Analysis of flow patterns in an activated sludge reactor

Transcription

1 Analysis of flow patterns in an activated sludge reactor A.B. Karama\ O.O. Onyejekwe*, CJ. Brouckaert^ and C.A. Buckle/ Department of Civil Engineering, University ofdurban-westville, * Department of Chemical Engineering, University ofnatal, Abstract Physical model tests to determine ways of improving the performance of sedimentation tanks and clarifiers are limited by the cost of constructing adequate models in a laboratory setup. Most of the standard tests carried out in the past were restricted by their inability to simulate actual flows. The study of flow characteristics are of considerable interest in many fields of engineering and physical sciences. In most approaches of simulation of activated sludge, the goal is to reach a good simulation of the mean value of COD, BOD etc. In our approach, we used a commercial computational fluid dynamics package to simulate velocity flows in an anaerobic zone of activated sludge reactor. Experimental results were obtained from oxidation reduction potential and tracer studies of residence time distribution. The experimental results were then verified by the computational fluid dynamics techniques. Introduction The demand for water has greatly increased while the available amount of water is limited. As a result, water and waste-water treatment plants have to be designed or operated to attain higher efficiencies. The main reasons for treating waste-water are to prevent pollution in order to protect the environment and to protect public health by safeguarding water supplies against the spread of water-borne diseases. With the increasingly stricter demands on discharged waste-water quality, the modelling and control of activated sludge (herein referred to as AS) processes have become a subject of numerous studies with growing interest. AS systems have been studied from different modelling approaches with the objective of a better understanding and prediction of the processes. One of the

2 164 Computer Methods in Water Resources IV approaches was made by Kabouris and Georgakakos* in estimating model parameters, kinetic and stoichiometric parameters, in an AS process. However, although their method did have plausible results, it is not suited for the detection of sudden parameters variation having time constants of minutes and identification of toxic inputs. Another contribution of a steady-state model for the single sludge AS system with or without nitrogen was proposed by Argaman^. The model addresses both fractions, the soluble fraction of the waste-water organics and the particulate fraction, which can be used for design of a new system or for analysing the performance of an existing one. However, the model is restricted to constant flow and load conditions and requires a set of 23 equations to be solved simultaneously and 35 design and operational variables. Garcia-Olivares and Becares^ developed a mathematical model and calibrated for a two stage AS pilot plant treating chemical-pharmaceutical industry waste-waters. They used two series of simulation having constant and variable parameter values in predicting the dynamic behaviour of the substrate and reactor suspended solids. Fairly accurate results were obtained. Development in AS processes have led to a number of researchers finding and predicting ways to improve on treating waste-waters. Dalmacija et al.^used AS in treating high salinity waste-water. Zhao et al/ improved on nitrogen removal using AS. Tian et al.* investigated on the accumulation of AS at low temperatures. A dead zone in AS reactor has an important function because it maintains a wide variety of fauna and flora in the reactor. Accurate prediction of velocity flow near the dead zone must therefore be determined for effective management of the reactor before constructing artificial structures for improvement. A number of studies have investigated sediment distribution and flow patterns in sedimentation tanks and clarifiers. Schamber and Larock^ and Abdel-Gawad and McCorquodale developed a mathematical model that can be used to predict flows satisfactorily in sedimentation tanks. Adams and Rodi^ used k-e model in their analysis to determine the hydrodynamic behaviour of flow in sedimentation tanks with simple geometries. To validate their model, they used dye concentrations to obtain flowthrough curves. DeVantier and Larock^ presented a model to simulate the formation and behaviour of density currents and the surface return flow in a circular secondary clarifier. Imam and McCorquodale^^ developed a model to simulate flow in a rectangular clarifier operating at neutral density conditions. Reasonable results were obtained when the model results were compared with experimental data. Sediment concentration and eddy viscosity were further introduced into the model to study the effect of reaction baffle submergence on solid removal and to simulate the settling of discrete particles. Experimental results showed reasonable agreement with the model when used for steady and unsteady solid transport. In a later work, Zhou and McCorquodale^ presented a numerical model to predict concentration distribution and the associated flow pattern for non uniform solids with high influent concentration in a secondary rectangular settling tank. The model was verified using three field applications and gave fairly close results.

