IP Library › Granted Patent US 12,510,449
Granted Patent B2
US 12,510,449 · App. 17/195,228 · Granted Dec 30, 2025

Quantification of shear stress field using tuft visualization and deep learning

Inventors: Morteza Gharib (Altadena, CA); Emile K. Oshima (Pasadena, CA); Pablo Hermoso Moreno (London, GB)
Assignee: California Institute of Technology
G01N3/12G01N3/068G06F30/15G06F30/27G06F30/28G06N3/08G01N2203/0246G01N2203/0641G06F2119/14
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,510,449
App. No.
17/195,228
Granted
Dec 30, 2025
Kind
B2
Abstract

A system and method for estimating the shear wall stress of an aerodynamic surface using a tuft visualization technique combined with a physics-informed neural network. The tuft visualization technique is a simplified method of generating velocity profile data of an aerodynamic model that can subsequently be used to generate a shear wall stress profile of the model. Systems and methods described herein also provide for additional input data using an augmented tuft and taps inputs for the physics-informed neural network to generate the shear wall stress profile.

Claims (35)

1 . A method for determining shear wall stress comprising:

providing a model of an aerodynamic surface;

determining a first set of input variables from spatial coordinates of the model;

disposing the model within a testing system;

determining a second set of input variables from spatial coordinates of the testing system;

connecting a plurality of tufts to the model in a manner such that the tufts can flow freely in a direction of an applied flow;

applying a generally horizontal dynamic flow force to the model thereby creating movement of the plurality of tufts;

using a camera visualization system to generate input data from the movement of the tufts for use as a third set of input variables;

applying the first, second, and third sets of input variables to a physics informed neural network to generate a set of output flow variables;

utilizing the output flow variables to determine velocity profile of the model; and

utilizing the velocity profile as well as a set of pressure outputs of the model to estimate the shear walls stress of the model.

2 . The method of claim 1 , further comprising:

performing automatic differentiation on the set of output flow variables to generate a set of residual data wherein the set of residual data is used to generate a loss function and wherein the loss function is used to augment the sets of input variables to improve the output flow variables.

3 . The method of claim 1 , further comprising:

disposing a plurality of pressure taps on the model wherein each of the plurality of pressure taps can generate pressure data with respect to the dynamic flow force and wherein the pressure data is used as a fourth set of input variables;

and wherein the fourth set of input variable is utilized to generate the output flow variables.

4 . The method of claim 1 , wherein the model is an airfoil.

5 . The method of claim 1 , wherein the model is a scaled model of an aircraft.

6 . The method of claim 1 , wherein the model is disposed at an angle with respect to the horizontal direction of the dynamic flow force.

7 . The method of claim 1 , further comprising a plurality of pressure taps disposed within the testing system, wherein the pressure taps are used to generate and provide pressure data with respect to the dynamic flow force within the testing system.

8 . A system for estimating the shear wall stress of an aerodynamic surface comprising:

a testing apparatus wherein the testing apparatus is configured to generate a dynamic flow force;

a model disposed within the testing apparatus in such a manner that the model is exposed to the dynamic flow force;

a plurality of tufts disposed on an outer surface of the model such that the plurality of tufts move in response to exposure to the dynamic flow force;

at least one camera visualization system configured to capture image data of the movement of the tufts when exposed to the dynamic flow force; and

estimating the shear wall stress of the model using at least the captured image data in a physics-informed neural network.

9 . The system of claim 8 , wherein the plurality of tufts and image data can be used to generate a first set of input variables that can be used to determine the wall shear stress of the model.

10 . The system of claim 9 , wherein the testing apparatus comprises a plurality of spatial coordinates that can be used as a second set of input variables, wherein the first and second set of input variables can be applied to the physics-informed neural network to estimate the wall shear stress of the model.

11 . The system of claim 9 further comprising a plurality of pressure taps disposed on the model wherein each of the plurality of pressure taps can generate pressure data with respect to the dynamic flow force and wherein the pressure data is used as a third set of input variables.

12 . The system of claim 8 , wherein the model is an airfoil.

13 . The system of claim 8 , wherein the model is a scaled version of an aircraft.

14 . The system of claim 8 , wherein the model is positioned at an angle with respect to the horizontal flow of the dynamic air flow.

15 . The system of claim 8 , wherein the apparatus is a wind tunnel.

