Turbulence is prevalent in nature and industry, from large-scale wave dynamics to small-scale combustion nozzle sprays. In addition to the multiscale nonlinear complexity and both randomness and coherent structures in its dynamics, practical turbulence is often nonhomogeneous and anisotropic, leading to great modeling challenges. In this paper, an efficient model is proposed to predict turbulent jet statistics with high accuracy. The model leverages detailed knowledge of readily available velocity signals from idealized homogeneous turbulence and transforms them into Lagrangian trajectories of a turbulent jet. The resulting spatiotemporal statistics are compared against experimental jet data showing remarkable agreement at all scales. In particular, the intermittency phenomenon is accurately mapped by the model to this inhomogeneous situation, as observed by higher-order moments and velocity increment probability density functions. Crucial to the advancement of turbulence modeling, the transformation is simple to implement for the jet configuration, with possible extensions to other inhomogeneous flows such as wind turbine wakes and canopy flows, to name a few.

The ingestion of turbulence can cause additional noise sources of rotor blades, which should be considered for multirotor powered urban air mobility vehicles encountering atmospheric turbulence. In this work, a turbulence grid was installed in the exit of an open-jet anechoic wind tunnel to generate turbulent flows. The grid turbulence was characterized using hot-wire anemometry, showing that turbulence intensity decays with the streamwise locations downstream of the grid, following a power law of $-5/7$. The power spectral properties of the grid turbulence were also assessed, and it agrees well with the von Kármán turbulence spectrum in the inertial subrange. Then, the aerodynamic force and noise of a rotor with a diameter of 217.2 mm were measured under both clean and turbulent flows. Force measurements show that the thrust and torque coefficients decrease with the advance ratio $J$. Noise measurements show that the tonal noise at the blade pass frequency (BPF) is more significant at the upstream locations under high advance ratios, and high-order BPF harmonics can also be amplified. Moreover, the turbulence ingestion noise mainly dominates the broadband contents from 10 to 50 BPF harmonics. The broadband noise can be scaled by the Mach number scaling of $M_{\infty }^2{M}_c^4$, where $M_{\infty }$ is the freestream Mach number and $M_c$ is the corresponding Mach number of the rotating speed at the blade tip.

This research article delves into the natural convection of liquid metal in a three-dimensional cavity with varying magnetic fields and wall conductivities using direct numerical simulation. Two opposite sidewalls are heated/cooled, while the magnetic field is perpendicular to the main circulation. Our primary focus is on examining flows within the Grashof number, $\text{Gr}⩽{10}^8$, the Hartmann number, $\text{Ha}⩽400$, and the wall conductance ratio, $C_w=0.01$–1. It is found that weakly conducting walls experience a significant enhancement in convection within a specific range of magnetic field strength, whereas highly conducting walls exhibit pronounced flow attenuation. The applied horizontal magnetic field alters the plume's topology and dynamics, generating a more coherent and energetic large-scale flow structure, while it weakens convection by consuming buoyant potential energy through Joule dissipation. This results in a competition between the rectifying effect and the damping effect to determine whether the magnetic field has a positive or negative feedback on heat transfer, with the quasi-two-dimensional state serving as a critical point. Additionally, varying wall conductivities transform the current distribution within the parallel layer, influencing the flow's response to changes in field strength. The formation of corner vortices can be considered by the curvature of the boundary layer that undergoes turning at corners. Furthermore, the effects of wall shear and plume transport on heat transfer are systematically investigated. The analysis reveals that while the plume area remains almost constant, the condensation of coherent structures facilitates greater horizontal heat transport per unit area of the plume, contributing significantly to the overall heat transfer enhancement. Finally, the computed Nusselt number Nu and Reynolds number Re can be correlated as functions of ${\text{Ha/Gr}}^{1/3}$. The critical conductance ratio referring to the complete suppression of convection conforms to the scaling of $2.31{\left({\text{Ha/Gr}}^{1/3}\right)}^{-5}$, where heat transfer occurs solely by thermal conduction once it exceeds this value.

