Use of three-dimensional computational fluid dynamics model for a new configuration of circular primary settling tank

Primary settling tanks (PSTs) are typically used in a wastewater treatment plant (WWTP) to remove suspended solids from the influent raw wastewater. Since the PSTs are an initial step in the treatment, the efficiency of their operation influences the subsequent biological and sludge treatment units (Gernaey & Vanrolleghem 2001), as well as the biogas and electric energy production in systems with anaerobic digestion and cogeneration (Patziger et al. 2016). Consequently, PSTs strongly influence the efficiency of the entire wastewater treatment plant (Asgharzadeh et al. 2011; Patziger et al. 2016).

The design of PSTs is usually based on either surface overflow rate or hydraulic retention time (Droste 1997; Patziger & Kiss 2015) with the assumption of uniform velocity distribution over the depth and width of the tank. Typically, the particle velocities are treated as constant within specified mass fraction classes. These assumptions are not always realistic and may lead to design and performance problems (Tamayol et al. 2008). For example, the flow in real tanks is highly non-uniform and can result in ‘dead’ zones and short-circuiting which tend to reduce the removal efficiency. Density currents, due to influent suspended solids, have different effects on the flow patterns in primary and secondary settling tanks; consequently, the design of inlets and baffles will differ. The process of flocculation or floc breakup can change the settling velocities and/or the fraction in a specific class.

The prediction of sedimentation in a PST is challenging since many factors affect the capacity and performance of a PST: the fraction of non-settleable particles (typically exceeding 40% of the total influent concentration), floating particles (Abdel-Gawad & McCorquodale 1985; Li et al. 2008), the sludge rheology and density coupled with solids removal mechanism (Griborio et al. 2014) and the geometry of the tank, especially, the location of the inlet and outlet baffles (Razmi et al. 2009). Additionally, PSTs are characterized by recirculation zones, bottom currents, turbulence, density currents (Tamayol et al. 2008; Razmi et al. 2009) and thermal density currents induced by variable influent temperatures as well as surface heat exchange (Griborio et al. 2014). Therefore, short-circuiting and dead areas are created, which reduce the effective sedimentation volume of the PST and solids may be resuspended, affecting the performance of these tanks (Tamayol et al. 2008; Razmi et al. 2009).

Several mathematical models have been proposed to describe the behavior of PSTs. These models include 1D models, but because more information about the different phenomena and their interactions in the sedimentation tank are required to accurately predict the tank performance (Balemans 2014), 2D and 3D models, also referred to as Computational Fluid Dynamics (CFD) models (WEF 2005), have been developed. The 2D models can predict sedimentation tank performance but cannot resolve asymmetric flows and geometry effectively. In contrast, a 3D model can fully resolve geometry and complex phenomena that are non-symmetrical. This way, a 3D model can be confidently used to simulate non-symmetric influent configurations, swirl effects induced by rotating scrapers and inlet vanes, hydrodynamic circulation and turbulence issues, which are relevant for the accurate simulation of the flow field, beyond the capabilities of some 2D models (Gong et al. 2010). In the last decade, commercial CFD solution-codes, such as ANSYS FLUENT, have been widely used for 3D secondary settling tank (SST) and PST modeling due to their advantages of easy-learning, friendly interface, and stability (De Clercq 2003; Goula et al. 2008; Patziger et al. 2016; Zhang 2017; Gao & Stenstrom 2019).

