transactions A. 2005; 36A:34713476.
 8. Aghazadeh Mohandesi J, Shahosseinie MH. and Parastar Namin R. Tensile Behavior of Functionally Graded Steels Produced by Electroslag Remelting. Metallurgical and Materials Transactions A. 2006; 37A:21252132.
 9. Nazari A and Aghazadeh Mohandesi J. Modelling impact resistance of functionally graded steels with crack divider configuration. Materials Science and Technology 2010; 26:13771383. http://dx.doi.org/10.1179/174328409X405652
 10. Nazari A and Aghazadeh Mohandesi J. Impact Energy of Functionally Graded Steels with Crack Divider Conûguration. Materials Science and Technology 2009; 25(6):847852.
 11. Nazari A, Aghazadeh Mohandesi J and Riahi S. Modeling Impact Energy of Functionally Graded Steels in Crack Divider Configuration Using Modified StressStrain Curve Data. International Journal of Damage Mechanics 2010; 21(1):2750. http://dx.doi.org/10.1177/1056789510397073
 12. Nazari A, Aghazadeh Mohandesi J, Hamid Vishkasogheh M and Abedi M. Simulation of impact energy in functionally graded steels. Computational Materials Science 2011; 50:11871196. http://dx.doi.org/10.1016/j.commatsci.2010.11.019
 13. Nazari A and Aghazadeh Mohandesi J. Impact Energy of Functionally Graded Steels in Crack Arrester Configuration. Journal of Materials Engineering and Performance 2010; 19:10581064. http://dx.doi.org/10.1007/s1166500995784
 14. Nazari A and Milani AA. Modeling ductile to brittle transition temperature of functionally graded steels by artificial neural networks. Computational Materials Science. 2011; 50:20282037. http://dx.doi.org/10.1016/j.commatsci.2011.02.003
 15. Nazari A and Milani AA. Ductile to Brittle Transition Temperature of Functionally Graded Steels. International Journal of Damage Mechanics 2010; 21(1). http://dx.doi.org/10.1177/1056789511398270
 16. Nazari A and Milani AA. Modeling Ductile  to  Brittle Transition Temperature of Functionally Graded Steels by Gene Expression Programming. International Journal of Damage Mechanics 2011; 22. http://dx.doi.org/10.1177/1056789511406561
 17. Nazari A and Milani AA. Ductile to brittle transition temperature of functionally graded steels with crack arrester configuration. Materials Science and Engineering A. 2011; 528:38543859. http://dx.doi.org/10.1016/j.msea.2011.01.105
 18. Nazari A and Milani AA. Modeling ductile to brittle transition temperature of functionally graded steels by fuzzy logic. Journal of Materials Science 2011; 46:60076017. http://dx.doi.org/10.1007/s108530115563z
 19. Nazari A, Aghazadeh Mohandesi J and Riahi S. Modified Modeling Fracture Toughness of Functionally Graded Steels in Crack Divider Configuration. International Journal of Damage Mechanics 2011; 20:811831. http://dx.doi.org/10.1177/1056789510382851
 20. Aghazadeh Mohandesi J, Nazari A, Hamid Vishkasogheh M and Abedi M. Modelling Simul. Materials Science and Engineering 2010; 18:075007.
 21. Nazari A, Aghazadeh Mohandesi J and Riahi S. Fracture Toughness of Functionally Graded Steels. Journal of Materials Engineering and Performance 2011. http://dx.doi.org/10.1007/s1166501199459
 22. Nazari A, Aghazadeh Mohandesi J and Riahi S. Modeling fracture toughness of functionally graded steels in crack arrester configuration. Computational Materials Science 2011; 50:15781586. http://dx.doi.org/10.1016/j.commatsci.2010.12.019
 23. Nazari A and Aghazadeh Mohandesi J. Modeling tensile strength of oblique layer functionally graded austenitic steel. Computational Materials Science 2011; 50:14251431. http://dx.doi.org/10.1016/j.commatsci.2010.11.029
 24. Nazari A and Riahi S. Effect of layer angle on tensile behavior of oblique layer functionally graded steels. Turkish Journal of Engineering & Environmental Sciences 2010; 34:1724.
