Journal Information
Vol. 9. Issue 1.
Pages 1025-1031 (January - February 2020)
Download PDF
More article options
Vol. 9. Issue 1.
Pages 1025-1031 (January - February 2020)
Original Article
DOI: 10.1016/j.jmrt.2019.11.041
Open Access
Utilization of color change and image processing to evaluate the Waste Foundry Sand reclamation level
Mohammad Reza Saboura,b,
Corresponding author

Corresponding author.
, Mohammadamin Akbaria, Ghorbanali Dezvareha
a Civil Engineering Faculty, K. N. Toosi University of Technology, Tehran, Iran
b Civil Engineering Faculty, K. N. Toosi University of Technology, No. 1364, Valiasr Street, Mirdamad Intersection, Tehran, Iran
Article information
Full Text
Download PDF
Figures (9)
Show moreShow less
Tables (4)
Table 1. Design of experiments and ΔRGB values.
Table 2. Statistical parameters used for modeling.
Table 3. Verification of the model (ΔRGB values of sample 14).
Table 4. TCLP analysis results of WFS samples.
Show moreShow less

Waste Foundry Sand (WFS) is the by-product of casting industries and is utilized to make molds. Some minerals, such as clay particles or chemical admixtures such as phenolic resins, are used to bond silica sand grains and shape the mold. Preventing environmental impacts such as WFS landfilling, many researches have been conducted to find an appropriate way for reclamation and evaluation of its cleaning. Due to the color changes during this process, image processing is helpful for an estimation of reclamation progress. In this paper, thermal reclamation method and Response Surface Methodology (RSM) are used as the most appropriate way of reclamation and design of experiments, respectively. Samples are heated at different temperatures for a variety of time durations. The color changes of samples are measured using image processing technique by detection of RGB (Red, Green, Blue) color coordinate system parameters changes (ΔRGB). The results of TCLP (Toxicity Characteristic Leaching Procedure) tests reveal that phenolic compounds are removed during the reclamation process, which causes WFS color changes from black to almost white. On this basis, a linear model is developed to predict RGB values relative to time or temperature variations. Results approve that an investigation on ΔRGB values provides a precise estimation of WFS grains reclamation level (less than 10 % error) to guide its reuse within the foundry industries.

Waste Foundry Sand
Response surface methodology
Waste management
Image processing
Full Text

Foundry industries produce different ferrous and non-ferrous metals. Most products are used in several industries such as aerospace and automobile. In order to produce these objects, different mold designs (to shape external surfaces) and core designs (to form internal void spaces) are required. Also, due to high temperatures during the casting process, the molding material should be able to resist the remarkable heat of the molten metal. Considering abovementioned issues, one of the most common materials to make molds in the casting process is uniform-size-distributed high-quality silica sand. Silica sand can take different shapes by mixing with some minerals or chemical admixtures, because the sand grains cannot naturally bond to each other [1].

Waste Foundry Sand (WFS) is the by-product of ferrous and non-ferrous metals casting process. During the casting process, foundry sand is recycled and reused several times, but eventually it cannot be utilized anymore. The result is a vast amount of a left over as WFS [2].

Landfilling, as the most common way to dispose WFS, causes considerable cost and environmental impacts. Consequently, researches were conducted towards introducing new ways of this material utilization in different industries such as road and building construction, foundry industry, and concrete production [3]. Also, it will be difficult to access new resources of raw sand in the near future, and regulations will restrict the extraction of sand from the land. As a result, the use of new sands in the casting industry will be reduced or prohibited, and industries are to use reclaimed sand for mold making shortly [4].

1.1Types of foundry sand

Referring to the casting methods, various chemical compositions are used to bond sand grains. Thus, different kinds of WFS can be produced during the casting process, being clay-bonded (green), air-setting bonded, and thermosetting bonded sands the most dominant produced ones [5].

