This paper presents an experimental study of the flow of an elasto-viscoplastic fluid (Carbopol) in a rectangular duct with an abrupt contraction of an aspect ratio of 4:1. At the corners of the contraction, a dead zone appears with no recirculation zone. The hydrodynamics of the flow is studied using high-resolution laser Doppler velocimetry (LDV). We focus, in particular, on velocity and shear rate profiles within and above the dead zone. The position of the solid–liquid interface is provided for different flow rates. The velocity profiles show a line where they overlap at the same point, known as the “slip line.” The position of this line is independent of the flow rate, but its velocity evolves with the flow rate, for which we have proposed a correlation. The position of this line aligns with the level of contraction for all flow rates. Shear rate profiles show a non-monotonic evolution: from zero at the solid–liquid interface to a peak, then decreasing to zero again in the plug zone at the core of the pipe, indicating the presence of different layers between the solid–liquid interface and the slip line. The line of maximum shear rate is independent of the flow rate, suggesting that the width of the zone between the slip line and this line remains constant at all flow rates. The study is completed by establishing correlations between the friction factor (f) and the Euler number (Eu) as a function of the generalized Reynolds number Regen upstream, downstream, and at the level of contraction.
This experimental study focuses on exploiting laser Doppler velocimetry (LDV), a non-intrusive technique, for rheological characterization based on analyzing flows of Carbopol solutions at two distinct concentrations and two different temperatures within a pipe. The velocity profiles obtained using LDV and the pressure drops associated with each flow rate were exploited to establish the behavioral law of aqueous Carbopol solutions. Two approaches were used: the first was an analytical velocity model to fit the experimental profile, and the second used the first derivative of the experimental velocity profile and the pressure drops to reconstruct the flow curve. In addition, a third reference characterization was carried out using a rotary rheometer equipped with a vane geometry. This study’s three rheological characterization methods showed excellent agreement concerning the Herschel–Bulkley model. Finally, all the laws resulting from these three methods were validated using an empirical law relating to Darcy’s coefficient of friction.
15 October 2025 in Journal of Non-Newtonian Fluid Mechanics
This comparative experimental and numerical study investigates the first hydrodynamic instability and flow transitions in the annular gap between concentric vertical cylinders, with the inner cylinder rotating and the outer cylinder stationary. The study contrasts the behavior of a non-Newtonian fluid, carboxymethyl cellulose (CMC), with that of a Newtonian fluid, potassium sulfate (K2SO4). The influence of inner cylinder rotational speed, represented by Reynolds (Re) and Taylor (Ta) numbers, and filling rates (Γ) on the first instability thresholds and flow modes is examined for both fluids. Experimental visualization and numerical simulations illustrate distinct flow characteristics, emphasizing the significant impact of CMC's non-Newtonian properties on flow structure formation and instability thresholds. Rheological analysis identifies instability thresholds specific to CMC at a 0.2% concentration. The study further analyzes azimuthal and axial wave numbers (n) and (λ) relative to Reynolds number (Re) and filling rates (Γ) to characterize flow dynamics comprehensively.
We experimentally and numerically investigate the flow of a viscoplastic fluid, namely a Carbopol gel, in a benchmark rectangular channel with a 4:1 abrupt contraction. Laser Doppler Velocimetry (LDV) measurements were performed in a two-dimensional plane to examine the evolution of the longitudinal and transverse velocity components far upstream, along the centerline, and near the contraction. Three-dimensional numerical simulations using OpenFOAM were then used to analyze the effects of inertia and geometry on the flow dynamics and to extend some observed phenomena to the three-dimensional case. The experimental results reveal the presence of a plug region upstream of the contraction, for which correlations are proposed to describe its velocity and width evolution. Near the contraction, a dead zone is observed, within which the velocity remains zero over a finite transverse distance. The interface separating this zone from the flowing region is identified for different flow rates. A slip line is also detected in all cases, with a position that remains independent of the flow rate and is located exactly at the contraction plane. Shear rate profiles exhibit a non-monotonic behavior, increasing from nearly zero values, reaching a maximum near the contraction, and then decreasing to zero at the center of the pipe. An elongation peak is observed at the contraction plane, and a correlation is proposed for its maximum value. The flow remains predominantly unidirectional along the centerline. The numerical results are used to propose correlations for the friction factor and Euler number. The simulations show that the slip line persists across different spanwise planes, with its position remaining unchanged and coinciding with the contraction level. This study provides new experimental and numerical insights into viscoplastic flow in a benchmark geometry and offers useful reference data for future modeling and validation efforts.
In this study, machine-learning (ML) methods are used to predict the experimental flow of an elasto-viscoplastic fluid (EVP) in a rectangular channel featuring an abrupt contraction with an aspect ratio of 4:1, as measured by laser Doppler velocimetry (LDV). Two distinct Python-based ML techniques, namely convolutional neural networks (CNNs) and artificial neural networks (ANNs) with the Rectified Linear Unit (ReLU) activation function, were used to predict the experimental LDV velocity profiles. 80% of the dataset is used for training, while 20% is used for internal model assessment. The excellent performance of both the CNN and ANN models in predicting EVP flow through a 4:1 contraction validates their predictive accuracy. When it comes to global and low-flow v-component prediction, CNN has a slight advantage over ANN, particularly for mean square error (MSE) values between 10⁻⁹ and 10⁻⁶, coefficient of determination R² = 0.9995 to 0.9971, and MAE/RMSE in 10⁻⁵ to 10⁻³. These low error values validate the reliability of the proposed ML-based method by demonstrating strong agreement with the experimental results. ML models can represent intricate EVP flows, reducing the need for computationally intensive simulations. CNN and ANN results accurately predicted several rare phenomena reported in the literature, including the slip line's position and velocity, as well as the dead zone.