Proportional Flow Control Valve

• PFCV is a valve providing continuous flow modulation by precisely mapping control input to effective opening area, ensuring predictable flow behavior.
• It employs diverse architectures—from microfluidic pilot-operated systems to electrically actuated and motorized designs—to accommodate various industrial applications.
• Key challenges include mitigating thin-gap throttling, managing hysteresis and dry friction, and coupling fluid–structure interactions for stable operation.

A proportional flow control valve (PFCV) is a regulating valve intended to produce a monotonic, predictable mapping between a control input and either the effective opening area $A(x)$ or the flow rate $Q$, with minimal hysteresis and no oscillations. In the cited literature, that function is realized in several distinct embodiments: pilot-operated microfluidic valves actuated by shape memory alloy (SMA) wires, gas-processing valve models with an effective orifice area state, electrically actuated throttle and butterfly valves, pneumatic regulating valves with positioners, cross-port hydraulic leakage valves for actuator position control, and motorized ball-valve electronic regulators for rocket propulsion systems (Berraies et al., 2024, Brüggemann et al., 2022, Daoudi et al., 2024, Chi et al., 2020, Wrat et al., 25 Sep 2025, Lee et al., 2024).

1. Definition and characteristic architectures

The defining feature of a PFCV is continuous modulation of a flow restriction rather than binary opening and closing. One formulation states the intended mapping explicitly as control input $\rightarrow$ pilot displacement $\rightarrow$ pilot pressure drop $\rightarrow$ main membrane lift $x$ $\rightarrow$ seat-gap area $A(x)$ $\rightarrow$ flow rate $Q$, with the desired behavior that $Q$0 increases smoothly with input over the operating pressure range (Berraies et al., 2024). A closely related control-oriented gas-processing model treats the valve as a component whose effective area $Q$1 is driven by a first-order actuator and whose flow is a function of upstream pressure, downstream pressure, and area (Brüggemann et al., 2022). In throttle-valve experiments, the same proportional concept appears as a PWM command $Q$2 driving a butterfly plate angle $Q$3, which changes the downstream flow area and yields direction-dependent behavior because of asymmetric hysteresis and dry friction (Daoudi et al., 2024).

Several physical architectures recur. The pilot-operated microfluidic configuration has two medium-separated stages, both normally closed: a pilot chamber with a small diaphragm and a main chamber with a larger diaphragm. The pilot opening reduces pressure in the pilot chamber, lowers the net seating force on the main membrane, and permits the main valve to lift from a rigid seat (Berraies et al., 2024). A conventional process-regulating configuration uses a pneumatic servo-motor, a positioner, and an electro-pneumatic transducer to convert an electrical command into stem displacement and thus into valve opening area (Chi et al., 2020). In hydraulic actuation, a PFCV can also be inserted directly between the two chambers of a cylinder to create an adjustable artificial leakage path, while the proportional directional control valve is held fully open (Wrat et al., 25 Sep 2025). In pressure-fed rocketry, a motor-driven commercial full-port ball valve serves as an electronic regulator whose effective $Q$4 is controlled by a cascaded pressure and position loop (Lee et al., 2024).

These architectures share the same functional objective but differ sharply in dominant physics. Microfluidic implementations emphasize thin gaps, strong added-mass effects, and membrane compliance (Berraies et al., 2024). Gas-network models emphasize port-based interconnection and conservation of mass (Brüggemann et al., 2022). Mechanical throttle and process valves emphasize hysteresis, stiction, and actuator uncertainty (Daoudi et al., 2024, Chi et al., 2020). Hydraulic and propulsion applications emphasize energy dissipation, relief-valve avoidance, high pressure capability, and coupled system dynamics (Wrat et al., 25 Sep 2025, Lee et al., 2024).

