Symmetry to asymmetry: innovative evolvement of a gas–liquid–solid (GLS) separator in UASB/EGSB reactors on a new perspective

Anaerobic digesters offer significant energy potential and cost-effectiveness for wastewater treatment, generating up to 1.0 kWh of electrical energy per kilogram of removed chemical oxygen demand (COD), compared to the 1.0 kWh consumed by aerobic processes (Ngwenya et al. 2022; Rattier et al. 2022; Kemausuor et al. 2018). Among anaerobic processes, up-flow anaerobic sludge blanket (UASB) reactors dominate globally, accounting for approximately 51–60% of installations (Latif et al. 2011; Deng et al. 2017). Expanded granular sludge bed (EGSB) and internal circulation (IC) anaerobic reactors represent about 11% of installations (Deng et al. 2017). A crucial element in the success of high-rate reactors is the gas–liquid–solid (GLS) separator, which effectively separates biomass from the mixed liquid (Lettinga & Hulshoff 1991; Latif et al. 2011).

The widely adopted traditional symmetrically structured GLS separator for wastewater treatment, developed with the UASB reactors, operates on the principle of gravity settlement, utilizing sedimentation-tank-like baffles and low up-flow velocities (0.5–1.5 m/h) (Latif et al. 2011). However, as applications and technology advances, several limitations have emerged: (1) GLS separators occupy 16–25% of the reactor height (Hashemian & James 1990), significantly increasing construction costs; (2) the effectiveness of original GLS separators becomes questionable when up-flow velocity (UFV) exceeds 4.0 m/h in EGSB scenarios (Seghezzo et al. 1998). Many existing structures are patented and restricted to certain corporations (Guo & Guo 2003; Pan et al. 2017; Hao & Shen 2021). Limited research has been published on specific mechanisms, structure configuration, and engineering design methods. Bastiani et al. (2023) identified that only 24 design-related papers published from 2007 to 2022, with most studies focused on the angle of baffles for GLS separators as a contributor to separation efficiency in EGSB (Pan et al. 2017), IC (Tu 2012), and UASB (Quan 2012). Ning (2019) optimized the settling zone height in UASB reactors for improved gas separation, while Zi et al. (2021) analyzed gas outlet configurations in an anaerobic reactor. As a powerful tool, these studies employed the CFD-solving technique to simulate GLS flow patterns, to discern how the GLS phases flow under various process and structure settings. The CFD technique has shown promise in enhancing the design of fluid-related engineering structures through reliable simulations to partly substitute the expensive laboratory tests (Nopens et al. 2020; Bastiani et al. 2023).

Based on this finding, this paper proposes an asymmetrical structure design for the GLS separator. As illustrated in Figure 1(b), asymmetrically configured baffles are more suitable for forming stable and regular vortex circulation than the traditional symmetric design. The moderate and uniform vortex/circulation generated by the asymmetrical design avoids severe collisions and excessive hydraulic head loss under the angled plates of the symmetric model. Furthermore, this paper evaluates the performance of the new model through laboratory experimental tests and CFD simulations.

The new structural configuration features a two-layered arrangement with asymmetrically positioned baffling plates. These design elements promote a uniform and moderate liquid flow vortex while reducing excessive inflexion and collision within the flow path. Additionally, the number of narrow gaps along the liquid flow path through the GLS separator has been reduced from three to two. As shown in Figure 1(b), the liquid flow path (under ideal conditions) of S2 through the GLS separator is shorter than that of S1, resulting in reduced hydraulic friction loss and thus a higher proportion of flux being split to S2. In this design, the first gap encountered by S2 is set at a minimum of 95 mm (Hao & Shen 2019), while the first gap for S1 is increased to offset the additional hydraulic friction loss experienced by S1. This adjustment aims to achieve equitable flux distribution between S1 and S2.

The new solid separation mechanism was effectively exemplified in the current design shown in Figure 1(b). The upward-flowing mixture, represented by S1 and S2, was initially directed by the asymmetrically arranged baffling plates to create a series of counter-clockwise vortexes. It then encounters head-on streams where the first coagulation of sludge fragments occurs through mutual collision. Guided by the top angled plates, the mixture formed the critical vortex, transported the sludge fragments onto the top sides of the baffling plates, where they are ready for further agglomeration and eventual separation from the liquid when the density required for them to settle by gravity was met.

