Thermal regeneration of waste foundry phenolic sand in a lab scale fluidized bed

Improper disposal of sand used in molding processes after casting increases logistical costs and environmental impact because of the presence of the phenolic resin in its composition. The regeneration process of waste foundry phenolic sand (WFPS) aims to recycle this material. As mechanical regeneration methods are not efficient to guarantee 100% cleaning of the sand grains and their use again in the molding process, this work investigated the efficiency of a method of thermal regeneration of this type of residue that can be employed as a complementary procedure. A laboratory-scale fluidized bed reactor was designed and built to regenerate WFPS that was previously treated by a mechanical method. The methodology used to design and construct the fluidized bed prototype is described, as well as the characterization of the residual, the standard clean sand and the regenerated sand. The results of the thermal regeneration in the fluidized bed were very satisfactory with respect to the regeneration efficiency. For the nine process conditions tested, loss on ignition values were reduced when compared to standard clean sand. This study presents the advantages of a combination of SEVERO, J.A.; MODOLO, R.C.E.; MORAES, C.A.M.; ZINANI, F.S.F. revista Matéria, v.23, n.1, 2018. two processes, mechanical and thermal regeneration, which allows to reduce the time and eventual temperature of resin removal due to the partial removal of the resin layer or its weakening during the mechanical regeneration process. Of the nine process conditions tested, six had loss on ignition values below the CSS. Thus, the thermal regeneration in the fluidized bed results was quite satisfactory in relation to the regeneration efficiency.

two processes, mechanical and thermal regeneration, which allows to reduce the time and eventual temperature of resin removal due to the partial removal of the resin layer or its weakening during the mechanical regeneration process.Of the nine process conditions tested, six had loss on ignition values below the CSS.Thus, the thermal regeneration in the fluidized bed results was quite satisfactory in relation to the regeneration efficiency.

INTRODUCTION
Requirements of environmental laws have forced foundries to increase the costs associated with the disposal of molding sands in specialized landfills.The foundry sector generates several types of wastes, including foundry sands agglomerated with phenolic resins.Waste foundry phenolic sand (WFPS) consists of uniformly sized, high-quality silica sand or lake sand that is bonded to form molds for ferrous (iron and steel) and nonferrous (copper, aluminum, brass) metal castings.This silica sand is coated with a thin film of burnt carbon, residual binder (bentonite, and sea coal, or organic resins by chemically bonding like in the present work).
The raw sand is normally of a higher quality than the typical bank run or natural sands used in fill construction sites.In the casting process, molding sands are recycled and reused multiple times.According to some authors, eventually the recycled sand degrades to the point that it can no longer be reused in the casting process [1,2].Studies about recycling examining the application of foundry wastes have been carried out worldwide.According to several studies, these wastes can be recycled or reused in different ways, in foundry manufacturing itself or in other industrial processes as building materials [3][4][5][6][7].The results thereof are useful to conceive strategies to avoid disposal costs and reduce raw material costs.
Research has also investigated WFPS as partial replacement of fine aggregate in concrete.These findings suggest that WFPS may be effectively utilized as a partial replacement of fine aggregates in making concrete of considerable quality, with no adverse effects in terms of mechanical, environmental, and micro-structural impacts [8][9].Laboratory studies that looked into physical, geotechnical, and leaching properties of flowable fills consisting of WFPS, cement, and fly ash mixed to different water contents were performed by Deng and Tikalsky (2008) [10].The authors observed that most of the physical properties data fell within narrow ranges, although values of copper/aluminum-based WFPS samples might exceed these ranges.Geotechnical properties of samples of flowable fills containing WFPS in both fresh and hardened phases were evaluated and observed to be similar to the features of specified flowable fills.The material leaching analyses indicated that the toxicity of WFPS flowable fills was below regulated criteria, but it should be emphasized that Brazilian Environmental Regulation Agencies are quite strict concerning WFPS recycling because of the presence of phenol compounds.Nevertheless, thermomechanical methods to regenerate foundry sands have been investigated in the effort to mitigate the effects of environmental degradation as much as possible, improve conservation of sand extraction sites, considering in-house recycling as the best solution for valuation of a waste according to the cleaner production concepts [11,12].Park et al. (2012) [13] tested two processes for the recycling and residue stabilization of waste foundry sands, considering the dry mechanical process for recycling, and the stabilization process for powdered residue.The results showed that coal refuse and sodium silicate stabilize heavy metals better than other processes may lead to the development of a cost-effective solution for stabilizing heavy metals in residues.Joseph et al (2017) [14] explored the possibility of reuse foundry sand considering the economic involved costs and the sand quality.The results showed that the major challenge remains the cost investment required to implement reclamation foundries units, mainly the small ones.
Following the perspective of thermal regeneration of WFPS, this work was carried out in three stages.The first was the design and construction of a lab scale fluidized bed.The second was WFPS characterization in order to define performance parameters of thermal regeneration efficiency.The third included experiment sand evaluation of regeneration efficiency considering various combinations of operational parameters.