3 Computer Methods in Water Resources IV 165 Using circular clarifiers, Zhou and McCorquodaie^ formulated a numerical model to predict the velocity and concentration distribution for anon uniform flocculated particle for turbulent density stratified flow in the clarifier. They used the eddy viscosity concept and the k-e turbulence model to calculate turbulent stresses. The model gave satisfactory results on density variations and solids concentration distribution when tested with experimental data. The main objective of this research is to use computational fluid dynamics techniques (herein referred to as CFD) to predict flow patterns in the anaerobic zone of an activated sludge reactor. CFD techniques compliment experimental and theoretical fluid dynamics by providing an alternative cost-effective means of simulating real flows in the reactor. This is as a continuation of a study made by Brouckaert^ using CFD techniques to model water and waste-water treatment systems. In this study, the 5-stage (anaerobic, primary anoxic, primary aerobic, secondary anoxic and secondary aerobic) Bardenpho reactor is used. The reactor is 4.5 m deep with a radius of 23.5 m and consists of a centrally placed clarifier of radius 10.5m as shown in Fig. 1. A large volume of industrial waste-water is treated by the reactor which consists of approximately 70 % textile effluent, 25 % chicken abattoir and 5 % domestic waste. The research will concentrate on the anaerobic zone where raw sewage and return sludge are fed into the reactor. Return sludge ^^ Return sludge Raw sewage (a) Figure 1 (a): Plan view of the 5-stage Bardenpho reactor.

4 166. Computer Methods in Water Resources IV INI (b) Figure 1 (b): Isometric view of the anaerobic zone. INI is a 400mm dia. raw inlet pipe. IN2 and INS are 75mm dia return sludge inlet pipes. Outline of laboratory tests Two laboratory tests were conducted at the anaerobic zone of the reactor viz. the oxidation reduction potential (ORP) and the residence time distribution (RTD) tests. Oxidation reduction potential test This test was conducted by BelP in order to provide more information on the uniformity of mixing in the anaerobic zone. He injected a dye at the inlet of the reactor and selected 12 different points to draw out samples and measure their ORP at the electrode which was fitted to a flow cell.

5 Residence time distribution test Computer Methods in Water Resources IV 167 To have better understanding of the dead zone in the anaerobic zone of the reactor, tracer studies were carried out to determine the RTD. Lithium chloride tracer was dosed at the inlet of the reactor and samples taken at different time intervals at the outlet. Concentration of lithium in the samples was determined and the results were then modelled by using a computer program, IMPULSE, developed by Brouckaert et al.^. IMPULSE is a computer program that allows modelling of systems using curves from tracer tests. The user assumes a flow model for the system and the program determines the theoretical response curve for the model to fit the experimental curve. Outline of the computational fluid dynamics model The CFD model that was used to predict flow patterns is a commercially available package known as PHOENICS. PHOENICS was developed by Concentration, Heat and Momentum** (CHAM) and can simulatefluidflow,heat transfer and chemical reactions. The user normally feeds in the data into the model e.g. geometry,fluidproperties, boundary conditions etc. and then runs the program. The model solves the Navier-Stokes equation by integrating the differential equation over a finite volume of a computational cell and (for transient problem) over a finite time and approximating the resulting volume, area and time by way of interpolation assumptions. Results and discussion From the ORP experiment, all the points monitored were found to have an average of ORP between -320 mv and -340 mv. Based on previous studies made by Carliell^, these values are way up above the average value of-450 mv required for rapid decolourisation of the highly coloured textile effluent. Hence, there was no full decolourisation and this shows that there might be stagnant regions in the reactor. With the RTD experiment, results indicate that the residence time for lithium after simulation with IMPULSE was found to be 57 min and that the volume used in tracer studies was 64 % of the actual volume of the anaerobic zone. This means that the 36 % of the volume was not utilised by tracer movement. Hence, there is a dead zone in the anaerobic compartment where hardly any flow is taking place. Velocity vectors of the flow simulation are presented in Figs. 2 (a), (b) and (c). The presence of the outlet is as shown in Fig. 2 (a). Note the shape of the velocity profiles in this region. A recirculation zone is formed near the outlet and this could be due to the high velocity of the return sludge. In Fig. 2 (b), strong currents of velocity are felt at the solid boundary and a recirculating flow occurs closer to the

6 168 Computer Methods in Water Resources IV (a) (c) Figure 2: (a) Simulated velocity flow pattern at the bottom of tank (b) Simulated velocity flow pattern at the centre of tank (c) Simulated velocity flow pattern on the surface of tank. inlets. This shows fairly good mixing is taking place though the area of recirculation seems to be dead. Fig. 2 (c) shows strong velocity flow next to the outlet wall and this could be attributed to the return sludge impinging on the wall surface and forcing the flow to move upward. A recirculation zone is also formed similar to that at the centre of the zone. Contour plots of vertical velocities are shown in Figs. 3,4 and 5. Fig. 3 shows negative velocity at the solid boundary next to the raw sewage inlet as expected. The velocity gradient are largest at the solid boundaries away from the inlet and this is due to the high inlet velocity of the return sludge which is forming recirculating flows. Fig. 4 shows a large portion where hardly any flow is taking place at the centre of the tank. This when compared with Figs. 2 (b) and 2 (c), clearly shows the presence of a dead zone at the centre. As expected, Fig. 5, shows a negative flow at the outlet. The velocity distribution is not uniform and is mostly pronounced at the centre.