 25. Nazari A, Aghazadeh Mohandesi J and Tavareh S. Microhardness profile prediction of a graded steel by strain gradient plasticity theory. Computational Materials Science 2011; 50:17811784. http://dx.doi.org/10.1016/j.commatsci.2011.01.014
 26. Nazari A, Aghazadeh Mohandesi J and Tavareh S. Modeling tensile strength of austenitic graded steel based on the strain gradient plasticity theory Computational Materials Science 2011; 50:17911794. http://dx.doi.org/10.1016/j.commatsci.2011.01.016
 27. Nazari A and Mojtahed Najafi SM. Prediction Charpy impact energy of bcc and fcc functionally graded steels in crack divider configuration by strain gradient plasticity theory. Computational Materials Science 2011; 50:31783183. http://dx.doi.org/10.1016/j.commatsci.2011.05.047
 28. Nazari A and Mojtahed Najafi SM. Prediction impact behavior of functionally graded steel by strain gradient plasticity theory. Computational Materials Science 2011; 50:32183223. http://dx.doi.org/10.1016/j.commatsci.2011.06.004
 29. Nazari A. Modeling Charpy impact energy of functionally graded steel based on the strain gradient plasticity theory and modified stressstrain curve data. Computational Materials Science 2011; 50(12):33503357. http://dx.doi.org/10.1016/j.commatsci.2011.06.029
 30. Nazari A. Application of strain gradient plasticity theory to model Charpy impact energy of functionally graded steels. Computational Materials Science 2011; 50(12):34103416. http://dx.doi.org/10.1016/j.commatsci.2011.06.039
 31. Nazari A. Strain gradient plasticity theory to predict the input data for modeling Charpy impact energy in functionally graded steels. Computational Materials Science 2011; 50(12): 34423449. http://dx.doi.org/10.1016/j.commatsci.2011.07.007
 32. Nazari A. Simulation of impact energy in functionally graded steels by mechanismbased strain gradient plasticity theory. Computational Materials Science 2012; 51(1):1319. http://dx.doi.org/10.1016/j.commatsci.2011.07.010
 33. Nazari A. Simulation Charpy impact energy of functionally graded steels by modified stressstrain curve through mechanismbased strain gradient plasticity theory. Computational Materials Science 2012; 51(1): 225232. http://dx.doi.org/10.1016/j.commatsci.2011.07.027
 34. Nazari A. Application of strain gradient plasticity theory to model Charpy impact energy of functionally graded steels using modified stressstrain curve data. Computational Materials Science 2012; 51(1):281289. http://dx.doi.org/10.1016/j.commatsci.2011.07.057
 35. Nazari A. Modeling fracture toughness of ferritic and austenitic functionally graded steel based on the strain gradient plasticity theory. Computational Materials Science 2011; 50:32383244. http://dx.doi.org/10.1016/j.commatsci.2011.06.008
 36. Nazari A. Strain gradient plasticity theory for modeling JIC of functionally graded steels. Computational Materials Science 2011; 50(12): 34033409. http://dx.doi.org/10.1016/j.commatsci.2011.06.038
 37. Nazari A and Riahi S. Computeraided prediction of physical and mechanical properties of high strength cementitious composite containing Cr2O3 nanoparticles. Nano 2010; 5(5):301318. http://dx.doi.org/10.1142/S1793292010002219
 38. Nazari A and Riahi S. Prediction split tensile strength and water permeability of high strength concrete containing TiO2 nanoparticles by artificial neural network and genetic programming. Composites Part B: Engineering. 2011; 42:473488. http://dx.doi.org/10.1016/j.compositesb.2010.12.004
 39. Nazari A and Riahi S. Computeraided design of the effects of Fe2O3 nanoparticles on split tensile strength and water permeability of high strength concrete. Materials and Design. 2011; 32:39663979. http://dx.doi.org/10.1016/j.matdes.2011.01.064
 40. Nazari A and Didehvar N. Modeling impact resistance of aluminumepoxy laminated composites. ANFIS 2011; 42:19121919.
 41. Jang JSR. ANFIS: adaptivenetworkbased fuzzy inference system. IEEE Transactions on Systems, Man and Cybernetics. 1993; 23(3):66585. http://dx.doi.org/10.1109/21.256541
 42. Sar1demir M. Predicting the compressive strength of mortars containing metakaolin by artificial neural networks and fuzzy logic. Advances in Engineering Software. 2009; 40(9):9207. http://dx.doi.org/10.1016/j.advengsoft.2008.12.008
 43. Topcu IB and Sar1demir M. Prediction of mechanical properties of recycled aggregate concretes containing silica fume using artificial neural networks and fuzzy logic. Computational Materials Science 2008; 42(1):7482. http://dx.doi.org/10.1016/j.commatsci.2007.06.011
 44. Jang JSR and Sun CT. Nurofuzzy modeling and control. Proceedings IEEE. 1995;83(3). http://dx.doi.org/10.1109/5.364486
 45. Guzelbey IH, Cevik A and Erklig A. Prediction of web crippling strength of coldformed steel sheetings using neural Networks. Journal of Constructional Steel Research 2006; 62:962973. http://dx.doi.org/10.1016/j.jcsr.2006.01.008
 46. Guzelbey IH, Cevik A and Gögüs MT. Prediction of rotation capacity of wide flange beams using neural Networks. Journal of Constructional Steel Research 2006; 62:950961. http://dx.doi.org/10.1016/j.jcsr.2006.01.003
 47. Cevik A and Guzelbey IH. Neural Network Modeling Of Strength Enhancement For Cfrp Confined Concrete Cylinders. Building & Environment. 2008; 43:751763. http://dx.doi.org/10.1016/j.buildenv.2007.01.036
 48. Cevik A and Guzelbey IH. A Soft computing based approach for the prediction of ultimate strength of metal plates in compression. Engineering Structures 2007; 29(3):383394. http://dx.doi.org/10.1016/j.engstruct.2006.05.005
 49. Ramezanianpour AA, Sobhani M and Sobhani J. Application of network based neurofuzzy system for prediction of the strength of high strength concrete. Amirkabir Journal of Science and Technology. 2004; 5(59C):7893.
 50. Ramezanianpour AA, Sobhani J and Sobhani M. Application of an adaptive neurofuzzy system in the prediction of HPC compressive strength. In: Proceedings of the 4th international conference on engineering computational technology; 2004; Lisbon. Lisbon: CivilComp Press; 2004.
 51. Topcu IB and Sar1demir M. Prediction of compressive strength of concrete containing fly ash using artificial neural network and fuzzy logic. Computational Materials Science. 2008; 41(3):30511. http://dx.doi.org/10.1016/j.commatsci.2007.04.009
Publication Dates

Publication in this collection
05 Apr 2012 
Date of issue
June 2012
History

Received
15 Oct 2011 
Reviewed
07 Feb 2012