Sei sulla pagina 1di 17

Applied Mathematical Modelling 38 (2014) 55925608

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

Parameters optimization of selected casting processes using


teachinglearning-based optimization algorithm
R. Venkata Rao , V.D. Kalyankar 1, G. Waghmare 1
Department of Mechanical Engineering, S.V. National Institute of Technology, Surat, Gujarat 395007, India

a r t i c l e

i n f o

Article history:
Received 16 August 2012
Received in revised form 31 December 2013
Accepted 15 April 2014
Available online 9 May 2014
Keywords:
Parameter optimization
Squeeze casting
Die casting
Continuous casting
Mathematical models
TLBO algorithm

a b s t r a c t
In the present work, mathematical models of three important casting processes are considered namely squeeze casting, continuous casting and die casting for the parameters optimization of respective processes. A recently developed advanced optimization algorithm
named as teachinglearning-based optimization (TLBO) is used for the parameters optimization of these casting processes. Each process is described with a suitable example which
involves respective process parameters. The mathematical model related to the squeeze
casting is a multi-objective problem whereas the model related to the continuous casting
is multi-objective multi-constrained problem and the problem related to the die casting is
a single objective problem. The mathematical models which are considered in the present
work were previously attempted by genetic algorithm and simulated annealing algorithms.
However, attempt is made in the present work to minimize the computational efforts using
the TLBO algorithm. Considerable improvements in results are obtained in all the cases and
it is believed that a global optimum solution is achieved in the case of die casting process.
2014 Elsevier Inc. All rights reserved.

1. Introduction
Casting is one of the oldest manufacturing processes and with the progress of time, lot of developments took place in the
casting process. It is used for production of various complicated shapes that cannot be easily manufactured by any machining
process. Keeping in view the involvement of high temperature and safety aspects of the operators, various manual activities
of the casting processes are getting replaced by the high end automatic technologies. Based on the type of mold and way of
lling the molten metal, various types of casting processes are now available such as sand casting, die casting, continuous
casting, squeeze casting, investment casting, etc. and each type of casting process is having its particular application area.
The major success of all these casting processes depends upon the close control over all the input parameters and the proper
control over the metal solidication. Due to involvement of advanced technologies in the casting process, a small variation in
any of the input parameters affects the process output and produce defective castings. Hence the designers and foundry personnel are making lot of efforts to develop various mathematical models in the form of inputoutput relations so as to
achieve exact parameters setting instead of going for trial attempts. However, this can be achieved by using advanced optimization techniques as tools for obtaining the optimum parameters setting for the casting processes under consideration.

Corresponding author. Tel.: +91 261 2201661; fax: +91 2612227334.


1

E-mail address: ravipudirao@gmail.com (R.V. Rao).


Tel.: +91 261 2201661; fax: +91 2612227334.

http://dx.doi.org/10.1016/j.apm.2014.04.036
0307-904X/ 2014 Elsevier Inc. All rights reserved.

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

5593

In this work, efforts are carried out to prove the importance of advanced optimization techniques in the eld of parameters optimization of various casting processes so that the designers and foundry personnel can achieve their objectives
along with satisfying various constraints and limits of the respective process models. In the past, use of very few optimization techniques was involved in this eld. Three important casting processes are considered in this work namely squeeze
casting, continuous casting and die casting process. Optimization of mathematical models of these processes is carried by
using a recently developed advanced optimization algorithm named as teachinglearning-based optimization (TLBO) algorithm. The contribution of this paper is the application of the TLBO algorithm to the selected casting processes and to prove
the effectiveness of the algorithm.
In the next section the TLBO algorithm is explained followed by the optimization aspects of squeeze casting, continuous
casting and die casting processes using the TLBO algorithm. The importance of parameters optimization of respective
processes is described with detailed literature survey, problem description and result comparison.
2. Teachinglearning-based optimization algorithm
Teachinglearning-based optimization algorithm is a teachinglearning process inspired algorithm recently proposed by
Rao et al. [1,2] and Rao and Patel [3] based on the effect of inuence of a teacher on the output of learners in a class. In this
algorithm a group of learners are considered as population and different subjects offered to the learners are considered as
different design parameters and a learners result is analogous to the tness value of the optimization problem. The best
solution in the entire population is considered as the teacher. The design parameters are actually the parameters involved
in the objective function of the given optimization problem and the best solution is the best value of the objective function.
The working of TLBO algorithm is divided into two parts, Teacher phase and Learner phase. Working of both these
phases is described in detail by Rao et al. [1,2]. The same explanation of teacher phase and learner phase is referred here
for the working of TLBO algorithm. Fig. 1 represents the owchart of TLBO algorithm [2]. The TLBO algorithm has been
already tested on several constrained and unconstrained benchmark functions and proved better than the other advanced
optimization techniques [3]. It is also proving better in various eld of engineering such as those reported by Niknam
et al. [47] in the eld of electrical engineering, Togan [8] in the eld of civil engineering. Similarly, Krishnanand et al.
[9] used it for the problems related to economic load dispatch, Rao and Kalyankar [1012] used it for various elds related
to manufacturing processes such as machining processes, modern machining processes, laser beam welding process, etc. and
Rao and Patel [13,14] used it to attempt multi-objective mathematical models in the eld of thermal engineering. Even
repinek et al. [15] raised some doubts about the algorithm-specic parameter less concept of TLBO algorithm
though C
and some other issues, however, Rao and Patel [3] had already cleared all those issues and justied that the TLBO algorithm
is an algorithm-specic parameter less algorithm.
In the literature, it is observed that, the TLBO algorithm is not yet used in the eld of optimization of mathematical models of casting process. Hence the same is now used for the parameters optimization of various casting processes under consideration. In the next three sections, application of TLBO algorithm is presented for the optimization of mathematical
models of squeeze casting, continuous casting and die casting process respectively.
3. Parameters optimization of squeeze casting process
Various automotive parts are now getting produced by magnesium, aluminum and their alloys due to their improved
mechanical properties and suitability to various advanced and complicated parts. Products belonging to these materials
along with other ferrous materials can be easily manufactured by casting processes. However, the products of squeeze casting process are comparatively stronger due to better grain size and less metallic shrinking. Squeeze casting process has number of advantages such as elimination of gas and shrinkage porosities, high ductility, reduction of metal wastage due to the
absence of feeders or risers, etc. over the other casting processes. Various important input parameters involved in the
squeeze casting process are: squeeze pressure, melt temperature, die preheating temperature, squeeze time, melt volume
and quality, time delay before pressurization, etc.
Proper setting of these input parameters will not only help to increase the production but it also minimizes the defects
and rejection level. However, this proper parameter setting should not be based on trial attempts. In the literature, some
researchers had carried out research to study the effects of various input parameters on the process output and also tried
to achieve effective parameter setting. Hu [16] reviewed the progress in squeeze casting process and presented the effect
of process variables on the cast structure and properties of magnesium alloys and magnesium based composites. Design
of experiments was also discussed in their work for optimization of squeeze casting process. The various important input
process parameters discussed were melt volume and quality, magnitude and duration of applied pressure, die temperature,
pouring temperature, time delay before pressurization, lubrication, etc. Kim et al. [17] analyzed the microstructure of a
squeeze cast product. Comparisons of microstructure for a squeeze cast billet and the gravity cast billet were carried out
and analysis suggested the squeeze cast process as a best choice. In their other work, Kim et al. [18] carried out the experimental investigation to study the effect of die geometry on the microstructure of squeeze cast component.
Zhou et al. [19] compared the results of identical alloy component produced by squeeze cast and high pressure die cast and
showed that the porosity problem was very less in case of squeeze casting. Their study also showed signicant improvement

5594

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

Initialize number of students (population), termination criterion


Calculate the mean of each design variable
Identify the best solution (teacher)
Teacher
Phase

Modify solution based on best solution


Difference_Meanj,k,i = ri (Xj,kbest,i - TFMj,i)

Reject

No

Yes

Is new solution
better than existing?