Fiber suspensions flowing in structured media are encountered in many biological and industrial systems. Interactions between fibers and the transporting flow as well as fiber contact with obstacles can lead to complex dynamics. In this work, we combine microfluidic experiments and numerical simulations to study the interactions of a rigid fiber with an individual equilateral triangular pillar in a microfluidic channel. Four dominant fiber dynamics are identified: transport above or below the obstacle, pole vaulting, and trapping, in excellent agreement between experiments and modeling. The dynamics are classified as a function of the length, angle, and lateral position of the fibers at the channel entry. We show that the orientation and lateral position close to the obstacle are responsible for the fiber dynamics and we link those to the initial conditions of the fibers at the channel entrance. Direct contact between the fibers and the pillar is required to obtain strong modification of the fiber trajectories, which is associated to irreversible dynamics. Longer fibers are found to be more laterally shifted by the pillar than shorter fibers that rather tend to remain on their initial streamline. Our findings could in the future be used to design and optimize microfluidic sorting devices to sort rigid fibers by length.

Microchannels have become prevalent as an integrated part of microfluidic devices in biochemistry and electronics applications. In such devices, the small scale results in a characteristically low Reynolds number laminar flow. The small scale also results in an associated high flow resistance. A design concept has been developed that reduced the flow resistance by featuring geometrically modified microchannels with cavities. Compared to an unmodified microchannel, the modification reduces flow resistance at low Reynolds numbers but conversely leads to higher flow resistance at high Reynolds numbers: i.e., a reversal of flow resistance occurred. Thus far, plausible fluidic mechanisms underlying such reversal have remained largely unstipulated. Based upon detailed pressure and flow field measurements, we stipulate that flow progression from laminar flow slippage to rotational vortices in cavitied microchannels is the main mechanism causing the reversal. We further clarify that the earlier transition of initial laminar flow to turbulent flow is triggered by instabilities generated along shear layers, formed between the mainstream flow and rotational vortices in each cavity.

To tackle the challenging near-wall turbulence modeling while preserving low computational cost, the near-wall nonoverlapping domain decomposition (NDD) method is proposed, incorporating the machine-learning technique. Using recently proposed implicit NDD (INDD), the solution can be calculated with a Robin-type (slip) wall boundary condition on a relatively coarse mesh and then corrected in the near-wall region at every iteration through an estimated turbulent viscosity profile obtained from a neural network. To maintain a reasonable complexity with acceptable accuracy, only the near-wall field properties, i.e., wall-normal distance, near-wall velocities, streamwise pressure gradients, and one near-wall scale parameter, are employed as the input features for the neural network. The benefit of incorporating machine learning is twofold. First, the near-wall turbulent viscosity is predicted more accurately than by the traditional algebraic functions used in the approximate approach. Second, similarly to the conventional NDD, the present simulations save one order of computational cost over the fully resolved one-block simulations. The accuracy and efficiency of the method are demonstrated on test cases with the $k\text{−}\varepsilon$ model, including turbulent flows in a channel and an asymmetric diffuser at different Reynolds numbers. Results compare favorably with the one-block benchmark solutions and show a better agreement when compared with approximate NDD predictions. In the latter case, a variation of diffuser geometry is also considered to test the model performance on engineering design tasks with another turbulent model (i.e., Spalart-Allmaras), showing good generalization capability on different turbulence closures.

Image-based modeling of mineral dissolution poses challenges due to its multiscale nature, requiring the consideration of multiphase reactive flow and transport at both the resolved pore scale (macropores/fractures) and the unresolved Darcy scale (micropores). The existing hybrid-scale simulation methods pose difficulties in handling the multiscale fluid-rock interactions and temporal structural evolution. In this study, we propose a multiscale compressive continuum species transfer (MC-CST) scheme to address the limitations of the standard CST scheme, which exhibits numerical diffusion issues at the gas-liquid interface and thereby suffers from inaccuracies in reactive transport simulations. The proposed scheme incorporates an additional compressive term derived from volume-averaging principles for the advection and diffusion fluxes in a single-field framework. To ensure the impermeable species transport condition at the solid boundary, a concentration extrapolation algorithm is developed. Four validation cases are conducted to demonstrate the model's capability in accurately simulating multiphase reactive flow and transport at various scales, including pore scale, continuum scale, and hybrid scales. Special attention is given to accurately modeling the thermodynamic conditions at the gas-liquid interface, particularly with respect to the concentration jump under conditions of large local Péclet numbers. Furthermore, we present a case study simulating calcite dissolution in a porous medium to underscore the importance of multiscale fluid-rock interactions for an in-depth comprehension of the dissolution regime.

