Pith. sign in

REVIEW 4 major objections 6 minor 30 references

Methodology for Online Estimation of Rheological Parameters in Polymer Melts Using Deep Learning and Microfluidics

T0 review · 4 major / 6 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A recurrent neural network trained on synthetic signals from a one-dimensional hydraulic circuit can estimate polymer-melt viscosity parameters from pressure and flow measurements in real time.

desk verdict A plausible simulation workflow that overreaches its evidence; the central claim of online estimation for real polymer melts is unsupported, but the methodology is worth a serious referee with major revisions. read the letter →

arxiv 2412.04142 v1 pith:326SYXWD submitted 2024-12-05 physics.flu-dyn cs.AI

Juan Sandubete-López , José L. Risco-Martín , Alexander H. McMillan , Eva Besada-Portas This is my paper · ORCID
classification physics.flu-dyncs.AI
keywords rheologypolymermeltsmicrofluidicsdeeplearningrecurrentneuralnetworkhydraulicRCcircuitpower-lawfluidonlineestimation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper proposes an end-to-end pipeline for online rheological estimation: a one-dimensional hydraulic RC circuit model generates synthetic pressure-drop and flow-rate signals for a generalized Newtonian fluid, and a recurrent neural network learns to map those signals back to the fluid's parameters. The target parameters are the zero-shear viscosity $\eta_0$, the power-law index $n$, and the relaxation time $\lambda$ of a polymer melt. If the approach works, microfluidic devices could monitor fluid properties continuously from signals that are already easy to measure, without offline sampling or a prescribed input waveform. The authors support the claim with simulated experiments in which the network's predictions, when re-simulated, reproduce the original signals, and they report that $n$ is the most reliably identified parameter.

What carries the argument

The central object is the hydraulic RC circuit: an input resistance $R_1$, an air capacitance $C$, and an output resistance $R_2$, modeled in one dimension as an electrical analog. The key mechanism is the weighted flow-rate expression $Q_i = (1-\alpha)Q_{v,i} + \alpha Q_{e,i}$, where $Q_{v,i}$ is the Newtonian Hagen-Poiseuille flow, $Q_{e,i}$ is the semi-analytic power-law flow, and the weight $\alpha$ is a generalized logistic function of the Deborah number, approximated here as $De_i \approx \lambda \dot\gamma_i = 6\lambda Q_{v,i}/(w h^2)$. The capacitance volume $V$ is the only state variable, integrated from $dV/dt = Q_2 - Q_1$ and converted to pressure through Boyle-Mariotte. The logistic blend gives the simulator a smooth Newtonian-to-elastic transition, and rejection-sampling conditions ensure each training run contains both regimes, which is what makes the synthetic signals informative about $\eta_0$, $n$, and $\lambda$.

What would settle it

Build a physical microfluidic prototype with the same resistance geometry and air capacitance, drive a polymer melt with independently known $\eta_0$, $n$, and $\lambda$ through the same random pressure sequences, and compare the measured pressure-drop and flow-rate signals with the simulator's predictions; systematic deviations in the transient response would show that equation (6)'s logistic blend is not faithful, so a network trained on it would not transfer to real fluids.

Watch

Extended reading notes

Core claim

The central claim, stated in the authors' terms, is that a bidirectional gated recurrent unit trained exclusively on synthetically generated data can identify the rheological parameters of a polymer melt from dynamic pressure and flow measurements in a microfluidic circuit. The circuit is designed as two equal rectangular hydraulic resistances separated by an air capacitance, so the transient charging of the capacitance carries information about viscosity. The simulator randomly composes steps, ramps, and sine pressure inputs and keeps only runs in which the flow spends enough time in both the Newtonian and the non-Newtonian regimes, defined through the Deborah number. After training on 5,500 such runs, the network takes 250 samples (12.5 s at 20 Hz) of two pressure drops and two flow rates and outputs estimates of $\eta_0$, $n$, and $\lambda$. The paper's verification shows that estimated parameters, when fed back into the simulator, produce curves close to the originals, and it uses Pearson correlations between parameter errors and signal errors as evidence that the measurement setup makes the parameter-to-signal mapping identifiable.

