You are on page 1of 10

ITERATIVE STANDARD STRATEGY FOR NON-LINEAR OPTIMIZATION IN MELTING

FURNACE CHARGE CALCULATIONS

Amin Jafari-Ramiani , Dorsa Yousefi and Mahdiyeh Mohajeri


Faculty of Materials and Metallurgical Engineering, Amirkabir University of Technology (Tehran Polytechnic), No. 424
Hafez St, Tehran 15916-34311, Iran

Copyright Ó 2021 American Foundry Society


https://doi.org/10.1007/s40962-021-00699-0

Abstract
This paper describes an approach to furnace charge cal- were introduced to the model, with different amounts of
culation for melting processes, to achieve a specified target non-metallic contaminations and different ratios of element
melt in crucible or induction furnaces with minimum loss in each. Also, 7-ton out-of-range initial melt was
material cost. There has been a challenge with the problem considered in the furnace to be corrected. The matrix of
regarding material loss, contaminations, and unwanted coefficients was build according to the numerical algorithm
physiochemical reactions during the process. Those put a of the model. Optimizations were successfully performed in
high degree of complexity, non-linearity, and uncertainty 3–5 iterations with Excel solver. For the test case, the
on the calculations. The current study presents a model calculations showed the optimum mass fractions of charge
that takes three important complexities of the problem into burdens and predicted to give 9649 kg melt with * 262 kg
account, including non-homogeneous element loss during of total materials loss. An optimality analysis was con-
melting, non-metal contaminations in scraps/charges, and ducted and showed that the solution has reached the
correction of possible initial melt in the furnace. The model minimum possible cost. The non-linear iterative algorithm
was based on the mass balance of chemical elements along revealed a convergent and fast performance. This feature
with the optimization of weights of charge materials. It provides a backbone for reliable and fast optimization in
presents a re-arrangement of non-linear mass balance melting operations which is of significant benefit for
formulations into an iterative standard linear-program- industrial automation.
ming framework. To evaluate the performance of the
model, an industrial-scale test case was introduced. The Keywords: furnace charge calculation, optimization,
test problem was to find an optimum combination of charge melting, non-linear programming, material loss, mass
materials to achieve target brass alloy C47940 in a 10-ton balance
induction furnace. Eight different types of charge material

contaminations, refractory, slag, gas/dust, energy/fuel/ex-


penses, and a target melt, finally. Melting is widely con-
Introduction sidered to be responsible for a couple of critical issues.
First, melting operations are highly energy-intensive, tak-
Melting is not only changing a solid to a liquid. In almost ing more than 55% of the total energy consumed in the
all of the basic metal productions and foundry processes, metallurgical sector of the global industry. (The metallur-
melting is a major process operation that deals with several gical sector itself makes about 20% of the global indus-
other phenomena, e.g., evaporations, oxidations, chemical try.)1–3 Second, melting is the origin of the formation of the
reactions, etc., alongside solid–liquid transformation. The chemical composition of metals which directly affects the
process also involves several parameters and a variety of quality of final products. A proper computer model for
input/output objects, e.g., metal scraps, ingots, melting furnace charge calculation would be a solution to
both of the issues. Despite this crucial importance, melting
has been poorly appreciated and received inadequate
Received: 26 July 2021 / Accepted: 15 September 2021

