Method for determining the flowback rate of fracturing fluid considering gel degradation during hydraulic fracturing Authors: Chengming Ma, Zhaobing Hao, Fan Yang, Qi An, Hanlie Cheng, Linqiang Zhang, Zhiping Li Published: 2026 ABSTRACT AbstractFracturing flowback plays a vital role in coalbed methane (CBM) reservoir development, but field flowback rates are still commonly selected based on operational experience and are therefore difficult to generalize across wells. To move beyond a field-application description, this study develops a mechanics-based and parameterized framework for estimating the allowable flowback-rate window after hydraulic fracturing. The model incorporates the post-break viscosity of the fracturing fluid and formulates the critical particle-mobilization condition from the force balance acting on coal fines and proppant before and after fracture closure. Instead of calibrating an empirical correlation for a single block, the method uses measurable or literature-constrained inputs, including particle size and density, fluid density and viscosity, fracture pressure differential, wetting-phase saturation, and contact-mechanics coefficients. Field parameters from two CBM wells are used as demonstrative cases to implement the workflow and to examine whether the calculated operating window is consistent with realistic field flowback-rate magnitudes. The calculated critical proppant flowback velocity ranges from 3.6 to 498 m³/day before fracture closure and 36–890 m³/day after closure, while the minimum velocity required to mobilize coal fines increases with particle size. Sensitivity analysis identifies particle radius, post-closure pressure differential, particle density, fluid viscosity, and wetting-phase saturation as the dominant controlling factors. Larger and denser particles increase the required flowback velocity, whereas higher fluid viscosity reduces it by enhancing drag. Capillary forces promote particle mobilization, while friction becomes the primary resisting force after fracture closure. Because no new laboratory flowback experiment is conducted in this work, the model is not claimed to be fully experimentally validated; instead, its reliability is supported by transparent parameterization, literature-constrained coefficients, order-of-magnitude comparison with reported flowback rates, and systematic sensitivity analysis. The proposed workflow provides a transferable screening tool for selecting flowback rates that balance early cleanup efficiency and long-term fracture conductivity in CBM and other propped-fracture operations.Keywords: Coalbed methane, Fracturing flowback, Flowback rate, Theoretical modelSubject terms: Energy science and technology, Engineering, Solid Earth sciences Introduction Hydraulic fracturing is a widely used technique to enhance CBM production1,2.The efficiency of the post-fracturing flowback process plays a key role in determining the success of the treatment and the productivity of the well3,4.Selecting an appropriate flowback rate is critical for preserving fracture conductivity and achieving rapid production. The ideal flowback rate should approach the maximum velocity that prevents proppant movement, ensuring efficient cleanup without compromising fracture structure5,6.Exceeding this limit can lead to proppant backflow, reduced conductivity, fracture closure, and, in severe cases, treatment failure7,8.In practice, engineers often rely on field experience, using sand-free flowback as a benchmark, though this approach lacks a robust theoretical foundation9–11. Research on proppant backflow began in the 1980 s, with early studies focusing on its mechanisms and predictive methods12,13.One common hypothesis suggests that sand arches form near the wellbore and collapse under fluid flow, releasing loosely packed particles14.Another view attributes backflow to unfilled spaces at the top of the fracture, where proppant is more likely to be dislodged15.Additional studies point to the development of channels within the proppant pack, which can cause fluidization under high flow velocities16,17. In the 1990 s, researchers developed both experimental and numerical models to better predict proppant backflow18,19.Empirical models quantified the effects of closure stress, proppant layering, and drag forces, and proposed correlations to estimate critical flow velocities20,21.Numerical models further explored how these variables influence proppant behavior22,23.Semi-theoretical approaches based on pressure gradients added new insight into the mechanics of proppant migration24.Collectively, these studies identified three essential conditions for sustained proppant backflow: initiation of motion, continued displacement, and sufficient fracture conductivity along the return path25. Research in China began later but has progressed rapidly, especially in response to the specific geological characteristics of domestic CBM reservoirs26,27.Recent efforts include laboratory investigations of how flowback rate, post-gel viscosity, and fracture geometry influence proppant movement28–30.Mechanical models have been developed to describe the forces governing proppant initiation and to estimate critical flowback rates31,32.Additional studies have examined proppant settling, migration, and their combined effect on fracture conductivity33–35.Despite these advances, most existing models account for only a few dominant forces and often overlook the complex mechanical behavior of proppant in CBM formations36,37.Many models focus solely on proppant movement, without considering the need to balance flowback rates to avoid both sand production and inadequate removal of coal fines38–40. Effective flowback strategies in CBM operations must maintain fracture conductivity while maximizing early production. Achieving this balance requires a more comprehensive understanding of the forces acting on particles during flowback. The present study aims to address this need by developing a flowback velocity model that reflects the unique conditions of CBM reservoirs and incorporates the post-break rheological condition of the fracturing fluid rather than a full gel-degradation model. The proposed model incorporates eight distinct forces acting on particles during the flowback process and considers both pre-closure and post-closure stages of the fracture. Separate analyses were conducted for coal fines and proppant under each condition. In the early phase of flowback, fracture pressure exceeds closure pressure, keeping fractures open and particle arrangements loose. Proppant is driven by pressure and viscous drag but constrained by gravity and the weight of overlying particles. In the later phase, fracture closure increases contact among proppant particles, and movement becomes increasingly restricted by overburden pressure, fluid drag, cohesion, and friction, while still driven by capillary and pressure forces. The model calculates critical flowback velocities based on these varying force conditions. Sensitivity analysis examines how factors such as particle radius, density, pressure differential, and wetting phase saturation affect flowback performance. The analysis also includes the influence of gel breaking, offering a more complete understanding of how fluid behavior impacts flowback efficiency in CBM wells. New insights of this study are threefold. First, we develop a unified force-balance framework that integrates eight particle-scale forces and explicitly distinguishes