We investigate drag reduction mechanisms in flows past two- and three-dimensional cylinders controlled by surface actuators using deep reinforcement learning. We investigate 2D and 3D flows at Reynolds numbers up to 8000 and 4000, respectively. The learning agents are trained in planar flows at various Reynolds numbers, with constraints on the available actuation energy. The discovered actuation policies exhibit intriguing generalization capabilities, enabling open-loop control even for Reynolds numbers beyond their training range. Remarkably, the discovered two-dimensional controls, inducing delayed separation, are transferable to three-dimensional cylinder flows. We examine the trade-offs between drag reduction and energy input while discussing the associated mechanisms. The present paper demonstrates discovery of transferable and interpretable control strategies for bluff body flows through deep reinforcement learning with limited computational cost.

A novel mechanism is identified, through which a spanwise-invariant surface feature (a two-dimensional forward-facing step) significantly stabilizes the stationary crossflow instability of a three-dimensional boundary layer. The mechanism is termed here as reverse lift-up effect, inasmuch as it acts reversely to the classic lift-up effect; that is, kinetic energy of an already-existing shear-flow instability is transferred to the underlying laminar flow through the action of cross-stream perturbations. To characterize corresponding energy-transfer mechanisms, a theoretical framework is presented, which is applicable to generic three-dimensional flows and surface features of arbitrary shape with one invariant spatial direction. The identification of a passive geometry-induced effect responsible for dampening stationary crossflow vortices is a promising finding for laminar flow control applications.

In extreme environments, the Rayleigh-Taylor instability (RTI) may occur under large variations in density and temperature and with fluid transport properties strongly dependent on temperature. Direct numerical simulations of the 3D fully compressible RTI are conducted, examining the idealized configuration of a hotter, less dense fluid pushing against a colder, denser fluid. Various temperature ratios and transport property configurations are explored to examine how heat conduction, large variations in transport properties, and sudden changes in transport properties can affect the evolution of the mixing layer. Nonuniform fluid expansion and contraction induced by heat transfer can significantly affect local density differences and overall instability growth, causing profile asymmetries about the initial interface for flow and mixing statistics. The departures from classical self-similar development of the instability along with misalignment between regions of mixing and regions of most intense turbulent activity caused by both heat transfer and transport property contrasts are examined. After sudden changes in fluid transport properties, which may occur as a result of rapid heating (e.g., in inertial confinement fusion), the flow quickly responds and begins to relax towards quasi-self-similar late-time evolution. For many dynamical quantities such as vorticity and dissipation, this late-time evolution resembles that of the configuration that already started with the final transport property magnitudes, suggesting that these quantities depend only on the transport properties and not on past flow history, provided that the density field distributions for the flows remain similar. On the other hand, the mixing evolution after the transport property change is unique, implying that both property magnitudes and previous history are impactful on the mixing. These simulations demonstrate how various temperature-related effects are extremely important to consider in compressible RTI flows with large temperature variations.

Taylor's hypothesis of frozen flow is revisited in homogeneous turbulent shear flow by examining the cancellation properties of Eulerian and convective accelerations at different flow scales. Using results of direct numerical simulations, vector-valued flow quantities, including the Lagrangian, Eulerian, and convective accelerations, are decomposed into an orthogonal wavelet series and their alignment properties are quantified through the introduction of scale-dependent geometrical statistics. Joint-probability density functions of the Eulerian and convective accelerations show antialignment at small scales of the turbulent motion, but this observation does not hold at large scales. Similarly, the angles of the scale-wise contributions of the Eulerian and convective accelerations were found to prefer an antiparallel orientation at small scales. Such antialignment, however, is not observed at the largest scales of the turbulent motion. The results suggest that Taylor's hypothesis holds at small scales of homogeneous turbulent shear flow, but not for large-scale motion. The Corrsin scale is proposed as a measure for the applicability of Taylor's hypothesis in such flows.