Load-bearing premise

The whole approach stands on the premise that the simplified one-dimensional hydraulic RC model, especially the smooth logistic blending rule between viscous and elastic flow, faithfully represents how a real polymer melt moves through the microfluidic circuit.

Editorial extensions

If this is right

  • If the central claim holds, polymer-melt viscosity can be monitored in real time from pressure-drop and flow-rate sensors that are already common in industrial lines, removing the need for offline sampling.
  • Training data can be produced cheaply by one-dimensional simulation instead of expensive computational fluid dynamics or physical experiments, so the methodology shortens the design cycle for microfluidic rheometers.
  • Because the network does not require a fixed input waveform, it could estimate parameters from arbitrary process-driven pressure sequences, reducing interference with normal operation.
  • The same simulator-plus-neural-network loop could be adapted to other generalized Newtonian fluid models and other microfluidic circuit geometries by changing the forward model and retraining.
  • The verification protocol, which re-simulates estimated parameters and checks signal errors, serves as a simulation-based test of whether a chosen measurement setup can identify the parameters at all.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the simulator is later refined to include neglected tubing, connectors, or two-dimensional effects, the same training loop could be rerun without physical prototypes, so the method's practical ceiling is set by simulator fidelity rather than by the network architecture.
  • The Pearson-correlation verification is effectively an identifiability screen; it could be reused on other sensor layouts to decide in advance whether a planned microfluidic measurement is informative enough for inverse estimation.
  • Because the network emits one parameter vector per 12.5-second window, the approach could be extended to track slow changes in fluid composition or temperature by sliding the window continuously, although the paper does not demonstrate time-resolved tracking.
  • A natural next test is whether a network trained on a range of simulated circuit geometries and fluid parameters can generalize to unseen chips, which would eliminate per-device retraining; the paper does not claim this.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. The paper proposes a methodology for estimating the rheological parameters (zero-shear viscosity η0, power-law index n, and relaxation time λ) of polymer melts from pressure-drop and flow-rate signals in a microfluidic circuit. The authors model the circuit as a one-dimensional hydraulic RC network, blend a Newtonian expression with a power-law expression via a logistic weight based on the Deborah number, generate synthetic training data by simulating random pressure inputs and applying rejection sampling, and train a bidirectional GRU network to predict the three parameters from four measured signals. The model is evaluated on held-out synthetic test data by fitting error distributions and computing Pearson correlations between parameter errors and simulated curve errors. The paper concludes that the behavior index n is the most reliably estimated parameter and lists physical prototype validation as future work.

Significance. If the claimed capability were established, the approach could be valuable for inline rheological monitoring in industrial and microfluidic settings. The methodological pipeline is clearly described, and the use of a one-dimensional hydraulic circuit abstraction combined with recurrent neural networks is a reasonable strategy for reducing the cost of training-data generation. However, the significance as presented is limited: the evaluation is entirely in silico, using test data generated by the same unvalidated simulator used for training, and the paper reports no quantitative accuracy metrics for the estimated parameters. The reported Pearson correlations are weak and internally inconsistent with the text's interpretation. Furthermore, the simulator contains an internal circuit-modeling error and an ad hoc constitutive transition that are load-bearing for the central claim. The paper's promise of online estimation of real polymer-melt properties is therefore not supported by the evidence.

