Environ. Sci. Technol. 2002, 36, 3335-3339
Wet Oxidation Lumped Kinetic Model for Wastewater Organic Burden Biodegradability Prediction SVETLANA VERENICH* AND JUHA KALLAS Department of Chemical Technology, Lappeenranta University of Technology (LUT), P.O. Box 20, Lappeenranta, FIN-53851 Finland
In many cases, treatment of wastewaters requires a combination of processes that very often includes biological treatment. Wet oxidation (WO) in combination with biotreatment has been successfully used for the treatment of refractory wastes. Therefore, information about the biodegradability of wastewater solutes and particulates after wet oxidation is very important. The present work proposes a model that can describe the oxidation process via organic concentration characteristics such as chemical oxygen demand (COD), biochemical oxygen demand (BOD), and immediately available BOD (IA BOD) and so can allow the prediction of biodegradability (i.e., BOD/COD ratio). The reaction mechanism includes the destruction of nonbiodegradable substances by two pathways: oxidation to carbon dioxide and water and oxidation to larger biodegradable compounds with their further degradation to smaller ones measured via IA BOD. The destruction of small biodegradable compounds to end products is also included in the model. The experiments were performed at different temperatures (170-200 °C) and partial oxygen pressures (0.5-1.5 MPa) in a batch stainless steel highpressure autoclave. The model of concentrated thermomechanical pulp circulation water was selected for the experiments. The proposed model correlates with the experimental data well and it is compared with other WO models in the literature.
Introduction and Background As a result of concerns about global increase in water pollution, industrial plants are trying to maximize the use of water inside the plant and minimize the discharge of water effluents into the environment (1, 2). To meet such ends, many treatment processes and combinations of processes are used (3, 4). The combination of processes provides a more complete degradation or elimination of pollutants from the water phase (4-7). A combination of oxidative methods with biological treatment has been found to be very effective, especially when oxidation techniques are used as the pretreatment process for the biological step (8-10). Wastewaters that are too dilute to be incinerated or too toxic to be biologically treated, for instance membrane filtration concentrates at pulp and paper mills (11), can be directed to wet oxidation (WO). This method requires elevated temperatures (125-320 °C) and pressures (0.5-20 MPa) to degrade refractory and toxic organics by oxygen to a state where they are biologically treatable. The process of wet * Corresponding author phone: + 358 5621 2147; fax: +358 5621 2199; e-mail:
[email protected]. 10.1021/es010244z CCC: $22.00 Published on Web 06/25/2002
2002 American Chemical Society
oxidation is rather complicated, especially if it involves the oxidation of industrial wastewaters containing enormous numbers of organic compounds. Therefore, many lumped kinetic models have been proposed to predict optimum operating conditions. Several successful attempts have been made to lump parameters in order to predict organic degradability for different types of the wastewater. These are general lumped kinetic model (GLKM) (12), lumped kinetic model (LKM) (13), extended lumped kinetic model (ELKM) (14), and lumped kinetic model for oil wastes (LKM-OW) (15). GLKM describes oxidation in three pathways that includes oxidation of organics (except acetic acid) to the end products of the reaction and the subsequent oxidation of acetic acid to carbon dioxide and water. LKM eliminates one of the reaction pathways and includes only the formation of carbon dioxide, water, and partially oxidized compounds from the organic compounds initially present in the wastewater. Lo´pez Beranal et al. have developed LKM-OW and they assume that WO has three steps and that the last two are parallel. The mechanism of the reactions can be expressed as
where A is oil/greases, B is degradable compounds, C is the end products of oxidation, and R is refractory products. Recently, Belkacemi et al. have proposed another lumped model for catalytic WO (CWO). Their ELKM is elaborated for heterogeneously catalyzed processes and involves six elementary steps, as shown,
where A is aqueous solutes, B is soluble intermediates produced by CWO, and C is all solid and gaseous end products. A*, B*, and C* are chemisorbed pseudo-species during surface reactions. Such a model was shown to be capable of predicting the time profiles of various lumped as well as the kinetic and adsorption/description equilibrium parameters. These briefly described models do not include changes in the biodegradability of treated liquid wastes during WO. This is important for subsequent biotreatment, which is the cheapest way to reduce the organic matter content in wastewaters. Therefore, the present work proposes a model that includes organics degradation to end products. It employs commonly used parameters to characterize wastewater, including chemical oxygen demand (COD), biochemical oxygen demand (BOD), and immediately available BOD (IA BOD). These characteristics are then used for biodegradability evaluation.