Accept

Keep previous solution

Select two solutions randomly X'total-P,i and X'total-

No

Is X'total-P,i better than


X'total-Q,i

X''j,P,i=X'j,P,i+ri(X'j,Q,i - X'j,P,i)

No
Reject

Yes

X''j,P,i=X'j,P,i+ri(X'j,P,i -X'j,Q,i)

Is new solution
better than existing?

Yes

Student
Phase

Accept

Keep previous solution

Is termination
criteria satisfied?

No

Yes
Final value of solutions
Fig. 1. Flowchart of TLBO algorithm [2].

in mechanical properties of squeeze cast component over that of die cast component. Baek and Kwon [20] studied the effect of
process parameters on the uidity of a squeeze cast AlSi product and showed that the uidity of a selected material increased
with the silicon content. Yang et al. [21] investigated the effects of various process parameters such as applied pressure, pouring temperature and die temperatures on the macrostructure of a squeeze cast alloy. Run-xia et al. [22] studied the effect of
specic pressure on the microstructure and mechanical properties of ZA27 squeezed castings. The experimental investigation
suggested that the ne microstructure can be obtained with the increase of pressure. The micro structural study also revealed
that the strength and plasticity of squeeze casting can be increased by homogeneous distribution of Al and Cu elements in the
matrix of squeeze casting ZA27 alloy.
Moosa et al. [23] studied the effect of various input process parameters on the ultimate tensile strength of squeeze cast
components made of carbon ber AlSi composites. Various parameters considered in their experimental study were
squeeze pressure, die preheating temperature, pouring temperature, squeeze time and delay time. A set of input process
parameter were suggested in their study for obtaining the maximum tensile strength for given combination of material. Zang
et al. [24] carried out the experimental investigation to study the inuence of applied pressure on the tensile behavior and
microstructure of a squeeze cast magnesium alloy. The results of their investigation showed that the fracture mode of the
alloy changes from brittle to ductile as the applied pressures increases. Senthil and Amirthagadeswaran [25] used the
squeeze casting process for preparing the aluminum alloy castings of a non symmetrical component and carried out experimental investigation to study the inuence of various process parameters on mechanical properties of the casting. Taguchis

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

5595

orthogonal array was used to conduct the required number of experiments. Various process parameters considered were
squeeze pressure, melt temperature, die preheating temperature, die insert material and compression holding time. Mathematical models were developed for hardness and tensile strength using MINITAB 14 software. However, no optimization
technique was involved in their work.
It is observed from the literature that even though some research work was carried out related to the parameters optimization of squeeze casting process, but no optimization technique was used by previous researchers. Hence there is a scope
for using the advanced optimization techniques in the eld of squeeze casting process parameters optimization. To prove
this, an application example related to squeeze casting process is taken from the literature and the TLBO algorithm is applied
to it to get improvement in the result. The example is described in the following subsection along with the result
comparison.
3.1. Application example
Senthil and Amirthagadeswaran [25] presented the parameters optimization of squeeze casting process using Taguchi
method. Non symmetrical AC2A aluminum alloy was used in their experimental investigation. Various input process parameters considered by Senthil and Amirthagadeswaran [25] were squeeze pressure (MPa), melt temperature (C), die preheating temperature (C), die insert material and compression holding time (seconds). Each parameter was considered at four
levels and Taguchis L16 orthogonal array was used to conduct 16 experiments. The inuence of these input process parameters were studied on the important mechanical properties such as hardness and tensile strength. MINITAB software was
used by Senthil and Amirthagadeswaran [25] for the analysis purpose and the relationship between the input and output
process parameters was presented in the form mathematical model and the same is reproduced below by Eqs. (1) and (2).

Average hardness; H 3:82542 0:8787  A 0:46587  C 0:30411  E  0:00393  A2  0:00116  C 2


0:00097  E2 0:00051  A  C  0:00333  A  E  0:00018  C  E;

Average tensile strength; TS 11:2606 2:5778  A 1:3316  C 0:7552  E  0:0116  A2  0:0034  C 2


0:0031  E2 0:0015  A  C  0:0097  A  E  0:001  C  E;

where, A = squeeze pressure, B = melt temperature, C = die preheating temperature and E = compression holding time. One of
the input parameters, die insert material, was not involved in both the equations and it was also not described about how to
choose the die insert material. Hence the same die insert material suggested by Senthil and Amirthagadeswaran [25] i.e. hot
die steel is continued in the present work. Same range of process parameters as used by Senthil and Amirthagadeswaran [25]
is considered in the present work and is produced below:

Squeeze pressure 50125 MPa;


Melt temperature 675750  C;
Die preheating temperature 150300  C;
Compression holding time 1560 s:
Senthil and Amirthagadeswaran [25] had shown the optimum level of process parameters as squeeze pressure = 100 MPa,
melt temperature = 725 C, die preheating temperature = 200 C and compression holding time = 45 s. This set of input
parameters gives the average hardness of 100.76 BHN and average tensile strength of 278.45 MPa. However, Senthil and
Amirthagadeswaran [25] obtained this optimum set of process parameters by using MINITAB software and no advanced
optimization technique was involved in their work.
In the present work, same mathematical models are used as given by Senthil and Amirthagadeswaran [25] for hardness
and tensile strength and the process parameters optimization is carried out by using the TLBO algorithm. Initially the mathematical models of hardness and tensile strength are attempted separately to nd out the optimum parameter setting for
each of them and then both are attempted simultaneously as a combined objective function and a common parameter setting is also obtained. The population size and number of generations required to run the TLBO algorithm are decided by conducting trials at the beginning to check for the consistency of result and nally a population size of 10 and number of
generations of 20 are used for running the algorithm for this problem. The results obtained by TLBO algorithm for hardness
and tensile strength along with the comparison with the previous results are shown in Tables 1 and 2, respectively.
The maximum hardness and tensile strength reported by Senthil and Amirthagadeswaran [25] are 100 BHN and 278 MPa,
respectively. However, the TLBO has increased the hardness from 100 BHN to 103 BHN and tensile strength from 278 MPa to
290 MPa which is always benecial for the metal industry. The convergence of hardness and tensile strength with respect to
generations is shown in Figs. 2 and 3, respectively. This shows that the use of advanced optimization algorithm like TLBO will
denitely help the casting industries to improve their performance. In order to check for consistency of result, each algorithm is run for 50 times and the average result and standard deviation is obtained. In case of hardness, the average result

5596

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

Table 1
Result obtained by TLBO algorithm for hardness.
Parameters

Expt. result [25]

Squeeze pressure (MPa)


Melt temperature (C)
Die preheating temperature (C)
Compression holding time (s)
Maximum hardness (BHN)

100
725
200
45
100.76

TLBO result
Best

Average

Std. deviation

119
686
225
15
103.068

102.738

0.438

Best

Average

Std. deviation

119
675
220
15
290.30

289.22

1.59

Bold indicates the best optimum value.

Table 2
Result obtained by TLBO algorithm for tensile strength.
Parameters

Expt. result [25]

Squeeze pressure (MPa)


Melt temperature (C)
Die preheating temperature (C)
Compression holding time (s)
Maximum tensile strength (Mpa)

100
725
200
45
278.45

TLBO result

Bold indicates the best optimum value.

104
103

Hardness (BHN)

102
101
100
99
98
97
1

10 11 12 13 14 15 16 17 18 19 20

Number of generaons
Fig. 2. Variation of hardness with generations.

292

Tensile strength (MPa)

290
288
286
284
282
280
278
1

8 9 10 11 12 13 14 15 16 17 18 19 20
Number of generations

Fig. 3. Variation of tensile strength with generations.