pre-closure and post-closure regimes, enabling a transparent mechanistic interpretation of how the controlling resistance shifts from gravity-dominated constraints to friction-dominated constraints during fracture closure. Second, we introduce a dual-criterion operational window by simultaneously determining the minimum velocity required for coal-fines mobilization and the maximum allowable velocity to prevent proppant backflow, thereby formalizing the practical trade-off between early cleanup efficiency and long-term fracture conductivity. Third, we represent early flowback rheology using a measurable post-break viscosity and provide a quantitative sensitivity ranking that identifies the dominant controllable parameters (particle size, viscosity, and post-closure pressure differential ΔP) for field choke management. It should be emphasized that the objective of this study is not to develop a flowback schedule that is only applicable to a specific CBM block. Rather, the field cases are used to demonstrate how the proposed force-balance framework can be parameterized using measurable engineering inputs. The governing equations and calculation workflow are independent of a particular well, and the required parameters can be updated for different proppant types, fracturing fluids, closure-pressure conditions, and particle-size distributions. Therefore, the methodology provides a general mechanistic criterion for flowback-rate design, while the case calculations serve as examples of model implementation. It should also be noted that the present study focuses on the development and reproducible implementation of a mechanics-based model rather than on new laboratory experimentation. Therefore, the term “validation” is used cautiously in this work. The model is examined through three complementary checks: (i) all governing equations, assumptions, and input parameters are explicitly provided to ensure reproducibility; (ii) the predicted critical flowback rates are compared with the order of magnitude of field flowback rates and published experimental observations; and (iii) sensitivity analysis is used to verify whether the model produces physically consistent responses to changes in particle size, density, fluid viscosity, pressure differential, and wetting-phase saturation. This approach provides a transparent basis for model verification, while dedicated experimental validation is identified as a necessary direction for future work. Methods Analysis of coal fines generation and solid particle transport Coal fines primarily form due to geological stress conditions and mechanical disturbances during drilling and completion. Shear forces between fault blocks, tensile stress concentrations at anticline crests, and compressive stress at syncline cores all contribute to coal rock fragmentation. Highly active tectonic zones further promote coal breakage. During perforation, the combined impact of fracturing fluids and proppants generates tensile stresses that cause surface shear and compaction damage. In addition, variations in production intensity may lead to coal erosion, stress-induced damage from matrix shrinkage during gas desorption, and pressure fluctuations caused by intermittent production—all of which can contribute to coal fines generation. Most studies on flowback velocity under gel degradation conditions focus on the single-phase water flow stage, which occurs before reservoir pressure falls below the critical desorption threshold. During this early phase, the coal matrix still retains gas, and only water is produced. As flowback continues, gel degradation progresses, reservoir pressure drops, and effective stress increases. These changes accelerate fracture closure, reduce permeability, and increase resistance to gas and water flow. Coal fines located farther from the wellbore generally fail to migrate, so production during this stage is dominated by medium to coarse particles near the well. The particle size distribution typically shows a bimodal pattern with a peak skewed toward larger diameters. As pressure continues to decline, the reservoir enters the gas–water two-phase flow stage. Strong hydrodynamic forces and relatively low effective stress keep the pore structure partially open, allowing larger coal fines to be mobilized and transported. This stage consists of two sub-phases: initial gas breakthrough and increasing gas production. Once pressure near the wellbore falls below the critical desorption level, gas begins to desorb, though flow remains intermittent and gas volumes are low. Matrix shrinkage effects remain limited, and permeability continues to decline under rising effective stress. Water supply from the formation becomes unstable, requiring frequent production adjustments to maintain a consistent drawdown. Excessive production rates can disturb solid particles, triggering the release of coal fines and proppant, and resulting in a three-phase flow of gas, liquid, and solids. As the pressure depletion cone expands, desorption intensifies and a continuous gas stream forms. Gas slippage effects increase gas-phase permeability, while matrix shrinkage becomes the dominant factor driving coal structure evolution. At the same time, water-phase permeability declines, leading to peak gas production and gradually decreasing water output. Production strategies and fluid properties stabilize during this phase, but the ability to carry coal fines weakens. The coal fines produced tend to be finer, with lower concentrations. However, elevated concentrations can still disrupt stable production. Eventually, the reservoir transitions to the late-stage gas-phase flow. Gas output steadily declines, and reservoir pressure approaches abandonment levels. By this point, most formation water has been removed, and recoverable methane has largely been produced in free form. Only a few areas continue to desorb gas. Production of both gas and water slows or ceases entirely, resulting in long-term low output. Most coal fines settle within fractures, and only a small amount is produced. These particles are typically under 1 μm and are carried out with the gas as airborne dust. Although gel degradation is considered in the modeling framework, the present study does not incorporate a full viscosity–time function because detailed kinetic parameters of gel breaking under CBM-relevant pressure, temperature, and breaker concentration are not available from field operations. Instead, a representative post-break viscosity is used to approximate the rheological state of the fracturing fluid during the early flowback stage after breaker activation. This treatment reflects common field practice, as gel breakers are designed to reduce viscosity rapidly after activation, after which the fluid rheology tends to approach a near-Newtonian state with reduced time dependency. Therefore, the model does not quantify the full gel-degradation kinetics; rather, it accounts for the post-break fluid behavior through a measurable viscosity input. The influence of viscosity variability is further evaluated in the sensitivity analysis. To clarify this assumption, the selected viscosity value represents the measured post-break viscosity at reservoir temperature, typically ranging from 1.2 to 2.0 mPa·s, consistent with breaker performance curves provided by suppliers. The influence of viscosity variability is further analyzed in the sensitivity analysis, demonstrating that the model captures the correct trend of