Professional Documents
Culture Documents
To cite this article: M.S. Kulkarni & A. Subash Babu (2005) Optimization of Continuous Casting
Using Simulation, Materials and Manufacturing Processes, 20:4, 595-606, DOI: 10.1081/
AMP-200041874
Article views: 50
Download by: [University of Nebraska, Lincoln] Date: 09 September 2015, At: 20:34
Materials and Manufacturing Processes, 20: 595–606, 2005
Copyright © Taylor & Francis, Inc.
ISSN: 1042-6914 print/1532-2475 online
DOI: 10.1081/AMP-200041874
M. S. Kulkarni
Operations Management Group, NITIE, Mumbai, Maharashtra, India
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015
A. Subash Babu
Industrial Engineering and OR, IIT Bombay, Powai, Mumbai,
Maharashtra, India
This paper describes the development and use of an integrated system of metamodels
and simulation experiments for managing quality in a continuous casting system.
A systematic procedure was developed to satisfy the necessary process conditions
connected to one or more process parameters. The values of these parameters are
to be determined such that all the process conditions are satisfied simultaneously to
ensure that the product will have the desired quality. This procedure utilizes a set
of metamodels, simulation experiments, and an index called the “undesirability index”
for quality, which establishes a functional relationship between the input variables or
process parameters and the quality criteria. This system, when applied to the data
pertaining to a case, demonstrated that the methodology developed could be used to
effectively manage product quality.
Key Words: Casting speed; Continuous casting; Critical process conditions; Critical quality
conditions; Metamodels; Mold flux; Multi-response optimization; Process models; Quality model;
Robust solution; Sensitivity analysis; Simulation; Simulation-based optimization; Undesirability index.
1. INTRODUCTION
The basic operation of a continuous casting system is to convert liquid steel
of a given composition into a strand of desired shape and size through a group of
operations such as mold operation, spray cooling zone, straightener operation, etc.
Researchers and practitioners have done numerous studies and developed models to
understand the interaction between the steel being cast and the process parameters
[1–10]. In these studies, much emphasis has been placed on determining how
different steels behave at elevated temperatures under conditions similar to those
during actual casting. This includes prediction of solidus temperature, prediction of
zero ductility temperature, liquid impenetrable temperature, prediction of thermal
strain, and prediction of critical stress. All these past efforts have significantly helped
595
596 KULKARNI AND SUBASH BABU
and improve the quality of products. To ensure that the product that comes out
of the continuous casting system is of desired quality, the system was developed
using the concept of critical process conditions, allowable process parameter ranges,
simulation, experimentation, and heuristic-based procedures, the details of which are
presented in this paper.
1 ≤ · Vc ≤ 3 (1)
Tsol
1157 for crack-sensitive grades (2)
00472
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015
Tsol
1050 for sticker-sensitive grades (3)
00072
where
where
where
05
s
mark depth (in microns) = 600 (11)
f
Condition 10. As the shell leaves the mold, it should be thick enough to
withstand the ferrostatic pressure of liquid steel in the core. If the stresses arising
due to bulging and thermal strains exceed a critical value, then cracking may occur
at the solidus front, leading to formation of internal cracks. To avoid this, the strand
should not leave the mold with a shell thickness smaller than a minimum value so
that the total stress is less than the critical stress for cracking. The criterion that was
used to prevent excessive stress in the shell is
where, T = total stress in shell in MPa, c = critical stress for cracking in MPa and
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015
is a function of steel composition. The multiplying factor 0.95 gives a safety factor
for selecting casting parameters.
Conditions 11 and 12. During continuous casting, the mold may get
distorted and change shape due to internally generated thermal stresses. These
stresses arise due to local thermal expansion of the mold wall, particularly near
the meniscus. Since most of the mold materials exhibit high distortion and rate of
wear above 300 C due to low strength and hardness, the mold hot face temperature
(Thot should be kept below 300 C. On the cooling channel side of the mold, i.e., at
the cold face, it is desirable to keep the cold face temperature (Tcold below 150 C
for preventing nucleate boiling heat transfer. The conditions to be satisfied were
stated as
The hot and cold face temperatures were calculated as a function of mold heat flux
through the mold, cooling water temperature, mold thermal conductivity, cooling
water velocity, and hydraulic diameter of the cooling channels.