The clay-bonded sand system is the oldest method for mold making, in which water and clay are mixed to bond sand grains. Clay-bonded (green) sand consists of high-quality silica sand (85–95 %), bentonite (4–10 %), water (2–5 %), and hydrocarbon compounds (2–10 %). This hydrocarbon admixture is coal dust, which produces lustrous carbon under the influence of high temperature while exposing to molten metal during the casting process. Lustrous carbon protects sand grains from entering to molten metal and improves the final surface. Therefore, a high-quality final surface of the product is obtained [5].

Chemically bonded materials are developed as a result of technology progress and introduction of new bonding materials. Chemically bonded sands consist of silica sand (93–99 %) and chemical pastes (1–3 %). Different kinds of chemical pastes, including phenolic-urethanes and epoxy resins, are utilized. Some materials, which are mainly named as thermosetting resins, are hardened by heating at a temperature above 200℃. This kind of mold making materials is used in shell molding process. Another method for mold making, named as Cold-Box method, uses a substance such as sodium silicate solution or phenolic resins to mix with sand, and the mixture is hardened by blowing gas (CO2 or amine). This process produces a weak acid, which hydrolyzes sodium silicate and produces amorphous silica. This amorphous silica bonds sand grains to each other [6].

1.2Physical and chemical properties of WFS

Referring to previous sections, WFS consists of uniform-size-distributed grains [1]. According to the presence of different hydrocarbon compounds in different types of foundry sand, this material has a variety of colors (e.g.: red, brown, and black). After the casting process, hydrocarbon compounds are mainly burned and the color turns into dark brown or black [7].

The chemical composition of WFS is related to the type of chemical admixture. Research articles revealed that different heavy metals (e.g.: Cr, As, Cd, Hg), phenol, formaldehyde, and other PAHs are found in WFS [8,9]. Different experimental methods are introduced to identify chemical composition and its quantity within the sample. Toxicity Characteristic Leaching Procedure (TCLP) [8,10], Synthetic Precipitation Leaching Procedure (SPLP) [8,11], and Extraction Procedure Toxicity (EP) [12,13] are the most common methods in this regard. In aforementioned methods, the sample is washed with a solution, then the concentration of each chemical composition is determined through Gas-Chromatography/Mass-Spectrometry (GC/MS) or other similar methods [10,11,13].

1.3Thermal reclamation method

WFS is mainly categorized as a hazardous waste based on the presence of some harmful compounds such as phenol and formaldehyde in many kinds of resins, [3]. Phenol is a solid or dense liquid, which melts down at 45℃. Its boiling point is 181.8℃ and has an 87g/l solubility in water at 25℃. It can remain about 2–5 days in soil without any change and more than 7 days in water in a high concentration [14]. As a result, these compounds should be removed before landfilling or any other usage [3]. Thermal reclamation is a method in which WFS is heated up to 500–900℃, and chemical compounds are evaporated or burned. This heat is produced by electricity or ignition of different fossil fuels such as gas and oil via rotary kiln, fluidized bed, and shaft furnaces as the most common equipment for this purpose [15].

Andrade et al. [16] drew the mass reduction percentage diagram of WFS due to the decomposition of phenolic resins. They concluded that this process is optimal by heating between 450℃ and 550℃ in 60min, while higher temperatures lead to more energy consumption. Another research on thermal reclamation of WFS is implemented by Lucarz [17], indicating that this method can fully substitute the mechanical reclamation. The author investigated the efficiency of thermal reclamation method for 30g WFS samples of furan molding sand through heating at 5 different temperatures for 2h. The results show that thermal treatment at 600–700℃ temperature range has an acceptable efficiency.

1.4Image processing

Colorimetric analysis of a special point is carried out by determining RGB (Red, Green, Blue) color coordinate system (RGB−CCS) values. These values are determined through an appropriate software (e.g.: Photoshop) by calculating differences between the values of each component before and after the reclamation process (ΔR, ΔG, ΔB). RGB−CCS uses 256-bit color scale with white and black corresponding to color intensity of “255,255,255” and “0,0,0”, respectively [18]. Many researches are conducted using this system to determine color changes in order to demonstrate changes in organic compounds or removal of contaminants from a substance [18–20].