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

5597

of 50 runs is 102.738 and the standard deviation obtained is 0.438. Similarly, in the case of tensile strength, the average
result and standard deviation of 50 runs obtained is 289.22 and 1.59, respectively. In both the cases the average result of
50 runs is observed to be very nearer to the best result and also the deviation is comparatively small.
Effort is also carried out here to obtain the common parameter setting which satises both the objectives and gives the
improved result. For this purpose, a combined objective function is prepared by normalizing both the objectives and assigning some weightage to each of them. The combined objective function used in the present work is given below by Eq. (3).

MaxZ w1  H=Hmax 1  w1  TS=TSmax ;

where, Hmax and TSmax are the maximum values of hardness and tensile strength respectively which can be obtained by
attempting an individual objective function and w1 is the weightage assigned to the rst objective. In the present case,
the weightage w1 is varied in the steps of 0.05 in order to obtain Pareto optimal points. The combined objective function
approach was not used by Senthil and Amirthagadeswaran [25]; however a common parameter setting was produced in
their work which was suggested for both the objectives. Now a population size of 10 and 20 generations are used for running
the TLBO algorithm for the combined objective function and a set of Pareto optimal solutions obtained is presented in the
form of a graph as shown in Fig. 4.
4. Parameters optimization of continuous casting process
Due to low production cost and higher production rates, continuous casting process is now getting widely used for the
mass production of various billets, blooms, thin sheets, etc. Regularly used metals in industries such as steel, copper and aluminum can be easily casted by using this process. In continuous casting process, molten metal is poured from one end and
the solidied slabs come out from other end through various rollers. However, control over various process parameters is
very important in this process to achieve the desired quality product. Various input process parameters involved in continuous casting process are: casting speed, ux viscosity, ux density, ux solidication temperature, mold thickness, mold
stroke length, cooling water temperature, cooling water ow rate, etc. Each of these parameters directly affects the process
outputs. Non-metallic inclusion in the form of oxides is one of the major problems in continuous casting of steels that can
lead to excessive casting repairs or rejected castings. To avoid such problems, close control over the casting speed, cooling
water ow rate, etc. is very important. By setting these various important parameters by trial may lead to more rejections
and thereby decreasing the production rate and increasing the cost. Selection of optimum process parameters will help to
avoid all these problems and this can be achieved by using advanced optimization algorithms.
In the literature it is observed that few researchers had carried out research on the continuous casting process. Zhou et al.
[26] used nondominated sorting genetic algorithm for multiobjective optimization of a continuous casting process. Two
objective functions were attempted in their work which involved maximization of the average value of the monomer conversion of the product and minimization of the length of the lm reactor. Seven decision variables were used in the study
which included the temperature, feed ow rate, the lm thickness, the monomer conversion at the output, and three coefcients describing the wall temperature used in the lm reactor. Effect of various decision variables was discussed and the
result in the form of Pareto set was presented. Cheung and Garcia [27] carried out optimization of the quality of a SAE 1010

292

Tensile strength (MPa)

290

288

286

284

282

280
101

101.5

102

102.5

103

103.5

Hardness (BHN)
Fig. 4. Pareto optimal points for combined objective function of squeeze casting process.

5598

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

steel billet by using heuristic search technique. A knowledge base was proposed in their work by interacting a numerical
heat transfer model and an articial intelligence heuristic search method. The work was focused on obtaining the optimized
cooling conditions for the continuous casting process which results in defect-free billet production.
Chakraborti et al. [28] used pareto-converging genetic algorithm for optimizing the casting velocity. Two different objectives were considered for casting velocity subjected to three constraints due to a number of physical considerations. The
work was restricted only to the mold region. Subsequently, Chakraborti et al. [29] extended their work on the spray and radiation cooling regions. In another work, Chakraborti et al. [30] used nite volume approach along with genetic algorithm to
solve the pertinent transport equations of a continuous caster mold and reported improvement in casting velocity and solidied shell thickness.
Santos et al. [31] used genetic algorithm and a knowledge base of operational parameters in order to develop a mathematical model and a computational algorithm to maximize the quality of steel billets produced by a continuous casting process. The optimization strategy was concerned with achieving a set of optimal cooling conditions in order to attain highest
product quality. In another work, Santos et al. [32] developed a numerical code based on genetic algorithm for an industrial
data of a continuous caster machine and determined optimum settings of water ow rates in different sprays zones.
Ghosh et al. [33] discussed the problems occurred in a continuous casting of a thin slab due to meniscus-level uctuation.
Genetic algorithm was used to nd the optimum combination of spray cooling. The total bulging was minimized and also the
surface temperature at the slab exit was also maintained minimum. Kulkarni and Babu [34] attempted the multiobjective
optimization problem for producing quality products in a continuous casting process. Simulated annealing algorithm was
used in their work to obtain the optimum process parameters in order to satisfy 17 critical quality conditions. Cheung
et al. [35] determined an improved cooling condition of mold and spray cooling zones of a continuously billet casting
machine and suggested suitable modication in the secondary cooling zone by proposing provision of two spray zones. A
heuristic search technique supported by a Knowledge Base and a heat ow mathematical model was used for determining
the results and the suggested modication resulted in a shorter metallurgical length and lower surface reheating compared
with the set of original spray zones.
Filipic et al. [36] used differential evolution for multi-objective optimization technique for optimizing the coolant ows in
continuous casting of steel in order to satisfy multiple objectives associated with temperature and core length. Miettinen
[37] described an interactive classication-based multiobjective optimization method: NIMBUS which converts the original
objective functions together with preference information coming from the decision maker into scalar-valued optimization
problems. A real-life problem related to continuous casting of steel was attempted in their work in order to minimize the
defects in the nal product.
Bhattacharya et al. [38] used genetic algorithm to obtain the optimum parameters of the mold oscillation system in the
continuous casting process. The important parameters considered were casting speed, stroke, frequency and deviation from
sinusoid. The main focus of the work was to maximize the lubrication index and minimize depth of oscillation marks and
friction. The comparison of results was made with the conditions described by the original equipment manufacturer. Subsequently, Bhattacharya and Sambasivam [39] used differential evolutionary algorithm for the same problem and compared
the results with those obtained by genetic algorithm.
Ye et al. [40] described an Engineering-Driven Rule-Based Algorithm (ERD) and various challenges related to the bleeds
detection were explained. An attempt was made to solve bleed detection problem using a real case study in which miss
detection rate and false alarm rate were optimized by considering the four important process variables related to the geometry and two important coefcients of the process. A full factorial three-level experimental design was used for determining
the process parameters to conduct the experiments. Lopez et al. [41] created a simulator for the continuous casting process
using a computational algorithm based on the numerical method. The results obtained by the simulator were validated for
the three different steel casters produced in an industry. Jabri et al. [42] used particle swarm optimization for tuning the
process parameters of a real plant in order to improve the bulging effect rejection.
It is observed from the literature that among the various optimization techniques available, simulated annealing (SA)
[34], genetic algorithm (GA) [38], differential evolutionary (DE) [39] and particle swarm optimization (PSO) [42] algorithms
were used by some researchers for the parameters optimization of continuous casting process. However, subsequently it is
proved by many researchers of different elds that the results given by SA, GA, DE, PSO, etc. are not always optimum. Thus it
can be concluded that there is a good scope for making use of recently developed advanced optimization techniques for
optimizing the continuous casting process parameters.Hence, in the present work, an attempt is made to optimize the
parameters of continuous casting process by using the TLBO algorithm.
4.1. Application example
An example of continuous casting process attempted by Kulkarni and Babu [34] is considered in the present work for the
process parameters optimization. 17 critical conditions were presented by Kulkarni and Babu [34] as an individual objective
and nally a combined total loss function was given which represents all the conditions. However, various equations given
by Kulkarni and Babu [34] involved certain process constants, and in some cases the required process constant values were
not mentioned in their work. Moreover, those required values cannot be assumed in order to make the comparisons of result.
Hence, in the present work, 10 critical conditions are considered out of 17 whose data is available and the entire result comparison is made with respect to these 10 conditions only. This approach will not affect the complete meaning of the process

5599

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

objectives as the individual conditions consist of respective mathematical models and those models are independent of the
conditions. Moreover, result of each condition was given by Kulkarni and Babu [34] and the comparison of results obtained in
the present work is strictly made for those conditions only. The conditions considered in the present work are conditions 1 to
9 and condition 13. The same equations of these conditions as given by Kulkarni and Babu [34] are considered and explained
below by Eqs. (4)(15).