decreasing critical flowback velocity as viscosity increases. Flowback velocity calculation model To determine an appropriate flowback velocity for CBM wells, a mechanical model was developed based on particle force analysis. The model accounts for two distinct fracture states: before and after closure. Prior to fracture closure, particles remain loosely arranged and are primarily supported by the fracturing fluid. As closure progresses under increasing formation stress, particles become more compact and experience higher confinement pressures. The model assumes uniform particle size, with no deformation or fragmentation, to simplify the mechanical calculations. During flowback, the pressure differential between the fracture and the formation drives fluid movement, while the stability of solid particles depends on the balance of competing forces. Some of these forces resist particle motion, while others act as the primary drivers of particle migration. The initiation of particle movement is governed by the rolling mechanism: particles begin to move when the driving moment exceeds the resisting moment, causing rotation about the contact point. Based on a detailed force balance, the model defines the critical conditions for particle displacement. The maximum critical flowback velocity is derived from the condition at which proppant particles remain stationary, while the minimum critical velocity is determined from the threshold at which coal fines begin to move. Using these two limits, the model establishes a safe flowback-rate window—high enough to remove coal fines efficiently, yet low enough to prevent proppant backflow and maintain fracture conductivity. Force analysis of sand particles To ensure reproducibility of the force-balance model, the key empirical constants used in the friction, film pressure, and cohesive force terms are briefly explained here. The friction coefficient () reflects the resistance at quartz–quartz contacts under closure stress, and a value of 0.75 is commonly adopted based on rock–proppant friction tests reported in Andrews & Kjorholt (1998)41. The liquid film coefficient () describes the thickness of the lubricating water film remaining between contacting particles during early fracture closure, and a representative value of 0.213 μm is taken from laboratory observations of quartz particle contacts (Hu, 2007)44. The cohesion coefficient () represents inter-particle cementation and water-bridging forces under closure stress; for quartz contacts, a typical value is 2.56 dyn/cm, consistent with measurements reported in Fu (2006) and subsequent CBM fracture-closure testing45. These constants provide reasonable physical estimates under CBM fracturing conditions and allow consistent application of the model across field cases. (1) Net Gravitational Force. The net gravitational force accounts for the particle’s own weight minus buoyancy. It is calculated as: (2) Overburden Particle Pressure. Particles in the lower layers experience vertical pressure from the weight of overlying particles. The total pressure is proportional to the number of overlying layers: (3) Pressure Gradient Drag Force. The pressure gradient drag force arises from differential pressure across the fracture, which drives particle migration. In a uniform pressure field, particles experience varying forces across their cross-sections. Integrating these forces yields the total drag force. Applying the Perkins–Kern–Nordgren model (PKN) model for elliptical fractures under Newtonian laminar flow conditions, the pressure gradient is expressed as38: Substituting into the drag force equation gives: (4) Viscous Drag Force. Viscous drag arises from the velocity difference between the particle and the surrounding fluid, exerting a force that encourages the particle to move with the flowing fluid. The viscous drag force is given by: where, , 41,42. The drag coefficient varies with flow regime:NRe≤2, k = 24, τ = 1; 2500, k = 0.44, τ = 0. (5) Frictional Force. Upon fracture closure, particles are compressed against one another. Since fluid pressure within the fracture only supports part of the closure stress, the remaining stress is transferred to the particle framework. If the particles are not crushed, they must overcome inter-particle friction to initiate movement43: (6) Liquid Film Force. Cementation between particles creates thin fluid films. Because these films cannot transmit hydrostatic pressure, a net downward force is generated44: (7) Inter-Particle Cohesive Force. During fracture closure, cementation also introduces cohesive forces between contacting particles: (8) Equivalent Capillary Force. During flowback, the saturation of fracturing fluid within the fracture gradually decreases, while gas or oil saturation increases. This change generates capillary resistance, which restricts fluid movement. The resulting reaction force acts on solid particles and promotes their mobilization and backflow. The equivalent capillary force, defined as equal in magnitude but opposite in direction to capillary resistance, quantifies this effect. Its calculation follows standard capillary pressure methods. The equivalent capillary force depends on particle radius, interfacial tension, contact angle (𝜃), and wedge angle (𝛼). The wedge angle reflects the saturation level of the wetting phase. Traditional models often estimate the wetting phase volume using two-dimensional planar approximations45, which can introduce inaccuracies. To improve precision, the current study adopts a three-dimensional integration method21 that better captures the spatial distribution of fluid around particles. Capillary pressure is governed by the Laplace equation. In the fracture environment, the wetting phase forms an annular film around the solid particles, with its center of curvature on the negative side (R2 < 0), while the non-wetting phase fills the central pore space and has a positive radius of curvature (R1 > 0). Using this configuration, the capillary force acting at the point of contact between spherical particles is expressed as: As illustrated in Fig. 1, the radius R1 represents the curvature of the interface between the wetting and non-wetting phases. The center of curvature lies on the side of the non-wetting fluid. Geometric diagram for equivalent capillary force. The radius R2 corresponds to the distance from the curved meniscus to the y-axis, with its center of curvature located on the side of the wetting fluid. This radius is calculated as the horizontal distance from the contact point between the non-wetting phase and the particle to the y-axis, minus the offset △r: Substituting R1 and R2 into the capillary force equation for spherical particles (previously defined as Eq. 10) yields the capillary pressure Pc. Multiplying this pressure by the cross-sectional area of a single particle gives the magnitude of the equivalent capillary force acting on an individual particle46: Flowback velocity analysis The critical flowback velocity represents the condition where all forces acting on a particle are balanced. At the onset of flowback, the fluid pressure within the fracture remains above the formation’s closure pressure, allowing the fracture to stay open. Although particles are motionless and loosely deposited, they do not form tight contacts or exhibit significant cementation. During this phase, each particle is influenced by four primary forces: net gravitational force, pressure from overlying particles, pressure-gradient