MATERIAL
The WFPS analyzed in this work was a waste from a foundry in Rio Grande do Sul, a state in south Brazil.Before any analysis, the WFPS (Figure 1) was mechanically treated (treatment by regeneration) by a dry attrition process in which a pneumatic conveying system thrusts a given mass of sand against a stationary plate, thereby partially separating or at least weakening resin layer bonding to the sand surface.To facilitate separation of the resin, the equipment is provided with a cyclonic separator and a bag filter.
Conventional siliceous sand was tested as comparative material.It is called clean standard sand (CSS).According to the Brazilian Standard Legislation for Solid Wastes, sands containing phenolic resins used in foundry have been classified as hazardous waste [15].

Sand samples characterization
Thermogravimetric analysis (TG-DTA) of the WFPS samples was carried out in a BP RB -3000 -20 simultaneous analyzer.A sample weighing approximately 20 g was placed in a microbalance and heated at a rate of 10 ºC/min from the room temperature to a final temperature of 1000 ºC under nitrogen flow.The particle size distribution and density were evaluated according to specific methods for this type of material.Loss on ignition was evaluated with maximum temperature at 950 ± 10 °C.The microstructure and sphericity of particles were analyzed using scanning electronic microscopy (SEM) (Zeiss, model EVO LS 15) and Image-J software.Mean particle sphericity was obtained by the estimation method proposed by Peçanha and Massarani (1986) [16], according to which roundness is calculated based on a ratio of diameters measured in the microscopic image and using equation ( 1), with an error of around 7%.
where ip d is the particle diameter measured, and cp d is the limited particle diameter.

Sizing of the fluidized bed chamber
i.

Pressure loss calculation in the fluidized bed
To calculate the pressure drop along the fluidized bed filled with settled particles, the classical Ergun equation was employed: where ∆P fr is the pressure drop by friction, L m is the height of the packed bed, ε m is the porosity of the packed bed, μ is the dynamic viscosity of the fluid, u 0 is the fluid surface velocity, Ф s is the mean particle sphericity, d p is the mean particle diameter based on size distribution and ρ g is the gas density. ii.

Minimum fluidization velocity
The minimum fluidization condition was calculated based on a simple balance of forces, assuming that the set of particles is fluidized at the moment the friction forces achieve the apparent bed weight, as described in Equations 3 and 4 [17]. where t is the bed cross-sectional area L mf is the minimum height of the bed under fluidization ε mf is the bed porosity under minimum fluidization ρ s is the particle density, ρ g is the gas density g is the acceleration of gravity.
As observed in Equation 4, it is possible to determine the minimum fluidization velocity (u mf ) using Equations 2 to determine the longitudinal pressure drop in Equation 4: Equation 5 can be used to determine the minimum fluidization velocity if properties of the gas (viscosity and density), bed particles (spherical shape, average diameter and density), and bed porosity under minimum fluidization are known.The bed porosity under minimum fluidization is usually determined by empirical correlations.Yet, in some cases the porosity under minimum fluidization assumes a value slightly higher than the porosity of the packed bed.Given the small discrepancy between these values, they were considered equivalent.(Kunii and Levenspiel, 1991) [17].)lists some experimental values obtained by Leva (1959) [18] for some sand particles under minimum fluidization.iii.