Endo et al. [21] developed a colloidal crystal-based chemical sensor with a reversibly tunable structural color. They detected a volatile organic compound using colorimetric methods. Indeed, color changes indicate the type of target molecules and their concentrations. Thus, chemical sensors to detect the type and color changes of a specific material are able to be widely used for environmental pollution evaluation.

In a recent research on detection of mercury ion (Hg2+) contamination, a digital image scanner based on the change in RGB values of a bio-functionalized gold nanoparticle (A-GNP) solution upon the addition of different Hg2+ concentrations is used. Poornima et al. [22] reported that a Digital Image Scanometer (DIS) simplifies contamination detection, produces results with good precision, and involves the use of fewer amounts of chemicals. Therefore, this system reduces the cost project and improves analysis accuracy.

2Testing procedure

This study is conducted on the casting unit WFS of IKCO, which is one of the most prominent Iranian automakers. The waste is the by-product of Cold-box casting process in which a phenolic resin is mixed with sand, and the mold is made by blowing amine. Due to the use of a phenolic resin, this waste has the potential of being hazardous for the environment. As a result, it is necessary to implement reclamation on it prior to any application or landfilling. The presence of phenol or other hydrocarbon compounds was assessed by TCLP test method. In this regard, sample 11 and a sample before reclamation process are tested.

Gallenkamp electrical kiln is utilized for thermal reclamation process, while temperatures raised up to 1000℃ with an increment of 20℃.

A set of experiments are designed using Design Expert V.7 and Response Surface Methodology (RSM) based on the color changes during the heating procedure. RSM is a combination of different statistical and mathematical techniques, which explores the relationship between several input parameters and the responses. Central Composite Design (CCD) method, which was first introduced by Box and Wilson in 1951 [23], is used to design experiments. Time and temperature are defined as input variables in the range of 10minutes–120minutes and 150℃–600℃, respectively, and ΔR, ΔG, and ΔB of each sample are defined as responses. The results are 13 different situations of experiments, which are reported in Table 1.

Table 1.

Design of experiments and ΔRGB values.

ΔRGB values obtained by MATLAB R2014aΔRGB values obtained by photoshop CS3 portableDesign of experiments 
ΔB  ΔG  ΔR  ΔB  ΔG  ΔR  Time (minute)  Temperature (℃)  No. 
15.9  20.3  26.4  35.3  39.4  47.5  65  375 
1.7  1.6  1.9  1.1  1.8  1.9  10  375 
5.0  7.2  9.9  4.1  5.4  8.3  26  216 
9.0  6.3  7.0  7.1  5.0  5.1  104  216 
29.3  32.2  38.1  27.1  30.1  36.3  65  375 
55.5  59.9  65.6  55.8  60.7  66.6  26  534 
37.3  42.9  50.2  21.8  24.0  29.1  65  375 
1.9  2.3  0.2  13.9  17.2  21.6  65  375 
54.5  67.6  73.5  49.6  66.1  73.8  120  375 
73.0  84.6  94.4  69.8  87.3  96.6  104  536  10 
77.8  90.1  106.0  80.5  93.8  108.8  65  600  11 
18.9  20.1  25.1  37.0  39.2  43.1  65  375  12 
0.0  0.9  0.6  3.9  2.3  3.1  65  150  13 
19.0  21.7  27.1  17.3  20.2  25.1  65  375  7″ 
17.2  20.9  25.7  19.1  23.2  27.6  65  375  8″ 

In order to evaluate color changes on the surface of samples before and after the heating process, a CANON EOS 70D camera is used to take photos of the samples upper surface. Photos are taken in a closed room with a constant light, which is not exposed to external light.

Adobe Photoshop CS3 Portable and MATLAB R2014a are utilized to extract average R, G, and B values of the whole surface of sample before and after reclamation. This process is performed manually using Photoshop by gaining R, G, and B values of 16 randomly selected points, and automatically through scanning and gaining R, G, and B values of all pixels using MATLAB. Finally, the average value of each parameter and differences of R, G, and B values before and after reclamation process are calculated and reported as ΔR, ΔG, and ΔB responses for each sample.