Hazen (1904) provided what can be considered the earliest PST model. The Hazen model introduced the concept of surface overflow rate and the simulation of discrete settling that is typically expected to occur in PSTs; a uniform horizontal flow was assumed, and turbulence was neglected (WEF 2005). Although later studies tried to improve upon the Hazen model by applying vertical mixing and they were aware of the importance of turbulent mixing and recirculation zones (Dobbins 1944; Camp 1946; Tebbutt & Christoulas 1975; Alarie et al. 1980; Paraskevas et al. 1993), they were not able to provide adequate solutions due to the lack of suitable hydrodynamics and turbulence models (Stamou 1995; Tamayol & Firoozabadi 2006). Thus, advanced numerical models were developed to simulate a realistic non-uniform velocity profiles by including a classification of influent solids in which a nonsettleable class was considered (Abdel-Gawad & McCorquodale 1985; Stamou et al. 1989; Tarpagkou et al. 2013; Tarpagkou & Pantokratoras 2014; Bachis et al. 2015). Also, the effect of variations in flocculation state on the settling process was studied (Torfs et al. 2017). Moreover, for a better understanding of the flow pattern in both SST and PST, a k-ɛ turbulence model was developed and improved (Celik et al. 1985; Adams & Rodi 1990; Stamou 1991; Patziger & Kiss 2015; Das et al. 2016; Rodi 2017; Gao & Stenstrom 2018). In the same way, other models accounted for the density currents effects on particle settling as a result of both the suspended solids and temperature gradients in the settling tanks (Devantier & Larock 1986; Zhou et al. 1994; Goula et al. 2008; Kriss & Ghawi 2008).

Over the last three decades, CFD modeling has been extensively used to explore the flow pattern of rectangular PSTs and to optimize their design, but fewer studies have been carried out in circular PSTs (Liu & García 2007, 2011; Goula et al. 2008). The results obtained from all studies (rectangular and circular PSTs) suggest that inlet design of PSTs should focus on dissipating the kinetic energy or velocity head of the sewage, avoiding short-circuiting, alleviating the density currents effects and minimizing the blanket disturbances. One applicable method to reduce the volume of the dead zones and increase the performance of the sedimentation tanks is to use a proper baffle configuration. Therefore, various models have been developed to study the optimal position and size of the baffles for PSTs (Tamayol & Firoozabadi 2006; Tamayol et al. 2008; Razmi et al. 2009; Liu & García 2011; Rostami et al. 2011; Shahrokhi et al. 2011, 2012, 2013; Esping et al. 2012; Zhang 2017). Furthermore, the inlet pipe arrangement, tank side water depth and feedwell depth have also been identified as important parameters affecting the PST behavior (Liu & García 2011). However, these models used relatively simple PST geometries and the solids removal was considered as an outlet for sludge on the tank floor; these simplifications do not reproduce some of the real behavior of PST. Furthermore, these models neglected the impact of the rotating scraper on sludge blanket height and on the tank hydrodynamics.

It can be concluded from the above referenced research that the current way in which circular PSTs are designed could and should be improved. Through the use of CFD as a tool that allows understanding, quantifying and visualizing the major processes dominating the tank performance, this research generated a new configuration that might lead to circular PST optimization. To achieve such an objective, field experiments and CFD modeling were conducted. The CFD model was developed for one of the circular PST located at the Cañaveralejo WWTP of Cali, Colombia. Additionally, the applied model included 3D geometry and the use of a rotational sludge scraper. Furthermore, the influence of continuously sludge removal is considered in the numerical simulation.

The first objective of this research was to obtain, through mathematical modeling supported by field and laboratory data, a thorough understanding of the behavior of the PSTs found at the Cañaveralejo WWTP in Cali, Colombia. Accordingly, a CFD model was developed with the general components of good modeling practice (Wicklein et al. 2015): (1) clearly defining the objective of the modeling task with all project parties involved; (2) collection of data, defining the model configuration (mesh generation) and simulation settings; (3) running the simulations and (4) post-processing, calibration, and validation. The CFD model was developed to simulate and analyze the flow patterns and solid settling with the current PST configuration to identify the root cause of the apparent underperformance. The CFD package FLUENT 16.1 is used in this project as a modeling platform. Finally, a new configuration is proposed to improve the PST performance after evaluating different alternatives using the developed CFD model.

DesignModeler, a 3D computer-aided design program included in ANSYS Workbench, was used to digitize the computational domain of a single PST of the Cañaveralejo WWTP, which has a rated capacity of 7.6 m³/s. The Cañaveralejo WWTP has eight identical circular PSTs. Individual tank dimensions are: 47.5 m for internal diameter and 4.2 m for the side water depth.