major comments (4)
  1. [Section 5 and Section 3.4] The central claim of the paper is that the trained network can estimate rheological parameters of real polymer melts from pressure and flow measurements. The only evaluation, however, is on synthetic test sets generated by the same one-dimensional simulator used to produce the training data (Section 3.4). Section 5 explicitly defers building a physical prototype to validate the simulations to future work. This demonstrates interpolation within the simulator's output distribution, not estimation of physical parameters; no comparison to experimental data, CFD, or an independent model is provided. This gap is load-bearing for the abstract's promise of 'online estimation of fluid properties' in real polymer melts.
  2. [Section 3.1] The Thevenin time constant is incorrect. For the declared circuit, with R1 in series with the parallel combination of R2 and the capacitance C, the resistance seen by the capacitance is Rth = R1 R2 / (R1 + R2), not R1 + R2. With R1 = R2, the value τ = (R1 + R2) C used in the model is four times the correct time constant. Since the transient response of the capacitance is the only state variable in the simulator (Eq. (7)), this error affects every generated training and test signal.
  3. [Section 3.1, Eqs. (3)-(6)] The transition model is physically unjustified. Equation (5) is a steady power-law (generalized Newtonian) flow-rate expression, not an elastic or viscoelastic correction; it contains no elasticity parameter other than λ entering through the Deborah number. Equation (6) blends Eq. (3) and Eq. (5) with a generalized logistic weight α(De) that is not derived from any constitutive model. The Deborah criterion De > 1/2 is used to signal elastic effects, but the power-law model does not represent elasticity. Thus the 'non-Newtonian regime' in the simulated data is an ad hoc interpolation rather than a faithful representation of polymer-melt viscoelasticity. The rejection sampling in Eqs. (9)-(10) then deliberately selects data from this unvalidated transition, so the training distribution inherits the modeling error.
  4. [Section 4 and Table 2] The paper reports no quantitative accuracy metrics for the estimated parameters. Figure 6 shows only qualitative normalized error distributions, and Table 2's Pearson correlation coefficients are not consistent with the text's interpretation: the largest magnitude is 0.533 (E(n) vs. E(ΔP2)), while the text states that 'correlation coefficients close to the unit imply high correlation' and concludes that n 'correlates strongly' with the output signals. The correlations for λ are below 0.12 in magnitude. The results therefore do not support the claim that the network reliably estimates η0, n, and λ, nor the assertion that the mapping from parameters to simulated curves is injective in the sense required by the verification procedure.
minor comments (6)
  1. [Section 4] The Figure 6 caption and the text describe the input signal configurations inconsistently: the caption says 'only step sequences' for panel (a), while the text says the first dataset is 'a sequence made out of 8-10 sinusoidal signals.'
  2. [Section 4] The text refers to 'the 10s experiment' after the simulation duration was earlier specified as 12.5 seconds; this should be corrected.
  3. [Section 3.2] The rejection-sampling thresholds and percentages in Eqs. (9)-(10), namely αv,th, αe,th, pv, and pe, are never given numerical values, which hinders reproducibility of the dataset generation.
  4. [Section 3.3] The claim that real-time operation is enabled by an input sliding window is not demonstrated; the reported experiments use a single 250-sample window and produce one estimate per sequence.
  5. [Section 3.1] The sentence explaining why tubing and connectors are neglected is unclear: it says their characteristic dimensions are 'around two orders of magnitude above' without specifying above what, and the relation to the third-power scaling of hydraulic resistance should be stated more carefully.
  6. [Section 3.4 and Table 2] The phrase 'a correlation coefficient for a particular estimated parameter closer to one half' is ambiguous; the authors presumably mean close to zero or close to 0.5, but the threshold for a useful correlation is not defined.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the pipeline is a train-on-simulator, test-on-simulator workflow; the unvalidated physical model is an external-validity limitation, not a circular step.