Kinetic Model In our previous work (16), the lumped kinetic model simplified the reaction system into three lumps, namely, (i) lump A is for unstable and refractory compounds that can VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3335
be expressed by the subtraction of BOD from COD, (ii) lump B represents all biodegradable compounds incorporated in the BOD characteristic, and (iii) lump C denotes all gaseous end products formed and can be simply expressed via difference in COD at time “zero” and COD observed during the oxidation process (i.e., COD0 - COD). In this paper, this lump will subsequently be marked in the Figures as CO2. Oxidation of the refractory organic compounds, lump A, is assumed to proceed in two parallel ways: in one of them, they are oxidized to the end products of the reaction, lump C; in the other, the wastewater compounds are partially oxidized and turn into biodegradable form, lump B, which is followed by further oxidation to the end products (i.e., carbon dioxide and water). Our present approach proposes a modification in the model by including one more component in the series of the reactions, namely, IA BOD. Therefore, the reaction network includes the two new lumps defined as follows: (i) lump BS designates the small biodegradable compounds (17) as carboxylic acids, aldehydes, alcohols and so forth and characterizes with IA BOD; (ii) lump BL incorporates rest of the biodegradable compounds which are not a part of lumps A, BS, or C and expressed as difference of BOD and IA BOD characteristics. The pathways of the reactions would be the following:
The refractory compounds, A, undergo two parallel reactions: one of them leading to the formation of the oxidation end products, lump C, and the second reaction to organic compounds degraded to biologically oxidizible large molecules, lump BL. The latter group undergoes further transformation to smaller biodegradable compounds or BS. The last step of the model considers the oxidation of BS to the oxidation end products. The rate equations of the oxidation pathways for each category of the organics can be expressed via a system of four ordinary differential equations (ODEs) with initial conditions of A[0] ) A0, BL[0] ) BL0, BS[0] ) BS0, and C[0] ) 0:
d[A] ) (k1 + k2)[A] dt
(4)
-
d[BL] ) k3[BL] - k2[A] dt
(5)
-
d[BS] ) k4[BS] - k3[BL] dt
(6)
-
d[C] ) -k4[BS] - k1[A] dt
(7)
-
where k1, k2, k3, and k4 equal to
( )
ki ) Ai exp -
Ei [O2]γi RT
(8)
It should be noted that the reaction for organic matter presented in this model was assumed to be first-order according to Li et al. (12) and Zhang and Chuang (13). In the parameter estimation, the best values for the parameters in the model are searched so that the fit given by the model is as close to the experimental data as possible. The objective function to be minimized consists of the square sum of the lumped values at each step of the reaction network 3336
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
Fobj )
(
∑
)
Ci,exp - Ci,calc Ci,exp
2
(9)
where Ccalc denotes the concentrations of lumped values predicted by the model, Cexp represents the lumped values obtained during the experimental runs, and i designates A, BS, BL, or C lumps. The Ccalc values of each lump for each iteration cycle are obtained by numerical integration of eqs 4-7. The ODEs were solved with automatic method for solving stiff and nonstiff ODE problems implemented in LSODE software (18, 19). A nonlinear constrained global optimization method with random search developed by Palosaari et al. (20) was applied for the parameters (rate constants) estimation. Afterward, the reaction orders with respect to oxygen and the activation energies were derived from the optimized rate constants obtained with the independent data set at varied oxygen pressures or temperatures.