Condition 1 :

Q 1 gV c ;

where g is the viscosity in poise and V c is the casting speed in m/min. In order to have a fair comparison of results, same
value of casting speed i.e. V c = 1.1 m/min is considered in this work as that suggested by Kulkarni and Babu [34] and the constraint for Q 1 was 1 6 Q 1 6 3.

Condition 2 :

Q 2 T sol =g0:0472 for crack sensitive grades;

Q 2 T sol =g0:072 for sticker sensitive grades;

where T sol is mold ux solidication temperature in C. Sticker sensitive grade is considered here and hence Q 2 is calculated
subsequently in this work by using Eq. (6). The constraint for condition 2 as given by Kulkarni and Babu [34] is considered as
1025 6 Q 2 6 1075.

Condition 3 :

Q 3 1:801  0:2461V c  0:044g  0:00107T sol ;

It is observed that the equation of Q 3 given by Kulkarni and Babu [34] was having slight error mainly related to the constant term. The same is rectied now and the constant term is changed from 1.952 to 1.801 and the correct equation of Q 3 is
given by Eq. (7) which now gives the exact value of L3 = 0.09 as reported by Kulkarni and Babu [34] with their parameters
settings and also satises the condition of 0.15 6 Q 3 6 0.45.

Condition 4 :

Q 4 0:70  2=s0:3 60=f gV c 2 


Q 4 0:70  2=s0:3 60=f gV c 2 

0:5

0:5

0:17 For 0:08 < %C < 0:16;

0:22 For 0:16 < %C < 0:08;

where Q 4 is related to the powder consumption in kg/m which is having a constraint of 0.15 6 Q 4 6 0.45 [34], s is the stroke
length in mm and f is the frequency in cpm. Even though it is not clear about which equation was used by Kulkarni and Babu
[34], but it can be seen that the Eq. (9) is misleading about its limit and also it violates the constraint limit of Q4 if the parameter limit given by Kulkarni and Babu [34] is substituted in it. Hence, Eq. (8) is considered in this work for condition 4 with
slight correction in it which satises the results of Kulkarni and Babu [34] also.

Condition 5 :

Q 5 V c =f ;

10

where, V c is the casting speed, as stated earlier, but it should be taken in mm/min (i.e. 1100 mm/min) and the constraint for
condition 5 is Q 5 6 25 mm [34].

Condition 6 :

0:5

Q 6 600 s=10  f :

11

In this case, the stroke length (s) should be taken in cm and frequency (f) should be taken in cycles per second as mentioned by Kulkarni and Babu [34]. The condition to be satised while parameter design is: Q 6 < 400 lm [34].

Condition 7 :

Q 7 V m =V c ;

12

where V m is the mold velocity (m/min) which was not given by Kulkarni and Babu [34]. However, it can be obtained by
attempting several equations related to mold velocity given by Kulkarni and Babu [34] and it is obtained as V m = 2.40 m/
min which satises the necessary constraints and their results. The constraint to be followed for condition 7 is Q 7 > 1.2 [34].

Condition 8 :

Q 8 Rp  q=V c volume=surface area;

13
3

where Rp is the pool drain rate in m/min, q is the liquid ux density in kg/m . The volume is in m and surface area is in m2.
The volume and surface area were not clearly given by Kulkarni and Babu [34], however, it is obtained by doing reverse calculations of appropriate equations related to condition 8 so as to get the same value as that of Kulkarni and Babu [34] in
order to have fair comparison of result. The constraint for condition 8 is 0.15 6 Q 8 6 0.45 [34].

Condition 9 :



0:25
Q 9 K  T surf =T sol Lm =V c g0:5 s0:25 f V c 0:25 =7:7 ;

14

where T surf is the strand surface temperature in C, T sol is the powder solidication temperature in C, Lm is the mold length in
meters, s is the stroke length in meters, f is the frequency in cpm, V c is the casting speed in m/min, g is the viscosity in poise
and K is a constant (K = 0.251). In this case also the strand surface temperature T surf was not given by Kulkarni and Babu [34].
However, it is obtained by doing reverse calculations of appropriate equations related to condition 9 so as to get same value
as that of Kulkarni and Babu [34]. The constraint for condition 9 is 0.15 6 Q 9 6 0.45 [34].

5600

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

Condition 13 :

0:3

Q 13 Lm =wide face dimension ;

15

where, Lm is the mold length in meters and wide face dimension (0.269) is in meter. The constraint for condition 13 as given
by Kulkarni and Babu [34] was Q 13 P 1.
All these conditions were having respective limits and constraints. Kulkarni and Babu [34] used quadratic loss functions
to derive the undesirability index for each condition. Finally the total undesirability index was obtained by them by summing all the individual loss function values and the optimum values for the process parameters was correspond to minimum
value for the total loss function. The individual loss functions given by Kulkarni and Babu [34] for these above mentioned 10
conditions are reproduced below by Eqs. (16)(25) and the total loss function is given by Eq. (26).

L1 Q 1  22 ;

16

L2 0:0016 Q 2  10502 ;

17

L3 44:5 Q 3  0:32 ;

18

L4 44:5 Q 4  0:32 ;

19

L5 0:0016 Q 5 2 ;

20

L6 6:25  106 Q 6 2 ;

21

L7 1:44 1=Q 7 2 ;

22

L8 44:5 Q 8  0:32 ;

23

L9 44:5 Q 9  0:32 ;

24

L13 1=Q 13 2 ;

25

Total loss function; Z L1 L2 L3 L4 L5 L6 L7 L8 L9 L13 :

26

The individual loss function values vary from 0 to 1 and if any loss functions value is larger, then the respective undesirability is higher and the objective is to minimize the undesirability. The loss function values obtained by Kulkarni and Babu
[34] for the above 10 conditions is given in Table 3. Simulated annealing algorithm was used by Kulkarni and Babu [34] to
obtain the optimum process parameters for the continuous casting process under consideration. Now the same models as
explained above are attempted by the TLBO algorithm to get improvement in the result. The total loss function is of minimization type subjected to those 10 conditions with their limits and constraints. The process variables involved in the equation consists of certain range and the same is used in the present work as given by Kulkarni and Babu [34]. As number of
conditions is to be considered simultaneously along with large number of input parameters, hence initially trials are carried
out by running the algorithm with different population sizes and number of generations to decide. Finally consistent results
are obtained by using the population size of 100 and the number of generations as 50. All the conditions are handled carefully and consistent results are obtained by TLBO algorithm which is given in Table 3 along with the comparison with the
results of simulated annealing.

Table 3
Results obtained by TLBO algorithm for the 10 loss functions under consideration.
Loss function

SA result [34]

L1
L2
L3
L4
L5
L6
L7
L8
L9
L13
Total loss function (Z)

0.24
0.80
0.09
0.12
0.19
0.15
0.30
0.67
0.80
0.87
4.23

Bold indicates the best optimum value.