Since the PSTs were identical, the modeling effort was confined to one PST. The CFD model included the major clarifier features and characteristics such as the influent column, the feedwell or center well (CW), the peripheral effluent weir, the sludge hopper, the rotational sludge scraper, and the sludge pipe.

After the geometry of the PST was generated, the computational domain was meshed into a finite number of grid cells. Great efforts, including sensitivity analyses, were dedicated to find a proper mesh. The mesh density was chosen such that the grid was finest where higher velocity gradients occur and where velocity is expected to have a greater impact on the sedimentation process. A mesh-independency study was performed to eliminate errors due to the coarseness of the grid and to determine the best compromise between simulation accuracy, numerical stability, convergence, and computational time. The mesh-independency was investigated by comparing grids with 373,360 (coarse), 500,252 (medium) and 644,428 (fine) elements of hexahedral form. A qualitative analysis of the velocity and volume fraction patterns, and a quantitative analysis of effluent total suspended solids (TSS) concentration were used to compare the three meshes. The simulation results with the medium mesh did not diverge significantly from the case with the fine mesh, but the difference between the prediction of the coarse mesh to other two meshes was significant and thus the medium mesh was selected.

The sludge scraper movement was incorporated into the model by means of a sliding mesh. To incorporate this approach, it was necessary to split the domain in two parts: stationary frame and rotatory frame, which are connected through an interface called the contact region.

The computational domain and its components as well as the mesh are presented in Figure 1. The scraper contact region is shown in green in the plan view.

Settling column tests were conducted during the initial experimental stage to identify the settling model to be used; however, it was decided to take advantage of the two-phase model already available in the software ANSYS Fluent. In this study, a mixture model was used as the multi-phase model; this is an alternative to the commonly used Vesilind settling model. This model has successfully been used in the past to simulate fluids in sedimentation tanks (De Clercq 2003; Al-Sammarraee et al. 2009; Wicklein & Samstag 2009; Yeoh & Tu 2009; Liu & García 2011; Zhu et al. 2012; Zhang 2014; Samstag et al. 2016; Guo et al. 2017). The mixture model can be applied in flows for a wide range of velocity differences, particle sizes and densities as long as the force equilibrium is achieved (Manninen et al. 1996).

Based on the discretization of the measured particle size distribution while maintaining a sufficient but not excessive number of equations to solve, three secondary classes of solids were implemented in the mixture model each with a different particle size (10, 50 and 80 μm). The density of the secondary phase was 1,396.5 kg/m³ and the properties of primary phase (water) were a density of 997.5 kg/m³ and a viscosity of 0.000924 kg/m.s. The particle size distribution was measured during five different events with averages of 14.9, 56.4 and 28.7% for the 10, 50 and 80 μm particle sizes, respectively. These events are described in the calibration and validation section along with the calibration results. These results indicate that three classes are suitable to simulate with a high degree of accuracy the dynamic behavior of the Cañaveralejo PSTs.

The purpose of choosing turbulence models is to find the most suitable model that fits the physical reality and gives reliable results. The computational cost in terms of CPU time and memory of the individual models was also considered. Gao & Stenstrom (2018) suggest that more observations inside the inlet zone are needed to achieve better model calibration and correct application of the turbulence model, which can be crucial to optimizing the geometry of the inlet structure and sludge hopper. Because of its robustness, economy and reasonable accuracy to better describe low Reynolds numbers flows in sedimentation tank, the standard k-ɛ model (SKE) was used to account for turbulence (Stamou et al. 1989; Zhou & McCorquodale 1992; Zhou et al. 1994; De Clercq 2003; Fan et al. 2007; Liu & García 2011; Stamou 2008; Gong et al. 2011; Shahrokhi et al. 2012, 2013; Ramin et al. 2014).