Terminal velocity
Terminal velocity is reached when the drag force on the particles is greater than the weight force exerted by gravity on them.
For non-spherical particles having sphericity between 0.5 and 1, an experimental correlation introduced by Kunii and Levenspiel (1991) [17] was employed to determine the terminal velocity.This correlation is summarized by Equations 6, 7 and 8.

Thermal regenerator (lab scale fluidized bed)
The thermal regenerator was designed to be a fluidized bed operating in a bubbling regime.A stainless steel A240 TP 304 cylindrical tube [19] with 101.6 mm of outer diameter and 1.5 mm of wall thickness was used as the fluidized bed chamber.The chamber height was of 410 mm.The air distributor plate was manufactured using a stainless steel 75-μm mesh.The distributor was held to the tube by means of two screwed flanges, as shown in Figure 2(a).A coil made of a copper tube 8 mm in internal diameter was built around the fluidized bed chamber, as shown in Figure 2(b).Four infrared ceramic plate LPG burners were placed around the chamber.The power of each burner was 2.6 kW.The test bench is depicted schematically in Figure 3.It is possible to see that the air flow was measured by a flow meter (FT) and controlled by a flow regulator valve.The air flows around the fluidized bed through the coil where it is preheated and then enters the fluidized bed through the distributor, fluidizing the material and promoting the combustion of the phenolic resin around the WFPS grains.ii.
Temperature and computer-based data acquisition Mineral insulation thermocouples (type K -chromium (NiCr) -aluminum (NiAl)) were used to measure temperature.They were positioned next to distributor plate (T1) and close to the burner ceramic plate (T2).The maximum uncertainty for temperature range from 0 °C to 1260 °C was of ± 0.75%, according to E 230 -Standard Specification and Temperature -Electromotive Force (EMF) Tables for Standardized Thermocouples [20].The system was monitored by a microcomputer linked to an Agilent data logger system model 34970. iii.

Regeneration tests and operating conditions
In order to check the performance of the fluidized bed, some tests were carried out with varying air flow rate and retention time.Nine operating conditions were tested (Table 2. Samples code submitted to the thermal regeneration and their respective flow rate and treatment time.).All tests were performed with samples of 1.5 kg of contaminated sand.The sequence of operations for the test was as follows: f.Cooling the sample to equalization system under conditions of temperature and humidity; g. Sample withdrawal.