Experimental Section Materials. For the present research, a model of concentrated thermomechanical pulp (TMP) circulation water was prepared using TMP pulp obtained after second refining from a pulp and paper mill. The preparation of the water studied is a laboratory-scale imitation of the real industrial process, and the preparation steps are described in one of our studies (21). A concentrate with the following properties was obtained: pH of 5.2-6.0, 6000-8000 mg of O2/L of COD, 1200-1900 ppm of soluble total organic carbon (TOC), 300350 mg/L of lignin/tannin, 0.2-0.35 of wastewater biodegradability defined as a ratio of BOD/COD, and 0-200 mg of O2/L of IA BOD. Experimental Procedure. The batch experiments were carried out in a 0.2 L stainless steel high-pressure reactor (Parr Instrument Co., Moline, IL) equipped with an electromagnetic stirrer for mixing. A thermal sensor and an external heating element were provided to control the reaction temperature. The impeller operated at 900 rpm, at which no mass transfer limitation was encountered. For all experiments, the cold reactor was loaded with 175 mL of wastewater. After preheating to the desired temperature, the oxidation process was initiated by dispersing pure oxygen obtained from an oxygen cylinder into the preheated solution. Six liquid samples were withdrawn at designated time intervals during the whole reaction period. After a 2-h run, the reaction was stopped by rapidly cooling the reactor with cold water. As no thermal degradation was observed during preheating period, COD, BOD, and IA BOD analyses were undertaken only for the samples extracted from the reactor during oxidation process in order to determine the extent of the oxidative reaction. Analytical Methods. COD was analyzed by the closed reflux dichromate method (22) using a COD reactor (Hach Company, Loveland, CO) and direct-reading spectrophotometer DR/2000 (Hach Company). The inaccuracy in the assays, calculated as relative standard deviation of two separate measurements, did not exceed 2%. BOD values for the organic components of the wastewater was determined with a methodology and device developed by Dr. Lange & Co. (Germany) that allows obtaining BOD values identical to BOD5 (23). The method is based on breaking down large molecules by the use of a thermostat (LT100) at a temperature of 148 °C and SensorBOD with a measuring cell (see the BOD measurement, Supporting Information).
Results and Discussion The wet oxidation experiments were carried out at a temperature range of 170-200° C and an oxygen partial
TABLE 1. Evaluated Kinetic Parameters as Oxygen Reaction Orders, Activation Energies, and Frequency Factors for Each Reaction Pathway Defined in the Model number of reaction
reaction order (γ)
regression coefficient ( r 2)
activation energy, E (kJ mol-1)
frequency factor, A (L mol-1 min-1)γ
regression coefficient (r 2)
1 2 3 4
1.8 1.5 0.6 0.8
0.99 0.98 0.97 0.90
87.5 42.5 89.9 20.0
3.98 × 1011 2.31 × 105 5.79 × 109 34.1
0.94 0.99 0.99 0.91
FIGURE 2. Plot for determination of activation energies and frequency factors for each reaction pathway of the proposed model.
FIGURE 1. Effect of oxygen partial pressure on concentration of biodegradable compounds at a reaction temperature of 170 °C during 120 min of oxidation process. Panel A indicates the changes in biodegradability of the concentrated TMP water. Panel B shows the increase of small molecules in the water measured as IA BOD/ COD ratio. pressure of 0.5-1.5 MPa in order to investigate the influence of these parameters on the oxidation rate of the wastewater studied. Influence of Oxygen Partial Pressure. The results obtained at the same temperature, 170° C, and different oxygen partial pressures of 0.5, 0.75, 1, and 1.5 MPa indicate the positive effect of oxygen partial pressure on the removal of organic mater and formation of biodegradable organics in the wastewater, as shown in Figure 1. As can been seen, the increase in oxygen partial pressure increased the biodegradability of the treated waste. At a pressure, PO2, of 1.5 MPa, the BOD/COD ratio reached about 90% already after 90 min. It is worth noting that about 65% of organics present in the water after a 2-h treatment correspond to small biodegradable compounds measured via IA BOD (Figure 1B). Increasing oxygen concentration in the water phase should increase the amount of radicals in the wastewater and, consequently, the amount of split biodegradable fragments (12). To evaluate the order of the reaction with respect to oxygen, the rate constants of the lumped biodegradable compound kinetic model (LBCKM ), k1, k2, k3, and k4, obtained
by solving system ODEs (4-7) with further optimization of their values were applied to the logarithmic form of eq 8. Plotting ln ki against ln[O2] gave a straight line with slope γi and Table 1 presents the orders of the reactions with respect to oxygen, γ1, γ2, γ3, and γ4. The obtained values indicate the strong influence of oxygen at the beginning of the oxidation process. Despite the “zero” value for the order of the reaction with respect to oxygen reported by other researchers and assumptions made in the models, our data indicate a value of 0.6-1.8. However, the subsequent oxidation of organic material from group A to BS via BL makes it more resistant to chemical oxidation. Therefore, it was found reasonable to expect some decrease in the dependency on oxygen partial pressure at each step of the reaction chain (24). Influence of Temperature. An increase in temperature increased the degradation of COD and improved biodegradability (i.e., BOD/COD) of the wastewater. Rate constants, k1, k2, k3, and k4, obtained from the experiments at identical oxygen partial pressure, 1 MPa, and different temperatures were used to evaluate the parameters for the Arrhenius equation as activation energies and frequency factors. To evaluate these constants, the logarithmic form of the eq 8 can be represented in the following way:
ln ki - γi ln[O2] ) ln Ai -
Ei 1 R T
(10)
Thus, the activation energies and frequency factors could be found from the graph of ln ki - γi ln[O2] versus 1/T (Figure 2) as the slope of the lines and intersection of the lines with the axis, respectively. The estimated parameters for LBCKM are presented in Table 1. As can be seen, the studied wastewater has shown resistance toward the direct oxidation to the carbon dioxide and water, and the oxidation process proceeded via intermediates of more biodegradable compounds. As was observed earlier (25, 26), the oxidation of the organic compounds proceeds via consequence-parallel reactions with formation of carbon dioxide and more resistant substances. Therefore, the last estimated parameter, E4, VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3337
TABLE 2. Values of Biodegradability, BOD/COD Ratios, Observed during the Experimental Runs and Predicted by the Model BOD/COD: Experimental/Predicted pressure (MPa)
1
1
1
T (°C) time (min)
200
180
170
0 10 60 120
0.24/0.24 0.39/0.38 0.89/0.91 0.98/0.99
0.36/0.36 0.41/0.43 0.71/0.71 0.88/0.89
0.29/0.29 0.31/0.31 0.41/0.45 0.62/0.59
pressure (MPa)
1.1
0.75
0.5
time (min)
186
170
170
0 10 60 120
0.28/0.28 0.36/0.35 0.59/0.62 0.82/0.82
0.32/0.32 0.34/0.37 0.46/0.46 0.54/0.58
0.19/0.19 0.20/0.20 0.28/0.28 0.37/0.37
T (°C)
FIGURE 4. Comparison of data prediction by LBCKM with the published earlier models: LKM and GLKM.
FIGURE 3. Model’s performance for the experimental data observed under following conditions: (A) at a temperature of 190 °C and 1 MPa of oxygen partial pressure; (B) 186 °C and PO2 of 1.1 MPa; (C) 170 °C and PO2 of 1.5 MPa. represents the lumped activation energy for disappearance of the BS. If the last lump consists of compounds that are able to release CO2 with the formation of carboxylic acids, the model described in this work will predict the WO process truly. As can been seen from the model scheme, the oxidation of recalcitrant compounds lumped in A proceeds in two parallel reactions which allows us to introduce the “selectivity point”, R, as a k2/k1 ratio (12). Values in the range 0.15-0.38 were obtained for this parameter, the value increasing with decreasing temperature. The rise in temperature shifts the reaction to complete oxidation with formation of carbon dioxide and water. A similar phenomenon was observed in the case of different oxygen partial pressures. The less pressure applied, the higher the value for the selectivity point observed. At a temperature of 170° C and oxygen partial pressure of 0.5 MPa, selectivity was found to be about 0.75. Generally, however, the reaction rates of each oxidation 3338
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
pathway lessened with a decrease in temperature and pressure. Model Validation. Once the activation energies and reaction orders have been obtained, the concentration profiles against time can be obtained. Figure 3 depicts some experimental results at temperatures of 190, 186, and 170° C and oxygen partial pressures of 1, 1.1, and 1.5 MPa as well the results of modeling to this set of experimental data. The BOD and COD model data were attained by adding lumped values of BS to BL and the sum of BS and BL was then added to the A component, respectively. As can been seen, data generated by LBCKM correlate with the experimental results well. The ratios of BOD/COD achieved during the experiments and predicted by the model are presented in Table 2. As it is difficult to obtain data for model validation published by other researchers with the data set from BOD test results, the already existing kinetic models were employed to predict our set of data and also to compare the accuracy of data prediction by LBCKM with selected models. LKM and GLKM were chosen for this purpose. However, as these models cannot predict the concentration of biodegradable matter only COD profiles were generated for all three models. The simulation data produced by the proposed model, LKM and GLKM are depicted in Figure 4. Examination of the models’ performances shows that LKM describes the data slightly better than LBCKM and GLKM. Good precision in data prediction was found at low temperatures for the proposed model with an insignificant decrease in accuracy
following at higher temperatures. The decrease in the precision at 200° C indicates the limitation for the use of the model at higher temperatures, where more intensive oxidation of the resistant compounds such as carboxylic acids begins. The last group BS consists of a mixture of organic materials, including also carboxylic acids, and the treatment at higher temperatures increases their fraction in BS. As a consequence, the disappearance rate of the BS lump decreases and this can cause some inaccuracy in prediction of the COD profile. Therefore, one must be aware of additional regrouping in the model at higher temperatures. Despite the model’s limitations, LBCKM has the advantage of BOD prediction and determination of the optimum time for the WO treatment in order to provide the best ratio of BOD to COD (usually up to 0.6) for successful subsequent biotreatment.
