Bio-Inspired Flow Control Using Fluid-Structure Interaction Modeling and Machine Learning

Nair, Nirmal Jayaprasad

Permalink

https://hdl.handle.net/2142/117778 Copy

Description

Title

Bio-Inspired Flow Control Using Fluid-Structure Interaction Modeling and Machine Learning

Author(s)

Nair, Nirmal Jayaprasad

Issue Date

2022-11-30

Director of Research (if dissertation) or Advisor (if thesis)

Goza, Andres

Doctoral Committee Chair(s)

Goza, Andres

Committee Member(s)

• Ansell, Phillip
• Bodony, Daniel
• Tran, Huy
• Wissa, Aimy

Department of Study

Aerospace Engineering

Discipline

Aerospace Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• fluid-structure interaction
• strongly-coupled
• coverts
• flow control
• machine learning
• reinforcement learning
• active
• passive

Abstract

Birds have a remarkable ability to perform complex maneuvers at post-stall angles of attack, partly due to the lift augmenting capabilities of self-actuating covert feathers. Covert-feathers-inspired flaps can be therefore leveraged to develop flow control strategies for next-generation unmanned- and micro-aerial vehicles that require high agility and maneuverability to operate in adverse environments in defence and commercial sectors. This thesis focuses on simulating and characterizing the aerodynamic benefits of such covert-inspired flow control methods via fluid-structure interaction (FSI) modeling and machine learning. Firstly, to computationally model covert-inspired flow control, we develop an efficient numerical algorithm for strongly-coupled FSI problems in an immersed boundary (IB) framework. Existing strongly-coupled IB methods are plagued by a severe computational bottleneck related to how the large-dimensional equations embedded withing a small-dimensional fluid-structure coupling matrix are evaluated. We present a remedy for this bottleneck wherein we precompute a modified matrix that encapsulates the large-dimensional operations to achieve significant computational savings. We also present a parallel implementation of the algorithm that scales favorably across multiple processors. Using the FSI solver, we simulate the flow past a covert-inspired passive flow control system. Most studies involving covert-inspired passive flow control model the feathers as a freely moving or a rigidly attached flap on a wing. A flap mounted via a torsional spring enables a configuration more emblematic of the finite stiffness associated with the covert-feather dynamics. The performance benefits and flow physics associated with this more general case remain largely unexplored. In this work, we model covert feathers as a passively deployable, torsionally hinged flap on the suction surface of a stationary airfoil. We numerically investigate this airfoil-flap system at a low Reynolds number of Re=1,000 and post-stall angle of attack of 20 deg. A parametric study is performed by varying the stiffness of the spring, mass of the flap and location of the hinge. The lift-enhancing FSI mechanisms are then analyzed in detail. Finally, we describe a covert-inspired hybrid active-passive flow control strategy as an extension of the passive counterpart to achieve even greater aerodynamic benefits. This method consists of actively varying the stiffness of the hinge in time to passively control the flap motion. The hinge stiffness is varied via a reinforcement learning (RL)-trained closed-loop feedback controller. The performance of the hybrid controller is analyzed in steady freestream conditions as well as in the presence of vortex gusts. In this hybrid method, to address the issue of practical unavailability of off-surface flow-field data, we also propose a state estimation framework that can estimate the full flow-field from limited sensor measurements located on the body surface using deep learning. We demonstrate the accuracy of this state estimation approach on a canonical problem of a flow past a flat plate, but emphasize its utility in the hybrid method.

Graduation Semester

2022-12

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Nirmal Jayaprasad Nair

Owning Collections