Three-dimensional direct numerical simulations are used to revisit the experiments on multibubble cavitation performed by Bremond et al., Phys. Fluids 18, 121505 (2006); Phys. Rev. Lett. 96, 224501 (2006). In particular, we aim at understanding the asymmetry observed therein during the expansion and collapse of bubble clusters subjected to a pressure pulse. Our numerical simulations suggest that the asymmetry is due to the force applied by the imposed pressure pulse and it is a consequence of the finite effective speed of sound in the liquid. By comparing our numerical results to the experiments, we found that the effective speed of sound under the experimental conditions was smaller than that of degassed water due to microbubbles in the system which resulted from prior cavitation experiments in the same setup. The estimated values of the effective speed of sound are consistent with those derived from the classical theory of wave propagation in liquids with small amounts of gas. To support this theory, we also present evidence of tiny bubbles remaining in the liquid bulk as a result of the fragmentation of large bubbles during the prior cavitation experiments. Furthermore, we find that this asymmetry also alters the direction of the liquid jet generated during the last stages of bubble collapse.

The interaction between a microscale collapsing bubble and a nearby viscoelastic solid is investigated numerically. The bubble initiates rapid expansion with high initial pressure, and the stand-off parameter is set to be near 1.0. The viscoelastic properties of the solid are described using the Kelvin-Voigt model, with its elastic modulus $E$ and viscosity $\eta$, varying several orders of magnitude ($E=20–2000$ kPa and $\eta =2–1000 \mathrm{Pa}\mathrm{s}$). The influence of solid viscoelasticity on bubble behavior and solid deformation are analyzed throughout the entire life cycle of the bubble, from its initial expansion to the moment of liquid jet impingement. The Deborah number, which quantifies the relative timescales of solid deformation and bubble expansion, is employed to characterize the bubble–solid interaction. As the Deborah number increases from the order of ${10}^{-2}$ to ${10}^2$, the dynamics of the bubble converge toward those observed in the cases of a rigid solid; while the solid deformation displacement reduces to a value close to zero, the maximum expansion radius of the bubble increases by up to 8%, and the speed of the liquid jet decreases by approximately 80%. The temporal distribution of energy components within the fluid domain reveals that the maximum expansion radius of the bubble and the kinetic energy of the liquid jet are inversely related to the energy transferred from the fluid to the solid. The imbalance in pressure surrounding the contracting bubble and the narrow region of the liquid jet are responsible for the enhancement of the liquid jet speed for small solid viscosity. Upon the impingement of the liquid jet onto the solid surface, the width and depth of the crater formed by the jet become greater for large jet speed and small solid viscosity.

In this work we apply a formulation for capturing detuned secondary instabilities. This formulation, based on two-dimensional stability analysis coupled with a Bloch wave formalism originally described by Schmidt et al. [Phys. Rev. Fluids 2, 113902 (2017)], allows us to consider high-dimensional systems resulting from several repetitions of a spatially periodic unit by solving an eigenproblem of much smaller size. Secondary instabilities coupling multiple periodic units thus can be retrieved. The method is applied on the secondary stability of a swept boundary-layer flow subject to stationary cross-flow vortices. Two distinct detuned secondary instabilities are retrieved. The first one, obtained for a detuning factor $\epsilon =0.35$ and reaching a maximum growth rate for streamwise wave number ${\alpha }_v=0.087$, was already found in the work of Fischer and Dallmann [Phys. Fluids A: Fluid Dyn. 3, 2378 (1991)]. The second instability is obtained for streamwise independent modes and small detuning factor $\epsilon =0.08$. The corresponding mode presents large-wavelength oscillations, arising from a characteristic beating phenomenon. The physical origin of these two secondary instabilities has been investigated by varying the amplitude of the primary disturbance: for the latter instability, reminiscent of a type III mode, the unstable branch continuously deforms as the amplitude is increased, whereas a change of topology of the spectrum is observed for the ${\alpha }_v=0$ mode.