drag, and viscous drag from the fluid. Gravitational and overburden forces act to restrain particle movement, while pressure-gradient and viscous drag forces promote mobilization. The balance between these opposing forces determines whether particles remain stable or begin to move with the fluid. Figure 2 provides a detailed illustration of the force interactions during early-stage flowback. Force analysis of loosely packed particles. Among the dominant forces acting on particles, both the pressure-gradient drag and the viscous drag depend directly on the flowback velocity. The critical velocity can be determined by equating the driving torque and the resisting torque, representing the critical velocity at which particle motion begins45,47: where, L1——moment arm of the driving force, µm; L2——moment arm of the resisting force, µm. Geometric relationships define L1 and L2 based on the configuration and contact conditions of the particles. , . During the later stage of flowback, fracture pressure falls below the formation closure pressure. Fracture apertures remain open only due to the mechanical support of the proppant. Particles become tightly packed under stress and form cemented contacts, resulting in more complex force interactions. Resisting forces include net gravitational force, overburden pressure, fluid film pressure, interparticle cohesion, and friction. In contrast, pressure-gradient drag, viscous drag, and equivalent capillary force contribute to particle mobilization. Figure 3 illustrates the detailed force distribution acting on particles under these conditions. Force analysis of tightly contacted particles. The corresponding maximum fracturing fluid flowback velocity is calculated using the following equation, which serves as the basis for determining the appropriate flowback rate. where, L1——moment arm of the driving force, ; L2——moment arm of the resisting force, . Generalized implementation workflow and reproducibility The proposed model is designed as a parameterized calculation framework rather than a field-specific empirical correlation. Its application requires four categories of inputs: (i) particle properties, including proppant and coal-fines size distributions and densities; (ii) post-break fluid properties, including density and viscosity under reservoir temperature; (iii) fracture-state variables, including fracture width, closure pressure, fracture-fluid pressure, and the pre-closure or post-closure state; and (iv) contact and interfacial parameters, including friction coefficient, cohesion coefficient, liquid-film coefficient, interfacial tension, contact angle, and wetting-phase saturation. These variables are not unique to the field cases considered in this study and can be obtained from laboratory testing, field operation records, or literature-based ranges. The generalized workflow contains five steps. First, the flowback stage is identified according to the relationship between fracture-fluid pressure and closure pressure. Second, the corresponding force set is selected for the pre-closure or post-closure regime. Third, the critical velocity for coal-fines mobilization is calculated as the lower bound of the flowback-rate window, while the critical velocity for proppant mobilization is calculated as the upper bound. Fourth, sensitivity analysis is conducted to identify the dominant controllable variables for choke management. Finally, the recommended operating window is updated when field pressure, fluid viscosity, or particle-size information changes during flowback. To improve reproducibility, all force terms, moment-balance equations, empirical coefficients, and case-specific input parameters are explicitly reported in the manuscript. The calculation is deterministic: for a given set of particle, fluid, fracture-state, and contact parameters, the critical flowback velocity can be reproduced by substituting these inputs into Eqs. (1)–(15). No well-specific fitting coefficient or hidden calibration parameter is introduced. Because independent laboratory flowback experiments were not conducted in this work, tthe plausibility of the model is examined through reproducibility checks, literature-constrained parameter selection, order-of-magnitude comparison with published field and experimental results, and physically interpretable sensitivity trends. This verification strategy does not replace direct experimental validation, but it provides a transparent and repeatable basis for evaluating the engineering plausibility of the proposed framework. The parameters listed in Table 1 are sufficient to reproduce the case calculations and sensitivity trends reported in this study. For other wells, the same equations can be used after replacing these parameters with site-specific measurements or appropriate literature-constrained values. Table 1Main input parameters and sources for model reproduction.Parameter groupParameterValue or range used in this studySource or basisProppant propertyProppant radius53, 106, 150, 212.5, 300, 425 μmField proppant designProppant propertyProppant density2.65–2.67 g/cm³Field proppant dataCoal-fines propertyCoal-fines density1.49 g/cm³Field CBM coal-fines densityCoal-fines propertyCoal-fines diameter25–200 μmSensitivity and mobilization analysisFluid propertyPost-break viscosity1.2, 2.8, 15.2 mPa·sField post-break viscosity dataFluid propertyFluid density1.0 g/cm³Representative gel-broken fracturing fluidContact parameterFriction coefficient0.75 Andrews & Kjorholt, 1998Contact parameterLiquid-film coefficient0.213 μm Hu, 2007Contact parameterCohesion coefficient2.56 dyn/cm Fu, 2006Fracture stateIncipient closure threshold10 PaNumerical transition threshold, not reservoir closure stressFracture statePost-closure ΔP range10⁻⁶–10⁻³ MPaSensitivity analysis Main input parameters and sources for model reproduction. Results and Discussion Field-case demonstration of the generalized model The field cases are used to demonstrate how the generalized model can be implemented with realistic CBM input parameters. They are not used to calibrate a well-specific empirical equation. Once the particle-size distribution, particle density, post-break viscosity, fluid density, fracture-state variables, and contact parameters are provided, the same force-balance equations can be solved for other wells. Wells LX55 and LX-8-4D were selected because they provide representative combinations of proppant size, post-break fluid viscosity, and CBM flowback conditions. These parameters are used as illustrative inputs to examine the magnitude and physical consistency of the calculated flowback-rate window under realistic engineering ranges. To examine the engineering plausibility of the calculated results, field-derived parameters from two CBM wells (LX55 and LX-8-4D) were used as input examples. The calculated critical flowback velocities fall within the same order of magnitude as reported early-stage flowback rates in similar CBM reservoirs, generally between 5 and 600 m³/day. (1) Maximum Flowback Velocity. The maximum flowback velocity corresponds to the critical sand production rate, defined as the minimum velocity required to initiate proppant displacement. To maintain long-term fracture conductivity in Well LX55, proppants of 70/140 mesh, 40/70 mesh, and 20/40 mesh were selected, with corresponding particle radii of 53, 106, 212.5, and 425 μm. The bulk density ranged from 1.46 to 1.54 g/cm3, and the true density from 2.65 to 2.67 g/cm3. Two types of