Treatment of tannery wastewater for reuse by physico-chemical processes and a membrane bioreactor

The leather and footwear industry is an important economical branch in Vietnam. Tanneries, which need up to 40 m³ of fresh water per tonne of raw hides (EIPPCB 2013), are a part of this branch. Since the tanning industry has been categorized as one of the most polluting industries, Vietnamese environmental regulations require that each company must install wastewater treatment systems (Thanh 2011). However, there is no differentiation between the effluent standards for different industries, thus chemical oxygen demand COD = 150 mg/L, biochemical oxygen demand BOD₅ = 50 mg/L, ammonium-nitrogen NH₄-N = 10 mg/L, total nitrogen N_tot = 40 mg/L and total phosphorus P_tot = 6 mg/L as major parameters are required for all discharges into receiving waters (MONRE 2011). Since data on actual effluent concentrations are not available, it is difficult to evaluate the current state of industrial wastewater treatment. In 2010, about 65% of the tanneries did officially comply with the regulations. On the other hand, almost all tanneries have been fined for various environmental violations. This indicates that there is still a lack of appropriate and reliable technologies (Thanh & Truong 2013).

The conventional way to treat tannery wastewater is to apply chemical precipitation and aerobic biological treatment in an activated sludge stage (EIPPCB 2013). It has been shown that sequencing batch reactors (SBR) might be better suited for the removal of organic matter measured as COD and nitrogen compounds (Murat et al. 2002). As discussed by Lin et al. (2012), the application of membrane bioreactors (MBRs) was also proven to be technically feasible in laboratory-scale studies. According to Munz et al. (2008), a MBR showed higher COD removal efficiencies (+4%) and provided more stable and complete nitrification than an activated sludge system. Wang & Liu (2011) obtained 88% COD removal using a MBR at a volumetric loading rate of 4.8 kg COD/(m³*d). Ganesan & Muthulakshmi (2015) varied the mixed liquor suspended solids concentration in their MBR. They found that COD removal increased from 75% at 8 g/L to 90% at 12 g/L, but did not give a detailed explanation. There is still a lack of studies with large-scale MBRs and tannery wastewater (Fazal et al. 2015).

The combination of MBRs with an adsorption process for enhanced removal of refractory organic matter from tannery wastewater has been investigated by adding powdered activated carbon to the bioreactor (Munz et al. 2007; Ng et al. 2013). In both studies, small improvements in COD removal were observed, although the main purpose was to improve fouling control and achieve a better filtration performance. Another approach to remove refractory organics was suggested by Giacobbo et al. (2015), who combined a MBR with a photoelectron-oxidation stage. However, the energy demand for the oxidation process in order to achieve considerable COD removal efficiencies was quite high.

Besides the concentration of organic matter in treated tannery effluents, the high salt content can limit the options for reuse. Both nanofiltration (Galiana-Aleixandre et al. 2005) and reverse osmosis (Scholz et al. 2005; De Gisi et al. 2009) have been applied on a small scale in order to produce a permeate that can be reused by the tannery for all process steps.

As a part of the joint Vietnamese–German research project EWATEC-COAST, a treatment concept comprising improved physico-chemical pre-treatment, a MBR as the central stage, and post-treatment by granular activated carbon (GAC) adsorption has been studied on a pilot scale at the Dang Tu Ky tannery in the Nhơn Trạch 1 Industrial Zone (Đồng Nai Province). The treated wastewater should be suitable for reuse purposes in the beamhouse of the factory. Therefore, no salt removal was required whereas a target value of 250 mg COD/L was set for the organic matter. The objective of the study was to demonstrate the feasibility of the three-stage treatment concept in reaching this target.

The feed to the pilot plant was withdrawn from the supernatant of the sedimentation basin of the tannery's technical-scale plant after straining and batch coagulation. Sulphuric acid, ferrous sulphate, poly-aluminium chloride and an organic polymer are added for pre-treatment in order to adjust pH, precipitate both chromium and sulphide, and remove particulate matter by flocculation. Mean values of some important parameters of the pre-treated wastewater are given in Table 1.

According to the data, the conductivity is quite high because of salts that are both washed out from raw hides and used as process chemicals. Compared with the level of organics in raw wastewater, COD is reduced significantly in the pre-treatment stage due to the removal of particulate organic material. This is in accordance with the findings in European tanneries. Moreover, both chromium and sulphide are removed efficiently. The remaining load of the wastewater is primarily made up of dissolved organic matter and nitrogen compounds.

Due to the field conditions encountered, the parameter BOD could not be measured regularly. Therefore, description of the MBR's performance had to be based on COD data. In the laboratory, both the short-term (5 days) and long-term (20 days) BOD was determined. The ratios obtained were COD/BOD₅ = 4.4 and COD/BOD₂₀ = 2.8, respectively. These results are in agreement with the literature data (EIPPCB 2013); however, the difference between both ratios indicates that a certain proportion of the organics was not easily biodegradable.