3Results and discussion

ΔRGB values are reported in Table 1. Results show that by increasing both time and temperature, ΔRGB values increase (Fig. 2). As discussed, high and low values of RGB represent light and dark colors, respectively. Color changes indicate that the reclamation method applied to WFS is very effective and hydrocarbon contaminants are removed at the exposure of high temperatures or increasing heating duration, although each of the two conditions leads to higher energy consumption.

Scrutinizing ΔRGB values of samples 1,5,7,8, and 12 highlights the following 2 issues.

  • 1

    ΔRGB values obtained by Photoshop show a great difference to the same values obtained by MATLAB. Fig. 1 shows the photos of samples 11, 5, and an unreclaimed sample (left to right). Sample 5 was heated at 375℃ in 65min and hydrocarbon contaminants are removed out of some WFS particles, while other particles remain contaminated. As a result, manual point selection through a software such as Photoshop leads to an error. This error is the result of selecting just white points (the points related to reclaimed particles) or black points (contaminated points). Thus, if the black points are chosen, ΔRGB values report an unreclaimed surface, and if the white points are chosen, ΔRGB values report a completely reclaimed surface. In this case, although reclamation process occurred partly, the results show an unreclaimed surface or a completely reclaimed surface.

    Fig. 1.

    WFS samples (left to right: sample 11, sample 5, unreclaimed sample).

  • 2

    The difference among ΔRGB values of sample 8 (which are low and approximately 0), 7, 1, 5, and 12, (Fig. 2) is the result of the location of each sample inside the kiln. Sample 7 was close to the source of heat, so it has faced higher heat compared to samples 1, 5, and 12, which were at the midpoint of the kiln, while sample 8 was far from the source of heat. Fig. 3 shows 2 photos of color changes occurred for sample 8. Obviously, there is no significant color change after the reclamation process.

    Fig. 2.

    ΔRGB values of samples 1–13.

    Fig. 3.

    Surface of sample 8 before (left) and after (right) reclamation process.


In order to correct errors in samples 7 and 8, thermal reclamation process was repeated again. Samples 7″ and 8″ were heated at 375℃ in 65min and ΔRGB values were measured. The results are reported in Table 1.

The values obtained by MATLAB are used in Design Expert V.7 for extracting a model on the basis of precision and accuracy of results. ΔRGB values of samples 7″ and 8″ were used in order to generate the model instead of samples 7 and 8. This model predicts the result of reclamation process based on temperature and heating duration. In Central Composite Design method, P-value and F-Value parameters are investigated. Lower P-value and higher F-value indicate significance of the results, validity of population, which is tested, and the sensitivity of input variables. The lower the P-value and the higher the F-value, the stronger the model.

Results obtained by Design Expert show that ΔR, ΔG, and ΔB values follow a great linear equation. Statistical parameters, P-value, and F-value are reported in Table 2, and the related linear equations are presented as Eq. 1,2, and 3, as follow.

Table 2.

Statistical parameters used for modeling.

F-Value  P-Value  Std. Dev.  Mean  Maximum  Minimum   
–  –  124.808  375  600  150  Temperature(̊C) 
–  –  30.509  65  120  10  Time (min) 
29.43  <0.0001  33.794  38.562  106.0  0.6  ΔR 
26.85  <0.0001  30.178  33.338  90.1  0.9  ΔG 
28.96  <0.0001  25.937  28.985  77.8  0.0  ΔB 

These equations are applicable within the introduced time and temperature ranges.

In order to analyze the differences between real values and the values produced by the model, “Predicted vs. Actual” graphs are drawn. Figs. 4–6 show “Predicted vs. Actual” graph for ΔR, ΔG, and ΔB values, respectively. As demonstrated, in lower values of time and temperature, error increases. Hence, the model functions accurately in higher ranges of time and temperature.

Fig. 4.

Predicted vs. actual graph of ΔR values.

Fig. 5.

Predicted vs. actual graph of ΔG values.

Fig. 6.

Predicted vs. actual graph of ΔB values.


