Global-in-time optimal control of stochastic third-grade fluids with additive noise

This paper establishes the global-in-time well-posedness of stochastic third-grade fluid equations on a two-dimensional torus with additive noise and proves the existence, uniqueness, and first-order optimality conditions for a velocity tracking control problem by transforming the system into an equivalent pathwise deterministic one.

The Gist

Imagine you are trying to steer a very stubborn, sticky boat through a choppy ocean. This isn't just any boat; it's made of a special, complex fluid (like thick honey mixed with rubber bands) that doesn't flow like water. This is what mathematicians call a Third-Grade Fluid.

Now, imagine the ocean isn't just choppy; it's being hit by unpredictable, random waves (the "noise"). Your goal is to find the perfect steering strategy (the "control") to keep this boat on a specific path, despite the fluid's stubbornness and the random waves.

This paper is the mathematical blueprint for solving that exact problem. Here is how the authors did it, broken down into simple concepts:

1. The Problem: A Sticky, Wobbly Boat

Most fluids (like water) are "Newtonian," meaning they flow predictably. But Third-Grade fluids are "non-Newtonian." They are like a mix of ketchup, toothpaste, and slime. If you push them, they might get thicker or thinner, and they have "memory" (viscoelasticity).

The math describing these fluids is incredibly messy. It involves equations with high-order derivatives (think of them as measuring not just speed, but how the speed is changing, and how that is changing, and so on). When you add random noise (like sudden gusts of wind or random waves), the equations become even harder to solve.

2. The Trick: Splitting the Problem

The authors faced a massive wall: How do you control a system that is both wildly complex and randomly chaotic?

They used a clever trick called decomposition. Imagine you have a car driving on a bumpy road.

• Part A (The Noise): They isolated the part of the motion caused purely by the random bumps. They treated this as a separate, simpler "noise car" that just bounces around.
• Part B (The Real Control): They subtracted this "noise car" from the total picture. What was left was a "deterministic" car. This car still has the sticky, complex fluid physics, but the random bumps are gone. It's now a predictable, albeit difficult, puzzle.

By solving the "noise car" first, they could prove that the "real car" (the actual fluid) behaves well enough to be controlled over a long period (globally in time), rather than just for a split second.

3. The Goal: The "Target" Path

The authors wanted to find the best control (an external force, like a thruster) to make the fluid's velocity match a desired target (like a specific speed or pattern).

They set up a "scorecard" (called a Cost Functional):

• Penalty 1: How far off the fluid is from the target path.
• Penalty 2: How much energy you used to push the fluid.
• The Goal: Minimize the total penalty. You want the fluid on track, but you don't want to burn all your fuel getting there.

4. The Solution: The "Shadow" Guide

To find the perfect steering strategy, the authors used a concept called the Adjoint System.

Think of this like a Shadow Guide:

• The State Equation is the fluid moving forward in time.
• The Adjoint Equation is a "shadow" that runs backward in time.

The Shadow Guide starts at the end of the journey (where you want to be) and works backward to tell you what you should have done at the beginning to get there. It calculates the "sensitivity" of the system: If I push here, how much will it help me reach the target later?

By comparing the "Shadow" with the actual fluid movement, they derived a set of rules (Optimality Conditions) that tell you exactly how to apply the force to get the best result.

5. Why This Matters

Why do we care about controlling sticky, random fluids?

• Nanofluids: These are tiny particles suspended in liquids, used in high-tech cooling systems for computers, car engines, and even medical treatments.
• Efficiency: If we can mathematically predict and control how these fluids behave, we can design better engines, more efficient manufacturing processes for plastics, and safer medical devices.

The Big Takeaway

Before this paper, mathematicians could only solve this problem for a very short time or under very strict, unrealistic conditions.

This paper is the first to say: "Yes, we can control these messy, random, sticky fluids for a long time, and here is the exact mathematical recipe to do it."

They proved that:

1. A solution exists (the boat won't sink).
2. The solution is unique (there's only one right way to steer).
3. We can find the best way to steer it using the "Shadow Guide" method.

In short, they turned a chaotic, sticky, random mess into a solvable, controllable system, opening the door for better engineering in the real world.

Technical Summary

Here is a detailed technical summary of the paper "Global-in-time optimal control of stochastic third-grade fluids with additive noise" by Kush Kinra and Fernanda Cipriano.

1. Problem Statement

The paper addresses the optimal control problem for a class of stochastic non-Newtonian fluids, specifically third-grade fluids, defined on a two-dimensional torus ($T^2$).

• Governing Equations: The system is described by the stochastic third-grade fluid equations perturbed by infinite-dimensional additive white noise. The equations involve complex nonlinear terms containing third-order derivatives (due to the constitutive law of third-grade fluids), which distinguishes them from the Navier-Stokes or second-grade fluid equations.
• Control Objective: The goal is to find an optimal distributed random external force (control) $f$ that minimizes a cost functional $J$. The cost functional represents a velocity tracking problem, aiming to minimize the deviation of the fluid velocity $v$ from a desired target field $v_d$, while penalizing the magnitude of the control effort.
• Key Challenge: The primary mathematical difficulty arises from the strong nonlinearity and the presence of third-order derivatives in the nonlinear terms. Previous works on stochastic third-grade fluids (e.g., with multiplicative noise) could only establish local-in-time optimal control results due to the lack of global regularity estimates for the sample paths. This paper aims to solve the problem globally in time without additional restrictions on physical parameters.