TLBO result
Best

Average

Std. deviation

0.23
0.002
0.01
0.05
0.14
0.13
0.30
0.0004
0.77
0.91
2.54

2.57

0.057

5601

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

4.5

Total loss function value

3.5

2.5

2
0

10

20
30
Number of generations

40

50

Fig. 5. Variation of total loss function value with generations.

The TLBO algorithm has suggested the following parameter settings which give the optimum results. Viscosity, g = 1.38
poise, frequency, f = 116.64 cpm, stroke, s = 11.46 mm, ux solidus temperature, T sol = 1075.86 (C), drain rate, Rp = 2.11
mm/min, ux density, q = 2349.23 kg/m3 and mold length, Lm = 704.47 mm. All these parameter values are within the
respective limits as that given by Kulkarni and Babu [34]. In case of individual loss function L2 and L8, there is considerable
improvement in result and the undesirability has become almost zero from 0.8 and 0.67, respectively. Overall, almost in all
cases, the loss function is improved except in L13 where it is slightly increased. This is quite possible because a common
parameter setting is obtained in the present work which satises all the conditions simultaneously. However the L13 function
was also on higher side as given by Kulkarni and Babu [34]. The TLBO algorithm has given a signicant improvement in the
result and the total loss function is improved from 4.23 to 2.54 thereby achieving the improvement by above 60%. The average result of 50 runs obtained is 2.57 which is very nearer to the best result and the standard deviation obtained is 0.057. The
convergence of result with generations is shown in Fig. 5. The comparison between the results for SA and TLBO is clearly
made only for the 10 conditions which are considered in this work. Even though Kulkarni and Babu [34] had attempted their
work for total 17 conditions, but keeping in view the present improvement in results and success rate, it may be stated that
the TLBO algorithm can also handle all those 17 conditions simultaneously provided the required data is made available.

5. Parameters optimization of die casting process


Die casting is a versatile process for producing various engineering parts by forcing molten metal under high pressure into
reusable steel molds. Die cast parts are now becoming very popular and nding its wide application ranging from very complicated parts used in automobiles to simple toys, but the overall equipment and metal die cost is very high. Also the accuracy level required to set the proper input parameters in a die casting process is very high which may leads to high volume of
rejection. Hence proper setting of the input process parameters is very important in die casting process. There are large number of input process parameters involved in this process which can be categorized as die casting machine related parameters,
shot sleeve related parameters, die related parameters and cast metal related parameters [43]. Out of these, the few important parameters which can signicantly affect the process output are melt temperature, injection pressure, injection time,
die temperature, holding pressure, pressure holding time, etc. Any small deviation in the setting of these parameters will
affect the product quality. Hence these parameters must be decided by making use of advanced computational techniques.
Some researchers had attempted to study the effect of various input parameters on the process output and tried to optimize the situation. Syrcos [43] presented the effects of process parameters on the casting density of aluminum alloy casting.
Various inuential process parameters considered were piston velocity, metal temperature, lling time and hydraulic pressure. However, any advanced optimization technique was not involved in their work and the optimum parameter setting
presented was obtained by using Taguchis method. Yarlagadda and Chiang [44] used articial neural network to predict
the process parameters for the given conditions of a pressure die casting. An industrial data was taken and the effects of four
input parameters was discussed in their work namely injection time, injection pressure, melt temperature and die temperature. In another work, Yarlagadda [45] applied the articial neural network for the physical model of a die cast product of
zinc alloy.
Lin and Tai [46] used simulated annealing algorithm for the optimization of various aspects of die casting process such as
optimization of runner design, optimization of position for the injection gate was carried out by Tai and Lin [47] and similarly
optimal selection of gate location was carried out by Lin [48]. Krimpenis et al. [49] carried out the simulation of pressure die

5602

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

casting using ProCAST software and nite element approach. Neural network model was then developed and parameters
optimization was carried out using genetic algorithm. Tsoukalas [50] used the combination of multivariable linear regression
and genetic algorithm for minimizing the porosity level of aluminum alloy die casting. Taguchis L27 orthogonal array was
used and experiments were conducted by considering ve input parameters namely holding furnace temperature, die temperature, plunger velocities in the rst and second stage and multiplied pressure. Optimized parameter setting obtained by
using genetic algorithm was produced in their work, however, it is observed that the computational efforts involved in their
work to obtain the optimum parameter setting was comparatively very high. Verrana et al. [51] used the concept of design of
experiments for analyzing the inuence of injecting parameters on the internal quality of die cast part. The important injection parameters considered were slow shot, fast shot and upset pressure and nally studied the effect of these parameters on
the casting density.
Kong et al. [52] studied the die temperature prole using infrared thermograph technology and attempt was made to
optimize the internal cooling system in order to provide even cooling to the components and the die. Commercial computational uid dynamics code was used in their work for optimizing and redesigning the internal cooling system. Wong and
Pao [53] emphasized various losses and difculties faced by foundries due to dependency on skilled foundry men and several
trial attempts carried out by them to achieve appropriate heat transfer and directional solidication in die casting. Use of
evolutionary algorithms in general and genetic algorithm in particular was highlighted in their work for optimizing the conditions of die casting process. Campatelli and Scippa [54] presented a case study of an automotive component, produced by
high pressure die casting, and optimized the tolerance level of the die geometry. However only the concept of nite element
simulation was involved in their work and any type of heuristic algorithm was not used by them for the optimization purpose. Zhang and Wang [55] used combination of articial neural network and genetic algorithm for optimizing the process
parameters of low pressure die casting in order to improve the quality of product. However, as any experiments were not
conducted and only the software simulated data was used, hence the level of relative error was very high.
Even though several advanced optimization techniques were available in the literature, but only genetic algorithm and
simulated annealing were used by few researchers in the past for the parameters optimization of die casting process. Hence
attempt is made here to use the TLBO algorithm to achieve the optimum solution for the die casting process. An application
example is considered for this purpose from the literature which is described below in detail along with the result.
5.1. Application example
This example is taken from the literature which is based on the research work of Tsoukalas [50]. In this work, Tsoukalas
[50] had attempted to minimize the porosity level of an aluminum alloy component produced by pressure die casting process. A combined approach of multivariable linear regression and genetic algorithm was used in their work to determine the
optimum condition for the situation under consideration. The input variables considered by Tsoukalas [50] were holding furnace temperature F (C), die temperature D (C), plunger velocity in the rst stage S (m/s), plunger velocity in the second
stage H (m/s) and multiplied pressure M (bar). L27 orthogonal array of Taguchi method was used in their work and experiments were conducted as per the experimental layout. Subsequently the relation between all the input parameters with the
response i.e. porosity was given in the form of a mathematical model. The same mathematical model as used by Tsoukalas
[50] is given below by Eq. (27).

Porosity; P 1:623  0:766  103 F  1:301  103 D  0:136S 0:029H  1:636  103 M:

27

The parameters involved in Eq. (31) were having certain ranges and the same range as used by Tsoukalas [50] is used in
this work and is given below by Eqs. (28)(32).

Holding furnace temperature F 610730  C;

28

Die temperature D 190270  C;

29

Plunger velocity in the first stage S 0:020:34 m=s;

30

Plunger velocity in the second stage H 1:23:8 m=s;

31

Multiplied pressure M 120280 bar:

32

Tsoukalas [50] used combination of multivariable linear regression and genetic algorithm to optimize the process parameters for the given conditions and compared his result with the experimental result. The result obtained by genetic algorithm
in their work had shown improvement over the experimental result. However, the numbers of generations taken by genetic
algorithm were 1000 and the population size was not mentioned. Also the required GA operator setting was also not given in
their work such as crossover rate, mutation rate, etc. Thus, it indicates that computational efforts taken by Tsoukalas [50] to
achieve the optimum parameters setting for the problem under consideration was comparatively more. Hence attempts are
carried out in the present work to nd out the optimum solution of the problem with minimum computational efforts.
The TLBO algorithm is now used for the parameters optimization of the same model of Tsoukalas [50] along with same

5603

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608


Table 4
Result obtained by TLBO algorithm and its comparison with previous results.
Parameters

GA result [50]

Holding furnace temperature F, (C)


Die temperature D, (C)
Plunger velocity in the rst stage S, (m/s)
Plunger velocity in the second stage H, (m/s)
Multiplied pressure M, (Bar)
Number of generation
Minimum porosity (%)

729.4
269.9
0.336
1.2
275.7
1000
0.251

TLBO result
Best

Average

Std. deviation

730
270
0.34
1.2
280
10
0.243

0.2496

0.02

Bold indicates the best optimum value.