Cultivated granular and flocculent sludge, purchased from Shandong Liboyuan Environmental Protection Materials Co., Ltd, were used to simulate fully developed sludge in UASB/EGSB bio-reactors. Based on Hao & Shen (2021) and laboratory observation, three factors were pre-selected for each type of sludge average UFV (m/h), sludge concentration in the bio-reactor (SCon in g/L), and the maximum size (diameter) of the sludge fragment (SFdia in mm). For UASB reactors, the accepted UFV is 0.5–1.0 m/h (Latif et al. 2011). The flow rate into the reactor was regulated using a valve in the pipeline, and the UFV was calculated by dividing the flow rate by the reactor's cross-sectional area. The size of the sludge fragments affects their operational density and drag force. A maximum diameter of 3.0 mm (0.5–3.0 mm, graded by stainless pore sieve) was selected based on a pre-screening of the sludge. The concentration of sludge inside UASB/EGSB is seldom reported, so an upper limit of 4.0 g/L was set based on the measurements from an operational laboratory USAB reactor.

The Response Surface Method with the Box–Behnken design type was used to design the experiment using Design Expert^® software. Each sludge type (granular and flocculent) has 23 runs with five groups and two blocks, as detailed in Table 1. Each run was carried out three times for better estimation.

In Table 1, ‘sludge size’ refers to the sieve pore diameter used for grading the sludge fragments after 4 hours of gravity sedimentation. A size of 0.5 indicates sludge fragments with diameters less than 0.5 mm, while a size of 1.0 indicates fragments between 0.5 and 1.0 mm. Tap water was used in the experiment to eliminate contaminants in the wastewater. Before each run, a specific type of sludge fragment with a defined diameter is pre-dosed into the up-flow reactor (Figure 2) at a specific concentration (suspended solid value: SS, according to the Chinese National Standard GB11901-89). The influent pump (submersible pump, QW25-8-15, China Success Pump Co., Ltd) initiated circulation by pumping filtered (0.5 mm) used tap water from the influent tank to the up-flow reactor through perforated distribution pipes at the bottom of the reactor (Figure 2). The flowmeter (Electromagnetic flowmeter, BEF6204-DN20, BELL Analytical Instruments (Dalian) Co., Ltd) was used to adjust the UFV to a predetermined value (refer to Table 1. After a 15-min pre-operation, at least three 500 mL effluent samples were collected from the up-flow reactor (with GLS separator) at 5-min intervals. The sludge concentration in the samples and the reactor (via the sampling point in Figure 2) was measured as a suspended solid value (SS, according to the Chinese National Standard GB11901-89).

To evaluate the gas phase and flow pattern inside the up-flow reactor (Figure 2), CFD simulation was used to analyze gas separation and flow patterns. Additionally, an industrial endoscope (GIC120-C, BOSCH) at sampling points was used to validate the fluid field simulated by the CFD software.

The wastewater inside the up-flow reactor consisted of gas, liquid, and solid phases. The Euler–Euler multiphase model, widely adopted to describe fluid behavior (Zi et al. 2021; Olaleye et al. 2022; Bastiani et al. 2023), was used to simulate the inhomogeneous flow of dispersed particles of gas and solid in a continuous liquid. Their behavior is governed by equations of momentum, continuous, volume conservation, pressure constraint, total energy, particle force, and turbulent model (Hao & Shen 2021; Ansys Inc.). These standard equations are omitted in this paper.

Experimental results indicated that smaller fragments are more prone to up-float than larger ones. Besides the morphology, which affects drag force, density plays a non-negligible role. This study selected an evenly mixed (in number) group of sludge fragments with diameters of 0.5, 1.0, 1.75, and 3.0 mm to mimic the sludge fragments in the fluid. The density of each size was determined by the average density of the sludge grade in Table 1, measured through experiments under wet conditions: vacuum-filtered sludge fragments (0.2 μm) balanced at ambient room temperature until filter dried. The average density is listed in Table 2. The CFD simulation was conducted using the Ansys CFX^® software package, with initial settings are summarized in Table 2.