2. Governing relations and proportionality

The basic flow law appearing across the literature is the orifice relation

$Q$5

with discharge coefficient $Q$6, effective flow area $Q$7, pressure drop $Q$8, and fluid density $Q$9 (Berraies et al., 2024, Wrat et al., 25 Sep 2025). For a circular seat with an annular gap, small lifts satisfy

$\rightarrow$0

so proportionality at small opening is immediately tied to the lift-area mapping (Berraies et al., 2024). In liquid systems, the regulating-valve literature also expresses turbulent non-choked flow as $\rightarrow$1, and laminar low-Reynolds operation as $\rightarrow$2 (Chi et al., 2020). In gas systems, the control-oriented model uses an isentropic-orifice-flow expression in which the mass flow is linear in the effective area $\rightarrow$3 and nonlinear in the pressure ratio $\rightarrow$4, under ideal-gas, isothermal-network assumptions with constant $\rightarrow$5 and compressibility factor $\rightarrow$6 (Brüggemann et al., 2022). For rocket pressurization, the same distinction appears as incompressible liquid flow for propellants and choked or subcritical compressible flow for gases (Lee et al., 2024).

Actuator dynamics enter directly into proportional behavior. In the gas-processing model, the effective area obeys the first-order law

$\rightarrow$7

with $\rightarrow$8, $\rightarrow$9, and $\rightarrow$0 the actuator time constant (Brüggemann et al., 2022). Linearization about an operating point produces

$\rightarrow$1

which gives a strictly proper path from command to flow and direct feedthrough from pressure perturbations to flow (Brüggemann et al., 2022). In process-regulating valves, a standard dynamic idealization is

$\rightarrow$2

with hydraulic, frictional, and actuator nonlinearities explicitly identified as sources of deadband, hysteresis, stiction, and saturation (Chi et al., 2020).

A central difficulty is that proportionality can break down when $\rightarrow$3, $\rightarrow$4, or hydrodynamic forces become strong functions of lift. The microfluidic pilot-operated study makes this explicit: in the thin-gap throttling regime, both $\rightarrow$5 and $\rightarrow$6 become strong functions of $\rightarrow$7 and local Reynolds number, undermining proportionality and potentially causing negative damping and self-excited oscillations (Berraies et al., 2024). A closely related implication appears in the throttle-valve benchmark, where the input-output map is degraded by asymmetric hysteresis and stochastic dry-friction effects, so that a nominally proportional command path is direction-dependent (Daoudi et al., 2024).

3. Internal flow physics, fluid–structure interaction, and instability

The most detailed internal-flow treatment among the cited works concerns a pilot-operated microfluidic valve analyzed with a strongly coupled partitioned fluid–structure interaction framework. The fluid solver is ANSYS Fluent in URANS/ALE form with approximately $\rightarrow$8 finite-volume cells and an SST $\rightarrow$9–$\rightarrow$0 model when local conditions become turbulent; the structural solver is ANSYS Transient Mechanical with Taylor–Hood tetrahedral elements, neo-Hookean hyperelasticity, Rayleigh damping $\rightarrow$1, $\rightarrow$2, and a distributed spring load (Berraies et al., 2024). Interface conditions enforce velocity continuity and traction equilibrium, the pilot membrane is treated with one-way FSI, and the main membrane with full two-way FSI (Berraies et al., 2024). Contact is frictionless normal contact with penalty plus Lagrange multiplier, while a finite separation $\rightarrow$3 is enforced numerically to avoid mesh collapse; when contact occurs, the fluid in the gap is replaced by a Darcy-type resistance (Berraies et al., 2024).

That model distinguishes clearly between satisfactory ON/OFF operation and unstable proportional operation. With $\rightarrow$4 bar, rapid pilot venting reduces pressure on the underside of the main membrane from $\rightarrow$5 bar to approximately $\rightarrow$6 bar in about $\rightarrow$7 ms, the main valve fully opens, streamlines settle quickly, and most flow passes through the main seat gap rather than the pilot channel (Berraies et al., 2024). Validation against experiment gave equilibrium flow rates of approximately $\rightarrow$8 l/h at $\rightarrow$9 bar and $x$0 l/h at $x$1 bar, with computed equilibrium flow differing by about $x$2 from experiment, while pressure trajectories near the gap matched experimental trends qualitatively (Berraies et al., 2024).