Condition 13. Since the mold plays an important role in heat extraction as
well as strand support, its length should not be smaller than a critical value. The
criterion for minimum mold length was stated as
where T = total stress in shell in MPa, c = critical stress for cracking in MPa.
Condition 15. In order to avoid stresses due to bulging, reheat after spray
zone should be minimized. Reheat beyond 100 C can cause formation of midway
cracks. This chance is even greater for those grades with low values of critical stress
600 KULKARNI AND SUBASH BABU
for crack formation. The limiting value of reheat that can be used as a condition to
be satisfied was stated as
Condition 17. From the point of view of controlling the hot face and cold
face temperatures, cooling water velocity plays an important role, as it is one of
the important parameters that control the heat extraction through the mold. The
conditions to be satisfied were as follows for shrinkage and crack-sensitive grades,
respectively.
1. Select all the process parameters of relevance and identify the range of
values, acceptable for each parameter. The process parameters and their ranges are
given in Table 1.
2. Select a large value of casting speed. For this casting speed, generate
scenarios represented by process parameter values using uniformly distributed
random numbers for each parameter, taking into account the respective range.
Using these generated values, check whether the associated scenario satisfies each
of the 17 critical conditions. If the process parameters related to a quality
criterion fulfill the respective critical condition, i.e., if they satisfy that criterion,
the undesirability index is 0 for that condition, otherwise it takes the value 1. The
total undesirability is the sum of all index values related to all 17 conditions. Thus,
the total undesirability index represents how well a generated scenario satisfied
all 17 critical conditions. This step is to be repeated a specified number of times
by generating a scenario during each simulation run. A number of cycles “c” of
simulation, with each cycle having a specified number of runs “r,” are carried out
by keeping the speed and range of values constant as set in steps 1 and 2.
3. At the end of r × c number of simulation runs, if no scenario leads to
0 value for the undesirability index, it means that the corresponding casting speed is
too high. Therefore, reduce the casting speed by multiplying the speed (set in step 2)
by a scale-down factor (say 0.95) and repeat steps 2 and 3 until a scenario is found
that satisfies all the conditions. Record this casting speed and the corresponding
parameter values and rerun the simulation.
Range of Range of
Parameters values Units Parameters values Units
Parameters 1 2 3 4 5 6 7 8 9 10
Viscosity 1.73 1.82 1.04 1.95 1 1.72 1.99 1.87 1.88 1.67
Frequency 173 182 104 195 100 172 199 187 188 167
Stroke 9.7 6.2 6.2 6.3 6 9.6 6.3 9.9 6.3 9.3
Flux solidus temperature 1082 1075 1072 1080 1050 1076 1079 1093 1079 1093
Drain rate 1.6 2.9 2.1 1.9 2.2 2.9 2.6 1.8 1.8 1.5
Flux density 2069 2053 2048 2064 2001 2057 2063 2093 2062 2093
Mold length 722 717 715 721 700 718 720 730 720 730
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015
Cooling water velocity 11.5 12.2 13.8 13.1 14.6 14.4 13.4 12.5 12.6 11
Hydraulic diameter 0.021 0.022 0.026 0.024 0.027 0.027 0.025 0.023 0.023 0.02
Mold thickness 12 13 15 14 7 12 14 14 14 12
Cooling water temperature 26 26 26 26 26 26 26 26 26 26
Spray zone length 5.2 5.1 5.1 5.2 5 5.1 5.2 5.2 5.2 5.2
Spray zone temperature 1029 1022 1020 1027 1000 1024 1027 1039 1026 1039
Spray end pitch 208 206 206 208 200 207 208 211 207 211
Simulation cycles 1 2 3 4 5 6 7 8 9 10
% Nonconformance 58.5 62.3 59.6 63.0 56.8 58.2 60.5 61.4 62.3 59.6
At the end of step 5, each parameter takes different ideal values as shown in
the rows of Table 2. Find the range (R) for each parameter, which is the difference
between the maximum and minimum of the ideal values of the respective parameter.