In sedimentation tanks, the hydraulic behavior near the wall can expect high gradient changes due to flow and particle collisions (Guo et al. 2017). Therefore, the near-wall region was modeled via standard semi-empirical logarithmic wall functions, which imposes the no slip boundary condition (Stamou 2008; Al-Sammarraee et al. 2009; Xanthos et al. 2011; Ghawi & Kriš 2012; Ramin et al. 2014; Guo et al. 2017).

The equations that govern the flow of an incompressible Newtonian fluid are unsteady Eulerian equations that describe the conservation of mass (continuity) and the momentum also known as the Navier–Stokes equation and energy equation. The details about the formulation could be found in ANSYS Fluent (ANSYS 2016). For closing the momentum equation, the standard k-ɛ model for calculation of Reynolds stress is added to these equations (De Clercq 2003; Dufresne et al. 2009; Wang et al. 2010).

In these equations, Gb, Y_M and G_k are production terms of turbulence kinetic energy due to buoyancy, compressibility and mean velocity gradients, respectively. C_1ɛ = 1.44 and C_2ɛ = 1.92 are dimensionless empirical constants (Rodi 2017). and are the turbulent Prandtl numbers (Stamou 1991; De Clercq 2003; Fan et al. 2007; Weiss et al. 2007; Amini et al. 2013; Ekmekçi 2015; Rodi 2017). Two source terms S_k and S_ɛ are used to take the effect of the sludge removal on k and ɛ into account (Weiss et al. 2007). The constant Cμ used to define the turbulent (or eddy) viscosity μ_t was 0.09 (Weiss et al. 2007; Ekmekçi 2015; Rodi 2017).

Moreover, to include the high concentration of solid particles that leads to interparticle collisions (sludge blanket), the ‘granular option’ was selected in the mixture model. In this option, molecular cloud theory is applied to the particle phase. This theory leads to the appearance of additional stresses in momentum equations which are determined by intensity of particle velocity fluctuations (ANSYS 2016). Consequently, the equations proposed by Schaeffer (1987) and Gidaspow et al. (1992) were selected to describe granular and collisional viscosity and frictional viscosity, respectively. Additionally, to define solids pressure, the equation presented by Lun et al. (1984) was applied (Cloete et al. 2011, 2012; Klimanek et al. 2015; Cloete & Amini 2016; Zhou et al. 2016; Zhang et al. 2017). Within these formulations, the radial distribution function was calculated according to Ogawa et al. (1980) (Cloete et al. 2011; Zhou et al. 2016) and the packing limit value of 0.3 was used. Kinetic energy associated with particle velocity fluctuations was represented by a granular temperature and inelasticity of the granular phase.

The modeling of the rotating flow such as those found around or induced by the sludge scraper is important because of the effect it has on the hydraulic behavior of the sedimentation tank. One method of explicitly calculating the flow field in a rotating flow scenario is the sliding mesh method. The sliding grid model explicitly calculates the mixing region, and then rotates this section of the grid relative to the rest of the domain. It is assumed the flow field is unsteady, and the interactions are modeled as they occur (Bridgeman et al. 2009). Consequently, the sliding mesh model was used to simulate the rotating motion of sludge scraper, which is available in ANSYS Fluent as ‘mesh motion option’. While the sliding mesh method is the most computationally demanding, it is also the most accurate method for simulating flows in multiple moving reference frames.

The governing equations were numerically solved using the finite volume method. The segregated pressure-based solver for incompressible flows was selected (De Clercq 2003; Goula et al. 2008; Carbone et al. 2014; Roy et al. 2014; Patziger & Kiss 2015). The pressure and velocity were coupled using the SIMPLE algorithm. The spatial discretization schemes were body force weighted (pressure), second order upwind (momentum, turbulent kinetic energy and energy), least square cell based (gradient) and QUICK (volume fraction). Temporal discretization of the governing equations was performed by a first-order implicit scheme.