full rationale

The claimed pipeline is: sample (η0, n, λ); simulate the 1D RC equations (3)-(8); reject uninformative runs via (9)-(10); train a BGRU on the synthetic signals; evaluate on held-out simulator data by re-simulating predicted parameters. None of these steps defines the target in terms of itself or fits a parameter and then reports a closely related quantity as a prediction. The test split is disjoint from the training split, so performance measures interpolation of the simulator's inverse map, not circularity. The rejection sampling in (9)-(10) and the Pearson checks in Table 2 are identifiability/experimental-design steps, not circular reductions. Section 5's statement that building a physical prototype 'to validate our simulations' is future work is an honest admission that real-melt predictive validity is not established, and the Section 3.1 Thevenin time constant τ=(R1+R2)C appears inconsistent with the stated R1-R2-C circuit; however, a modeling error or a missing external benchmark is not a self-referential derivation. There are no load-bearing self-citations, no imported uniqueness theorems, and no renamed known result. Therefore the circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on the fidelity of the 1D simulator, which involves several hand-chosen elements: the logistic transition, rejection sampling thresholds, capacitance volume, and parameter ranges. None of these are calibrated against experimental data, and the paper labels physical validation as future work.

free parameters (4)
  • Logistic transition function parameters
    The function α(De) used to weight equations (3) and (5) in equation (6) is described as a generalized logistic, but its parameters and the thresholds αv,th, αe,th are not specified. These are chosen by hand and directly affect the simulated signals.
  • Rejection sampling thresholds and percentages = αv,th, αe,th, pv, pe (values not given)
    Equations (9) and (10) require certain fractions of samples to be in Newtonian and non-Newtonian regimes. The thresholds and percentages are not specified, and they shape the training distribution by discarding simulations.
  • Capacitance volume V
    The capacitance is defined as C = K_fluid V, but V is never given, so the transient dynamics and time constants are not reproducible from the text.
  • Parameter ranges in Table 1 = η0 ∈ [1e-3, 1.6e-3], n ∈ [0.9, 1.1], λ ∈ [1e-4, 5e-4]
    These ranges are chosen so that both Newtonian and non-Newtonian regimes appear within 12.5-second experiments. They are not justified from measured properties of real polymer melts, and the viscosity range is closer to water than to typical polymer melts.
assumptions (5)
  • standard math Hagen-Poiseuille laminar flow in rectangular capillaries (equation 3)
    Used to compute Newtonian flow rates through the microfluidic resistances.
  • domain assumption Srivastava-Burns semi-analytic power-law flow approximation (equation 5)
    Used to compute elastic-flow contribution for non-Newtonian fluids; it is a cited approximation, not an exact solution.
  • ad hoc to paper The logistic interpolation α(De) is a valid smooth transition between Newtonian and elastic regimes (equation 6)
    The paper introduces this weighting without physical derivation; it is a modeling choice that is not validated against experiments.
  • domain assumption The hydraulic capacitance obeys Boyle-Mariotte with a single air volume state (equation 8)
    Assumes isothermal compression of air in the capacitance and neglects other sources of compliance.
  • domain assumption Tubing and connector resistances are negligible compared with R1 and R2
    Stated in Section 3.1; it is violated for high-viscosity or strongly shear-thinning fluids, which the authors acknowledge when they restrict n to 0.9-1.1.
invented entities (1)
  • Generalized logistic transition function α(De)
    purpose: To interpolate between the Newtonian and elastic flow-rate expressions in equation (6), producing a smooth regime transition
    The function is not derived from physics and its parameters are unspecified; it is a construct introduced to make the simulation work, with no external falsifiable prediction.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Methodology for Online Estimation of Rheological Parameters in Polymer Melts Using Deep Learning and Microfluidics." pith.science (2026). https://pith.science/paper/326SYXWD

@misc{pith2026241204142,
  author       = {Pith},
  title        = {Pith review of: Methodology for Online Estimation of Rheological Parameters in Polymer Melts Using Deep Learning and Microfluidics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/326SYXWD}},
  note         = {Machine review of arXiv:2412.04142}
}
read the original abstract

Microfluidic devices are increasingly used in biological and chemical experiments due to their cost-effectiveness for rheological estimation in fluids. However, these devices often face challenges in terms of accuracy, size, and cost. This study presents a methodology, integrating deep learning, modeling and simulation to enhance the design of microfluidic systems, used to develop an innovative approach for viscosity measurement of polymer melts. We use synthetic data generated from the simulations to train a deep learning model, which then identifies rheological parameters of polymer melts from pressure drop and flow rate measurements in a microfluidic circuit, enabling online estimation of fluid properties. By improving the accuracy and flexibility of microfluidic rheological estimation, our methodology accelerates the design and testing of microfluidic devices, reducing reliance on physical prototypes, and offering significant contributions to the field.