gel-broken fracturing fluids were used, with viscosities of 15.2 mPa·s and 2.8 mPa·s, and a fluid density of g/cm3 was assumed. For post-closure conditions, we consider an incipient-closure state, defined as the moment when the fracture-fluid pressure just falls below the minimum in-situ closure stress. In the base case, a pressure differential of 10 Pa is adopted solely as a numerical threshold to distinguish pre- and post-closure conditions in the friction term of Eq. (7). This value does not represent the true reservoir closure stress, which is typically in the MPa range for CBM formations at the studied depths. The novel aspect of our treatment is not the magnitude of 10 Pa itself, but its use as a methodological incipient-closure trigger that enables a smooth transition between the pre-closure and post-closure force formulations without introducing arbitrary large friction at the instant of closure. This approach preserves mechanical consistency, avoids discontinuities in the critical-velocity calculation, and allows systematic exploration of how friction progressively dominates as the pressure differential increases. To demonstrate that the conclusions are not tied to the incipient-closure threshold, Sect. 3.2 (Fig. 11) evaluates the post-closure pressure differential over 10⁻⁶−10⁻³ MPa. Across this range, the calculated critical flowback velocity increases monotonically with ΔP, while the directional sensitivities to particle size and fluid viscosity remain unchanged. This confirms that the 10 Pa threshold is used only as a conservative computational trigger and does not determine the observed parameter trends. Effect of frictional resistance on critical flowback velocity under varying post-closure pressure differentials. (Black line: proppant diameter 200 μm, density 2.5 g/cm³; Blue line: proppant diameter 100 μm, density 1.5 g/cm³). Note: 10 Pa is used only as an incipient-closure numerical threshold, not as the actual reservoir closure stress (typically MPa). Table 2 and the accompanying figure present the calculated critical velocities for various particle sizes and fluid conditions. Before fracture closure, the critical velocity ranges from 3.58 to 498.57 m³/day; after closure, it increases to 36.7–890.63 m³/day, depending on particle size and fluid viscosity Figs. 4 and 5. Table 2Calculated Critical Flowback Velocities for Proppants in Well LX55.Average particle radius (µm)True density (g/cm3)Fracturing fluid viscosity (mPa·s)Critical velocity before fracture closure (m³/day)Critical velocity after fracture closure (m³/day)4252.6615.2218.41425.17212.52.6615.255.50196.031062.6615.213.4385.93532.6615.23.5836.704252.662.8498.57890.63212.52.662.8227.36562.131062.662.874.29378.63532.662.817.90201.40 Calculated Critical Flowback Velocities for Proppants in Well LX55. Critical Proppant Flowback Velocities Calculated for Well LX55. (Blue indicates values before fracture closure; Yellow indicates values after fracture closure). Well LX-8-4D used proppants sized 70/140 mesh, 40/70 mesh, and 30/50 mesh, corresponding to particle radii of 53, 106, 150, 212.5, and 300 μm. The bulk density ranged from 1.46 to 1.54 g/cm3, and the true density ranged from 2.65 to 2.67 g/cm3. Two gel-broken fracturing fluids were applied, with viscosities of 15.2 mPa·s and 2.8 mPa·s. The fracturing fluid density was assumed to be 1 g/cm3. After fracture closure, the closure pressure exceeded the internal fracture pressure by 10 Pa, which was used as the incipient-closure input for calculating the post-closure critical flowback velocity. Under varying particle sizes and fluid viscosities, the critical sand production velocity ranged from 4.42 to 415.83 m³/day before fracture closure and increased to 45.34 to 863.72 m³/day after closure Table 3. Table 3Critical Proppant Flowback Velocities Calculated for Well LX-8-4D.Average Particle Radius (µm)True Density (g/cm3)Fracturing Fluid Viscosity (mPa·s)Critical Velocity Before Fracture Closure (m³/day)Critical Velocity After Fracture Closure (m³/day)3002.6615.2136.03374.91212.52.6615.268.57242.201502.6615.234.28158.151062.6615.216.59106.17532.6615.24.4245.343002.662.8415.83863.72212.52.662.8280.90694.521502.662.8184.69567.341062.662.891.79467.80532.662.822.12248.83 Critical Proppant Flowback Velocities Calculated for Well LX-8-4D. Calculated Critical Proppant Flowback Velocities for Well LX-8-4D. (Blue: Before Fracture Closure; Yellow: After Fracture Closure). (2) Minimum Flowback Velocity. The minimum flowback velocity refers to the threshold velocity required to mobilize coal fines. For the analysis, coal fines were assumed to have a density of 1.49 g/cm3, with particle diameters set at 25, 50, 75, 100, 125, 150, 175, and 200 μm. The gel-broken fracturing fluid was assigned a viscosity of 1.2 mPa·s and a density of 1 g/cm3. Figure 6 presents the calculated critical velocities for each particle size. Calculated critical flowback velocities for coal fines of various sizes. The results indicate that, during the initial flowback stage, the flow rate should exceed the critical velocity required to remove coal fines, but remain below the threshold that would cause proppant backflow. Once the fracture closes and the critical velocity for sand production increases, operators can adjust flow rates to prioritize other production considerations, thereby accelerating the transition to stable gas production. Although no new laboratory flowback experiment was performed in this study, the calculated results can be checked against published experimental observations and field-scale flowback behavior. First, the calculated critical proppant flowback velocities fall within the same order of magnitude as reported early-stage flowback rates in CBM and other hydraulically fractured reservoirs. For the two field cases, the predicted critical velocities range from several m³/day to several hundred m³/day before fracture closure and increase substantially after closure, which is consistent with the operational observation that early flowback is usually controlled by small choke sizes and that higher flowback rates are more acceptable after proppant packs become mechanically confined. Second, the model reproduces the main physical trends reported in previous experimental and numerical studies. Larger and denser particles require higher flowback velocities to initiate movement because gravitational and contact-related resistance increases. Higher fluid viscosity lowers the critical velocity because viscous drag is enhanced. After fracture closure, frictional resistance becomes dominant and the critical velocity increases with pressure differential. These trends are consistent with the established understanding of proppant backflow and particle transport in hydraulic fractures. Third, the sensitivity analysis provides an internal consistency check for the model. The directional responses of the model are physically interpretable and remain unchanged when the input parameters are varied over representative engineering ranges. Therefore, while the model still requires further validation using controlled laboratory flowback experiments, the present comparison supports its reproducibility and engineering plausibility as a screening-level calculation framework. Sensitivity analysis of key parameters To more clearly evaluate the influence of key parameters, the sensitivity analysis was organized into three categories: particle properties, fluid properties, and fracture-state/interfacial