The design flow rate was 500–2,000 l/h, corresponding to a flux of 3.4–13.6 L/m²*h. Since it was difficult to build up the activated sludge, the reactor was run solely at the lower limit of that range. During some weeks the plant was operated with only one module, and the flux could be doubled without any problems. Thus, the operation of the membranes had to be optimized further for a full-scale application.

The MBR effluent was used to study further removal of dissolved organics by GAC (Carbsorb 30, Calgon Carbon Corp.). This product is based on bituminous coal; the median particle diameter by weight of the material applied was 1.95 mm. The particle density was 0.704 g/cm³ and the filter density 0.420 g/cm³, corresponding to a bed porosity of 0.40. Altogether, 200 L/h of the MBR effluent were fed into a fixed-bed column with an inner diameter of 0.49 m and a bed depth of 1.26 m. The resulting empty bed contact time (EBCT) was 1.2 hr. Such a high value is not unusual for GAC columns operated with high inlet concentrations and small volumetric flow rates (Fettig et al. 1996).

Adsorption of the MBR effluent was also investigated by determining overall isotherms applying the bottle-point method, and by conducting both adsorption rate tests and a rapid small-scale column test (RSSCT). For the isotherm studies, the carbon particles were crushed in a ball mill. The resulting powdered activated carbon had particle diameters below 40 μm. For each test, between 0.01 g and 1.0 g of activated carbon were added to 0.1 L of solution with a pH of 6–7. After 96 hr of contact time on a shaker, samples were filtered through 0.45 μm membranes and analysed for dissolved organic carbon (DOC) and UV absorbance at 254 nm. The solid-phase concentrations were calculated from a mass balance.

The isotherm data were evaluated using the ADSA software (Johannsen et al. 1992). This approach assumes that multi-solute systems can be described as a mixture of a few competing fictive components, based on the IAST model, and a non-adsorbable fraction. Each component is characterized by the parameters K_i and n_i of the Freundlich isotherm and its initial concentration c_o,i. For the sake of simplicity, the n_i values are usually assumed being the same for all adsorbable components (Johannsen et al. 1992).

Granular carbon samples were taken for the rate tests. The short fixed-bed reactor technique was used in order to determine a mean external mass transfer coefficient for the organics’ uptake by GAC. Diluted solutions must be applied here, and UV absorbance was used as a surrogate for DOC. For MBR effluent samples a specific UV absorbance of 2.82 ± 0.8 was determined, thus UV data were considered being suitable as surrogates. From the results, an effective diffusivity of the organics as well as mean film transfer coefficients for other flow rates was derived. Details of this approach are given by Fettig & Sontheimer (1987). A mean intra-particle diffusivity of the organics was determined from a concentration-vs-time curve measured in a 2 L glass beaker. The film-homogeneous diffusion model was fitted to the experimental data assuming the same value of the intra-particle diffusivity for all adsorbable components. The model is described in detail by Sontheimer et al. (1988).

The RSSCT is intended to provide information about the breakthrough in technical-scale adsorbers. For that purpose, a glass column with an inner diameter of 8 mm was filled with 33 cm of 60 × 70 mesh carbon particles which were obtained by crushing and sieving a representative sample of Carbsorb 30. The resulting median particle diameter by weight was 0.23 mm. A peristaltic pump was used to feed the column at constant rate corresponding to a filter velocity of 2.28 m/h and an EBCT of 8.7 min. These operating data were determined according to Crittenden et al. (1991), who proposed to calculate the EBCT for a RSSCT from the EBCT of the large column by multiplying the latter with the ratio of the carbon particle diameters assuming that the intra-particle diffusivity is proportional to particle size. The breakthrough curve was compared with data predicted by the film-homogeneous diffusion model using the isotherm and kinetic parameters determined from batch tests.

COD, ammonium (NH₄-N), nitrite (NO₂-N) and nitrate (NO₃-N) were measured on-site with Merck Spectroquant test kits, and total suspended solids were determined according to DIN 38409 H2. DOC concentrations were measured with a Dimatoc 2000 analyser (DIMATEC), and UV absorbance was determined with a UviLine 9400 spectrophotometer (SCHOTT). pH was measured with a pH meter 539 (WTW).

Therefore, the data were evaluated by selecting one non-adsorbable fraction corresponding to the Freundlich coefficient K₁ = 0, and three adsorbable fractions corresponding to K₂ = 6, K₃ = 12 and K₄ = 22, respectively, with the Freundlich exponent being n = 0.3, and fitting the IAST model to both isotherms simultaneously. The resulting fractions of each component (c_o,i/c_o,total) listed in Table 2 reveal that the non-adsorbable fraction is about one-fifth (20.8%) while the major fraction (37.6%) is found for the Freundlich coefficient K₄ = 22. This means that a significant amount of the organics is fairly well adsorbable.