International Journal of Metalcasting


mathematical modeling investigations compared to the conclusion that oxidation and evaporation of metals are
other industrial operations. mostly responsible for the loss of alloying elements.15 The
rate of evaporation of an element depends on the vapor
Infrequent attempts have been reported with the purpose of pressure of the element and the temperature of the melt.
developing proper computer models for melting processes. The situation can be critical for metals with a low melting
Kungurtsev et al.4 worked on a software solution for point like zinc and magnesium16 in a high melting point
automation of weighing of materials to be charged into alloy like brass. A great portion of the material losses is
melting furnaces. Stankevich et al.5 conducted mathemat- believed to be the consequence of the oxidations and dross
ical modeling for heat and mass transfer analysis of melting generation. Some alloying elements, which have a high
of scraps in an electric arc furnace. Seidu and Onigba- relative thermodynamic potential for oxidation (e.g., Mg,
jumo,6 and Seidu et al.7 worked on an operational com- Al, Ti, Si, etc., according to Ellingham diagram),17 exhibit
puter program for melting in induction furnaces. Crucible more loss by oxidation. As an example, Omole and
and induction furnaces are often considered as non-reacting Oluyori18 studied carbon and silicon losses in cast iron
systems in which smelting is not the primary goal of the melting in a rotary furnace. The rate of oxidation of an
process. Their charge calculations models are commonly element depends on the partial pressure of oxygen in the
based on a mass balance of chemical elements without any system, and the temperature and surface area of the metal.
important chemical reaction8. For these types of furnaces, a The other important mechanism of materials loss is mixing/
set of mass balance equations is raised with each equation absorption by slag and refractory materials. It is not yet
tracking a main alloying element of the target melt9. The precisely predictable that how much loss occurs for each
solution of the model is expected to give a set of weights alloying element in a melting operation. Nevertheless, the
for all charge burdens by which the composition of the final issue of material loss is practically managed with the def-
melt would fall inside the specified range according to the inition of experimental coefficients/fractions of loss.19
grade of the alloy. The models are also required to mini- Those loss coefficients are statistically estimated for dif-
mize the cost of melting, i.e., it should find an optimum ferent elements, materials, and furnaces, and they are so
combination of the weights of charge materials that mini- likely to be dependent on the liquid metal temperature.
mizes an objective cost function. Such a charge calculation
model is recognized as a standard linear programming (LP) Other issues have also been addressed in melting furnace
problem10. Solving a standard LP problem is straightfor- charge calculations. Ziółkowski20,21 has pointed out the
ward by, for instance, a MATLAB code, Excel solver, or a uncertainties in the chemical compositions of input mate-
piece of code with the SIMPLEX algorithm. However, the rials. He proposed a fuzzy optimization model to tackle the
mass balance models are facing some serious uncertainties; deviations in chemical composition. Ziółkowski22 also
(1) the melting operation is entangled with unwanted indicated that furnace charges are sometimes lump-weight
physiochemical reactions such as oxidations, evaporations, parts that could be analyzed through integer optimization.
melt-slag, and melt-refractory interactions11. Those reac- Capuzzi and Timelli23 provided a review of the process of
tions cause sensible amounts of material losses; (2) it is melting aluminum scraps. They pointed out the issue of the
usually supposed that the chemical compositions of the accumulation of undesired elements in metal recycling and
charge materials are known as input parameters before the discussed possible technological opportunities to redress
calculations are started. Contrary to expectations, there is the issue. Li et al.24 investigated the kinetics of melting of
always a significant uncertainty in the composition of input solid scrap in the liquid metal. As a solution for the cor-
charge materials, since the charge materials, especially rection of the initial out-of-range melt, Ziółkowski25 sug-
metal scraps,12 are contaminated and recycled from dif- gested setting a second-order compromise objective
ferent sources.13 (3) Doing a charge calculation would be function to minimize the total mass of the final melt as well
straightforward with an empty furnace and a predefined as minimizing the cost function.
mass of the target melt as an input parameter. However,
practical cases sometimes concern a furnace that holds an Collecting all those ambiguities into consideration will
initial amount of melt with a composition that falls out of make a charge calculation model complex and highly non-
the specified range. The correction of an initial melt linear. That will fall out of the standard LP problems and
introduces other unknowns to the problem, e.g., the final would be impractical for industrial applications. Ignoring
weight of the melt will not be known anymore. It also puts the uncertainties, on the other hands, gives over-simplistic
uncertainty on the feasibility of the solution. calculations which lead to multiple melt corrections,
extensions of melting time, and further increases in the
Perhaps, the loss of alloying elements has been the first cost. In this context, the present study tries to re-examine
issue in the mass balance of melting since early investi- the formulation of the problem. It aims to present a mod-
gations were conducted on the process. Bassett14 had ified modeling approach that takes the materials loss and
observed that there was a considerable amount of mismatch the initial melt into account, and simultaneously, preserves
between the input and output zinc content of copper alloys the model in the reliable framework of standard linear
during the melting operation. It was turned out to the programming. Therefore, the main question of the paper is

International Journal of Metalcasting


how to carry out the non-linear complexities of the problem
into a linear framework. The following section of the paper
begins with problem identification. The concept of the
modeling approach is described with formulations. A new
algorithm is presented which integrates the model formu-
lations and completes the answer to the main question
raised by this study. The next following section provides a
test case demonstrating the implementation and calcula-
tions of the model. Then, the results are shown, evaluated,
and discussed. The conclusion is drawn in the final section.