Figures

Figures reproduced from arXiv: 2412.04142 by the authors.

Figure 1
Figure 1. Methodology diagram. The elements in blue, a), b) and d), operate together with the simulator [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Simulated microfluidic circuit. a) shows the fluidic system diagram composed by one input [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Diagram showing the simulation flow. a) generates set of parameters for which block b) creates [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Designed artificial neural network. It receives four inputs signals: [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Two simulations performed over 100s with the same random pressure input signal. The parameters [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Error distributions for the three estimated parameters ( [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 22 canonical work pages

  1. [5]

    Green gelled propellant highly throtteable rocket motor and gas generator technology: status and application

    “Green gelled propellant highly throtteable rocket motor and gas generator technology: status and application” https://doi.org/10.1051/eucass/201911091. Cho, K., B. van Merrienboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk et al

    work page arXiv

  2. [7]

    Simultaneous measurement of rheological properties in a microfluidic rheometer

    “Simultaneous measurement of rheological properties in a microfluidic rheometer”. Physics of Flu- ids 32(5) https://doi.org/10.1063/5.0006060. Del Giudice, F

    work page doi:10.1063/5.0006060

  3. [11]

    DeepDDM: A Compact Deep-Learning Assisted Platform for Micro-Rheological Assessment of Micro-V olume Fluids

    “DeepDDM: A Compact Deep-Learning Assisted Platform for Micro-Rheological Assessment of Micro-V olume Fluids”. IEEE Access 11:66467–66477 https://doi.org/10.1109/ACCESS. 2023.3290496. Keslerová, R. and K. Kozel

    arXiv 2023

  4. [12]

    Numerical modelling of incompressible flows for Newtonian and non-Newtonian fluids

    “Numerical modelling of incompressible flows for Newtonian and non-Newtonian fluids”. Mathematics and Computers in Simulation 80(8):1783–1794 https://doi.org/https://doi.org/10.1016/j.matcom.2009.12.005. Kirchberger, C., P. Kröger, M. Negri, and H. Ciezki

    work page doi:10.1016/j.matcom.2009.12.005 2009

  5. [16]

    Design and use of an online drilling fluid pipe viscometer

    “Design and use of an online drilling fluid pipe viscometer”. Flow Measurement and Instrumentation 87:102224 https://doi.org/https://doi.org/10.1016/j.flowmeasinst.2022.102224. McIntyre, D., A. Lashkaripour, P. Fordyce, and D. Densmore

    arXiv 2022

  6. [17]

    Machine learning for microfluidic design and control

    “Machine learning for microfluidic design and control”. Lab on a Chip 22(16):2925–2937 https://doi.org/10.1039/d2lc00254j. Meek, G. A., R. Williams, D. Thornton, P. Knapp and S. Cosser

    doi:10.1039/d2lc00254j

  7. [18]

    F2E - Ultra High Pressure Distributed Pump Common Rail System

    “F2E - Ultra High Pressure Distributed Pump Common Rail System”. SAE Mobilus https://doi.org/https://doi.org/10.4271/2014-01-1440. Mortensen, N. A., F. Okkels, and H. Bruus

    work page doi:10.4271/2014-01-1440 2014

  8. [20]

    Viscoelastic and shear-thinning effects of aqueous exopolymer solution on disk and sphere settling