Proportional mode exposes a different regime. At $x$3 bar, pilot opening increases bypass flow through the pilot channel to about $x$4 l/h, yet the main valve remains sealed and the device remains effectively closed at the main seat (Berraies et al., 2024). At $x$5 bar, a pilot opening of $x$6 still yields only about $x$7 l/h through the pilot conduit with the main valve closed, but at approximately $x$8 opening the total flow jumps from about $x$9 l/h to about $\rightarrow$0 l/h while the pressure drop across the main stage falls from about $\rightarrow$1 kPa to about $\rightarrow$2 kPa, and self-excited oscillations begin (Berraies et al., 2024). The mechanism is attributed to the sharp pressure drop in the thin seat-gap, the ensuing Venturi effect, and an effective negative damping that counteracts opening. Troughs in differential pressure coincide with peaks in flow rate and reclosure events; the oscillation is not purely harmonic because membrane motion stores and releases volume; and the solver eventually fails to converge near $\rightarrow$3 s because of the severity of the instability (Berraies et al., 2024).

This distinction is conceptually important. A PFCV does not become proportional merely because its actuator can be set to intermediate positions. The microfluidic results show that slow proportional ramps can dwell in a thin-gap throttling state that is bypassed during ON/OFF actuation, so the same valve may be adequate for binary operation yet unstable for continuous modulation (Berraies et al., 2024).

4. Control-oriented models and network interconnection

In gas-processing systems, the PFCV can be represented as either a static or dynamic component in a port-based state-space framework. The static model uses upstream pressure $\rightarrow$4 and downstream mass-flow input $\rightarrow$5 to produce downstream pressure $\rightarrow$6 and upstream mass-flow output $\rightarrow$7, with $\rightarrow$8, directly encoding steady-state mass conservation through the pass-through of flow (Brüggemann et al., 2022). The dynamic model augments that relation with the first-order actuator and the linearized orifice-flow equation, yielding a compact realization

$\rightarrow$9

suitable for model-based MIMO control design (Brüggemann et al., 2022).

The interconnection formalism is based on paired pressure and mass-flow variables at p-ports and q-ports. Internal series connections require consistent pairing of pressure inputs with pressure outputs and mass-flow inputs with mass-flow outputs, and the aggregate realization is obtained from the stacked component models and the wiring matrices $A(x)$0 and $A(x)$1 through

$A(x)$2

$A(x)$3

An equivalent assembly can be performed with Matlab’s connect function (Brüggemann et al., 2022). The same work emphasizes a zero-frequency property associated with conservation of mass: for a single pipe section, $A(x)$4 and $A(x)$5, and the static valve satisfies the same relation in the steady-state sense because $A(x)$6 (Brüggemann et al., 2022). For the dynamic valve, the paper draws a distinction: the standalone component does not itself encode a unity $A(x)$7 because it has no explicit mass-flow input in its linearized input vector; that property is recovered at the correctly interconnected network level (Brüggemann et al., 2022).

This control-oriented perspective differs from high-fidelity FSI, but the two are complementary rather than contradictory. The state-space representation is intended for network synthesis, controllability analysis, and regulator design (Brüggemann et al., 2022), whereas the FSI model resolves localized pressure losses, contact, and membrane motion that can destroy proportional operation altogether (Berraies et al., 2024). A plausible implication is that multiscale PFCV analysis often requires both abstractions: a reduced interconnection model for control design and a localized CFD–FSI model for trim geometry and stability assessment.

5. Control strategies, identification, and adaptation

Several control strategies are represented in the cited studies. In the throttle-valve benchmark, a baseline discrete-time PI controller is written in incremental form as

$A(x)$8

equivalently $A(x)$9 for slowly varying references, with plant identification from the ARX model

$\rightarrow$0

Sampling time is $\rightarrow$1 ms, PRBS length is $\rightarrow$2 samples centered at $\rightarrow$3 duty, and the tuned gains differ across the three valves, for example $\rightarrow$4, $\rightarrow$5, $\rightarrow$6, $\rightarrow$7, $\rightarrow$8 for Valve 1 (Daoudi et al., 2024). The command is constrained to $\rightarrow$9 for Valves 1 and 2 and to $Q$0 for Valve 3 (Daoudi et al., 2024). The baseline PI tracks setpoints but exhibits direction-dependent bias, overshoot, and oscillations, especially for Valve 1 and at low angles (Daoudi et al., 2024).

