Influence of geometrical and operational parameters on the axial dispersion in an aerated channel reactor

Abstract

Residence time distribution experiments have been performed on an activated sludge 3000 m³ channel reactor aerated by gas diffusion (for different liquid flowrates under constant aeration rate and constant water depth) and on a bench-scale channel reactor aerated from the bottom (for different liquid and gas flowrates and water depths) in order to characterize their hydrodynamics. Both units can be modeled as plug flow reactors with axial dispersion. A general correlation has been obtained to predict the axial dispersion coefficient as a function of the gas and liquid velocities and the geometrical parameters of the full-scale and bench-scale reactors. Finally, to facilitate the simulation of biological reactions in transient state, an equivalent model based on tanks-in-series with variable back-mixing flowrate is proposed.

Introduction

Activated sludge process is the most widespread technique for biological wastewater treatment. In large wastewater treatment plants the aerobic degradation is often taking place in channels or closed-loop reactors (Degrémont, 1991). The channel reactor with bottom aerators is the oldest type of systems and is particularly adapted to large plants (Fig. 1a). In closed-loop processes, the mixed liquor is circulating as in a “racetrack”. They are called “oxidation ditches” when the aerators are horizontal (Fig. 1b) and “carrousels” when they are vertical (Fig. 1c). From these basic designs many variations have been proposed by various manufacturers, by inclusion of anaerobic and anoxic zones equipped with mechanical mixing devices or as the Orbal system with several concentric ovaloid channels (Fig. 1d). Due to the shape and size of these units, there is definitely an effect of hydrodynamics on the efficiency of the pollution abatement, as concentration gradients are experimentally observed, for nutrients as well as for oxygen (Dudley, 1995). This situation can be extended to the case of facultative aerated lagoons (Dorego and Leduc, 1996). Classically channels and closed-loop reactors will have a depth of up to a few meters. An exception is the underground Henriksdal plant in Stockholm whose channels are 200 m long, 10 m wide and 12 m deep and where vertical gradients could be superposed to longitudinal effects. The effect of the flow behavior on the efficiency of wastewater treatment has been often pointed out (Metcalf and Eddy, 2002). Unfortunately, there is a limited number of correlations to predict the flow behavior. The design of such reactors is often based on land availability at a fixed mean residence time. Hourly data obtained on a full-scale plant (industrial data, unpublished) for the treatment of the same wastewater with the same activated sludge between 10 a.m. and 5 p.m. on a dry weather day, in two full-scale aerated channel reactors having identical mean residence times (1.75 h, C.V.=5%) but different flow behaviors (Peclet numbers of 16 (C.V. 6%) and 1.7 (C.V. 6%)), show that the soluble COD removal between their inlet and their oulet differs in average by 35% (C.V. 43%).

Efforts are being made to model multiphase bioreactors using Computational fluid dynamics (CFD) (Dhanasekharan et al., 2005). However, it is not yet possible to couple this approach with a complex kinetic biomodel (Henze, 2000) and simulate the behavior of full-scale plants over a long period of time. Znad et al. (2004) notice that little work has been performed on the mathematical modeling of airlift bioreactors or bubble columns. These reactors have many similarities with aerated wastewater channels, although the gas and liquid flow directions are parallel in the first case and transverse in the second case.

In channel reactors the wastewater flow behavior can be represented by two equivalent models:

• the classical plug flow reactor with axial dispersion (axial dispersion model or ADM) which contains two parameters, the mean residence time (τ) and the Peclet number (Pe) (Levenspiel, 1998)

• $$\mathit{Pe}=\frac{\mathit{uL}}{D}\text{.}$$

  The Peclet number represents the ratio of the convective flow to the diffusive one. The dispersion number

  $$d=D/\mathit{uL}=1/\mathit{Pe}\text{,}$$

  has been considered to quantify the degree of mixing in lagoons where the treatment efficiency is modeled through the Wehner–Wilhelm equation (Dorego and Leduc, 1996). The ADM approach produces a continuous model.

• the tanks-in-series model (TSM) which contains also two parameters; the mean residence time (τ) and the number of mixing cells (J). This discrete approach has been used by De Clercq et al. (1999) to describe the hydrodynamics in various wastewater treatment units.

  In most cases the following equivalence between both models (Villermaux, 1982) may be applied

  $$\mathit{Pe}=2j+1\text{.}$$

  Wastewater bioreactors design criteria include the food/biomass ratio (kg BOD/kg biomass day⁻¹), the BOD loading (kg BOD m⁻³ day⁻¹), the sludge age (in days) and some hydraulic data such as peak flow. The design of a reactor only based on the liquid and sludge residence times may result in large errors on the predicted efficiency of a full-scale wastewater treatment plant. Some other important features are also sensitive to hydrodynamics. For example, filamentous bulking should be limited to prevent the decrease of the settling rate in the clarifier. A plug flow behavior will favor zoogleal microorganisms and be detrimental to the overgrowth of filamentous bacteria (Chudoba et al., 1973).

  The purpose of this paper is to evaluate the possibility to define a general framework to model the hydrodynamics of aerated channel reactors, taking into account the operating parameters (gas and liquid flowrate) and the geometrical parameters (length, width and mixed liquor height). For this purpose, tracing experiments have been performed in a full-scale reactor and in a bench-scale reactor. In full-scale plants, the wastewater flow-rate changes continuously with time. A day to night flowrate ratio equal to 3 can be observed (Le Bonté, 2003). The sewage system and the rainwater network are still often connected, and depending on the climate, the flowrate may vary in a proportion (with respect to summer dry weather conditions) in the range of 1–7 during heavy rain periods. Plant managers are often reluctant to change the operating conditions for specific tracer tests during normal operations as they want to avoid any accidental discharge of pollutants in receiving waters. Consequently, it is often very difficult to carry out tracing experiments in steady-state conditions. Complex flow connections between units (De Clercq et al., 1999) can make the data interpretation delicate. In order to study the effect of a large range of operating conditions and geometrical parameters, it is necessary to build a bench-scale plant. The scale-down and scale-up of multiphase bioreactors are very complex tasks. As many of them are used for shear sensitive species such as mammalian or insect cells, shear rate is often the selected criterion (Merchuk et al., 1994; Maranga et al., 2004). Although shear has an effect on sludge flocs (Liu et al., 2005), it is certainly not as drastic as for mammalian cells and is not an issue in classical activated sludge systems. The efficiency of the aerated sludge channel reactor depends on

• the biological kinetics which depend only on the treated waste,

• the oxygen mass transfer coefficient, which depends on gas flowrate, gas hold-up, bubble size, distribution, bubble coalescence especially in deep reactors,

• the liquid residence time distribution.

Since the first two are fixed either by the biochemistry or by the production of air bubbles, the liquid residence time distribution seems to be an adequate criterion for scale-down. Obviously this choice is still debatable since it does not give any answer to the question about the conservation of the ratio of the bubble size to the characteristic dimension of the reactor.

The results obtained on a full- and a bench-scale reactor where similar liquid residence time distributions can be achieved are discussed, based on the plug flow model with axial dispersion. A semi-empirical correlation for the coefficient of dispersion has been determined to facilitate the scale-up of large reactors. To help incorporate the complex kinetics related to the bioreactions taking place in the mixed liquor, the fluctuations of the axial dispersion have been taken into account using a tanks-in-series model with back-mixing model with a constant number of cells and a variable backflow rate.