Materials and Method

Identification of the Problem

An induction furnace is considered for melting operation


with a capacity of mF regarding its maximum allowable
mass content (in kg or ton). The geometry and dimensions Figure 1. Schematic representation of an induction
of the furnace can be different depending on the process furnace with input, output, and materials loss.
requirements. This research is focused on non-reacting
systems, i.e., crucible or induction furnaces mainly, where (1) The melting equipment is considered to be an
smelting is not the goal of the operation. The desired output induction furnace and an open system. Accu-
of the process is a target melt with a specified range of mulations of mass, e.g., refractory buildups, are
chemical compositions with E elements count in the alloy. regarded as a part of output mass.
The last element of the alloy (number E) is supposed to be (2) The output melt is assumed to be entirely metal.
the base metal which is taken as the reminder element. The The input charge, however, includes both metal
furnace probably contains an initial amount m0 of melt with parts and non-metallic contaminations. The
a known chemical composition. A selection of N different percentage of the weight of contaminations in
materials is available and waiting to be charged into the each charge is known as LCj . The contaminations
furnace, e.g., scraps, ingots, returns, hardeners, etc. Each of are considered as a portion of the lost material.
the charge materials has been characterized by its purchase (3) Chemical composition of each charge material is
price, its chemical composition, and its percentage of non- denoted by Cij which only refers the metallic
metallic contaminations. However, inevitable phenomena part of the charge, i.e., the constituents of the
occur during the process. Evaporation, oxidation, splash- non-metallic portion are not included in the
ing, absorption, and other reactions lead to an undesired chemical composition. It should be reminded
element loss in the form of gas, dust, fumes, slag, and that semi-metallic elements, like carbon and
refractory build-up. In detail, those phenomena may be silicon that exist as alloying elements in the
neither investigated nor interested within the context of the metallic charge, are considered in the
present problem. Nevertheless, element loss is important composition.
and experimental data on the percentage of element loss in (4) A portion of the elements of the metallic part of
each charge material are likely to be available. Finally, the the charge materials will be lost through oxida-
problem asks to determine the weight (mass) of each of the tions, evaporations, dust, deslagging, etc. Each
charge materials in such a way that two following condi- material could have a different amount of loss
tions are satisfied; (1) the chemical composition of the final for each constituent element. That means, for
melt falls within the specified range; (2) total cost of the instance, there is a difference between zinc loss
consumed charge materials is kept at the minimum value. in swarf/cheeps charges and zinc loss in bulk
part scraps. Every element loss in each material
To establish a model, an open system is defined on the is characterized by Lij which denotes the weight
boundaries of the furnace. Figure 1 illustrates the melting percentage of i element loss in charge j corre-
furnace and input/output of the system. The input charge sponding to the total weight of the element i in
materials are denoted by j and have an unknown mass mj . that charge. The values Lij are assumed to be
The output includes two parts; the molten metal mM , and already known from experimental data or any
the total of the material loss evaluated by mL : The fol- other knowledge on the matter.
lowing assumptions are made: (5) The cost of the melting process includes mate-
rials, energy, labor, machinery, and processing
expenses. Evaluation of the total cost of the

International Journal of Metalcasting


melting process requires investigating, at least, Substituting Eqn. 3 and 4 into Eq. 2 gives the total mass of
heat balance and energy consumption26 which materials loss:
are out of the focus of the present model. This
model is aimed to only minimize the cost of raw E X
X N   X
N
materials. The other costs would be simply mL ¼ Lij Cij 1  LCj mj þ LCj mj
added to the final optimized materials cost. It is i¼1 j¼0 j¼0
! Eqn: 5
also assumed that there is no value for any of the X
N  X
E

lost materials. That means cost savings, for ¼ 1  LCj Lij Cij þ LCj mj
j¼0 i¼1
instance, from the recycling of slag, dust, etc.,
are not included in the present optimization. The
(6) The melting operation can be started with a mass P mass balance of each element would take the form
Cij mj ¼ Ci M mM if there was no material loss in the
of initial melt in the furnace to be corrected, system. However, the presence of materials loss requires
namely m0 . This is a probable case but is not the two corrections; first, the contaminations mCj should be
always case. The present model considers a subtracted from the left side, since the input
general case where m0 can take zero value for an contaminations are assumed to have no metallic
empty furnace or take a positive value for an elements. Second, the loss of the element mEi should be
initial melt to be corrected. considered on the right (output) side in addition to the melt
content. Therefore, the element mass balance equation is
Mathematical Formulation corrected as the following form:

N 
X  N 
X 
With the aforementioned assumptions, the open system of
melting furnace can be examined starting from an overall 1  LCj Cij mj ¼ Ci M mM þ 1  LCj Lij Cij mj
j¼0 j¼0
mass balance:
Eqn: 6
X
N
To have a normalized formulation, we would rather define
mTotal ¼ m0 þ mj ¼ mM þ mL Eqn: 1
|fflfflfflfflffl{zfflfflfflfflffl} mass fractions instead of masses and rewrite the mass
j¼1
|fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl} Output MeltþLoss balance equations in reduced form:
Initial MeltþInput Charge
mj mM mL
The term m0 stands for the probable initial melt in the xj ¼ ; xM ¼ ; xL ¼ Eqn: 7
furnace. When there is no initial melt, m0 will be null. For mTotal mTotal mTotal
convenience in formulation, we consider the initial melt as X
N
xj ¼ xM þ xL ¼ 1 Eqn: 8
an input charge with the index zero. The presence of m0 , j¼0
however, puts some extra constraints on the problem which
N 
X  
will be discussed later in this section. The total material 1  LCj 1  Lij Cij xj ¼ Ci M xM Eqn: 9
loss includes loss of allowing elements plus non-metal j¼0
contaminations: !
X
N  X
E
xL ¼ 1 LCj Lij Cij þ LCj xj Eqn: 10
mL ¼ mEL þ mCL j¼0 i¼1
|{z} |{z}
Total element losses ðMetal lossÞ Contaminations ðNonMetalsÞ In the case of the presence of an initial melt in the above set
XE X N
of relations, the weight of the target melt mM is usually
¼ mEi þ mCj
unknown. Moreover, xM and mTotal are also unknown,
i¼1 j¼0
which makes the system to be non-linear. To arrange the
Eqn: 2 equations in a standard linear programming form, Eqn. (9)
The contamination of each charge burden is evaluated by a can be written with xM ¼ 1  xL as:
coefficient (fraction) of contamination:   
X
N 1  LCj 1  Lij X
N
mCj ¼ LCj mj Eqn: 3 Cij xj ¼ C~ij xj ¼ Ci M Eqn: 11
j¼0
ð1  x L Þ j¼0
The loss of each metal element is obtained by summation
of the loss of the same element in all charges: Where the coefficients C~ij are the modified compositions
defined as:
X
N     
mEi ¼ Lij Cij 1  LCj mj Eqn: 4 1  LCj 1  Lij
j¼0 C~ij ¼ Cij Eqn: 12
ð1  x L Þ

International Journal of Metalcasting


^ _
Since the modified coefficients of Eqn. (12) are depending constraint limits x 0 and x0 would be equal to m0 =mT which
on xj ’s, this balance equation is a non-linear relation, but force x0 to that certain amount in the calculations.
now, it matches the form of standards linear programming.
Another variable to be expressed for the optimization
model is the objective function. The cost function is Summary of the Optimization Model
defined to be the total cost of the consumed charge
materials (see assumption 5): The mathematical model is now ready to be summarized in
the standard linear programming form. The problem is to
X
N find mass fractions xj to minimize the below objective
PM mM ¼ Pj m j Eqn: 13 function, with the following set of constraints:
j¼0

To derive the objective function in the standard form, the X


N
Minimize PM ¼ P~j xj
mass of the melt mM ¼ mTotal  mL is substituted, and the j¼0
sides of the equation are divided by total mass to obtain the 8 N
> P
following form: >
> xj ¼ 1
^ _
0  xj  1; x 0  x0  x 0
< j¼0

X
N
Pj X
N >
> P
N ^ _
>
: C~ij xj ¼ Ci M C i M  Ci M  C i M
PM ¼ xj ¼ P~j xj Eqn: 14
j¼0
ð1  xL Þ j¼0
j¼0

Eqn: 17
where the term P~j is defined as the gross price of each
charge material which is a bit more than the purchase price The modified compositions C~ij are expressed ine Eqn. (12),
of it regarding the cost of contaminations and metal losses. and the^ modified
_
prices P~j are defined in Eqn. (15). The
Attention should be paid to the point that the gross price is terms C i M and C i M are the minimum and maximum limits
depending on the materials losses in the process, which is of the specified range for the target melt, respectively.
also unknown and put more non-linearity to the problem.

Pj Iterative Solution Algorithm


P~j ¼ Eqn: 15
1  xL
Although the developed model summarized in Eqn. (17)
A set of constraints should be applied to the problem has been arranged in a standard linear programming for-
including the limit range of chemical composition of the mat, it is a non-linear system of relations. That is because
target melt, the conservation of overall mass, and the the modified prices P~j and the modified coefficients C~ij are
limitation of the capacity of the furnace for correction of depending on the solutions of mass fractions xj . To solve
initial melt. Since the equations have been arranged in the the non-linear system, an iterative algorithm is proposed
standard form, the chemical composition and overall mass according to the following steps:
conservation can be simply implemented. The limitation of
^
the capacity of the furnace puts a constraint on the total Step (1). Set the initial values to be xj ¼ 0, x0 ¼ 0 and
_ _
mass of the charged materials which entails m0  mT  mF . x0 ¼ 1 (if there is no initial melt m0 ¼ 0 set x 0 ¼ 0); ignore
Dividing m0 by the terms of this inequality gives: the loss coefficients Lj and Lij ; Thus xL ¼ 0, P~j ¼ Pj ,
C