RESULTS AND DISCUSSION
For each sample treated as described in Table 2, the regeneration efficiency was determined.The mass weight loss after thermal regeneration was compared to a standard.A sample of WFPS was subjected to loss on ignition analysis, which weight loss was taken as the standard for the sand resin conted.The percentage weight loss of each sample as compared with the standard was assumed as the thermal regeneration efficiency of that treatment.
The moisture content of all WFPS samples was less than 1%. Figure 4 shows the results for the Thermogravimetric analysis of the standard sample.The first noticeable mass loss occurred between 60 ºC and 200 ºC.In this stage, the mass loss is associated mainly with water evaporation.Because of the low moisture content, the mass loss in this stage was low, about 0.2%.The most important mass loss was recorded between 200 ºC and 740 ºC.In this stage, the mass loss was of approximately 1.04%.This fraction supposedly consists of the phenolic compounds that volatilize in this range of temperature.The total mass loss was approximately 1.45%.
The result of a DTA analysis of the standard WFPS sample is depicted in Figure 5.This result revealed two expressive peaks at 573 °C and at 594 °C.The first peak is characteristic of an endothermic reaction, detected at 573 °C.According to the literature, this reaction is probably the α-β quartz polymorphic transformation [21].At 594 °C, an exothermic reaction peak was already expected.According to the literature, this consists of the phenolic resin volatilization temperature [11,12].The DTA results were important in order to establish a minimum treatment temperature for the WFPS.In this work, the minimum treatment for thermal regeneration was set as 600 °C.for sand mass density, which was expected since the mass percentage of resin in the phenolic sand is about 1%, as detected in the DTA analysis.
The approximate sphericity of the WFPS used in the present work were determined using Scanning Electron Microscopy (SEM).The original image of a WFPS sample, as shown in Figure 7(a), was treated using Image-J software and converted to binary, as shown in Figure 7(b) and Figure 8.The mean particle sphericity, as defined by Equation 1, was calculated via software.The result for mean sphericity was 0.68.This value is lightly below the values usually found for round sand, which is about of 0.86 [17].
The theoretical value for porosity under minimum fluidization value adopted for calculating the minimum fluidization velocity was determined interpolating experimental values shown in Table 1.Using average particle diameter and sphericity of 0.233 mm and 0.68, respectively, a porosity under minimum fluidization of 0.52 was obtained.
The minimum fluidization velocity was obtained using Equation 5.The properties of atmospheric air according to ISO standard conditions (15°C, 101.3kPa), porosity under minimum fluidization of 0.52, sphericity of 0.68, bulk density of the solid particles of 2514 kg/m³, and mean particle diameter of 0.233 mm were employed.The value of 0.066 m/s was obtained for the minimum fluidization velocity.Using the same values, the terminal velocity was calculated using Equations 6, 7, and 8.A terminal velocity of 0.3111 m/s was obtained.
Using these results, it was concluded that the airflow rate range of operation was set to 1.8 m 3 /h (velocity equal to 0.066 m/s) and 8.6 m 3 /h (velocity equal to 0.31 m/s).This range comprises air velocities enough high to keep the system fluidized assuring that there would not be particle elutriation.In further tests, flow rates of up to 15 m 3 /h have been employed without expressive particle transport, showing that the theoretical terminal velocity was not a definitive value in the prediction of particle carrying by the air flow.11 that the temperatures on T1 and T2 stabilized after 30 minutes of test.At this moment, it may be assumed that the process achieved steady state.Among the treatments employed, as summarized in Table 2, A1, A4 and A7 lasted 15 minutes, i.e., did not achieve the condition of steady state.For samples A2, A5 and A8, the treatment time was of 30 minutes, and for samples A3, A6 and A9 a treatment time of 45 minutes was employed, which was longer than the time do achieve steady state.
In tests of loss on ignition, the regeneration efficiencies of the samples after regeneration treatment were obtained.The mass loss on the third column of Table 3 was compared to the mass loss of a nonregenerated sample, WPFS, which was equal to 1.17%, similar to the value obtained on the thermogravimetric test.The first observation of Table 3 indicates that 15 minutes is not enough time to achieve a good regeneration efficiency.The efficiencies were of 72.6%, 65.0% and 35.9% for the flow rates of 5, 10 and 15 m³/h, respectively.It is also possible to observe that between the treatments A1, A4 and A7, treatments with different air flow rates and low time (less than 30 minutes) the increase in flow rate lowers process efficiency.Before 30 minutes the process did not achieve steady state, and the increase in air flow rate may be causing lower temperatures inside the bed.It is possible to observe the lowest T1 temperature in Figure 9 for the highest air flow rate of 15 m³/h.For treatments of more than 30 minutes, i.e., in which steady state was achieved, the best efficiency was in the condition of air flow rate equal to 10 m³/h, corresponding to air velocity of 0,36 m/s.By Figures 10 and 11, it may be seen that the air enters at a higher temperature when its flow rate is equal to 10 m³/h than when it is of 15 m³/h, justifying a better process efficiency.When comparing treatment times, 30 minutes resulted in higher efficiency than 45 minutes of process, but this difference is of the order of 1 -2%, and it is possible to argue that it falls within measurement uncertainties.The highest efficiency was obtained for air flow rate equal to 10 m³/h for 30 min.This indicates that the maximum efficiency of the process depends on the ideal combination of time, air flow rate, and temperature.
Figure 12 shows the SEM images (150 x magnification) of the samples after the thermal regeneration process performed at the lab scale fluidized bed.The presence of the phenolic resin on particle surfaces in the samples treated for 15 minutes (A1, A4, and A7) is evident.As the residence time of the samples in the bed increases, the resin content on particle surfaces decreases, as observed.There are barely some traces of resin in the surfaces of particles treated for 30 and 45 minutes, as it may be observed in the two right-hand columns of Figure 12.