variables. Particle radius and density mainly control gravitational and contact-related resistance; fluid density and viscosity mainly affect buoyancy and viscous drag; and pressure differential and wetting-phase saturation control the post-closure frictional and capillary contributions. This classification allows the dominant controlling mechanisms to be distinguished rather than only presenting separate parameter-response curves. The initiation of particle movement primarily depends on the force balance acting on the particles. The factors influencing this balance include particle characteristics, fluid properties, and the magnitude of the applied forces. (1) Particle Radius. Figure 7 illustrates the variation in critical flowback velocity with particle radius. With proppant density fixed at 2.5 g/cm3, fracturing fluid density at 1 g/cm3, and viscosities of 1.2 mPa·s and 1.5 mPa·s, the critical velocity increases as the particle radius increases from 100 μm to 350 μm. Although the radius appears in all force equations, changes in particle size influence resistance more significantly than driving forces. Larger particles require higher flowback velocities to initiate movement. Effect of Proppant Radius on Critical Velocity. (Black line: Fracturing fluid viscosity 1.5 mPa·s; Blue line: Fracturing fluid viscosity 1.2 mPa·s;. (2) Particle Density. As shown in Fig. 8, increasing particle density leads to a higher critical flowback velocity. For a proppant diameter of 200 μm, fluid density of 1 g/cm3, and viscosities of 1.2 mPa·s and 1.5 mPa·s, the critical velocity increases significantly with density. For instance, with a viscosity of 1.5 mPa·s, a particle density of 2 g/cm3 yields critical velocities of 92.4 m³/day (pre-closure) and 287.95 m³/day (post-closure), while a density of 3 g/cm3 increases these values to 152.02 m³/day and 329.08 m³/day, respectively. This rise stems from the greater gravitational force resisting particle movement. Effect of Proppant Density on Critical Velocity. (Black line: Fracturing fluid viscosity 1.5 mPa·s; Blue line: Fracturing fluid viscosity 1.2 mPa·s;. Solid line: After fracture closure; Dashed line: Before fracture closure) Solid line: After fracture closure; Dashed line: Before fracture closure) (3) Fracturing Fluid Density. Figure 9 demonstrates that critical flowback velocity decreases with increasing fluid density. For a 200 μm proppant, a fluid density increase from 0.6 to 1.6 g/cm3 causes a reduction in critical velocity from 169.91 to 75.12 m³/day (pre-closure), and from 376.18 to 248.6 m³/day (post-closure). Higher fluid density strengthens buoyancy and viscous drag, promoting particle movement and reducing the required velocity. Effect of Fracturing Fluid Density on Critical Velocity. (Black line: Proppant diameter 200 μm, density 2.5 g/cm3; Blue line: Proppant diameter 100 μm, density 1.5 g/cm3; Solid line: After fracture closure; Dashed line: Before fracture closure). (4) Fracturing Fluid Viscosity. Among the variables, viscosity is dynamic over time. Field applications use fluids ranging from low (viscosity > 5 mPa·s), medium (> 20 mPa·s), to high (> 40 mPa·s). Post-gel-breaking viscosities drop below 5 mPa·s at 30 °C within 120 min. As shown in Fig. 10, higher viscosity reduces the critical flowback velocity. Since viscosity directly influences drag forces, more viscous fluids require lower velocities to mobilize particles. Effect of Fracturing Fluid Viscosity on Critical Velocity. (Black line: Proppant diameter 200 μm, density 2.5 g/cm3; Blue line: Proppant diameter 100 μm, density 1.5 g/cm3; Solid line: After fracture closure; Dashed line: Before fracture closure). (5) Pressure Differential Between Fracture and Formation. The magnitudes of net gravity, pressure-driven drag, viscous drag, liquid pressure, and interparticle cohesive forces primarily depend on the properties of the proppant particles and the fracturing fluid. These forces remain relatively stable during flowback. In contrast, frictional resistance—affected by the pressure differential between the fracture and the surrounding formation—and the effective capillary force—governed by wetting phase saturation—vary significantly throughout the process and require detailed analysis. The pressure differential, defined as the difference between formation closure pressure and fracture fluid pressure, represents the stress borne by the proppant framework. Friction acts as a major barrier to proppant mobilization and increases with the post-closure pressure differential because ΔP represents the stress transferred to the proppant framework. Once the fracture-fluid pressure falls below the minimum closure stress, friction becomes the dominant resisting force. As ΔP increases during continued flowback, the calculated critical flowback rate increases accordingly. For example, at the onset of closure (ΔP = 1 × 10⁻⁶ MPa), the calculated critical flowback rate is 124.6 m³/day; under otherwise identical conditions, increasing ΔP to 5 × 10⁻⁴ MPa raises the critical flowback rate to 4325.74 m³/day. This monotonic response is consistent with Eq. (7) and indicates progressively stronger resistance to particle mobilization at higher differentials. When ΔP approaches the closure-stress scale, the predicted critical rate may exceed practical operational limits, implying a markedly reduced likelihood of proppant movement under typical field flowback conditions. These results highlight the controlling role of ΔP-driven friction in post-closure flowback. (6) Wetting Phase Saturation. The effective capillary force, a key driver of proppant mobilization, depends on the saturation level of the wetting phase. As flowback progresses, the content of fracturing and formation fluids declines while oil and gas saturation increases. For proppant particles, the wetting phase includes both injected and formation water. A decrease in saturation leads to a stronger capillary force, with an accelerating rate of increase as saturation further declines. Under a constant pressure differential of 1 × 10− 3 MPa, the effective capillary force and corresponding critical flowback velocity can be determined. During early fracture closure, high wetting phase saturation minimizes capillary force. When saturation drops to 14%, the critical flowback rate decreases to 243.83 m³/day. At 13%, the rate further drops to 59.62 m³/day. From the onset of capillary influence to near-zero saturation, the critical velocity decreases by 184.21 m³/day. Although effective capillary force significantly lowers the critical flowback rate, its overall impact remains smaller than that of friction. Proppant mobilization begins when flowback reaches the critical velocity. Before fracture closure, particles are loosely arranged, requiring relatively low velocities to initiate movement. After closure, tighter particle packing necessitates higher velocities. Larger and denser particles require greater force to mobilize, while smaller particles move more readily. Low-viscosity and low-density fluids offer less drag, requiring higher velocities for mobilization. Conversely, high-viscosity or high-density fluids generate stronger drag forces and can move particles at lower rates. As flowback continues, both the effective capillary force and friction evolve, but friction consistently plays the dominant role. Consequently, the critical flowback velocity increases after fracture closure. Field operations commonly reflect this trend, with early-stage