0.45

Porosity (%)

0.4

0.35

0.3

0.25

0.2
1

4
5
6
7
Number of generations

10

Fig. 6. Variation of porosity with generations.

parameters range. The above mentioned mathematical model of porosity was attempted by TLBO algorithm by using a very
small population size of 10 and the number of generations used is 10. The results obtained by TLBO algorithm are given in
Table 4 along with the comparison with the results of genetic algorithm.
The same example is tried with the TLBO algorithm several times to check for any further improvement in the result and
it is observed that the result given in Table 4 is the global optimum solution for the problem under consideration. Even
though the number of generations used to run the TLBO algorithm is 10, but the global optimum solution of 0.243% is
obtained in fourth generation itself and the consistent result is observed in further generations. The algorithm is allowed
to run 50 times in order to check the average result and it is reported as 0.2496 which is almost nearer to the best result.
The standard deviation of 50 runs obtained is 0.02 which is also an indication of good and consistent result. The convergence
of result obtained by TLBO algorithm is shown in Fig. 6. In the case of results given by Tsoukalas [50], the number of generations used was 1000 and the result of porosity = 0.251 was converged only after 800 generations as shown by the graphical representation in their work. Also the parameter setting given by Tsoukalas [50] is very difcult to set on the machine,
whereas the TLBO algorithm has given such a setting which can be easily adjusted on the machine. Hence it is clear from this
work that the computational effort taken by TLBO algorithm is very negligible compared to genetic algorithm and proved its
capabilities in the eld of parameters optimization of die casting process. The TLBO algorithm has also given the believed to
be global optimum solution in this case which can be treated as success by the foundry personnel to make use of this
advanced algorithm to obtain the accurate optimum setting for the die casting process with less computational efforts.

6. Conclusions
Mathematical models in terms of inputoutput process parameters of squeeze casting, continuous casting and die casting
processes are optimized in this work using the recently developed TLBO algorithm. The processes identied are getting
widely used in various industries for producing very complicated parts of small and big size with variety of materials. These
processes involve various input parameters which affects the output of respective process. The effectiveness of these processes depends upon the relations of process parameters with each other and the similar relations in the form of various
mathematical models are used in the present work. The detailed literature survey has proved that there is a good scope
for the use of advanced optimization techniques like TLBO in the eld of parameters optimization of these casting processes.
Multi-objective mathematical model is considered in the case of squeeze casting process and initially the objectives are

5604

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

attempted individually where the hardness and tensile strength is improved considerably. A common parameter setting is
also obtained for satisfying both these objectives simultaneously. For continuous casting process, a problem having 10 individual loss functions and a total loss function are considered which was earlier attempted by using simulated annealing algorithm. Even though the original problem of continuous casting process was having 17 loss functions but only 10 loss
functions are considered in the present work whose complete information is available. Fair comparison of these 10 loss functions with the results of the previous researchers is justied by considering the 10 complete models of the respective loss
functions. The TLBO algorithm has effectively handled the problem and has given signicant improvement of above 60%
in total loss function compared to the previous result. In the case of two loss functions, the undesirability index is almost
reduced to zero.
For die casting process, the model under consideration is having ve input parameters and it was earlier attempted by
using genetic algorithm using 1000 generations. Whereas, in the present work, the same model is satisfactorily attempted
with only 10 generations, thereby drastically reducing the computational efforts. Moreover, the solution obtained is also
believed to be a global optimum solution for the die casting process. Thus, the TLBO algorithm have effectively handled
the various mathematical models and proved its capabilities in the eld of parameters optimization of casting processes with
less computational efforts. The algorithm can be attempted on other types of casting processes also.
Appendix A. Code of TLBO algorithm for the parameters optimization of selected casting processes
To run the TLBO code, user has to create separate MATLAB les for each function (i.e., separate .m le for main, mainline,
tlbo and fun) and then the main.m le is to be executed.
%%%%%%%%%%%%%%%% Code for Hardness problem %%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%% main %%%%%%%%%%%%%%%%%%%%%
clear all
clc
no_of_run = 50;
no_of_student = 10;
%specify population size
no_of_iteration = 20;
%specify number of iterations
tf = 1;
for i = 1:no_of_run
[bvf bvx]=mainline(no_of_student, no_of_iteration, tf);
bvf1(i,:)=bvf;
bvx1(i,:)=bvx;
end
bvf = bvf1(:,1);
[bvfmin,k0]=max(bvf)
bvxmin = bvx1(k0,:)
bvfmin = bvf1(k0,:)
%%%%%%%%%%%%%%%%%%% mainline %%%%%%%%%%%%%%%%%%%%%
function [bvf bvx]=mainline(no_of_student, no_of_iteration, tf)
no_of_variable = 3;
% specify number of variables
lowerlimitofa = 50;
% specify the lower bound
upperlimitofa = 125;
% specify the upper bound
lowerlimitofb = 150;
upperlimitofb = 300;
lowerlimitofc = 15;
upperlimitofc = 60;
for i = 1:no_of_student
% initialization of the variables
a = lowerlimitofa + rand(upperlimitofa-lowerlimitofa);
b = lowerlimitofb + rand(upperlimitofb-lowerlimitofb);
c = lowerlimitofc + rand(upperlimitofc-lowerlimitofc);
x(i,:)=[a b c];
end
% limit array
limit=[lowerlimitofa;
upperlimitofa;
lowerlimitofb;

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

upperlimitofb;
lowerlimitofc;
upperlimitofc];
parameter=[no_of_student;
no_of_iteration;
tf;
no_of_variable];
[bvf,bvx]=tlbo(limit,x,parameter);
%%%%%%%%%%%%%%%%%%%%% tlbo %%%%%%%%%%%%%%%%%%%%
function [bvf,bvx]=tlbo(limit,x,parameter)
no_of_student = parameter(1);
no_of_iteration = parameter(2);
tf = parameter(3);
no_of_variable = parameter(4);
for i = 1:no_of_student
variable = x(i,:);
[funxz]=fun(variable);
fxx(i,1)=funxz;
end
for ng1 = 1:no_of_iteration;
m = mean(x);
[sfx,k0]=max(fxx);
bt = x(k0,:);
for i = 1:no_of_student
k = 1;
for j = 1:no_of_variable
xs(i,j)=x(i,j)+rand(bt(1,j)-tfm(1,j));
if xs(i,j)<limit(k) || xs(i,j)>limit(k + 1)
x1(i,j)=x(i,j);
else
x1(i,j)=xs(i,j);
end
k = j + 2;
end
variable = x1(i,:);
end
for i = 1:no_of_student
variable = x1(i,:);
[funxz]=fun(variable);
fxx1(i,1)=funxz;
end
for i = 1:no_of_student
if fxx1(i,1)<fxx(i,1)
fxx1(i,:)=fxx(i,:);
x1(i,:)=x(i,:);
end
end
[sfx1,k1]=max(fxx1);
bs = x1(k1,:);
for i = 1:no_of_student
k = 1;
for j = 1:no_of_variable
xs(i,j)=x1(i,j)+rand(bs(1,j)-x1(i,j));
if xs(i,j)<limit(k) || xs(i,j)>limit(k + 1)
x(i,j)=x1(i,j);
else
x(i,j)=xs(i,j);
end
k = j + 2;