Unlike the other two reduction indices, the phase-separation efficiencies require further investigation. Therefore, laboratory-scale experiments and CFD simulations were conducted to evaluate this aspect.

The results of the 23 experimental runs for both the granular (left half) and flocculent sludge (right half) are illustrated in Figure 4. For comparison, the average solid–liquid separation efficiencies of the traditional GLS separator for both sludge types are also shown as a baseline (horizontal lines). The new GLS separator prototype achieved an average removal efficiency of 98.3% for granular sludge and 96.0% for flocculent sludge over the 23 runs. These figures surpass the maximum efficiencies of 93.5% (granular sludge) and 80.7% (flocculent sludge) achieved by the traditional model in Figure 1(a). This performance exceeds those reported in the literature. Such as Huang (2005), who measured a granular sludge-removal efficiency of 93.5–95% in an EGSB reactor; Guo & Guo (2003), who found 76–94% efficiency for a new GLS separator; Yasar et al. (2007), who reported a maximum 77% of sludge-removal efficiency for a UASB reactor treating combined industrial wastewater; and Caixeta et al. (2002), who tested an 81–86% removal efficiency for a new GLS separator in a UASB reactor treating slaughterhouse wastewater. The new structure developed in this study is suggested to be a reliable, efficient, and cost-effective alternative for solid separation in up-flow reactors.

The new design exhibited stable performance, withe coefficients of variation (CV) significantly lower at 0.02 for granular sludge and 0.05 for flocculent sludge, compared to 0.3 and 1.4 for the traditional model and 0.4 and 0.8 for the structure proposed by Hao & Shen (2021). Additionally, the newly developed module has a simpler architecture, resulting in an approximate 30% reduction in height and a 14.8% decrease in material consumption compared to the traditional structure shown in Figure 1(a). This means less construction material is required for the up-flow reactor and the GLS separator, making it superior to the three modules presented by Hao & Shen (2021).

For gas removal/separation, the angled plate effectively traps and captures rising gas bubbles. Experiment observations confirm that if the vortex is not intense and unsteady, almost all air bubbles can be trapped in the angled regions of the plates. Simulated results indicated that the new GLS separator gas separation/removal efficiency comparable to those reported in the literature, ranging from 95.5 to 99.9% (Hao & Shen 2012, 2019).

The variation in phase-separation efficiencies of the new model derives not only from sludge type but also from the three factors. Sludge concentration (p = 0.0087) and the combination of sludge concentration and size (p = 0.0320 for granular and 0.0210 for flocculent) are significant (p < 0.05) for both types of sludge. The behavior of the flocculent sludge is also susceptible to its size. Surprisingly, UFV dominates the expansion status of the sludge blanket inside the reactor. However, the p-values for both sludge types exceed 0.48 and are statistically nonsignificant. Experimental observation reveals that when UFV increases, the expansion height of the sludge blanket scales up, producing more small sludge flocs (they are secondarily produced and floc shaped; they are not the original small particles) that rise quickly. Some pass through the GLS separator to the effluent weirs. but their size is so small that the weight increment to SS can be statistically neglected until the UFV surpasses ∼12.0 m/h. In this respect, the recommended upper limit UFV of 7.0 m/h in the Chinese specification (HJ2023-2012 2012) is relatively safe.

Considering multiple factors, Figure 5(third row) shows the response surface of sludge-removal efficiency with the two statistically significant factors. The proposed GLS separator provides a robust and steady sludge removal for granular sludge through the 23 experimental runs, with a minimum removal efficiency of 95.31%. When sludge concentration exceeds ∼2.7 g/L (the yellow line on the bottom surface of the left of Figure 5), at least 99% sludge-removal efficiency can be attained. A combination of higher sludge concentration with smaller sludge fragment size is favorable preventing sludge from washing out of the UASB/EGSB reactor. Conversely, flocculent sludge fragments are more susceptible to variation in operational factors, with a wide range of 82.39–99.55% sludge-removal efficiency. Extra attention is needed to monitor sludge concentration and fragment size inside the bio-reactor. This finding is consistent with granular sludge, where higher sludge concentration with smaller fragments benefits sludge removal by the GLS separator (the red lines on the bottom surface of the right of Figure 5). The underlying reason is the stacking of factors: smaller fragments have higher densities or/and are subject to smaller fluid drag force; higher sludge concentration means more collision frequency among fragments, more intense coagulation , and stronger hindrance of the up-flowing sludge fragments . However, an exception applies to the correlation between sludge removal and fragment size. When the fragment size exceeds 2.5 mm, high sludge-removal efficiency remains achievable, although it is not as significant as when fragment size increases from 0.5 to 2.5 mm. This trend is observed in Figure 5 and the upper corner of the bottom surface of Figure 5(right), suggesting another behavior scenario for flocculent sludge in bio-reactors that needs further investigation.