16 . The system of claim 14 wherein the angle is at least 2 degrees off of horizontal.

17 . The system of claim 14 , wherein the angle does not exceed 12 degrees off of horizontal.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jun 30, 2023
From: GHARIB, MORTEZA; OSHIMA, EMILE K.; MORENO, PABLO HERMOSO
To: CALIFORNIA INSTITUTE OF TECHNOLOGY
Reel/Frame 064129/0528 →
Continuity (2)
Provisional Application 62986505 · Mar 6, 2020
Related Publication 20210278330A1 · Sep 9, 2021
References Cited (35)
US 6380535B1 · Wetzel · 2002 [cited by examiner]
US 10711604B2 · Johnson · 2020 [cited by examiner]
US 20120301296A1 · Greenblatt · 2012 [cited by examiner]
US 20190047678A1 · Jaiswal · 2019 [cited by examiner]
D. R. Troolin and E.K. Longmire; “Large Format PIV to Resolve Large and Small Structures Surrounding a Gurney Flap”; Sep. 11-14, 2007 ; 7th International Symposium on Particle Image Velocimetry (Year: 2007). [cited by examiner]
“Smoke and Tufts”, NASA, Sep. 5, 2009, Retrieved from the Internet https://www.grc.nasa.gov/www/k-12/airplane/tunvsmoke.html. [cited by applicant]
Azabi et al., “Artificial Intelligence to Enhance Aerodynamic Shape Optimisation of the Aegis UAV”, Machine Learning and Knowledge Extraction, Apr. 4, 2019, vol. 1, pp. 552-574; doi:10.3390/make1020033. [cited by applicant]
Bhargava et al., “An experimental study of near wall flow parameters in the blade end-wall corner region”, NASA Scientific and Technical Information Division, 1989, 313 pgs. [cited by applicant]
Brunton et al., “Machine Learning for Fluid Mechanics”, Annual Review of Fluid Mechanics, vol. 52, 2020, First published as a Review in Advance Sep. 12, 2019, pp. 477-508, https://doi.org/10.1146/annurev-fluid-010719-06… [cited by applicant]
Dupuis et al., “Surrogate Modeling of Aerodynamic Simulations for Multiple Operating Conditions Using machine Learning”, AIAA Journal, vol. 56, No. 9, Sep. 2018, doi: 10.2514/1.J056405. [cited by applicant]
Fang et al., “Neural Network Models for the Anisotropic Reynolds Stress Tensor in Turbulent Channel Flow”, Journal of Turbulence, vol. 21, No. 9-10, 2020, pp. 525-543, doi: 10.1080/14685248.2019.1706742. [cited by applicant]
Freymeth, , “Flow visualization in fluid mechanics”, Review of Scientific Instruments, vol. 64, No. 1, 1993, pp. 1-18, DOI:10.1063/1.1144433. [cited by applicant]
Genç et al., “A comparative evaluation of shear stress modeling based on machine learning methods in small streams”, Journal of Hydroinformatics, vol. 17, No. 5, Apr. 28, 2015, pp. 805-816, doi: 10.2166/hydro.2015.142. [cited by applicant]
Glorot et al., “Understanding the difficulty of training deep feedforward neural networks”, in the Proceedings of the 13th International Conference on Artificial Intelligence and Statistics, (AISTATS) 2010, Chia Laguna … [cited by applicant]
Hultmark et al., “Turbulent Pipe Flow at Extreme Reynolds Numbers”, Physical Review Letters, vol. 108, No. 9, Mar. 2, 2012, doi: 10.1103/PhysRevLett. 108.094501. [cited by applicant]
Jagtap et al., “Locally adaptive activation functions with slope recovery for deep and physics-informed neural networks”, Proceedings of the Royal Society, vol. A 476, Jul. 15, 2020, 20 pgs., https://doi.org/10.1098/rsp… [cited by applicant]
Khozani et al., “Shear Stress Distribution Prediction in Symmetric Compound Channels Using Data Mining and Machine Learning Models”, Frontiers of Structural and Civil Engineering, vol. 14, Dec. 2019, pp. 1097-1109. [cited by applicant]
Liu et al., “Pressure and Temperature-Sensitive Paints”, Berlin, Germain: Springer—Verlag Berlin Heidelberg, 2021, 552 pgs (presented in two parts). [cited by applicant]