These values are given in Table 3.
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015
6. Now reset the range for each parameter using the range of values obtained
in step 5. Keeping the casting speed as obtained at the end of step 5 (i.e.,
0.95 m/min), repeat the simulation experiment involving a number of cycles “c”
and for each cycle, a number of runs “r” as before. During each run, values for
parameters or scenarios are to be generated using uniform distribution. These values
are used to evaluate the critical condition values and to see whether the 17 critical
conditions are satisfied. For the example problem, considering all 10,000 runs in
each cycle, the percentage of runs in which all 17 conditions were not satisfied was
recorded for all 10 cycles considered, as shown in Table 4.
From the results given in Table 4, it can be seen that about 60% of runs
do not satisfy all 17 critical conditions. Therefore, the nonconformance percentage
needs to be reduced. To achieve this, it was felt necessary to reduce the range of
parameter values used for generating the simulated values of process parameters.
The middle point “M” of the range “R” for each parameter was determined and the
modified range of each parameter was set as M ± 025 (R). For the example under
consideration, the values of mid point and the modified range for each parameter
obtained in this manner are shown in Table 5.
Further, simulation experiments consisting of 10 cycles, of 10,000 runs each,
were carried out using the range of values of parameters derived by using the values
in Table 5. The corresponding percentages of nonconformance obtained at the end
of this experiment are shown in Table 6.
Results presented in Table 6 reveal that the nonconformance percentage
ranges from 6.7 (cycle 5) to 7.2 (cycle 6). Since it was felt necessary to further reduce
this nonconformance percentage, the ranges of parameters were taken as M ±020
Table 5 Mean and modified range of values for each of the parameters
Simulation cycles 1 2 3 4 5 6 7 8 9 10
% Nonconformance 6.8 7.1 6.8 6.9 6.7 7.2 6.9 6.7 7.0 7.1
Table 7 Mean and modified range of values for each of the parameters
(R) instead of M ±025 (R). The modified ranges of values of the parameters are
shown in Table 7.
Using these modified ranges of values, a simulation experiment was carried
out as before and the corresponding values of nonconformance percentage obtained
are shown in Table 8.
It can be seen in Table 8 that the percent nonconformance is about 1% and
if considered satisfactory, the simulation can be stopped. The corresponding casting
speed of 0.95 m/min and the process parameters values shown in Table 7 can be
considered the final range of values. However, during actual caster operation, the
casting speed varies during a cast. Therefore, we need to carry out a sensitivity
analysis using different casting speeds, as described in step 7 below.
7. Different casting speeds in the range 0.8 to 1.0 m/min were considered
and using the data given in Table 7, simulation experiments as explained above
were conducted to find out the percentage of nonconformance. This will indicate
the robustness of the solution. The results are given in Table 9.
From the results shown in Table 9, even if the casting speed varies from 0.85
to 0.95 m/min, the quality of the product is not likely to be affected significantly.
However, the casting speed at 1 m/min leads to nonconformance of about 30%,
whereas at 0.8 m/min it is about 90%, which certainly could be a very detrimental
situation. This also is in agreement with industry data [14] for casting speed, which
is 0.88 m/min for the same steel and cross section.