Because the governing equations are non-linear, the solution procedure is iterative. To achieve the convergence in the solutions, under-relaxation factors of 0.1 (slip velocity), 0.2 (granular temperature), 0.3 (momentum), 0.5 (volume fraction, turbulent kinetic energy and turbulent dissipation energy), 0.7 (pressure) and 1 (other variables) were adopted. Additionally, the convergence is indicated by the residual of key variants continuity, velocity components and turbulence parameters. The solution was considered converged when the normalized residual value for all variables was 10⁻³ (Ying & Sansalone 2011; Garofalo 2012; Yu et al. 2013; Rameshwaran et al. 2013). The time step selected was 3 s.

Prior to performing simulations, the system was defined by a set of operational characteristics and boundary conditions. In the model, five different boundary types were defined. Unidirectional constant velocity is specified at the inlet according to the flow rate and cross-section area of the inlet pipe. The flow rate and inlet velocity are presented in the following section. At the effluent weir (outlet) and sludge- withdrawal outlet, an outflow boundary condition is applied. For this type of boundary, no conditions need to be defined. Instead, all variables at the outlet are extrapolated from the computed variables in the cells near the outlet. The walls were configured as ‘stationary walls’ with the exception of those located in the moving area, which were configured as ‘moving walls’ and the ‘no slip’ option was selected.

The free water surface was considered as a symmetry plane where changes in the water surface positions are negligible. Additionally, the ‘Symmetry’ boundary condition assumes that there is no diffusion, so the normal velocity component and the normal gradients of all other variables are zero at the symmetry plane. Finally, the inlet turbulence boundary condition was specified using the estimated turbulence kinetic energy (TKE) (6.01 × 10⁻⁴ m²/s²) and turbulent dissipation rate (TDR) (3.6 × 10⁻⁵ m²/s³) (De Clercq 2003; Kim et al. 2005; Weiss et al. 2007; Razmi et al. 2009; Liu & García 2011; Kiss 2012; Shahrokhi et al. 2013; Balemans 2014; Cloete & Amini 2016).

The CFD model was calibrated and validated using a comprehensive field data set collected at the Cañaveralejo WWTP. The field data collected include clarifier influent flow, (b) primary sludge flow, (c) influent and effluent TSS concentrations, (d) primary sludge concentrations, and (e) particle size distribution (PSD).

PSD was analyzed for the influent and effluent of the PST using a laser diffraction particle analyzer. The PSD results indicate a higher proportion of small particles in the effluent compared to the influent, indicating that the largest particles were basically removed during the sedimentation process. The simulation results using influent particle sizes of 10, 50 and 80 μm showed a high degree of agreement with the observed data.

The model was calibrated with data collected over a one-day period (Date 1) of the PST and was validated using additional data over a four-day period (Dates 2 to 5). A sensitivity test on particle sizes was conducted to select the best set of three particle diameters to represent the effluent concentration on Date 1. Data from the subsequent days were used for validation. During these dates, the influent flow rate to the PST was maintained basically constant at 1.27 ± 0.01 m³/s (inlet velocity of 0.72 m/s). Even though some differences between the simulated and measured data were observed, the results presented in Table 1 revealed an acceptable agreement between the model and the field data both in terms of the prediction of effluent total suspended solids (ETSS) concentration and TSS efficiency. Note that the mean predictive errors in the ETSS and efficiency are less than typical errors reported by De Clercq (2003), McCorquodale et al. (2005), Fan et al. (2007) and Patel et al. (2016). Based on these verifications, the model was considered to possess reasonable predictive capability and suitable for application.

The nature of flow in the inlet is of critical importance to the PST performance, since it is generally here that most of the aggregation and energy dissipation process occurs (Griborio & McCorquodale 2006; Goula et al. 2008; Applegate et al. 2010; Ghawi & Kriš 2011; Liu & García 2011; Shahrokhi et al. 2012; Zhu et al. 2012; Das et al. 2016). The flow inside the CW is turbulent, having an influence on the particle aggregation process and consequently the size and density of aggregates that are formed, as well on the sedimentation process. Accordingly, the inlet zone should be designed to dissipate the kinetic energy of the influent; to provide a uniform radial flow pattern to maximize the use of the available settling area and to avoid inducing recirculation patterns in the bulk of the PST; to cope with variations in the flow rate or solids concentrations; and to be cost effective, easy to manufacture and retrofit (Nguyen et al. 2012).