Meinhart et al., “PIV measurements of a microchannel flow”, Experiments in Fluids, vol. 27, Oct. 1999, pp. 414-419, DOI: 10.1007/s003480050366. [cited by applicant]
Minelli et al., “Active Aerodynamic Control of a Separated Flow Using Streamwise Synthetic Jets”, Flow, Turbulence and Combustion, vol. 103, pp. 1039-1055, Aug. 30, 2019, doi: 10.1007/s10494-019-00058-4. [cited by applicant]
Omata et al., “Extracting quantitative three-dimensional unsteady flow direction from tuft flow visualizations”, Fluid Dynamics Research, vol. 49, No. 5, Sep. 5, 2017, doi: : 10. 1088/1873-7005/aa8551. [cited by applicant]
Oweis et al., “The mean velocity profile of a smooth-flat-plate turbulent boundary layer at high Reynolds number”, Journal of Fluid Mechanics, vol. 665, Dec. 2010, pp. 357-381, doi: 10.1017/S0022112010003952. [cited by applicant]
Raffel et al., “Particle Image Velocimetry: A Practical Guide”, Cham, Switzerland: Springer International Published AG, 2007, 460 pgs. [cited by applicant]
Raissi et al., “Hidden fluid mechanics: learning velocity and pressure fields from flow visualizations”, Science, vol. 367, No. 6478, Feb. 28, 2020, doi: 10.1126/science.aaw4741. [cited by applicant]
Raissi et al., “Hidden physics models: Machine learning of nonlinear partial differential equations”, Journal of Computational Physics, vol. 357, Mar. 15, 2018, pp. 125-141, https://doi.org/10.1016/j.jcp.2017.11.039. [cited by applicant]
Simpson et al., “Detecting vortices within unsteady flows when using single-shot PIV”, Experiments in Fluids, vol. 59, No. 125, Jul. 7, 2018, 13 pgs., https://doi.org/10.1007/s00348-018-2575-3. [cited by applicant]
Sin et al., “Methodology for Calculating Shear Stress in a Meandering Channel”, U. S. Department of the Interior Bureau of Reclamation, Jun. 2012, 168 pgs. [cited by applicant]
Steinfurth et al., “Tuft Deflection Velocimetry: a simple method to extract quantitative flow field information”, Experiments in Fluids, vol. 61, No. 146, Jun. 6, 2020, doi:10.1007/s00348-020-02979-7. [cited by applicant]
Tropea et al., “Springer Handbook of Experimental Fluid Mechanics”, Berlin, Germany: Springer-Verlag Berlin Heidelberg, 2007, 1570 pgs. (presented in three parts). [cited by applicant]
Vey et al., “Extracting quantitative data from tuft flow visualizations on utility scale wind turbines”, Journal of Physics: Conference Series, vol. 524, The Science of Making Torque from Wind 2014 (TORQUE 2014) Jun. 18… [cited by applicant]
Wang et al., “Physics Informed Machine Learning Approach for Reconstructing Reynolds Stress Modeling Discrepancies Based on DNS Data”, Physical Reviews Fluids vol. 2, 034603, Feb. 27, 2017, arXiv:1606.07987v2, 36 pgs. [cited by applicant]
Wang et al., “Understanding and Mitigating Gradient Pathologies In Physics-Informed Neural Networks ETWORKS”, arXiv:2001.04536v1 [cs.LG] Jan. 13, 2020, 28 pgs. [cited by applicant]
Westerweel et al., “Particle Image Velocimetry for Complex and Turbulent Flows”, Annual Review of Fluid Mechanics, vol. 45, No. 1, Jan. 2013, pp. 409-436, DOI:10.1146/annurev-fluid-120710-101204. [cited by applicant]
Willert, , “High-Speed Particle Image Velocimetry for the Efficient Measurement of Turbulence Statistics”, Experiments in Fluids, vol. 56, No. 17, Jan. 13, 2015, doi: 10.1007/s00348-014-1892-4. [cited by applicant]
Yang et al., “B-PINNs: Bayesian physics-informed neural networks for forward and inverse PDE problems with noisy data”, Journal of Computational Physics, vol. 425, Jan. 15, 2021, 109913, https://doi.org/10.1016/j.jcp.20… [cited by applicant]