2. Methodology

The authors employ a rigorous analytical framework combining stochastic analysis, functional analysis, and optimal control theory. The methodology proceeds through several critical steps:

A. Decomposition and Pathwise Deterministic Transformation

To handle the additive noise, the authors decompose the stochastic velocity field $v$ into two parts:
$$v = u + z$$

• $z$ (Ornstein-Uhlenbeck Process): A linear stochastic differential equation driven by the noise $W$. The authors prove that $z$ possesses high regularity (specifically $\dot{H}^5$ regularity) and satisfies crucial exponential integrability conditions (Eq. 1.5).
• $u$ (Random PDE): A pathwise deterministic partial differential equation with random coefficients (dependent on $z$). This transformation allows the application of deterministic PDE techniques to the stochastic system.

B. Global Well-Posedness of the State Equation

The authors establish the existence and uniqueness of a global-in-time strong solution to the transformed system for $u$.

• Regularity: They prove that the solution $v$ belongs to $L^p(\Omega; L^\infty(0, T; D(A^{3/2})))$, which corresponds to $H^3$ regularity in space.
• Exponential Integrability: A pivotal result is the proof that the solution satisfies:
  $$E\left[\exp\left\{\hat{c}\int_0^T \|v(s)\|_{\dot{H}^3}^2 ds\right\}\right] < +\infty$$
  This estimate is essential for handling the nonlinear terms and ensuring the well-posedness of the linearized and adjoint equations globally in time.

C. Linearized State and Adjoint Equations

To derive optimality conditions, the authors analyze the sensitivity of the state with respect to the control:

1. Linearized State Equation: They prove the existence and uniqueness of the solution to the linearized equation around the optimal trajectory. They show that the Gâteaux derivative of the control-to-state mapping coincides with this linearized solution.
2. Adjoint System: A backward stochastic PDE (the adjoint system) is formulated. The authors prove its well-posedness by transforming it into a forward problem via time reversal and applying similar regularity arguments used for the state equation.

D. Duality and Optimality Conditions

The core of the optimal control derivation relies on establishing a duality relationship between the linearized state equation and the adjoint equation. This duality allows the authors to express the derivative of the cost functional in terms of the adjoint state, leading to the formulation of first-order necessary optimality conditions.

3. Key Contributions

1. Global-in-Time Optimal Control: This is the first work to address the global-in-time optimal control problem for stochastic third-grade fluids. Previous studies were limited to local-in-time results due to regularity constraints.
2. Handling Additive Noise: The paper successfully converts the stochastic system into a pathwise deterministic one using an infinite-dimensional Ornstein-Uhlenbeck process, bypassing the difficulties associated with multiplicative noise in this specific context.
3. High Regularity Estimates: The authors derive novel exponential integrability estimates for the $H^3$ norm of the solution. This is a significant technical achievement that overcomes the analytical barriers posed by the third-order nonlinear terms.
4. Existence of Optimal Solution: They prove the existence of at least one optimal control pair $(\bar{f}, \bar{v})$ within the set of admissible controls.
5. Necessary Optimality Conditions: The paper rigorously derives the first-order necessary optimality conditions (variational inequalities) involving the adjoint state, providing a theoretical foundation for numerical algorithms.

4. Main Results

• Theorem 3.2: Proves the existence and uniqueness of a global strong solution to the stochastic third-grade fluid equation with additive noise, with sample paths in $L^\infty(0, T; D(A^{3/2}))$ and satisfying exponential integrability.
• Theorem 4.2 & 6.2: Establish the well-posedness (existence and uniqueness) of the linearized state equation and the adjoint system, respectively.
• Proposition 5.2: Demonstrates that the control-to-state mapping is Gâteaux differentiable, with the derivative given by the solution to the linearized equation.
• Theorem 2.5: The main result stating the existence of an optimal solution to the control problem and the validity of the first-order optimality conditions:
  $$E\left[\int_0^T (\psi(t) - \bar{f}(t), \bar{p}(t) + \nabla_f L(t, \bar{f}, \bar{v})) dt\right] \geq 0$$
  for all admissible controls $\psi$, where $\bar{p}$ is the adjoint state.

5. Significance

• Theoretical Advancement: The work bridges a significant gap in the mathematical theory of non-Newtonian fluids. By overcoming the "local-in-time" limitation, it opens the door to long-term control strategies for complex fluids.
• Industrial Applications: Third-grade fluids model various industrial materials (polymers, food processing, nanofluids). The ability to control these flows globally in time, even under stochastic perturbations (noise), is crucial for optimizing manufacturing processes, ensuring stability, and enhancing product quality in real-world scenarios.
• Methodological Impact: The technique of decomposing the system to handle additive noise and deriving specific exponential integrability estimates for high-order Sobolev norms provides a robust template for analyzing other complex stochastic fluid dynamics problems with strong nonlinearities.

In summary, this paper provides a comprehensive mathematical framework for the optimal control of stochastic third-grade fluids, proving that global-in-time solutions exist and that necessary optimality conditions can be rigorously derived, despite the inherent complexity of the governing equations.