The purpose of the present study is to extend the simple concept of apparent coefficient of restitution, widely approached in the literature for the case of a single contact point between a sphere and a wall, to the case of bouncing whose complexity is increased due to the shape of the contacting object. For this purpose, experiments are carried out with a finite-length cylinder, freely falling in a liquid at rest. A rigid tail is attached to the cylinder, allowing one to maintain a vertical trajectory and to keep the axis of the cylinder parallel to the bottom wall down to a small gap between them. Yet, more complex 3D motions of the cylinder with respect to the wall occur during bouncing, including multiple-contact-point bouncing between the cylinder and the bottom as well as cavitation. When the Stokes number St (the ratio between characteristic inertial forces experienced by the particle compared to viscous forces in the fluid) is increased, the experimental results suggest that the ratio of the apparent coefficient of restitution to the solid one increases from 0 (at low St) to 1 (at large St) with a critical Stokes number ${\mathrm{St}}_c$ below which no bouncing is observed, as usually obtained in the literature for a single-contact-point bouncing sphere. In a conceptual approach to understanding the observed experimental features using a cutoff length scale prior to contact and a contact timescale, we investigated numerical modeling of an idealized situation, a 2D infinite cylinder falling parallel to the wall. To this end, we carried out numerical 2D simulations where the fluid equations of motion were coupled to the particle equation of motion through an immersed boundary method. The particle equation of motion was coupled to an elastic force to model bouncing. This numerical model requires parametrization of the cutoff length (interpreted as a roughness) and the contact time (associated with the contact elasticity) used here to capture the experimental observations. The simulations confirmed that (1) the departure of the coefficient of restitution from 0 is strictly dependent on the apparent roughness and (2) the coefficient of restitution depends on the contact time. Finally, in an effort to rationalize the experiments and the simulations for such a conceptual approach, we modeled the coefficient of restitution as the product of two contributions to the mechanical loss of energy: the collision-to-terminal velocity ratio ($V_c/V_t$) of the approach phase and the rebound-to-collision velocity ratio ($-V_r/V_c$) of the contact phase. We then interpreted the experimental measurements in light of this model, showing evidence that the assumption of a global relationship between the contact time and an apparent roughness (all being linked to the bouncing complexity including multiple-contact-point and cavitation in experiments) leads to a reasonably good prediction of the coefficient of restitution in the intermediate regime in St. This suggests the relevance of lumping the complex details of physical phenomena involved during contact into a simple concept based on the contact apparent roughness and elasticity.

In this study, asymptotic scaling of near-wall streamwise turbulence intensity $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2$ ($u_{\tau }$ is the friction velocity) is theoretically explored. The three scalings previously proposed are first reviewed with their derivation and physical justification: (1) $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim ln{\text{Re}}_{\tau }$ (${\text{Re}}_{\tau }$ is the friction velocity); (2) $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim 1/U_{\infty }^{+}$ ($U_{\infty }^{+}$ is the inner-scaled freestream velocity in boundary layer); (3) $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim {\text{Re}}_{\tau }^{-1/4}$. A new analysis is subsequently developed based on velocity spectrum, and two possible scenarios are identified based on the asymptotic behavior of the outer-scaling part of the near-wall velocity spectrum. In the former case, the outer-scaling part of the spectrum is assumed to reach a nonzero constant as ${\text{Re}}_{\tau }\to \infty$, and it results in the scaling of $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim ln{\text{Re}}_{\tau }$, both physically and theoretically consistent with the classical attached eddy model. In the latter case, a sufficiently rapid decay of the outer-scaling part of the spectrum with Re is assumed due to the effect of viscosity, such that $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2<\infty$ for all ${\text{Re}}_{\tau }$. The following analysis yields $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim 1/ln{\text{Re}}_{\tau }$, asymptotically consistent with the scaling of $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim 1/U_{\infty }^{+}$. The scalings are further verified with the existing simulation and experimental data and those from a quasilinear approximation [Holford et al. J. Fluid Mech. 980, A12 (2024)], the spectra of which all appear to favor $\overline{u^{\prime }{u}^{\prime }}/u_{\tau }^2\sim 1/ln{\text{Re}}_{\tau }$, although new datasets for ${\text{Re}}_{\tau }\gtrsim O\left({10}^4\right)$ would be necessary to conclude this issue.

Liquid drops placed on substrates may vibrate, slide, take history-dependent shapes, and even detach or break up when subjected to external forces. Although CFD tools can nowadays reproduce these motions, they are affordable for at most a few drops at a time. This work provides a low-dimensional model of such drops, in which the drop shape is approximated by a rectangular cuboid. The model results in three ordinary differential equations per drop. It is sufficiently simple to allow closed-form solutions in a variety of configurations. By systematically comparing the cuboid predictions to experimental, numerical, and theoretical results previously obtained with real drops, we discuss the extent to which the cuboid approximation reproduces the order of magnitude and qualitative dependence on parameters of a series of phenomena. The latter include natural drop vibrations, the detachment of pendant drops, multiple drop shapes allowed by contact angle hysteresis, the nonlinear retraction of drops after spreading, the sliding or climbing of drops on inclined substrates, and drop-induced damping of substrate vibrations. The cuboid model therefore provides a low-cost representation of drop motion on a substrate in response to arbitrarily complex external forcing.