5605

5606

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

end
variable = x(i,:);
end
for i = 1:no_of_student
variable = x(i,:);
[funxz]=fun(variable);
fxx(i,1)=funxz;
end
for i = 1:no_of_student
if fxx(i,1)<fxx1(i,1)
fxx(i,:)=fxx1(i,:);
x(i,:)=x1(i,:);
end
end
end
[bvff,k2]=max(fxx);
bvf(1,:)=fxx(k2,:);
bvx = x(k2,:);
%%%%%%%%%%%%%%%%%%%% fun %%%%%%%%%%%%%%%%%%%%
function funxz = fun(variable)
no_of_variable = 3;
funxz = 3.82542 + 0.8787variable(1) + 0.46587variable(2) + 0.30411variable(3)  0.00393variable(1)variable(1)
 0.00116variable(2)variable(2) + 0.00097variable(3)variable(3) + 0.00051variable(1)variable(2)  0.00333
variable(1)variable(3)  0.00018variable(2)variable(3);;
end
%%%%%%%%%%%%%% Code for Tensile strength problem%%%%%%%%%%%%%%
The code is similar to that given above for the Hardness problem. The les of tlbo and main remain the same. However,
the fun le is to be replaced by the following le and also make the appropriate changes in mainline le by specifying variables and the bounds.
%%%%%%%%%%%%%%%%%%%% fun %%%%%%%%%%%%%%%%%%%%%%
function funxz = fun(variable)
no_of_variable = 3;
funxz = 11.2606 + 2.5778variable(1) + 1.3316variable(2) + 0.7552variable(3)  0.0116variable(1)variable(1)
 0.0034variable(2)variable(2) + 0.0031variable(3)variable(3) + 0.0015variable(1)variable(2)  0.0097variable(1)variable(3)  0.001variable(2)variable(3);
end
%%%%%%%%%%%%%%%%% Code for total loss function%%%%%%%%%%%%%%%
The code is similar to that given above for the Hardness problem. The le main remain the same. However, the fun le is
to be replaced by the following le and also make the appropriate changes in the tlbo le for minimization, and mainline le
by specifying variables and the bounds.
%%%%%%%%%%%%%%%%%%% fun %%%%%%%%%%%%%%%%%%%%
function funxz = fun(variable)
no_of_variable = 10;
L1 = (variable(1)-2)^2;
L2 = 0.0016 (variable(2)1050)^2;
L3 = 44.5 (variable(3)0.3)^2;
L4 = 44.5 (variable(4)0.3)^2;
L5 = 0.0016 (variable(5))^2;
L6 = 6.2510^6 (variable(6))^2;
L7 = 1.44 (1/variable(7))^2;
L8 = 44.5 (variable(8)0.3)^2;
L9 = 44.5 (variable(9)0.3)^2;
L13 = (1/variable(10))^2;
funxz = L1 + L2 + L3 + L4 + L5 + L6 + L7 + L8 + L9 + L13;

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

5607

%%%%%%%%%%%%%%%%%%% Code for porosity %%%%%%%%%%%%%%%%


The code is similar to that given above for the Hardness problem. The le main remains the same. However, the fun le is
to be replaced by the following le and also make the appropriate changes in the tlbo le for minimization, and mainline le
by specifying variables and the bounds.
%%%%%%%%%%%%%%%%%%%%%% fun %%%%%%%%%%%%%%%%%%%%
function funxz = fun(variable)
no_of_variable = 5;
funxz = 1.6230.76610^3variable(1)  1.30110^3variable(2)  0.136variable(3) + 0.029variable(4) 
1.63610^3variable(5);
end

References
[1] R.V. Rao, V.J. Savsani, D.P. Vakharia, Teachinglearning-based optimization: a novel method for constrained mechanical design optimization problems,
Comput. Aided Des. 43 (2011) 303315.
[2] R.V. Rao, V.J. Savsani, D.P. Vakharia, Teachinglearning-based optimization: an optimization method for continuous non-linear large scale problems,
Inf. Sci. 183 (2012) 115.
[3] R.V. Rao, V. Patel, An elitist teachinglearning-based optimization algorithm for solving complex constrained optimization problems, Int. J. Ind. Eng.
Comput. 3 (4) (2012) 535560.
[4] T. Niknam, A.K. Fard, A. Baziar, Multi-objective stochastic distribution feeder reconguration problem considering hydrogen and thermal energy
production by fuel cell power plants, Energy 42 (2012) 563573.
[5] T. Niknam, R.A. Abarghooee, M.R. Narimani, An efcient scenario-based stochastic programming framework for multi-objective optimal micro-grid
operation, Appl. Energy 99 (2012) 455470.
[6] T. Niknam, R.A. Abarghooee, M.R. Narimani, A new multi objective optimization approach based on TLBO for location of automatic voltage regulators in
distribution systems, Eng. Appl. Artif. Intell. 25 (2012) 15771588.
[7] T. Niknam, F. Golestaneh, M.S. Sadeghi, H-multiobjective teachinglearning-based optimization for dynamic economic emission dispatch, IEEE Syst. J.
6 (2) (2012) 341352.
[8] V. Togan, Design of planar steel frames using teachinglearning based optimization, Eng. Struct. 34 (2012) 225232.
[9] K.R. Krishnanand, B.K. Panigrahi, P.K. Rout, A. Mohapatra, Application of multi-objective teaching-learning-based algorithm to an economic load
dispatch problem with incommensurable objectives, in: SEMCCO (2011) Part I, LNCS, 7076, 2011, pp. 697705.
[10] R.V. Rao, V.D. Kalyankar, Parameter optimization of machining processes using a new optimization algorithm, Mater. Manuf. Processes 27 (9) (2012)
978985.
[11] R.V. Rao, V.D. Kalyankar, Multi-objective multi-parameter optimization of the industrial LBW process using a new optimization algorithm, in:
Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 226(6), 2012, pp. 10181025.
[12] R.V. Rao, V.D. Kalyankar, Parameter optimization of modern machining processes using teachinglearning-based optimization algorithm, Eng. Appl.
Artif. Intell. 26 (2013) 524531.
[13] R.V. Rao, V. Patel, Multi-objective optimization of heat exchangers using a modied teachinglearning-based optimization algorithm, Appl. Math.
Model. 37 (2013) 10471062.
[14] R.V. Rao, V. Patel, Multi-objective optimization of two stage thermoelectric cooler using a modied teachinglearning-based optimization algorithm,
Eng. Appl. Artif. Intell. 26 (2013) 430445.
[15] M. Crepinek, S-H. Liu, L. Mernik, A note on teachinglearning-based optimization algorithm, Inf. Sci. 212 (2012) 7993.
[16] H. Hu, Squeeze casting of magnesium alloys and their composites, J. Mater. Sci. 33 (1998) 15791589.
[17] S.W. Kim, G. Durrant, J.H. Lee, B. Cantor, The microstructure of direct squeeze cast and gravity die cast 7050 (Al-6.2Zn-2.3Cu-2.3Mg) wrought Al alloy, J.
Mater. Synth. Process. 6 (2) (1998) 7587.
[18] S.W. Kim, G. Durrant, J.H. Lee, B. Cantor, The effect of die geometry on the microstructure of indirect squeeze cast and gravity die cast 7050 (Al-6.2Zn2.3Cu-2.3Mg) wrought Al alloy, J. Mater. Sci. 34 (1999) 18731883.
[19] M. Zhou, H. Hu, N. Li, J. Lo, Microstructure and tensile properties of squeeze cast magnesium alloy AM50, J. Mater. Eng. Perform. 14 (2005) 539545.
[20] J.K. Baek, H.W. Kwon, Effect of squeeze cast process parameters on uidity of hypereutectic AlSi alloy, J. Mech. Sci. Technol. 24 (2008) 711.
[21] Y. Yang, L. Peng, P. Fu, B. Hu, W. Ding, B. Yu, Effects of process parameters on the macrostructure of a squeeze cast Mg-2.5 mass% Nd alloy, Mater. Trans.
50 (12) (2009) 28202825.
[22] L.I. Run-xia, L.I. Rong-de, B.A.I. Yan-hua, Effect of specic pressure on microstructure and mechanical properties of squeeze casting ZA27 alloy, Trans.
Nonferrous Met. Soc. China 20 (2010) 5963.
[23] A.A. Moosa, K.K. Al-Khazraji, O.S. Muhammed, Tensile strength of squeeze cast carbon bers reinforced AlSi matrix composites, J. Miner. Mater.
Charact. Eng. 10 (2) (2011) 127141.
[24] Q. Zhang, M. Masoumi, H. Hu, Inuence of applied pressure on tensile behaviour and microstructure of squeeze cast Mg alloy AM50 with Ca addition, J.
Mater. Eng. Perform. 21 (2012) 3846.
[25] P. Senthil, K.S. Amirthagadeswaran, Optimization of squeeze casting parameters for non symmetrical AC2A aluminium alloy castings through Taguchi
method, J. Mech. Sci. Technol. 26 (4) (2012) 11411147.
[26] F. Zhou, S.K. Gupta, A.K. Ray, Multiobjective optimization of the continuous casting process for poly (methyl methacrylate) using adapted genetic
algorithm, J. Appl. Polym. Sci. 78 (2000) 14391458.
[27] N. Cheung, A. Garcia, The use of a heuristic search technique for the optimization of quality of steel billets produced by continuous casting, Eng. Appl.
Artif. Intell. 14 (2001) 229238.
[28] N. Chakraborti, R. Kumar, D. Jain, A study of the continuous casting mold using a pareto-converging genetic algorithm, Appl. Math. Model. 25 (2001)
287297.
[29] N. Chakraborti, R.S.P. Gupta, T.K. Tiwari, Optimisation of continuous casting process using genetic algorithms: studies of spray and radiation cooling
regions, Ironmaking Steelmaking 30 (4) (2003) 273278.
[30] N. Chakraborti, K.S. Kumar, G.G. Roy, A heat transfer study of the continuous caster mold using a nite volume approach coupled with genetic
algorithms, J. Mater. Eng. Perform. 12 (2003) 430435.
[31] C.A. Santos, J.A. Spim, A. Garcia, Mathematical modelling and optimization strategies (genetic algorithm and knowledge base) applied to the
continuous casting of steel, Eng. Appl. Artif. Intell. 16 (2003) 511527.
[32] C.A. Santos, N. Cheung, A. Garcia, Application of a solidication mathematical model and a genetic algorithm in the optimization of strand thermal
prole along the continuous casting of steel, Mater. Manuf. Processes 20 (2005) 421434.