Griborio & McCorquodale (2006) reported that the CW promotes the aggregation of unflocculated particles which has a major effect on the clarifier performance. However, an even greater benefit of the CW is the improvement of the tank hydrodynamics by reducing the strength of entrainment and recirculation flows. An important difference between the functioning of the inlet zone in PSTs and SSTs is that concentration-induced density currents are less important in PSTs which affect the design of the inlet and CW. For example, the downward deflection of the influent jets is less pronounced in PSTs than in SSTs which significantly changes the impact on the CW wall. The proposed new configuration of PST takes advantage of the weakness of the density effect in PSTs.

Based on these understandings, different geometrical configurations were modeled in this study to select the one that could be used to effectively enhance the hydraulic and sedimentation efficiency. The proposed geometrical configurations included the modification of CW diameter, the width of influent ports and the location of a second baffle within the CW. This second CW baffle (also referred to as second center well) was placed inside of the first CW to avoid a very strong downward current toward the sludge blanket and sludge hopper. The best arrangement is showed in Figure 2.

The results are focused on three different zones, identified as the main zones to establish the PST hydraulic behavior: feedpipe, inlet region and slope (outside of the hopper) and sludge hopper. Considering the relevance of hydrodynamic behavior in the PST efficiency, a detailed analysis of the velocity vectors and contours predicted by the model was carried out, to identify whether the modifications made are likely to reduce the hydrodynamic limitations identified in the original PST (OPST).

Velocity vector profile of the feedpipe and inlet ports are presented in Figure 3. This figure demonstrates the influence of the duct curvature on the flow, with the presence of the bend causing a non-uniform distribution of the mean velocity in its vicinity and influent ports. Njobuenwu et al. (2013) showed that changes in the direction of a flow, as in a bend, generate velocity and pressure gradients with strong shear layers close to the geometry boundaries. As a result, a higher velocity inlet on the left side of OPST is observed. However, even though this event also occurs in the NPST (Figure 3(b)), it can be observed that the flow pattern of both left and right side of PST is almost similar after the flow leaves the inlet ports. This suggests that the installation of the second CW baffle permits a uniform flow distribution among the PST and consequently, the hydrodynamic problems associated with short-circuiting of the influent to the bottom of the tank should be reduced.

Figure 4(a) displays that on both sides of OPST the horizontal influent hits the CW and immediately turns downward. The downward current is forced in the opposite direction to form a larger recirculation zone below the inlets port. Although the recirculation zones at both sides of the inlet zone are similar in rotation, on the left side the recirculation zone invades the near-field zone above the sludge hopper. Then, the current splits up at the bottom PST, and part of the current is directed towards the hopper. The other part of the current changes its direction and flows as a horizontal bottom current outside the inlet zone.

On the right side of OPST the current exits the recirculation zone and flows as a horizontal current from the bottom lip of the CW outside the inlet zone. Also, an upward current is formed in the bottom hopper; this current travels over the hopper by running underneath the recirculation zone towards outside the inlet zone. Figure 4(b) shows that the flow patterns in the CW and hopper are significantly improved by the NPST.

After the horizontal current exits the inlet zone a second recirculation zone is formed at the near-field of sludge hopper (Figure 5(a)); this produces an upward current (similar in both sides). Additionally, the horizontal bottom current is still present and flows towards the peripheral wall in opposite direction to the flow induced by the scraper movement, which can potentially impact the scraper torque and efficiency. The NPST corrects these problems as indicated in Figure 5(b).

These findings suggest that the OPST efficiency is deteriorated by accumulation of settled solids in the slope (outside of the CW) and resuspension of settled solids, particularly in the sludge hopper. Thus, the solids transport toward the sludge hopper is compromised and may result in low withdraw sludge concentrations.