C~ij ¼ Cij . In these conditions, the problem is initialized so


m0 ^ _ m0 that the melting process would have no materials loss.
¼ x0  x0  x 0 ¼ Eqn: 16
mF m0
^ _ Step (2) Solve the optimization problem Eqn. (17) using
The terms x 0 and x0 are the minimum and maximum limits,
the most recent coefficients and parameters. Find mass
respectively, for the initial mass fraction. Those terms are
fractions xj and record the relative change in each one from
to be evaluated and applied as the constraints for x0 in
their previous values.
optimization. It should be noticed that there are two special
^ _
cases for the constraint limits x0 and x 0 according to the
Step (3) Check whether the last solution is feasible; in the
status of the initial melt and demand of the melting process.
case of initial melt m0 [ 0, calculate total mass
One special case occurs when there is no initial melt. In
^ _ mTotal ¼ m0 =x0 ; check if the condition mTotal  mF is sat-
that case both of x0 and x0 take zero value to force x0 to be
isfied; if the total mass exceeds the capacity of the furnace,
zero in the calculations. Another special case occurs when
i.e., mTotal [ mF , then update m0 ¼ x0 mF and report that
not only there is an initial melt to be corrected, but also the
the initial melt should be decreased to the new m0 for a
melting process demands to reach a certain weight of the
feasible solution in the capacity of the furnace.
target melt, i.e., mM is determined. In this case, the total
mass could be evaluated from Eqn. (1) and both of the

International Journal of Metalcasting


Step (4) Check if convergence is reached; i.e., if the out of the scope of the present study. With this problem, we
summation of Pthe absolute change of all xj is less than a aimed to investigate only the performance of the model and
small value jDxj j  e then stop calculations and report algorithm of optimization. Therefore, careful attention
the solution. Otherwise, continue the calculation with the should be paid to the realistic evaluation of the model
next step. parameters for implementation in a real process.

Step (5) Correct the parameters and coefficients; using the Based on the standard model summarized in Eqn. (17), the
most recent mass fractions xj , calculate xL from Eqn. (10), test problem was prepared in a spreadsheet. A matrix of the
P~j Eqn. (15) , and C~ij Eqn. (12), respectively. coefficients was built with ðE þ 2Þ  ðN þ 1Þ rows and
columns. The problem is then solved through iterative
Step (6) Return to step 2 and repeat the calculations. optimization processing according to the proposed algo-
rithm. With each iteration, the matrix of coefficient and
cost parameters were updated based on the values obtained
Model Evaluation: An Industrial Test Case from the previous iteration. The GRG-non-linear algorithm
of the Excel solver28 was used to reach optimal conditions
To evaluate the proposed model, a test problem was drawn. in
P each iteration. The convergence criteria being
The problem involves a 10-ton industrial induction furnace jDxj j  104 were reached after four iterations.
for melting brass alloys. The brass target alloy was chosen,
with seven alloying elements, and eight available charge
materials, so that the problem represents a relatively high- Results and Discussion
number and full-scale multicomponent system. This system
makes a complicated matrix of coefficients that is capable The solution of the test problem resulted in the mass
of conducting a full test on the performance of the model. fractions of the charge materials shown in Figure 2. The
The target melt has been demanded to be in the specifi- bar chart demonstrates the solutions for three iterations.
cations of C47940 alloy27. The furnace is also considered The first iteration yields the results as the melting process
to contain 7 tons of initial melt. The initial melt is chosen would have no materials loss. In the next two iterations, the
to be achieved from C26000 brass scrap27 which is quite mass fractions have been updated taking the loss of ele-
far from the composition of the target alloy. The problem ments and contaminations into accounts. A fast conver-
requires the initial melt to be corrected with a minimum gence has been achieved with the standard model. As
cost so that it falls in the range of the specified target alloy. indicated in the legend of Figure 2, the total weight of the
The final weight of the melt is not known, and we do not calculated charge has been predicted to be about 9909 kg.
know even if this melt correction is feasible or not. Input The total mass fraction of material loss has also been
data of the problem are listed in Table 1. The input data predicted to be xL ¼ 0:02624 for the test problem which
include available charge materials with their prices, non- corresponds to about 262 kg of material loss. The weight of
metal contaminations, all alloying element contents, and the final melt is about 9649 kg consequently from the
percentages of element loss. The prices have been roughly relation mM ¼ xM mT ¼ ð1  xL ÞmT . Regarding the capac-
estimated from different sources on the internet. The data ity of the furnace, the solution of the total charge was
of contaminations and element losses in materials have obtained to be feasible. However, if the furnace had less
been collected from experimental estimations and obser- capacity than 9909 kg, the solution would not be feasible.
vations made in industrial sites. There is about ± 20% or In that case, the algorithm would report the issue and
more uncertainty on the collected values. For some element would ask the operator to decrease the amount of initial
losses, e.g., evaporations of zinc, there is some information melt. Figure 3 shows the converged solution for input mass
available in the literature19. Generally speaking, a practical fractions of charge materials in conjunction with output
range of 0.5–7 wt.% was chosen for losses in the content of fractions of material loss related to each material.
all elements. Those elements with a low melting point and
a high tendency of oxidation, like Zn, possess a higher The chemical composition of the output melt is displayed
value in the range. A practical range of 0.5–2 wt.% of in Figure 4. Specified upper and lower limits for the
contamination was also chosen for all charge materials. alloying elements of the target melt have been indicated on
The charge materials which have a high surface area and the plot. The relative amount of each element loss has also
could be collected from unknown sources, like chips- been depicted on the diagram. The weight percentage of
scraps, were assumed to have a higher amount of con- each element loss can be multiplied by the total input mass
tamination. On the other hands, the charge materials which to obtain the mass of element loss. It should be noticed that
have a definite source and low surface area were supposed the reminder element has a considerable amount of loss as
to have a lower amount of contamination. Taken from any well as other elements. Therefore, it is necessary to take the
source of information, the coefficients of loss for materials reminder element into account to be able to evaluate the
elements will have some uncertainties. Accurate evaluation loss of the total material. However, the constraints on the
of loss coefficients requires separate investigation, which is reminder element might not be applied in the standard