7 Computer Methods in Water Resources IV 169 Figure 3. Longitudinal cross-section of simulated vertical velocity contours at the raw sewage inlet position. Figure 4. Longitudinal cross-section of simulated vertical velocity contours at the centre of the tank. Figure 5. Longitudinal cross-section of simulated vertical velocity contours at the outlet position.

8 170 Computer Methods in Water Resources IV Conclusion A CFD model has been used successfully to predict flows and the dead zone area of an anaerobic section of an AS reactor. In addition, two experimental results were also verified by the model. It can be deduced that the experimental data are in a fairly good agreement with velocity flow distributions of the model. Acknowledgement: This research was supported by a grant from the Water Research Commission of South Africa. References 1. Kabouris, J.C. and Georgakakos, A.P. Parameter and state estimation of the activated sludge process-i, Model development, Water Resources, 1996,30, Argaman, Y. A steady-state model for the single sludge activated sludge system-i, model development, Water Resources, 1995,29, Garcia-Olivares, A. and Becares, E. Calibration of a model for an A+B activated sludge pilot plant treating industrial wastewater, Water Resources, 1995, 29, Dalmacija, B., Karlovic, E., Tamas, Z. and Miskovic, D. Purification of high-salinity wastewater by activated sludge process, Water Resources, 1996,30, Zhao, H., Isaacs, S.H., S0eberg, H. and Kummel, M. A novel control strategy for improved nitrogen removal in an alternating activated sludge process - part I, process analysis, Water Resources, 1994, 28, Tian, S., Lishman, L. and Murphy, K.L. Investigations into excess activated sludge accumulation at low temperatures, Water Resources, 1994,28, S chamber, D.R. and Larock, B.E. Numerical analysis of flow in sedimentation basins, Journal of Hydraulics Division, 1981,107, Abdel-Gawad, S.M. and McCorquodale, J.A. Strip integral method applied to settling tanks, Journal ofhydraulic Engineering, 1984,110, Adams, E.W. and Rodi, W. Modelling flow and mixing in sedimentation tanks, Journal of Hydraulic Engineering, 1990, 116, DeVantier, B.A. and Larock, B.E. Modelling sediment-induced density currents in sedimentation basins, Journal ofhydraulic Engineering, 1986, 113, Imam, E. and McCorquodale, J.A. Simulation of flow in rectangular clarifiers, Journal of Hydraulic Engineering, 1983, 109,

9 Computer Methods in Water Resources IV Imam, E. and McCorquodale, J. A. Numerical modelling of sedimentation tanks, Journal of Hydraulic Engineering, 1983, 109, Zhou, S. and McCorquodale, J.A. Modelling of rectangular settling tanks, Journal of Hydraulic Engineering, 1992, 118, Zhou, S. and McCorquodale, J.A. Mathematical modelling of a circular clarifier, Canadian Journal of Civil Engineering, 1992, 19, Brouckaert, C.J., Baddock, L.A.D. and Buckley, C.A. The application of computational fluid dynamics for modelling of water treatment systems, pp , Extended Abstracts, Biennial Conf and Exhibition of the Water Institute ofsouthern Africa (WISA ), Port Elizabeth, South Africa, Bell, C.B., Barclay, S.J. and Buckley, C.A. Decolourisation of textile effluent using a modified 5-stage Bardenpho nutrient removal process, pp , Extended Abstracts, Biennial Conf and Exhibition of the Water Institute of Southern Africa ( WISA), Port Elizabeth, South Africa, Brouckaert, C.J., Baddock, L.A.D., Barnett, J.L. and Buckley, C.A. Impulse: A PC program for determination of residence time distribution of biological and chemical reactors, pp , Seventh International Symposium on Anaerobic Digestion, Cape Town, South Africa, Concentration, Heat and Momentum Ltd. The PHOENICS Reference Manual, CHAM Limited, Carliell, C.M. Biodegradation of Azo Dyes in an Anaerobic System, MScEng. Dissertation, University of Natal, South Africa, 1993.