As discussed, many researches imply that due to the presence of carbon and some other chemical contents, WFS has a variety of colors from black to medium tan [7,24,25]. Color changes during reclamation process prove the correctness of the hypothesis that those chemical contents are burned or evaporated while being heated. Consequently, investigating color changes could be a reliable and applicable method for evaluation concerning reclamation level.

In order to verify the obtained model, sample 14 is heated at 500℃ in 30min and ΔRGB values are compared with the result of the model (Table 3). Referring to Fig. 7, differences between actual values and predicted ones are less than 10 % and the accuracy of the model is quite acceptable.

Table 3.

Verification of the model (ΔRGB values of sample 14).

ΔB  ΔG  ΔR  No. 
44.6  49.1  57.9  14 
40.21  45.41  52.95  predicted 
9.8 %  7.5 %  8.54 %  % Error 
Fig. 7.

Predicted vs. actual graph of ΔRGB values for sample 14 (Model Verification).


The results of TCLP test indicate the presence of phenol and some other phenolic compounds before the reclamation process (Fig. 8). Fig. 9 illustrates that after the reclamation process on sample 11 at 600℃ in 65min pulses become weak. Thus, the concentration of phenolic compounds decreases to almost zero (Table 4).

Fig. 8.

TCLP analysis of WFS before the reclamation process (unreclaimed sample).

Fig. 9.

TCLP analysis of WFS after the reclamation process (sample 11).

Table 4.

TCLP analysis results of WFS samples.

Sample 11  Unreclaimed sample   
<25ppb  625ppb  Phenol 
<25ppb  <25ppb  2-chlorophenol 
<25ppb  <25ppb  2,4-dimethyl phenol 
<25ppb  <25ppb  4-chloro-3-methyl phenol 
<25ppb  <25ppb  2,4-dichlorophenol 
<25ppb  <25ppb  2-nitrophenol 
<25ppb  <25ppb  2,4,6-trichlorophenol 
<25ppb  49 ppb  4-nitrophenol 
<25ppb  <25ppb  Pentachlorophenol 
<10ppb  <10ppb  Formaldehyde 

Waste Foundry Sand (WFS) consists of different contamination based on the type of casting process and mold making method. Thermal reclamation process is a method for removing contaminants from the sand and preparing it to be reused in various industries. The results of TCLP tests before and after thermal reclamation process indicate that this method is an appropriate cleaning for WFS grains. During thermal reclamation process, burning or evaporation of hydrocarbon contaminants leads to color changes, from almost black to almost white. Quantification of these changes is a good index for determination of contaminants removal level. In order to quantify color changes, photos should be taken of the samples surface in a room with a constant light before and after the reclamation process and RGB values should be determined. MATLAB, Photoshop, and other software are useful for measuring ΔRGB values, but MATLAB is more accurate because it reports an average of all pixels RGB values. The enhancement of temperature leads to breaking chemical bonds through providing higher levels of energy. Therefore, hydrocarbon compounds burn or evaporate easier, and higher levels of reclamation process are achieved. As same as temperature, the enhancement of heating time duration increases the exposure of hydrocarbon contaminants to thermal energy. Thus, more hydrocarbon compounds burn or evaporate, and reclamation process progresses in a more efficient manner. Due to the costs of chemical tests (TCLP, SPLP, etc.) and the preparation time of their results, image processing helps to quickly evaluate the reclamation level. A linear model is obtained in order to predict ΔRGB values based on reclamation temperature and heating duration. The values out of the model indicate less than 10 % error in comparison with actual ones of ΔRGB.

Conflicts of interest

The authors declare no conflicts of interest.