Otherwise, as shown in Figure 4(b), a recirculation zone in the inlet zone of NPST was identified but at a smaller magnitude than that found in the OPST. The influent flow turns sharply upward as a current reaching the zone between the first and second CW baffles and largest portion of the currents from the inlet ports is directed radially outward and not directed toward the PST bottom. Consequently, the recirculation in the inlet zone is reduced and the scouring on the thickened sludge at the sludge hopper is prevented.

A higher re-entrainment of flow from outside of the CW into the inlet zone is observed accompanied by two additional recirculation zones (counter-rotating eddies). This may promote contact between the suspended particles and provide favorable conditions for the settling process.

Furthermore, with the installation of the second CW baffle and especially by the curved edge of this CW, recirculation regions are formed behind the first CW (Figure 5(b)). This recirculation pattern avoids short-circuiting of influent to the effluent weir. The combination of the recirculation flow and small velocities appear to provide the required quiescent flow conditions for the settling process of the suspended solids.

According to Figure 5(b), the flow is forced upward and redirected back to the PST bottom. When the flow hits the bottom, it splits up into two opposite-direction currents, one bottom current towards the peripheral wall (first current) and other one towards the sludge hopper (second current). The fact that the direction of the second current is in the direction of the current induced by the scraper movement, and that the first current is weak and far from the hopper, provides a favorable condition for the settled solids to be directed towards the sludge hopper. This also prevents the settled solids from staying on the slope or accumulate in the corners.

The first current flows along the bottom of the end wall and then turns upwards to effluent weir. Over this current, a third current is developed which is directed towards the effluent weir.

The velocity contours shown in Figure 6(a) and 6(b) reveal the high impact that the inlet has on PST behavior. In the case of OPST, a marked downward velocity contour travels from the CW to the PST bottom closely approaching the sludge hopper. This results in elevating the sludge blanket and dragging solids outside the hopper. In the NPST, at the zone between the first CW and second CW, the inflow energy is dissipated (Figure 6(d)) and the downward velocity is significantly reduced minimizing the impact on the sludge blanket. This flow pattern allows a higher solids concentration in inlet zone close to the sludge hopper (Figure 7). Outside of the inlet zone, the velocity magnitude is reduced gradually along the radial direction as the energy is dissipated.

As displayed in Figure 6(c) and 6(d), the turbulent kinetic energy is maximum at the inlet ports. The turbulent energy is confined and spread on the top of CW. Behind the CW, there is very low turbulence.

In a PST, an efficient CW is one that dissipates the initial kinetic energy to reduce solids dispersion (due to the relatively low suspended solid concentration), reduces entrainment and produces a radially uniform discharge flow at the CW exit. However, in the case of OPST no evidence of this was observed.

In NPST, at the zone between first CW and second CW, the TKE is still present, as the bulk of the flow currents pass through this region. Due to this, the inlet TKE will be fully dispersed and will not appear in the bottom current, therefore, short-circuiting is reduced in this region (Zhou & McCorquodale 1992). This indicates that when comparing the efficacy of OPST and NPST in dissipating the kinetic energy, the installation of second CW produces a better effect.

The density difference between a heavy fluid zone and a lighter fluid zone causes a stable density stratification (Gao & Stenstrom 2017) which contributes to a damping of vertical mixing (Samstag et al. 1992). When this happens, the density currents do not significantly affect the ETSS removal efficiency unless short-circuiting occurs and the upward currents flow directly to the exit (Gao & Stenstrom 2017). Consequently, as mentioned by several authors, because of the relatively low suspended solid concentration in PST, discrete particle settling prevails and the presence of solids does not greatly affect the flow pattern of the liquid but the mixture turbulence can be affected (Imam et al. 1983; Stamou et al. 1989; Liu & García 2011; Zhang 2017). Thus, in inlet zone of PST the concentration of the solids is required, and shear is needed to promote aggregation and to improve sedimentation process. Additionally, the inlet kinetic and potential energy must be satisfactorily dissipated within the inlet zone, which requires a design heavily based on hydrodynamic principles (satisfying volume and hydraulic retention time as well as optimal geometric design) (Patziger & Kiss 2015). In that respect, Figure 7 shows that unlike OPST, in the NSPT inlet zone a higher concentration of solids and sludge accumulation in the hopper were observed.