The same benchmark then augments PI with Reinforcement Learning with Guides. The combined policy is

$Q$1

where the learned perturbation is constrained to a reduced action subspace $Q$2 scaled by $Q$3; the state is $Q$4; the cost is $Q$5; and the optimization uses TD3 with deterministic policies and $Q$6 ReLU hidden layers (Daoudi et al., 2024). Episodes last $Q$7 steps, both pure RL and PI-RL agents are trained for $Q$8 episodes, and PI-RL displays superior sample efficiency relative to pure TD3 while often achieving lower MSE than the baseline PI under nominal and low-to-moderate noise conditions (Daoudi et al., 2024). The paper attributes the improvement to guided exploration and to learned compensation for asymmetric hysteresis and dry friction (Daoudi et al., 2024).

In rocket propulsion, the control problem is formulated differently but remains recognizably PFCV control. The motorized ball valve is regulated by a cascaded architecture: an outer pressure PID sets a target valve angle, and an inner position PID drives the motor to that angle using encoder feedback (Lee et al., 2024). Feedforward terms are built from a simple valve model $Q$9, gas-side choked-flow assumptions, liquid-side incompressible-flow assumptions, and dynamic gain scheduling $Q$00 with $Q$01, $Q$02, $Q$03 (Lee et al., 2024). The motivation is explicit: standard PID produced pressure oscillations greater than $Q$04 bar at flow start and poor tracking late in flow, whereas feedforward plus dynamic gains improved accuracy under changing ullage and upstream-pressure conditions (Lee et al., 2024).

In hydraulic actuator control, the cited work uses a PID controller tuned by a fuzzy controller. The valve command is

$Q$05

with fuzzy adaptation based on the error $Q$06 and error rate $Q$07 (Wrat et al., 25 Sep 2025). Quantitative tracking metrics are not reported, but the controller is used consistently in both the conventional proportional directional-control configuration and the cross-port PFCV leakage-compensation configuration to enable an energy comparison (Wrat et al., 25 Sep 2025).

6. Design trade-offs, force compensation, and diagnostics

One major design theme is the need to avoid disproportionate hydrodynamic penalties at partial opening. For the pilot-operated microfluidic valve, the recommended mitigations are geometrical and dynamical: chamfer or round the downstream edge of the valve seat, introduce a short diffuser immediately downstream of the seat gap, use a converging–diverging seat profile, shape the main membrane lip for smoother streamline curvature, enforce a minimum mechanical lift $Q$08, add restrictors or damping cavities, regulate the inlet pressure, increase effective structural damping, tune the spring stiffness, use slower SMA ramps with “step-skip” logic, and apply closed-loop feedback that detects pressure ripple and temporarily changes the opening trajectory (Berraies et al., 2024). The same work states the key stability criterion succinctly: avoid operating in ultra-thin-gap throttling where $Q$09 and $Q$10 are large and negative (Berraies et al., 2024).

A second design theme concerns steady-state flow-force compensation in spool-type hydraulic valves. The corrected analysis of flow force rejects the classical assumption that compensation arises downstream from a turbine-bucket-like exit profile. Instead, compensation is attributed to an upstream static-pressure imbalance acting on an inclined chamfer on the high-pressure land (Lugowski, 2013). For small openings, the annular metering area is approximated by $Q$11, the uncompensated closing force is written as

$Q$12

and the compensating opening force as

$Q$13

so that the net steady-state flow force is $Q$14 (Lugowski, 2013). The cited design guidance is that an upstream chamfer angle $Q$15 is essential, square land $Q$16 produces no compensation, and $Q$17 maximizes $Q$18 for fixed $Q$19 and $Q$20 (Lugowski, 2013). The paper’s interpretation is also a correction of a long-standing misconception: the exit profile angle $Q$21 does not materially contribute to steady-state compensation in the partially open case (Lugowski, 2013).

Hybrid modeling provides a third, data-driven design and diagnostic route. The regulating-valve study combines mechanistic equations with an LS-SVM block that identifies unknown nonlinear parameters or residual corrections, using the regression function

$Q$22

and the unbiased solution

$Q$23

with $Q$24 (Chi et al., 2020). On the DAMADICS benchmark, using features $Q$25 gave RMSE $Q$26 m$Q$27/h, MAPE $Q$28, Err$Q$29, while adding temperature $Q$30 improved accuracy to RMSE $Q$31 m$Q$32/h, MAPE $Q$33, Err$Q$34 (Chi et al., 2020). The same framework supports residual-based fault diagnosis for stiction, leakage, cavitation or critical flow, actuator degradation, and sensor drift (Chi et al., 2020).