5608

R.V. Rao et al. / Applied Mathematical Modelling 38 (2014) 55925608

[33] S. Ghosh, K. Mitra, B. Basu, Y.A. Jategaonkar, Control of meniscus-level uctuation by optimization of spray cooling in an industrial thin slab casting
machine using a genetic algorithm, Mater. Manuf. Processes 19 (3) (2004) 549562.
[34] M.S. Kulkarni, A.S. Babu, Managing quality in continuous casting process using product quality model and simulated annealing, J. Mater. Process.
Technol. 166 (2005) 294306.
[35] N. Cheung, C.A. Santos, J.A. Spim, A. Garcia, Application of a heuristic search technique for the improvement of spray zones cooling conditions in
continuously cast steel billets, Appl. Math. Model. 30 (2006) 104115.
[36] B. Filipic, T. Tusar, E. Laitinen, Preliminary numerical experiments in multiobjective optimization of a metallurgical production process, Informatica 31
(2007) 233240.
[37] K. Miettinen, Using interactive multiobjective optimization in continuous casting of steel, Mater. Manuf. Processes 22 (2007) 585593.
[38] A.K. Bhattacharya, D. Sambasivam, A. Roychowdhury, J. Das, Optimization of continuous casting mould oscillation parameters in steel manufacturing
process using genetic algorithms, IEEE Congr. Evol. Comput. (2007) 39984004.
[39] A.K. Bhattacharya, D. Sambasivam, Optimization of oscillation parameters in continuous casting process of steel manufacturing: genetic algorithms
versus differential evolution, Evol. Comput. (2009) 77102.
[40] E.P.L. Ye, J. Shi, T.S. Chang, On-line bleeds detection in continuous casting processes using engineering-driven rule-based algorithm, J. Manuf. Sci. Eng.
131 (2009) 610081610089.
[41] A.R. Lopez, G.S. Cortes, M.P. Pardave, M.A.R. Romo, R.A. Lopez, Computational algorithms to simulate the steel continuous casting, Int. J. Miner. Metall.
Mater. 17 (5) (2010) 596607.
[42] K. Jabri, D. Dumur, E. Godoy, A. Mouchette, B. Bele, Particle swarm optimization based tuning of a modied smith predictor for mould level control in
continuous casting, J. Process Control 21 (2011) 263270.
[43] G.P. Syrcos, Die casting process optimization using Taguchi method, J. Mater. Process. Technol. 135 (2003) 6874.
[44] P.K.D.V. Yarlagadda, E.C.W. Chiang, A neural network system for the prediction of process parameters in pressure die casting, J. Mater. Process. Technol.
8990 (1999) 583590.
[45] P.K.D.V. Yarlagadda, Prediction of die casting process parameters by using an articial neural network model for zinc alloys, Int. J. Prod. Res. 38 (2000)
119139.
[46] J.C. Lin, C.C. Tai, The runner optimisation design of a die-casting die and the part produced, Int. J. Adv. Manuf. Technol. 14 (1998) 133145.
[47] C.C. Tai, J.C. Lin, The optimal position for the injection gate of a die-casting die, J. Mater. Process. Technol. 86 (1999) 87100.
[48] J.C. Lin, Selection of the optimal gate location for a die-casting die with a freeform surface, Int. J. Adv. Manuf. Technol. 19 (2002) 278284.
[49] A. Krimpenis, P.G. Benardos, G.C. Vosniakos, A. Koukouvitaki, Simulation-based selection of optimum pressure die-casting process parameters using
neural nets and genetic algorithms, Int. J. Adv. Manuf. Technol. 27 (2006) 509517.
[50] V.D. Tsoukalas, Optimization of porosity formation in AlSi9Cu3 pressure die castings using genetic algorithm analysis, Mater. Des. 29 (2008) 2027
2033.
[51] G.O. Verrana, R.P.K. Mendes, L.V.O.D. Valentina, DOE applied to optimization of aluminium alloy die castings, J. Mater. Process. Technol. 200 (2008)
120125.
[52] L.X. Kong, F.H. She, W.M. Gao, S. Nahavandi, P.D. Hodgson, Integrated optimization system for high pressure die casting processes, J. Mater. Process.
Technol. 201 (2008) 629634.
[53] M.L.D. Wong, W.K.S. Pao, A genetic algorithm for optimizing gravity die castings heat transfer coefcients, Expert Syst. Appl. 38 (2011) 70767080.
[54] G. Campatelli, A. Scippa, A heuristic approach to meet geometric tolerance in high pressure die casting, Simul. Model. Pract. Theory 22 (2012) 109122.
[55] L. Zhang, R. Wang, An intelligent system for low-pressure die-cast process parameters optimization, Int. J. Adv. Manuf. Technol. 65 (2013) 517524.

Potrebbero piacerti anche