Simulation cycles 1 2 3 4 5 6 7 8 9 10
% Nonconformance 1.2 1.1 0.97 1.1 1.2 1.2 0.95 1.1 1.0 0.98
USING SIMULATION IN CONTINUOUS CASTING 605
Vc, m/min 1 2 3 4 5 6 7 8 9 10
1.00 31.5 30.6 31.8 30.4 31.3 30.3 32.1 31.5 31.0 32.6
0.99 4.6 4.6 4.5 4.6 4.2 4.8 3.9 5.0 4.6 4.3
0.95 1.2 1.1 0.97 1.1 1.2 1.2 0.95 1.1 1.0 0.98
0.90 1.1 1.1 0.97 1.2 1.1 0.98 1.3 1.1 1.3 0.95
0.85 1.2 1.3 1.2 1.1 0.98 1.2 1.3 0.95 0.98 1.1
0.81 11.9 11.7 12.1 12.0 11.3 11.5 12.6 12.4 11.3 12.9
0.80 89.9 90.1 89.8 89.9 90.0 89.9 90.2 89.8 90.1 89.8
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015
3. CONCLUSION
An integrated methodology developed for selecting optimum values of process
parameters has been presented in this paper. This system uses a systematic
procedure for selecting proper parameter values in a continuous casting system.
An example case study of En8 steel of 269 × 269 mm crosssection was considered
to demonstrate the methodology. It was observed that the solution obtained not
only satisfies all the conditions but also is reasonably robust. The robustness was
checked by varying the casting speed for the same scenario. The results show
that the solution obtained for one particular casting speed, using the integrated
methodology, was also valid for a reasonable range of casting speed. It was noticed
that the methodology could be successfully employed for managing quality of a
continuous casting process.
REFERENCES
1. Won, Y.M.; Han, H.N.; Yeo, T.J.; Oh, K.H. Analysis of solidification cracking using
the specific crack susceptibility. ISIJ International 2000, 40, 129–136.
2. Kim, K.H.; Yeo, T.J.; Oh, K.H.; Lee, D.N. Effect of carbon and sulphur in
continuously cast strand on longitudinal surface cracks. ISIJ International 1996, 36,
284–289.
3. Suzuki, H.G.; Nishimura, S.; Nakayama, Y. Improvement of hot ductility of
continuously cast carbon steels. Trans. ISIJ 1984, 24, 54–59.
4. Weinberg, F. The ductility of continuously cast steel near the melting point—hot
tearing. Met. Trans. B 1979, 10, 219–227.
5. Hannerz, N.E. Critical hot plasticity and hot cracking in continuous casting with
particular reference to composition. Trans. ISIJ 1985, 25, 149–158.
6. Nakagawa, T.; Umeda, T.; Murata, J.; Kamimura, Y.; Niwa, N. Deformation
behaviour during solidification of steels. ISIJ International 1995, 35, 723–729.
7. Rindler, W.; Kozescnik, E.; Buchmayr, B. Computer simulation of the brittle
temperature range (BTR) for hot cracking in steels. Steel Research 2000, 71, 460–465.
8. Clyne, T.W.; Wolf, M.M.; Kurz, W. The effect of melt composition on solidification
cracking of steel, with particular reference to continuous casting. Met. Trans. B 1982,
13, 259–266.
9. Kudoh, M.; Itoh, Y. Trial for the reduction of oscillation mark depth in continuous
casting. Steel Research International 2003, 74, 147–152.
606 KULKARNI AND SUBASH BABU
10. Hooli, P. Mould film thickness between mould and steel shell—effect on heat flux and
defect formation. Steel Research International 2003, 74, 480–484.
11. Kulkarni, M.S.; Subash Babu, A. A system of process models for estimating
parameters of continuous casting using near solidus properties of steel. Materials and
Manufacturing Processes 2003, 18, 287–312.
12. Kulkarni, M.S.; Subash Babu, A. Metamodels for continuous casting, CARE Technical
Report; Indian Institute of Technology: Bombay, 2001.
13. Kulkarni, M.S.; Subash Babu, A. Quality and productivity management of continuous
casting process, Second Ph.D Progress Report, Dept. of Mechanical Engineering,
Indian Institute of Technology Bombay, 2001.
14. Concast data sheets, Specification distribution sheets for thermal modeling, Concast (I)
Ltd., 1992.
Downloaded by [University of Nebraska, Lincoln] at 20:34 09 September 2015