Acknowledgments We would like to acknowledge the Graduate School in Chemical Engineering (Finland) for financial support of this research. We are also thankful to Veronica Garcia Molina MSc. from Universitat de Barcelona for carrying out most of the experiments.
Supporting Information Available A description of the BOD measurement procedure with Dr. Lange’s SensorBOD (Germany). This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Gleadow, P.; Hastings, C.; Richardson, B.; Towers, M.; Uloth, V. Tappi J. 1998, 81 (9), 199. (2) Allen, L.; Polverari, M.; Levesque, B.; Francis, W. Tappi J. 1999, 82 (4), 188. (3) Wastewater Engineering: Treatment, Disposal, and Reuse, 3rd ed.; Clark, B. J., Morriss, J. M., Eds.; McGraw-Hill: New York, 1991. (4) Dalan, J. A. Chem. Eng. Prog. 2000, (Nov), 71.
(5) Mantzavinos, D.; Hellenbrand, R.; Livingston, A. G.; Metcalfe, I. S. Can. J. Chem. Eng. 2000, 78, 418. (6) Hodgson, A. T.; Hitzroth, A. J.; Premdas, P. D.; Hodson, P. V.; Duff, S. J. Tappi J. 1998, 81 (2), 166. (7) Chen, W.; Horan, N. J. Environ. Technol. 1998, 19, 173. (8) Scott, J. P.; Ollis, D. F. Environ. Prog. 1995, 14 (2), 88. (9) Chakchouk, M.; Hamdi, M.; Foussard, J. N.; Debellefontaine, H. Environ. Technol. 1994, 15, 323. (10) Roy-Arcand, L.; Methot, M.; Archibald, F. S. Tappi J. 1996, 79 (6), 133. (11) Verenich, S.; Laari, A.; Kallas, J. Waste Manage. 2000, 20 (4), 287. (12) Li, L.; Chen, P.; Gloyna, E. F. AIChE J. 1991, 37 (11), 1687. (13) Zhang, Q.; Chuang, K. T. AIChE J. 1999, 45 (1), 145. (14) Belkacemi, K.; Larachi, F.; Sayari, A. J. Catal. 2000, 193, 224. (15) Lo´pez Bernal, J.; Portela Migule´z, J. R.; Nebot Sanz, E.; Martı´nez de la Ossa, E. J. Hazard. Mater. 1999, B67, 61. (16) Verenich, S.; Kallas, J. Presented at IWA 2nd International Conference on Oxidation Technologies for Water and Wastewater Treatment, Clausthal-Zellerfeld, Germany, May 2000; Paper R 028. (17) Manual for ECM SensorBOD, BDA325/edition 11.98; Dr. Lange & Co.: Du ¨ sseldorf, Germany, 1998. (18) Hindmarsh, A. C. Lawrence Livermore Laboratory of Scientific Computing; Stepleman, R. S., et al., Eds.; North-Holland Publishing Company: Amsterdam, The Netherlands, 1983; pp 55-64. (19) Petzold, L. R. Siam. J. Sci. Stat. Comput. 1983, 4, 136. (20) Palosaari, S.; Parviainen, S.; Hiironen, J.; Reunanen, J.; Neittaanma¨ki, P. Acta Polytec. Scand. 1986, 172, 1. (21) Verenich, S.; Kallas, J. Chem. Eng. Technol. 2001, 1183. (22) Standard Methods for the Examination of Water and Wastewater, 8th ed.; Eaton, A. D., Clesceri, L. S., Greenberg, A. E., Eds.; APHA: Washington, DC, 1995; Part 5000. (23) Reiter, C. Allgemeine Papier-Rundschau 1995, 18, 449. (24) Donlagic, J.; Levec, J. Ind. Eng. Chem. Res. 1997, 36, 3480. (25) Imamura, S. Ind. Eng. Chem. Res. 1999, 38, 1743. (26) Donlagic, J.; Levec, J. Environ. Sci. Technol. 1998, 32, 1294.
Received for review September 26, 2001. Revised manuscript received April 22, 2002. Accepted April 29, 2002. ES010244Z
VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3339