International Journal of Metalcasting


Table 1. The Specification of Charge Materials and Parameters of the Test Problem

Materials,j Spec. Price 1 Contamination LC


j Composition and loss of elements
No. ($/kg) (wt.%)
Pj Elements,i Pb Fe Sn P Ni Zn Total others Cu
(wt.%) (wt.%) (wt.%) (wt.%) (wt.%) (wt.%) (wt.%) (wt.%)

Initial melt C26000 0.00 0.5 Cij 0.040 0.050 0.001 0.001 0.001 35.80 0.50 rem
Lij 5 0.5 2 0.5 0.1 5 0 0.5
Leaded-Copper CuPb 1.91 1.5 Cij 16.00 0.300 0.001 0.002 0.000 1.00 0.50 rem
Lij 5 0.5 3 0.5 0.1 4 0 0.5

International Journal of Metalcasting


Copper-Phosphorous CuP 2.80 0.5 Cij 0.050 0.100 0.001 8.00 0.001 0.010 0.40 rem
master alloy Lij 2 0.5 3 1 0.1 3 0 0.5
CuSn trims C52100 2.35 0.5 Cij 0.050 0.100 10.00 0.020 0.001 0.200 0.40 rem
Lij 2 0.5 4 0.5 0.1 3 0 0.5
High Strength Copper trims C19400 2.21 0.5 Cij 0.100 2.60 0.001 0.100 0.001 0.150 0.40 rem
Lij 3 1 2 0.5 0.1 3 0 0.5
Zinc – 2.50 0.1 Cij 0.015 0.001 0.012 0.005 0 99.95 0.01 rem
Lij 2 0 3 0.5 0 6 0 0.2
Copper-Nickel alloy C70100 4.00 0.5 Cij 0.001 0.050 0.001 0.005 3.50 0.250 0.40 rem
Lij 2 0 2 0.5 0.5 3 0 0.5
Copper anode – 6.20 0.1 Cij 0 .0005 .0003 0.001 0.001 0.020 0.005 rem
Lij 1 0 2 0.5 0 2 0 0.5
Target Melt C47940 ? ? CiM C 1.00 C 0.1 C 1.2 C 0.02 C 0.1 C 27.96 C0 rem2
B 2.00 B 1.0 B 2.0 B 0.04 B 0.5 B 34.58 B 0.5
Other parameters
Furnace capacity mF 10,000 (kg)
Initial melt m0 7,000 (kg)
1
The prices have been estimated from averaging on data collected from the internet. These numbers were only used to evaluate the performance of the present model, and they do not
represent a real case study.
2
The specification of the alloy C47940 requires that copper (the reminder element) falls between 63 and 66 wt.%.
Figure 3. Calculated mass fractions of charge materials
Figure 2. Calculated mass fraction of charge materials
with the output mass fractions of materials loss.
with the proposed model.