Accordingly, the mean molecular weight is about 1,000 Da for the adsorbable fraction of the MBR effluent. This can be attributed to the occurrence of humus-type species formed during degradation processes in the MBR, and to refractory organics like tannins in raw wastewater (Munz et al. 2009). The effective diffusivity was also used to calculate a mean film transfer coefficient for the conditions of the RSSCT test.

The intra-particle diffusivity was derived from concentration-vs-time data of a rate test conducted with a sample of the original GAC particles. The test lasted for 192 hr, and more than 60% of the organics initially present in the water phase were adsorbed. The value of 0.9*10⁻¹³ m²/s found for the intra-particle diffusivity is quite small and resembles parameters obtained for humic substances (Sontheimer et al. 1988).

The adsorption data reveal that activated carbon has a limited potential for polishing the MBR effluent. In a tannery comparable to Dang Tu Ky, the average freshwater water demand is about 500 m³/d, and the demand for the beamhouse processes (soaking and liming) is about 210 m³/d. Thus the reuse potential can be estimated to 42% of the total freshwater demand. As mentioned earlier, the main reason for this limitation is the requirement for the discharge of salts from the water circuit when reusing water for other process steps in the tannery.

For the GAC stage, the operation was simulated with the adsorption model described in the previous section. It was proposed to set up three adsorption columns in series with 2 m of bed length each. Varying the filter velocity revealed that the carbon capacity could be utilized better when the filter velocity was <2 m/h. Based on the simulation results a value of 1.8 m/h was chosen that corresponds to a total EBCT of 3.3 hr. The column diameter will be 2.5 m and the total mass of carbon will be about 12,375 kg. Although the MBR effluent concentration was sometimes as high as 800 mg COD/L (Figure 2), an average influent value of 550 mg COD/L to the GAC stage was assumed. This number is based on the mean influent concentration to the MBR (Table 1) and the average COD removal efficiency. When the effluent concentration reaches the limit of 250 mg COD/L, the first column will receive fresh carbon, then it will be put in the last position. According to the simulation, a usable carbon capacity of at least 50 g DOC/kg should be obtained in this way.

During the operation of the pilot-scale adsorber the mean influent COD and DOC values were 489 mg/L and 149 mg/L, respectively, resulting in a factor of 3.3 between both parameters. Therefore, the carbon capacity of 50 g DOC/kg corresponds to about 165 g COD/kg. Under these conditions, the running time of one column after the starting-up phase will be about 33 d and the specific carbon demand on the order of 1.8 kg/m³. This value is quite high, therefore, it might be questioned whether GAC is the best process for polishing MBR effluent. However, there are not many alternatives to remove refractory organics. Advanced oxidation processes have been studied on a laboratory scale for tannery wastewaters (Lofrano et al. 2013), but they require expensive oxidation chemicals and are more complicated to operate than a simple GAC system. Therefore, it will depend on the local conditions whether the approach described herein to produce water for reuse is both economically and ecologically reasonable.

• (1)

  In the MBR, dissolved organic matter could be removed from pre-treated tannery wastewater by 81% on average. The reactor did not reach its maximum capacity at a volumetric loading rate of 1 kg COD/(m³*d). Meanwhile, nitrification was obviously an unstable process although more than 70% ammonium removal was obtained during some periods. The average removal of total nitrogen was 36%.

• (2)

  Adsorption onto activated carbon was suited to remove refractory organic substances from the MBR effluent. Adsorption equilibria could be described by applying the IAST model and a fictive component approach. More than 40% of the organics were either non-adsorbable or weakly adsorbable. The effective diffusivity estimated from short fixed-bed reactor data indicated an average molecular weight of about 1,000 Da. The intra-particle diffusivity was on the same order as kinetic parameters found for humic substances, thus the uptake rate on GAC was rather slow. The breakthrough curve determined for a pilot-scale adsorber could be predicted successfully by the modelling approach based on the film-homogeneous diffusion model while there were some deviations with the RSSCT.

• (3)

  For the outline of a technical-scale plant, a volumetric loading rate of 3 kg COD/(m³*d) was assumed for the MBR. Three GAC columns in series were to be applied for polishing purposes. However, the usable carbon capacity was estimated to be rather low, and as a result, the specific carbon demand would be about 1.8 kg/m³.

Project funding by the German Federal Ministry of Education and Research (BMBF) under grant no 02WCL1217B, and by the National University of Ho Chi Minh City is gratefully acknowledged. The authors would like to thank Thomas Lücking, Stefan Nordbruch and Nguyen Thi Thanh Phuong for their support, and Claudia Steinert, Tilman Steinert, Henning Zeich and Nguyen Cong Vu for their technical assistance and their dedication to the project.