N. Cruz, C. Briens, F. Berruti.
Green sand reclamation using a fluidized bed with an attrition nozzle.
Resour Conserv Recycl, 54 (2009), pp. 45-52
R. Alonso-Santurde, A. Coz, J.R. Viguri, A. Andrés.
Recycling of foundry by-products in the ceramic industry: Green and core sand in clay bricks.
Constr Build Mater, 27 (2012), pp. 97-106
Í. Navarro-Blasco, J.M. Fernández, A. Duran, R. Sirera, J.I. Álvarez.
A novel use of calcium aluminate cements for recycling waste foundry sand (WFS).
Constr Build Mater, 48 (2013), pp. 218-228
A. Ghosh, Modern Sand Reclamation Technologies for Economy, Environment Friendliness and Energy Efficiency, in 61st INDIAN FOUNDRY CONGRESS 2013, Kolkata, 2013. Date of Access: September 28, 2018.
J. Brown.
Sands and sand bonding systems.
Foseco non-ferrous foundrymen's handbook, Butterworth-Heinemann, (1999), pp. 149-166
A. Abedi.
Fundamentals of metal casting.
Shahid Rajaee Teacher Training University, (2011),
R. Siddique, G. Singh.
Utilization of waste foundry sand (WFS) in concrete manufacturing.
Resour Conserv Recycl, 55 (2011), pp. 885-892
R. Siddique, G. Kaur, A. Rajor.
Waste foundry sand and its leachate characteristics.
Resour Conserv Recycl, 54 (2010), pp. 1027-1036
S. Ji, L. Wan, Z. Fan.
The toxic compounds and leaching characteristics of spent foundry sands.
Water Air Soil Pollut, 132 (2001), pp. 347-364
Toxicity characteristic leaching procedure method.
Synthetic precipitation leaching procedure method.
Hazardous waste, HSE department.
National Iranian Oil Company, (2012),
Extraction procedure toxicity test method and structural integrity test.
M. Abdollahi, S. Hassani, M. Derakhshani.
Academic Press, (2014), pp. 871-873
J. Svoboda.
Foundry sand reclamation.
R.M. Andrade, S. Cava, S.N. Silva, L.E.B. Soledade, C.C. Rossi, E. RobertoLeite, C.A. Paskocimas, J.A. Varela, E. Longo.
Foundry sand recycling in the troughs of blast furnaces: a technical note.
J Mater Process Technol, 159 (2005), pp. 125-134
M. Lucarz.
Thermal reclamation of the used moulding sands.
METABK, 54 (2015), pp. 109-112
T. Soga, Y. Jimbo, K. Suzuki, D. Citterio.
Inkjet-printed paper-based colorimetric sensor array for the discrimination of volatile primary amines.
D. Kim, A. Jo, H.-M. Yang, B.-K. Seo, K.-W. Lee, T.S. Lee.
Colorimetric detection and removal of radioactive Co ions using sodium alginate-based composite beads.
J Hazard Mater, 326 (2016), pp. 69-76
R. Volinsky, M. Kliger, T. Sheynis, S. Kolusheva, R. Jelinek.
Glass-supported lipid/polydiacetylene films for colour sensing of membrane-active compounds.
Biosens Bioelectron, 22 (2007), pp. 3247-3251
T. Endo, Y. Yanagida, T. Hatsuzawa.
Colorimetric detection of volatile organic compounds using a colloidal crystal-based chemical sensor for environmental applications.
Sens Actuators B Chem, 125 (2007), pp. 589-595
V. Poornima, V. Alexandar, S. Iswariya, D. Parameshwari, R. Muthukumar, T. Uma.
Digital image based simple scanometric device for the express detection of aqueous contamination of Hg2+.
Sens Actuators B Chem, 274 (2018), pp. 472-480
J.P. Kleijnen.
Response surface methodology.
Handbook of simulation optimization, Springer, (2015), pp. 81-104
R. Siddique, G. de Schutter, A. Noumowe.
Effect of used-foundry sand on the mechanical properties of concrete.
Construct Build Mater, 23 (2009), pp. 976-980
A. Bakiyaraj, M. Subas Chandra Bos.
A novel high textured concrete mixture by using multiple used foundry black sand for the construction of architecurally designed green buildings - an alternate for natural river sand.
Journal of Materials Research and Technology

Subscribe to our newsletter

Article options
Cookies policy
To improve our services and products, we use cookies (own or third parties authorized) to show advertising related to client preferences through the analyses of navigation customer behavior. Continuing navigation will be considered as acceptance of this use. You can change the settings or obtain more information by clicking here.