model. Attention should be paid to P the point that the con-


straint of the overall mass balance xj ¼ 1 is equivalent
to the constraint of the reminder element. Thus, only one of
these two constraints should be applied in the model,
otherwise, the model would become ill-conditioned. That
would cause numerical instability in fine precisions. Nev-
ertheless, some alloy specifications require that the amount
of the reminder element falls in a particular range, like the
target melt C47940 in the present test problem that requires
63  CCu  66. In this case, there is probably another
unconstrainted element, zinc for instance, or the value of
total other elements, which could be conditioned with
upper and lower limits so that the reminder element falls
within the specified range. Therefore, a general rule can be Figure 4. Calculated chemical composition of the final
stated for charge calculations; always take ‘‘total other’’ melt and weight percentage of each element loss
and ‘‘reminder’’ elements into account, but do not put according to the input mass.
constraints on the reminder element. It is only used for the
calculation of total element loss. in online industrial charge calculations where a numeri-
cally fast and stable model is required. Furthermore, the
To evaluate the optimality of the obtained solution, it can standard form of the model in conjunction with the iterative
be checked that whether the objective function, i.e., the solution algorithm is capable of being generalized for more
calculated price, has found its minimum value with the complexities and non-linearities in applications.
obtained solution. In another word, if any of the obtained xj
is changed, the solution of the problem is expected to lead An important point to be discussed is considering the ele-
to a more cost value. Therefore, other rounds of calcula- ment loss in each material. The coefficient of loss or
tions were performed using the model with separate sometimes the coefficients of absorption (the complemen-
increments in the obtained solution. In each calculation, an tary of loss) of the alloying elements are usually evaluated
extra constraint was introduced into the model which held based on industrial and experimental estimations29. For
one xj at an incremental deviation from its solution. Then, instance, zinc has a relatively high amount of loss, taking
the calculation was repeated. If another feasible solution about 4–6% on average, because of its low boiling point
was obtained, the price was recorded. Figure 5 shows the and its high vapor pressure30 regarding the melting tem-
variations of cost value concerning the deviations of each perature of the brass alloys. Nevertheless, the element loss
mass fraction xj . The diagram reveals that any deviation in in different charge materials depends on a variety of
mass fractions will increase the cost function or will lead to parameters, e.g., specific surface of scraps. Therefore, the
an infeasible solution. This is a reason to the point that the value of element loss is not the same in different charge
objective function has reached its minimum feasible value. materials. It has been well substantiated in melting that
It is worth mentioning that since the proposed model has materials in the shape of swarfs and chips exhibit much
been organized in a standard LP form, it will be convenient more element loss than ingots and thick parts. An average
for the user to investigate the optimality conditions and to element loss, however, will strongly depend on how much
evaluate the numerical behavior of the model in various weight of swarfs, foils, chips, and sheets is present in the
cases of industrial application. That is of great importance

International Journal of Metalcasting


List of Symbols
Letters

C Composition
E Number of elements
L Coefficient of loss
m Mass
N Number of charge materials
P Price
x Mass fraction
Greek Letters

D Difference
Abbreviation
Figure 5. Variations of the optimized price of melt
versus deviations of mass fraction of charge materials.
LP Linear programming
charged burden. In the proposed model, the loss coeffi- Superscripts
cients Lij provide flexibility for considering more realistic
element loss in different materials. Another major source of C Contamination
the unreliability of the charge calculation models is the E Elements
presence of contaminations in charge material. In the pre- Maximum limit
sent model, the coefficients of non-metal contamination LCj Minimum limit
have been considered to help to reduce the uncertainties in * Modified coefficient
the chemical composition of the final melt. In this context,
Subscripts
the methodology of the evaluation of these coefficients is
also important, which falls out of the scope of the present
0 Initial melt
paper and needs more investigation in future studies.
i Element
j Charge material
Conclusion L Loss
T Total
In this paper, a charge calculation model for melting fur- M Melt
naces has been presented. The model is based on the mass F Furnace
balance of the alloying element and optimization of the
cost of the charge materials. The model considers non-
metallic contaminations in charge materials and element Acknowledgments
losses due to unwanted evaporations, oxidations, dust,
The authors would like to gratefully acknowledge
deslagging, etc. The model also considers the likely case of
Rahyaft Advanced Science and Technologies for infor-
an initial out-of-range melt in the furnace which should be
mation and advice about the process of melting. This
corrected according to a target alloy specification. The
research did not receive any specific grant from funding
main contribution of the presented model is that all of the
agencies in the public, commercial, or not-for-profit
non-linear formulations, emerged from considering con-
sectors. The authors declare that they have no compet-
taminations, material loss, and melt corrections, have been
ing interests.
arranged into a standard linear-programming framework of
optimization problems, which is summarized in Eqn. (17).
The model was tested, and the performance of the model REFERENCES
showed that non-linear phenomenological complexities can
be effectively considered through an iterative algorithm of 1. E. Stefana, P. Cocca, F. Marciano, D. Rossi, G.
the standard LP model. This feature provides a backbone Tomasoni, A review of energy and environmental
for reliable and fast optimization which is of significant management practices in cast iron foundries to
benefit for industrial automation. increase sustainability. Sustainability 11, 1–18 (2019).
https://doi.org/10.3390/su11247245
2. K. He, L. Wang, A review of energy use and energy-
efficient technologies for the iron and steel industry.