Townsend's attached eddy hypothesis (AEH), one of the most elegant models in incompressible wall turbulence, has been recently applied to compressible wall turbulence to explain the numerical observations and predict the scaling laws. Before a more profound extension can be established, a comprehensive investigation of the features of wall-attached eddies for streamwise velocity and temperature fields in compressible wall-bounded turbulence is required. In this work the AEH and the inner-outer interaction model [Marusic et al., Science 329, 193 (2010)] are combined to isolate the signature of attached eddies at a targeted length scale and then assess their inclination angles statistically based on the direct numerical simulation database. The inclination angle obtained in the streamwise velocity fluctuating fields, which approaches ${45}^{\circ }$ as the Reynolds number increases, shows a minor Mach-number influence within the Mach-number range included in this work. As for those in temperature fluctuations, a high statistical similarity can be seen to streamwise velocity fluctuations. This slight Mach-number effect indicates that a uniform model can be potentially developed for compressible wall-bounded turbulence with different Mach numbers.

The onset of droplet motion along a horizontal anodized aluminum surface arising from aerodynamic loading imposed by an accelerating wall-bounded shear flow was investigated for water droplet volumes ranging from 75 to $120\mu \mathrm{L}$. Two accelerating shear flows were considered, a flat plate boundary layer and an impinging jet with orientation angles spanning ${30}^{\circ }$ to ${90}^{\circ }$. The flows were linearly accelerated at three different rates, 1.2, 2.2, and $4.4\mathrm{m}/{\mathrm{s}}^2$, to a maximum flow speed of $20\mathrm{m}/\mathrm{s}$. Droplets in the flat plate boundary layer were observed to have a constant depinning threshold Weber number of ${\mathrm{We}}_{h,\mathrm{crit}}=7.5\pm 0.5$. By contrast, droplets in impinging jets exhibited lower thresholds in the range of $2⪅{\mathrm{We}}_{h,\mathrm{crit}}⪅4$. Droplet volume and flow acceleration had marginal influence on depinning threshold for the considered parameter range. Rigorous dimensional analysis, supplemented by statistical examination of the present data and results from previous studies, is employed to identify dominant dimensional groups governing ${\mathrm{We}}_{h,\mathrm{crit}}$. A new dimensionless group that encapsulates interrelated parameters connected to droplet shape and substrate wettability is proposed, producing good collapse of available data and allowing for a simplified empirical estimation of critical depinning velocity.

Bioclogging, the clogging of pores with living particles, is a complex process that involves various coupled mechanisms such as hydrodynamics and particle properties. The lack of sensitive methods to simultaneously measure the hydraulic resistance of a clog and the position of particles at high enough resolutions limits our understanding of this tight fluid-structure interplay. In this article, we explored bioclogging at the microscale level, where flow rates are typically extremely low ($<100\mathrm{nl}{\min }^{-1}$). We developed a highly sensitive method to precisely measure small flow rates ($<6.7\%$ error) with a low response time ($<0.2\mathrm{s}$) while retaining the capability to image the structure of a fluorescently labeled yeast clog at high resolution. This method employed a microfluidic device with two identical channels: a first one for a yeast suspension and a second one for a colored culture medium. These channels merged into a single wide outlet channel, where the interface of the colored medium and the medium coming from the yeast clog could be monitored. As a yeast clog formed in the first channel, the displacement of the interface between the two media was imaged and compared to a pre-calibrated image database, quantifying the flow through the clog. We carried out two kinds of experiments to explore the fluid-clog interaction: filtration under constant operating pressure and filtration under oscillating operating pressure (backflush cycles). We found that in both scenarios, the resistance increased with clog length. On the one hand, experiments at constant imposed pressure showed that the clog's permeability decreased with increased operating pressure, with no detectable changes in cell density as assessed through fluorescence imaging. On the other hand, backflush cycles resulted in an approximately four times higher permeability, associated with a significant nonmonotonic decrease in cell density with the operating pressure. Leveraging on our unique dual measurement of fluid flow and clog microstructure, our study offered enhanced insights into the intricate relationships between clog microstructure, permeability, and construction history. A better understanding of the fluid-structure interplay allowed us to develop a novel physical modeling of the flow in a soft and confined porous medium that challenged the empirical power-law description used in bioclogging theory by accurately replicating the measured permeability-pressure variations.