flowback conducted through small choke sizes, followed by high-rate flowback through larger chokes to enhance efficiency Fig. 12. Effect of Effective Capillary Force on Critical Velocity. (Black line: Proppant diameter 200 μm, density 2.5 g/cm3; Blue line: Proppant diameter 100 μm, density 1.5 g/cm3). To provide a clearer and more practical understanding of parameter sensitivity, a brief synthesis of the major controlling factors is summarized here. Among all variables, particle radius exhibits the strongest positive correlation with critical flowback velocity; larger particles require substantially higher drag forces to initiate motion. Particle density also shows a pronounced influence, as increased density enhances gravitational resistance and raises the flowback velocity threshold. In contrast, fracturing fluid viscosity acts in the opposite direction—higher viscosity strengthens viscous drag, effectively lowering the critical velocity needed to mobilize particles. Fluid density likewise reduces the required velocity by increasing buoyancy. In terms of reservoir conditions, the pressure differential between the fracture and formation exerts the most significant impact after fracture closure, sharply increasing frictional resistance and causing a substantial rise in critical flowback velocity. Meanwhile, decreasing wetting-phase saturation enhances capillary forces, which can partially offset friction and lower the critical velocity, though its magnitude remains secondary compared with friction. Within the parameter ranges examined in this study, particle size, post-closure pressure differential, and fluid viscosity show the strongest effects on the calculated critical flowback velocity. This ranking provides a mechanistic basis for prioritizing particle-size characterization, post-break viscosity monitoring, and choke-controlled pressure management during early flowback. Based on the magnitude of the response in Figs. 7, 8, 9, 10 and 11, the dominant factors can be ranked as: particle size (radius) ≳ post-closure pressure differential ΔP > fluid viscosity, followed by particle density and wetting-phase saturation, while fluid density is secondary under typical CBM flowback conditions. From an engineering standpoint, this ranking suggests that field operations should prioritize (i) controlling/monitoring ΔP through choke management after closure, and (ii) constraining the effective particle size distribution (proppant selection and fines control), while (iii) tracking post-break viscosity to anticipate shifts in the critical velocity window. These priorities provide a practical basis for designing a conservative early-stage flowback schedule that balances coal-fines cleanup and proppant stability. As summarized in Table 4, particle radius and post-closure pressure differential are the two most influential parameters because they directly control the dominant resisting forces before and after fracture closure, respectively. Fluid viscosity also has a strong influence because it directly modifies viscous drag. Particle density and wetting-phase saturation have secondary but still non-negligible effects, whereas fluid density shows a relatively weaker influence under the parameter ranges considered in this study. Therefore, the controlling factors can be ranked as follows: particle radius ≳ post-closure pressure differential > fluid viscosity > particle density ≈ wetting-phase saturation > fluid density. This ranking should be interpreted as a model-based sensitivity order within the studied parameter ranges, rather than as a universal ranking for all reservoirs. Table 4Summary of key parameter effects on critical flowback velocity.ParameterDirection of influenceMain force affectedPhysical interpretationOperational implicationParticle radiusPositiveGravity, overburden pressure, contact resistanceLarger particles require stronger hydrodynamic force to initiate motionCoarser proppant requires more conservative early flowbackParticle densityPositiveNet gravityDenser particles have greater resisting forceHigh-density proppant is less prone to flowbackFluid densityNegativeBuoyancy, dragHigher density reduces effective particle weightDense fluids can mobilize particles at lower velocityFluid viscosityNegativeViscous dragHigher viscosity enhances drag forceIncomplete gel breaking may increase particle mobilization riskPost-closure ΔPStrong positiveFrictionLarger stress transfer to proppant pack increases frictional resistanceChoke management should avoid abrupt drawdown changesWetting-phase saturationNegative when saturation decreasesCapillary forceLower wetting saturation enhances capillary force and promotes particle mobilizationLate-stage saturation change may alter proppant stability Summary of key parameter effects on critical flowback velocity. Generality, scalability, and applicable conditions The broader applicability of the proposed method lies in its parameterized mechanical structure. The model does not rely on empirical coefficients fitted to a particular CBM block. Instead, it calculates the critical flowback velocity from particle-scale force balance, in which the input parameters can be obtained from laboratory measurements, treatment design data, field monitoring, or literature-based ranges. These parameters include particle radius and density, post-break fluid viscosity, fracture width, pressure differential, wetting-phase saturation, interfacial tension, and contact-related coefficients. For different wells or reservoirs, the calculation procedure remains the same. Only the input parameters need to be updated according to local geological and operational conditions. For example, changes in proppant type or coal-fines size distribution mainly affect gravitational, drag, and contact-related forces; changes in fracturing-fluid formulation affect the viscous-drag term through post-break viscosity; and changes in drawdown strategy or closure condition affect frictional resistance through the post-closure pressure differential. Therefore, the proposed method provides a transferable framework for evaluating the allowable flowback-rate window under different engineering scenarios. The scalability of the model is reflected in both computational and engineering aspects. Computationally, the model is based on closed-form force-balance equations and therefore can be rapidly applied to multiple wells, proppant sizes, fluid systems, and pressure-differential scenarios without requiring large numerical simulations. From an engineering perspective, the same workflow can be scaled from a single-well calculation to a multi-well screening process by updating the input parameter set for each well. This makes the method suitable for preliminary flowback-rate design, sensitivity ranking, and choke-management optimization across a group of CBM wells. Nevertheless, the model should be applied within its mechanical assumptions. It is most suitable for early to mid-stage flowback in propped fractures where particle mobilization can be approximated by rolling initiation and where the fracturing fluid after gel breaking behaves approximately as a Newtonian fluid. For reservoirs with complex fracture networks, strong non-Newtonian flow, severe proppant crushing, or rapidly evolving fracture geometry, additional coupling with fracture-network flow models, proppant-transport models, or time-dependent rheological models may be required. Limitations Although the proposed methodology provides a generalized