International Journal of Metalcasting


Renew. Sust. Energy Rev. 70, 1022–1039 (2017). Metall. Mater. Trans. B 18, 733–740 (1987). https://
https://doi.org/10.1016/j.rser.2016.12.007 doi.org/10.1007/BF02672891
3. R.D. Naranjo, J.-Y. Kwon, R. Majumdar, W.T. 17. M. Hasegawa, Chapter 3.3 - Ellingham diagram, in
Choate, Advanced Melting Technologies: Energy Treatise on Process Metallurgy. (Elsevier, Amster-
Saving Concepts and Opportunities for the Metal dam, 2014), pp. 507–516
Casting Industry (BCS Incorporated, USA, 2005) 18. S. Omole, R. Oluyori, Study of carbon and silicon loss
4. A.B. Kungurtsev, Y.I. Senkevich, H.O. Zinovatnaya, through oxidation in cast iron base metal using rotary
N.O. Novikova, The algorithms for automated calcu- furnace for melting. Leonardo Electron. J. Pract. Tech.
lation of the furnace charge in smelting and refining 14, 59–64 (2015)
metal. Odes’yi Politech. Univ. Prat. 1, 61–71 (2017). 19. J.R. Brown, Foseco Non-Ferrous Foundryman’s
https://doi.org/10.15276/opu.1.51.2017.11 Handbook (Butterworth-Heinemann, Oxford, 1999)
5. Y. Stankevich, V. Timoshpolskii, N. Pavlyukevich, M. 20. E. Ziółkowski, Algorithm for burden calculation for
German, P. Grinchuk, Mathematical modeling of the foundry furnaces using charge materials with an
heating and melting of the metal charge in an electric- uncertain composition. Arch. Metall. Mater. 58,
arc steel-making furnace. J. Eng. Phys. Thermophys. 887–889 (2013). https://doi.org/10.2478/amm-2013-
82, 221 (2009) 0117
6. S.O. Seidu, A. Onigbajumo, Development of charge
21. E. Ziółkowski, Algorithms of furnace charge burden
calculation program for target steel in induction
optimisation in foundries. Arch. Metall. Mater. 52,
furnace. Leonardo Electron. J. Pract. Technol. 6,
487–495 (2007)
181–191 (2015)
7. S. Seidu, O. Adetunji, S. Olagunju, Development of 22. E. Ziółkowski, Burden optimisation of lump charge
scrap charge maximization program capable of melt materials for foundry furnaces. Arch. Metall. Mater.
prediction and modification in crucible furnace. Eur. (2017). https://doi.org/10.1515/amm-2017-0327
J. Appl. Eng. Sci. Res. 6, 1–9 (2018) 23. S. Capuzzi, G. Timelli, Preparation and melting of
8. E. Ekmekci, Y. Yetisken, U. Camdali, Mass balance scrap in aluminum recycling: a review. Metals (2018).
modeling for electric arc furnace and ladle furnace https://doi.org/10.3390/met8040249
system in steelmaking facility in Turkey. J. Iron Steel 24. J. Li, N. Provatas, G. Brooks, Kinetics of scrap
Res. Int. 14, 1–6 (2007). https://doi.org/10.1016/ melting in liquid steel. Metall. Mater. Trans. B 36,
S1006-706X(07)60064-8 293–302 (2005). https://doi.org/10.1007/s11663-005-
9. A.E. Morris, G. Geiger, H.A. Fine, Handbook on 0031-2
Material and Energy Balance Calculations in Mate- 25. E. Ziółkowski, Optimisation of correction of the
rials Processing (TMS Wiley, Hoboken, 2011) chemical composition of molten metal using charge
10. J.S. Arora, Introduction to Optimum Design (Aca- materials with fuzzy composition. Arch. Foundry Eng.
demic Press, UK, 2017) 8, 129–132 (2008)
11. B. Beeley, Foundry Technology (Butterworth-Heine- 26. M. Kovacic, K. Stopar, R. Vertnik, B. Sarler, Com-
mann, India, 2001) prehensive electric arc furnace electric energy con-
12. E. Worrell, M.A. Reuter, Handbook of Recycling, Iii sumption modeling: a pilot study. Energies 12, 2142
(Elsevier, Boston, 2014) (2019)
13. U.M.J. Boin, M. Bertram, Melting standardized alu- 27. J.R. Davis, ASM Specialty Handbook Copper and
minum scrap: a mass balance model for europe. JOM Copper Alloys (ASM international, Ohio, 2001)
57, 26–33 (2005). https://doi.org/10.1007/s11837-005- 28. R.W. Larsen, Engineering with excel, in Excel Master
0164-4 Series, 4th edn. (Pearson Education Inc., London,
14. W.H. Bassett, Zinc losses. J. Ind. Eng. Chem. 4, 2013)
164–167 (1912). https://doi.org/10.1021/ie50039a003 29. T.V.R. Rao, Metal Casting: Principles and Practice
15. O.S. Onyekwere, C.I. Orji, K.A. Uyanga, Determina- (New Age International, New Delhi, 2007)
tion of percentage zinc loss during melting of zinc 30. T. Iida, R.I.L. Guthrie, The Physical Properties of
scrap in a crucible furnace. Global J. Eng. Tech. Adv. Liquid Metals (Clarendon press, Oxford, 1988)
1, 22–26 (2019). https://doi.org/10.30574/gjeta.2019.
1.1.0004
Publisher’s Note Springer Nature remains neutral with
16. M.M. Collur, A. Paul, T. Debroy, Mechanism of
regard to jurisdictional claims in published maps and
alloying element vaporization during laser welding.
institutional affiliations.

International Journal of Metalcasting

You might also like