    “Viscoelastic and shear-thinning effects of aqueous exopolymer solution on disk and sphere settling”. Scientific Reports 9(1) https://doi.org/10.1038/s41598-019-44233-z. Mustafa, A., D. Haider, A. Barua, M. Tanyeri, A. Erten and O. Yalcin

    work page doi:10.1038/s41598-019-44233-z

Show all 30 references
  1. [21]

    Machine learning based microfluidic sensing device for viscosity measurements

    “Machine learning based microfluidic sensing device for viscosity measurements”. Sensors & Diagnostics 2(6):1509–1520 https://doi.org/10.1039/d3sd00099k. Oh, K. W., K. Lee, B. Ahn, and E. P. Furlani

    doi

  2. [23]

    Bidirectional recurrent neural networks

    “Bidirectional recurrent neural networks”.IEEE Transactions on Signal Processing45(11):2673– 2681 https://doi.org/10.1109/78.650093. Shampine, L. F

    doi

  3. [25]

    Oxford University Press https://doi.org/10.1017/S0022112006213855

    Introduction to Microfluidics . Oxford University Press https://doi.org/10.1017/S0022112006213855. Takken, M. and R. Wille

    doi

  4. [27]

    Physics for Scientists and Engineers . W.H. Freeman https://doi.org/10.1007/978-3-662-67936-4. Vinuesa, R. and S. L. Brunton

    doi

  5. [29]

    Tacotron: Towards End-to-End Speech Synthesis

    “Tacotron: Towards End-to-End Speech Synthesis”. In Interspeech https://doi.org/10.21437/Interspeech.2017-1452. Zhao, Q., S. Yan, B. Zhang, K. Fan, J. Zhang and W. Li

    2017 doi

  6. [30]

    An On-Chip Viscoelasticity Sensor for Biological Fluids

    “An On-Chip Viscoelasticity Sensor for Biological Fluids”. Cyborg and Bionic Systems 4 https://doi.org/10.34133/cbsystems.0006. AUTHOR BIOGRAPHIES JUAN SANDUBETE-LÓPEZ is an industrial PhD candidate by Universidad Complutense de Madrid, and at Microfluidic Innovation Center, P...

    doi

  7. [1997]

    Long Short-Term Memory

    “Long Short-Term Memory”. Neural Computation 9(8):1735–1780 https://doi.org/https: //doi.org/10.1162%2Fneco.1997.9.8.1735. Jarujareet, U., K. Wiratchawa, P. Panpisut, and T. Intharah

    1997

  8. [2004]

    Reexamination of Hagen-Poiseuille flow: shape dependence of the hydraulic resistance in microchannels

    “Reexamination of Hagen-Poiseuille flow: shape dependence of the hydraulic resistance in microchannels.”. Physical review. E, Statistical, nonlinear, and soft matter physics 71 5 Pt 2:057301 https://doi.org/https://doi.org/10.1103/PhysRevE.71.057301. Mrokowska, M. M. and A. Kr...

    doi

  9. [2005]

    Intrinsic Viscosity of Polymers and Biopolymers Measured by Microchip

    “Intrinsic Viscosity of Polymers and Biopolymers Measured by Microchip”. Analytical Chemistry 77(22):7137–7147 https://doi.org/10.1021/ac050932r. Sandubete-López, Risco-Martín, McMillan, and Besada-Portas Liu, N., H. Gao, Y . Xu, X. Chai, Y . Hu and L. Duan

    doi

  10. [2006]

    Analysis of Non-Newtonian Liquids Using a Microfluidic Capillary Viscometer

    “Analysis of Non-Newtonian Liquids Using a Microfluidic Capillary Viscometer”. Analytical Chemistry 78(5):1690–1696 https://doi.org/10.1021/ac0518046. Tabeling, P

    doi

  11. [2010]