7. Applications, reported performance, and limitations

The application range of PFCVs in the cited literature is unusually broad. In microfluidics, the target functions are precise dispensing, mixing, or dosing under relatively high pressure differences in miniaturized hydraulic systems (Berraies et al., 2024). In gas processing, the valve is one component in a larger network of pipes, compressors, branches, joints, and tanks, intended for aggregate MIMO control design in state-space form (Brüggemann et al., 2022). In throttle and butterfly valves, the emphasis is nonlinearity, asymmetric hysteresis, and stochastic dry friction in electromechanical actuation (Daoudi et al., 2024). In process-control regulating valves, the emphasis is flow and pressure prediction under strong nonlinearity, time variation, and parameter uncertainty (Chi et al., 2020). In mobile hydraulics, the PFCV appears as an artificial leakage path that reduces relief-valve losses (Wrat et al., 25 Sep 2025). In rocketry, it appears as an electronic regulator for both gaseous pressurant and cryogenic liquid propellant (Lee et al., 2024).

Reported performance varies with domain and metric. The hydraulic leakage-compensation study reports energy per cycle of $Q$35 kJ for the conventional proportional directional-control circuit and $Q$36 kJ for the PFCV circuit, corresponding to an energy saving of $Q$37 (Wrat et al., 25 Sep 2025). The rocket regulator study reports regulation of pressures within $Q$38 bar in the abstract, and in the detailed results approximately within $Q$39 bar for tank pressures and within $Q$40 bar for injector pressures during most of a static fire while simultaneously throttling; the reported system handles $Q$41 kg/s of liquid, $Q$42 kg/s of gas, upstream pressures up to $Q$43 bar, and throttles thrust from approximately $Q$44 kN down to approximately $Q$45 kN while maintaining $Q$46 (Lee et al., 2024). The throttle-valve learning study reports that PI-RL has better sample efficiency than traditional RL agents and outperforms the PI controller across the tested scenarios, although specific MSE values are not listed in the provided details (Daoudi et al., 2024). The hybrid regulating-valve study reports the DAMADICS accuracy figures given above (Chi et al., 2020). The microfluidic study reports acceptable ON/OFF behavior but instability in proportional mode because of flow-induced vibrations in the thin-gap throttling regime (Berraies et al., 2024).

The limitations are equally domain-specific. The microfluidic FSI model treats the pilot membrane with one-way FSI, prescribes rather than fully models the SMA thermo-electro-mechanical physics, and uses a prescribed $Q$47-gap and Darcy replacement that may miss squeeze-film effects; the paper suggests quasi-Newton IQN-ILS, Robin–Neumann coupling, or monolithic solvers as possible improvements (Berraies et al., 2024). The gas-processing model is linearized around an operating point, assumes ideal gas with constant $Q$48, constant temperature, and a first-order actuator, and does not include piecewise choked-flow logic within the model itself (Brüggemann et al., 2022). The throttle-valve control study focuses on angle regulation rather than direct flow measurement and notes that PI-RL can underperform pure TD3 in some high-noise cases when the PI guide is systematically biased (Daoudi et al., 2024). The hydraulic energy study neglects fluid inertia and ignores temperature and pressure dependence of fluid properties in simulation, while detailed stability analysis and quantitative tracking metrics are not reported (Wrat et al., 25 Sep 2025). The rocket regulator study does not report numerical stability margins, bandwidth, or explicit anti-windup design, though it recommends such measures for implementation (Lee et al., 2024).

Taken together, these results show that “proportional” in PFCV design is not merely a matter of actuator command continuity. It depends on the coupled properties of geometry, hydrodynamic losses, actuator dynamics, hysteresis, pressure-network interactions, and feedback design. The cited studies consistently indicate that stable proportional behavior requires a predictable $Q$49-to-input mapping, while deviations from that objective arise from thin-gap Venturi suction, flow-force imbalance, asymmetric hysteresis, relief-valve energy loss, or network-level coupling (Berraies et al., 2024, Lugowski, 2013, Daoudi et al., 2024, Wrat et al., 25 Sep 2025).