The sludge fragments are propelled and carried by the continuous water flow, creating similar motion patterns. Notably, sedimentation of sludge fragments occurs in the bottom, downward-flowing part of each of the nine vortexes in the GLS separator, where the sludge fragment's motion is typically perpendicular to the nearby plates. No discernible natural gravity-induced straight vertical downward settlement is found in Figure 6. This finding reaffirms that the primary mechanism for separating and removing sludge is the water-vortex circulation-induced settlement (Hao & Shen 2019, 2021) and the present design can effectively achieves this new separation mechanism. Attention should be paid to vortex-induced settlement rather than blocking and re-bouncing of sludge separation/removal when developing the new structures of GLS separators. In comparison, gas bubbles are slightly affected by the water vortexes. Although turbulence arises in the water-vortex regions, the flow patterns are irregular, the streamlines in Figure 5 suggest that most of the gas bubbles bypass and extricate the water vortexes, reaching the gas outlet (G1–G5) directly due to their significant density difference from water. This finding implies that for the gas phase, blocking and collecting by the angled structure is the principal separation/removal mechanism. It can effectively separate the gas bubbles from washing out the bio-reactor. However, if the UFV of water is high or the water flow is poorly configured (other structures), violent water currents can disperse the gas bubbles, leading to escapement.

The flow pattern of sludge observed in the laboratory experiments matches the simulated visualization in Figure 6. Nine vortexes consistently form and remain robust despite changes in sludge type and the three other factors. Only small sludge fragments flow into the GLS separation region, escaping the coagulation, capture, and hindrance of the sludge blanket suspended beneath the GLS separator. Most of these fragments differ from the original sludge fragments in size and morphology, suggesting that many small sludge fragments originate from larger ones through peeling and fragmentation. This phenomenon contributes to an average gap of 2.7% (absolute) between simulated and experimentally measured solid separation/removal efficiencies, as CFD simulation cannot approximate the fragmentation process. Within the GLS separation volume, experimental observations re-verify the proposed sludge separation/removal mechanism: (1) circulation: sludge fragments up-flow into the GLS separation volume, generating vortex flow the nines regions (Figure 6) under the blockage and guidance of the plates, causing sludge fragments to flow circularly with the water; (2) deposition: when the mixed flow encounters the plates' surface , flow speed declines, and most sludge fragments settle rather than bounce back onto the plates; and (3) agglomeration and return: settled sludge fragments on the plates' surface are compacted and pushed by the water current and ever-precipitating sludge fragments. When their height is sufficient at the outer edge of the plates, collapse occurs and large fragments with adequate density settle downwards into the bio-reactor below.

This study presents a novel perspective on the solid–liquid separation mechanism in GLS separators and introduces an innovative asymmetric GLS separator prototype. Limited experiments and CFD simulations have yielded several key findings:

The new GLS separator prototype features a simpler asymmetric structure, achieving approximately 30% reduction in height and 14.8% reduction in material usage. It demonstrates excellent solid–liquid separation performance, with averages of 98.3% for granular sludge and 96.0% for flocculent sludge. The asymmetric structure significantly enhances flocculent sludge separation efficiency and reduces the CV to 0.02 and 0.05, respectively.

Several factors significantly influence sludge separation efficiency, including sludge concentration and fragment size. Higher sludge concentration (>2.5 g/L) and smaller (<1.0 mm), denser fragments improve separation efficiency. Flocculent sludge is more susceptible to operational factors.

The morphology of sludge fragments, along with other physical–chemical factors, plays a crucial role in the agglomeration process and needs further investigation.

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

The authors declare there is no conflict.