    InTech https://doi.org/10.5772/545

    Numerical Simulations - Examples and Applications in Computational Fluid Dynamics, Chapter Numerical Simulation in Steady flow of Non-Newtonian Fluids in Pipes with Circular Cross-Section, 3–22. InTech https://doi.org/10.5772/545. Hochreiter, S. and J. Schmidhuber

    doi

  12. [2012]

    Design of pressure-driven microfluidic networks using electric circuit analogy

    “Design of pressure-driven microfluidic networks using electric circuit analogy”. Lab on a Chip 12(3):515–545 https://doi.org/10.1039/C2LC20799K. Schuster, M. and K. Paliwal

    doi

  13. [2013]

    Springer US https://doi.org/10.1007/978-3-642-27758-0_1129-4

    Non-Newtonian Fluids in Microfluidics, 1–12. Springer US https://doi.org/10.1007/978-3-642-27758-0_1129-4. ASTM

    doi

  14. [2014]

    Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

    “Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation” https://doi.org/10.3115/v1/D14-1179. Del Giudice, F

    doi

  15. [2016]

    Overview on the Gelled Propellants Activities of DLR Lampoldshausen

    “Overview on the Gelled Propellants Activities of DLR Lampoldshausen”. In 52nd AIAA/SAE/ASEE Joint Propulsion Conference, 2016 https://doi.org/10.2514/6.2016-4665. Krüger, T., H. Kusumaatmaja, A. Kuzmin, O. Shardt, G. Silva and E. M. Viggen

    2016 doi

  16. [2017]

    The lattice Boltzmann method

    “The lattice Boltzmann method”. Springer International Publishing 10(978-3):4–15 https://doi.org/10.1007/978-3-319-44649-3. Lee, J. and A. Tripathi

    doi

  17. [2019]

    A novel design of flow structure model for online viscosity measurement

    “A novel design of flow structure model for online viscosity measurement”. Insight - Non-Destructive Testing and Condition Monitoring 61(1):9–14 https://doi.org/10.1784/insi.2019.61.1.9. Boulais, E. and T. Gervais

    2019 doi

  18. [2020]

    ASTM International https://doi.org/10.1520/D2196-20

    Annual Book of ASTM Standards , V olume 06.01, Chapter Standard Test Methods for Rheological Properties of Non-Newtonian Materials by Rotational Viscometer. ASTM International https://doi.org/10.1520/D2196-20. Bie, Y ., X. Guo, P. Song, J. Yang and Z. Li

    doi

  19. [2021]

    The potential of machine learning to enhance computational fluid dynamics

    “The potential of machine learning to enhance computational fluid dynamics”. arXiv preprint arXiv:2110.02085:1–13. Wang, Y ., R. J. Skerry-Ryan, D. Stanton, Y . Wu, R. J. Weiss, N. Jaitly et al

    arXiv

  20. [2022]

    A Review of Microfluidic Devices for Rheological Characterisation

    “A Review of Microfluidic Devices for Rheological Characterisation”. Micromachines (Basel) 13(2):167– 186 https://doi.org/10.3390/mi13020167. Ferziger, J. H., M. Peri ´c, and R. L. Street

    doi

  21. [2023]

    The 2D microfluidics cookbook – modeling convection and diffusion in plane flow devices

    “The 2D microfluidics cookbook – modeling convection and diffusion in plane flow devices”. Lab on a Chip 23(8):1967–1980 https://doi.org/10.1039/D2LC01033J. Caldas Pinto, P., H. Ciezki, K. Naumann, J. Ramsel, P. Kröger, H. Niedermaier et al

    1967 doi

  22. [2024]

    Accelerated Computational Fluid Dynamics Simulations of Microfluidic Devices by Exploiting Higher Levels of Abstraction

    “Accelerated Computational Fluid Dynamics Simulations of Microfluidic Devices by Exploiting Higher Levels of Abstraction”. Micromachines 15(1):129 https://doi.org/10.3390/mi15010129. Tipler, P. and G. Mosca

    doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.