CONCLUSIONS
The data obtained in the lab tests using fluidized bed bench scale enable quantifying the increase in efficiency (by 96%) in the regeneration of phenolic sands generated in casting process, when compared to the conventional process of mechanical regeneration.This methodology will allow not only the valuation of a material considered as a hazardous waste according to ABNT NBR 10004/2004 [15] to be recycled in the same process, in which it was generated.This application will allow the preservation of a non-renewable natural resource, and avoiding its disposal in in- This study also shows the advantages of a combination of two processes, mechanical regeneration and thermal regeneration, which allows reducing time and eventually temperature of resin removal, because of the partial removal of the resin layer or its weakening during the mechanical regeneration process.

Figure 2 :
Figure 2: Fluidized bed design (a) and fluidized bed prototype (b) designed and built for the experiment.
compressed air flow rate, a control valve was employed, with a calorimetric transducer model SD 6050 with response of 4 to 20 mA, capacity of measuring from 0.2 to 75 m³/h calibrated in accordance with ISO 2533 -Standard Atmosphere (ISO 1975) at standard conditions (101.3 kPa, 15 °C, and relative humidity of 0%).
a. Insert the sample in the fluidizer chamber; b.Air flow adjustment (using flow transducer display); c.Ignition of the heating system (LPG infrared burners); d.Startup data acquisition program in interval defined to test; e. End test;

Figure 6
Figure6shows the result for the granulometric distribution of the WFPS based on the percentage of retained mass.Here, 99% of WFPS particles were retained between meshes of 0.595 mm and 0.074 mm openings.The mean particle diameter (Sauter diameter) for the WFPS was calculated as 0.233 mm.The mass density of the WFPS, determined by picnometry, was of 2514 kg/m 3 .This is a typical value

Figures 9 ,
Figures 9, 10 and 11 show how the temperatures on points T1 and T2 changed over time, as measured by the respective thermocouples.The temperatures were measured at intervals of 1 s.It is possible to see in Figures 9, 10and 11 that the temperatures on T1 and T2 stabilized after 30 minutes of test.At this moment, it may be assumed that the process achieved steady state.Among the treatments employed, as summarized in Table2, A1, A4 and A7 lasted 15 minutes, i.e., did not achieve the condition of steady state.For samples A2, A5 and A8, the treatment time was of 30 minutes, and for samples A3, A6 and A9 a treatment time of 45 minutes was employed, which was longer than the time do achieve steady state.
Figures 9, 10 and 11 show how the temperatures on points T1 and T2 changed over time, as measured by the respective thermocouples.The temperatures were measured at intervals of 1 s.It is possible to see in Figures 9, 10and 11 that the temperatures on T1 and T2 stabilized after 30 minutes of test.At this moment, it may be assumed that the process achieved steady state.Among the treatments employed, as summarized in Table2, A1, A4 and A7 lasted 15 minutes, i.e., did not achieve the condition of steady state.For samples A2, A5 and A8, the treatment time was of 30 minutes, and for samples A3, A6 and A9 a treatment time of 45 minutes was employed, which was longer than the time do achieve steady state.

Figure 7 :
Figure 7: SEM image of WFPS particles (a) and the WFPS binary image (b).

Figure 12 :
Figure 12: Aspect of WFPS samples after bench tests in the fluidized bed under different conditions (by SEM).

Table 2 :
Samples code submitted to the thermal regeneration and their respective flow rate and treatment time.

Table 3 :
Loss on ignition of tested samples