According to the results shown in Figure 8, the particle size had a significant effect on the sedimentation efficiency. Particles of 80 μm displayed a satisfactory settling performance, and most of these particles settled to the slope bottom (OPST) or to the sludge hopper (NPST) (Figure 8(c)). However, the behavior of particles of 50 μm was diverse. A fraction of these particles became suspended in the flow field by the recirculation zone behind the first CW (in OPST) and second CW (in NPST), while others escaped into the outlet flow (Figure 8(b)).

In NPST, particles of 80 μm and 50 μm were prone to back mixing due to the recirculation zones and upward flow, eventually returning to the inlet zone and hopper. This event led to sludge accumulation at the hopper, consequently a higher sludge concentration in hopper was obtained and better sludge-withdrawal efficiency can be expected (Zhou et al. 2005). Unlike OPST where sludge accumulation at the slope bottom was observed, even though this sludge accumulation was high, its movement to the hopper was not fully achieved. Therefore, the sludge withdrawal concentration was low and, over time, a major fraction of accumulated solids is expected to reach the tank outlet, affecting the sedimentation efficiency.

Otherwise, particles of 10 μm tend to come out with the effluent because of their small diameter and low settling velocity. Thus, the removal efficiency of these particles was almost zero (OPST) or lower than that of larger particles (NPST) (Figure 8(a)). However, a difference is observed between both PSTs, which consists of the ability to accumulate a fraction of these particles in the NPST hopper (Figure 7(b)).

Notice that in NPST the sludge inventory is observed as a sludge blanket and an accumulation of a thick sludge at the hopper, unlike OPST where the sludge inventory is observed as an accumulation on the slope bottom (Figures 7 and 8). This implies that the second CW of NPST allows the formation of a consistently more concentrated sludge blanket layer and promotes the accumulation and thickening of sludge in the hopper, reducing the risk of solids leaving the PST outlet.

To compare the performance between OPST and NPST, TSS concentration and efficiency are assessed (Table 2). It is apparent that in NPST the efficiency of the TSS removal is better than in the case of OPST by 8%. The advantage of the NPST configuration is also evident when the sludge-withdrawal concentration is evaluated. The OPST results in a diluted concentration because of the hydrodynamic limitations observed for this configuration. The NPST produces a thickened primary sludge with a high TSS concentration.

The results in Table 2 are presented for the calibration influent flow rate of 1.27 m³/s, which is equivalent to a surface overflow rate of 61.92 m³/m²/d. While this surface overflow rate is typical for primary clarifiers, it is recommended that that the removal efficiency, sludge-withdrawal sludge concentration, and overall tank hydrodynamics of NPST are evaluated for different influent flow rates to determine the advantages of this configuration at different influent conditions.

A 3D CFD model has been applied in this study to investigate and improve the performance of PST. Based on the CFD analysis of the OPST, a second CW baffle was proposed (NPST). The second CW baffle has been found to result in an improvement in PST performance with a more even distribution and symmetrical flow pattern.

In general, the second CW appears to provide better hydraulic behavior that than in the OPST, thus enhancing sedimentation. The second CW acts as an obstacle to suppress the downward current since influent TKE is effectively dissipated and forces the particles to move toward the bottom of the PST. In addition, the second CW allows, in the inlet zone, the formation of a consistently more concentrated sludge blanket layer and promotes the accumulation and thickening of sludge in the hopper, reducing the risk of solids leaving the PST outlet. The main advantage of second CW baffle lies in its ability to generate a thicker sludge. The second CW baffle could be easily installed in an existing PST as a solution to dilute sludge-withdrawal problems at minimal cost compared to other solution.

All relevant data are included in the paper or its Supplementary Information.