mechanical framework for evaluating particle mobilization during flowback, its quantitative prediction depends on the accuracy of reservoir-specific input parameters. Therefore, the method should be interpreted as a transferable screening and design tool rather than a universal closed-form rule. The governing equations and calculation workflow can be transferred to other wells, but the calculated critical velocities must be recalculated using site-specific particle, fluid, pressure, and fracture-state parameters. Accordingly, the numerical results obtained for LX55 and LX-8-4D should not be directly used as universal flowback rates for other CBM reservoirs. A major limitation is that the proposed model has not yet been validated using dedicated laboratory flowback experiments under controlled HP/HT conditions. The present results should therefore be regarded as mechanically consistent predictions and screening-level estimates rather than fully experimentally validated flowback criteria. Although the field cases and literature comparison provide order-of-magnitude support, they cannot fully quantify prediction errors or uncertainty bounds. Future work should include propped-fracture flowback experiments with controlled particle size, closure stress, fluid viscosity, and pressure differential, so that the predicted critical velocities can be directly compared with observed particle-mobilization thresholds. Parameter measurability under CBM-relevant HP/HT conditions also remains limited. Several force components, such as cementation-induced cohesion, liquid-film pressure under closure stress, and the evolution of interparticle friction, cannot be directly measured or continuously monitored in situ. These parameters are therefore represented using literature-based values or representative ranges. In addition, the model adopts constant contact-related coefficients, including the friction coefficient, film coefficient, and cohesion coefficient, to maintain analytical tractability. In practice, these coefficients may vary with proppant type, fines content, surface roughness, fluid chemistry, and evolving wetting conditions, which may lead to deviations in predicted critical velocities. The rheology of the fracturing fluid is represented by a single post-break viscosity rather than a full viscosity–time function. This approximation is reasonable for screening-level calculations when the fluid has largely completed gel breaking and behaves approximately as a near-Newtonian fluid. However, if incomplete gel breaking, delayed breaker activation, or temperature-dependent degradation causes strong time-dependent viscosity changes, the drag-related terms and the resulting flowback-rate window may vary with time. In addition, the model is formulated for a representative planar, propped fracture during early to mid-stage flowback and does not explicitly capture fracture propagation, fracture intersections, complex fracture networks, proppant crushing, or severe fracture-width heterogeneity. These effects should be addressed in future work by coupling the present force-balance model with transient fracture-flow, proppant-transport, or discrete-fracture-network models. Conclusions (1) This study proposes a generalized force-balance method for estimating the allowable flowback-rate window during CBM fracturing. The method incorporates post-break fluid viscosity to represent the rheological state of the fracturing fluid after gel breaking and distinguishes particle mobilization mechanisms before and after fracture closure. (2) A comprehensive force-based model calculates critical flowback velocities. It considers eight major forces: net gravity, overburden pressure, pressure drag, viscous drag, friction, lubrication pressure, particle cohesion, and capillary forces. The model assumes uniform proppant size, no deformation, and closest packing. A novel 3D volumetric integration improves capillary force calculation accuracy. (3) Proppant and fluid properties exert relatively stable effects on flowback. Proppant diameter and density show a positive correlation with the critical velocity, whereas fluid density and viscosity exhibit a negative correlation. Capillary and friction forces change significantly during flowback, especially post-closure. Capillary forces lower the required flowback velocity, whereas friction increases it. (4) Before fracture closure, loosely arranged proppants require lower flowback velocities. After closure, tight packing increases resistance. Larger, heavier particles are harder to mobilize. High fluid viscosity or density enhances drag and promotes proppant return. Friction consistently dominates over capillary forces, indicating that post-closure flowback should begin at a low rate and then gradually increase to maintain efficiency and prevent potential damage. (5) The proposed workflow can be implemented as a transferable field-screening tool for estimating the allowable flowback-rate window using site-specific post-break viscosity, particle-size distribution, and post-closure pressure differential. The two field cases are demonstrative rather than universal, and the workflow should be applied to other CBM wells after updating the relevant reservoir and treatment parameters. Further HP/HT propped-fracture flowback experiments are needed to quantify prediction uncertainty and refine the contact and interfacial parameters. Abbreviations net gravitational force, mN net gravitational force, mN gravity, mN buoyancy force, mN pressure from overburden particles, mN pressure drag force, mN viscous drag force, mN frictional force, mN liquid film force, mN cohesive force, mN equivalent capillary force, mN formation closure pressure, Pa internal fracture pressure, Pa equivalent capillary pressure, Pa particle radius, μm principal radii of curvature of the interface, μm moment arm of the driving force, μm moment arm of the resisting force, μm average fracture width, m particle density, g/cm3 fracturing fluid density, g/cm3 fluid viscosity, mPa·s flowback velocity, m/s drag coefficient Reynolds number constant coefficient constant coefficient friction coefficient (0.75 for quartz) film coefficient (0.213 μm for quartz) cohesion coefficient (2.56 dyn/cm for quartz) contact angle between particles and vertical axis (typically 30°) angle between the solid-liquid interface perimeter on the particle and the y-axis interfacial tension, mN/m contact angle gravitational acceleration, (9.8 m/s² μm for quartz) liquid column height, m number of particle layers Author contributions C.M.M. and F.Y. wrote the main manuscript text. Z.B.H. and Q.A. contributed to the methodology design and data analysis. H.L.C. and C.H. prepared the figures and tables. Z.P.L. supervised the project and provided critical revisions. All authors reviewed and approved the final version of the manuscript. Funding This work was supported by the financial supports from the Comprehensive Scientific Research Project of China National Offshore Oil Corporation (KJZH-2023-2303). Data availability The datasets generated and/or analysed during the current study are not publicly available due to confidentiality agreements but are available from the corresponding author on reasonable request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The datasets generated and/or analysed during the current study are not publicly available due to confidentiality agreements but are available from the corresponding author on reasonable request.