Coherent structures modeling in stenotic transitional flow via resolvent analysis

This study demonstrates that linear resolvent analysis based on mean flow data from a large-eddy simulation can effectively characterize the transitional dynamics in an axisymmetric stenosis, successfully reconstructing turbulent kinetic energy and wall shear stress fluctuations through low-rank, axisymmetric coherent structures.

The Gist

The Big Picture: A Clogged Pipe and a Flowing River

Imagine blood flowing through an artery like water rushing through a garden hose. Now, imagine someone pinches the hose in the middle, creating a narrow "choke point" (this is called a stenosis).

When water (or blood) squeezes through this pinch, it speeds up. But once it exits the narrow part and enters the wider pipe again, things get chaotic. It doesn't just flow smoothly; it swirls, wobbles, and creates turbulent eddies. This turbulence creates a "shaking" force against the pipe walls. In the human body, this shaking can damage the artery walls or cause dangerous clots.

This paper asks a simple question: Can we predict this chaotic shaking using simple math, without needing to simulate every single drop of water?

The Problem: Too Much Data, Too Little Understanding

Scientists have powerful computers that can simulate blood flow in incredible detail (like taking a high-speed video of every water molecule). However, these simulations produce mountains of data that are hard to read. It's like trying to understand a symphony by listening to a recording of every single instrument playing at once—you hear the noise, but you can't easily pick out the melody.

Furthermore, real medical scans (like MRI) don't give us that much detail; they only give us a "blurry" average of the flow. The researchers wanted to know: Can we use a simple, linear model based on that "blurry" average to predict the complex, chaotic shaking?

The Solution: The "Amplifier" Analogy

The researchers used a method called Resolvent Analysis. Think of the blood flow as a giant, complex sound amplifier.

1. The Input (The Noise): In a real artery, there are tiny, random jitters in the blood flow (like static noise on a radio).
2. The System (The Amplifier): The shape of the artery (the pinch) acts as the amplifier. It takes those tiny jitters and makes them huge.
3. The Output (The Shaking): The result is the big, chaotic waves that hit the artery wall.

The researchers' goal was to figure out exactly how this amplifier works. They wanted to know: "If we feed in a tiny wobble, what kind of big wave comes out?"

What They Found: Two Types of "Wobbles"

By analyzing the flow, they discovered two distinct ways the artery amplifies chaos:

1. The "Lazy Sway" (Low Frequency)
At very slow speeds, the flow acts like a pendulum. The researchers found a specific "stationary" mode (a wobble that doesn't spin) that makes the jet of blood sway back and forth, sticking to one side of the artery wall. This is similar to a Coanda effect, where a stream of water clings to a curved surface. This sway happens at very low frequencies and is driven by a specific instability in the flow.

2. The "Rolling Drum" (Medium Frequency)
At faster speeds, the flow behaves like a drum skin being hit. The researchers found that the most powerful waves are axisymmetric—meaning they roll around the pipe like a perfect ring or a donut, rather than twisting like a corkscrew.

• Why this is surprising: Previous studies on similar flows at lower speeds found that the "corkscrew" twists were the strongest. But at this higher speed (Reynolds number 4000), the "rolling rings" are the winners.
• These rings form in the shear layer (the boundary between the fast jet and the slow swirling water) and grow until they break apart into chaos.

The "Magic Trick": Predicting Chaos from a Blur

The most impressive part of the study is what they did next. They took the average flow (the "blurry" picture, like a long-exposure photo where the motion is smoothed out) and used their "amplifier" math to predict the chaos.

• The Test: They compared their simple math model against the "high-definition" computer simulation (the full video).
• The Result: The simple model was shockingly accurate.
  • It correctly predicted where the energy (the shaking) would be strongest.
  • It correctly predicted that the "rolling rings" (axisymmetric waves) were the main culprits in the area just after the pinch.
  • It could even reconstruct the Turbulent Wall Shear Stress (tWSS). Think of this as the "friction" the blood exerts on the artery wall. The model showed that just a few simple wave patterns could explain more than half of this friction in the immediate aftermath of the pinch.

The Conclusion: A Simple Map for a Complex World

The paper concludes that even though blood flow in a clogged artery looks incredibly messy and chaotic, it is actually driven by a few simple, predictable rules.

By treating the artery as a linear amplifier, the researchers showed that you don't need a supercomputer to understand the dangerous shaking forces. You just need the "average" flow data. This is like being able to predict the shape of a storm's waves just by looking at the average wind speed, without needing to track every single raindrop.

In short: The chaotic turbulence behind a narrowed artery is not random noise; it is a structured response to the shape of the artery. The researchers built a simple "decoder" that can read the average flow and accurately predict the dangerous shaking forces that threaten our health.

Technical Summary

Technical Summary: Coherent Structures Modeling in Stenotic Transitional Flow via Resolvent Analysis

Problem Statement
The transition to turbulence in blood flow past arterial stenoses generates wall shear stress (WSS) fluctuations that significantly influence cardiovascular disease progression, including plaque rupture and endothelial dysfunction. While computational fluid dynamics (CFD) provides high-resolution flow data, extracting physical insights and dominant coherent structures from complex, time-resolved datasets remains challenging. Furthermore, data-driven methods often lack physical interpretability and require high-resolution temporal data, limiting their applicability to clinical techniques like 4D-flow MRI, which have constrained resolution. There is a current gap in the linear modeling of the forced dynamics in stenotic flows at Reynolds numbers beyond the transition threshold, specifically regarding the amplification mechanisms driving turbulent stresses.

Methodology
The study investigates an axisymmetric stenosis with a 75% area reduction at a Reynolds number of 4000 ($Re=4000$), representative of peak systolic conditions in the aorta. The methodology integrates three primary components:

1. Large-Eddy Simulation (LES): High-fidelity flow data was generated using the $k-\omega$ SST detached-eddy simulation (DES) model in OpenFOAM. The mean flow was obtained via time- and azimuthal-averaging of the LES snapshots.
2. Linear Modeling Framework:
   • Global Linear Stability Analysis (LSA): Performed on the mean flow to identify intrinsic eigenmodes. To account for turbulent dissipation, a RANS-like eddy viscosity ($\nu_t$) was assimilated into the mean flow using a Physics-Informed Neural Network (PINN) that solves the Reynolds-averaged Navier-Stokes equations.
   • Resolvent Analysis (RA): Used to characterize the input-output response of the linearized system to harmonic forcing. The nonlinear terms were treated as external forcing. To ensure the system was asymptotically stable for RA, the eddy viscosity was tuned to stabilize the unstable stationary mode identified in LSA.
   • Pseudospectral Analysis: Employed to distinguish between modal resonance and non-normal pseudoresonance effects.
3. Data-Driven Validation: Spectral Proper Orthogonal Decomposition (SPOD) was applied to the LES snapshots to identify the most energetic spatiotemporal coherent structures. The linear model's predictions were validated against SPOD modes, and the alignment between optimal resolvent response modes and leading SPOD modes was quantified.

Key Contributions and Results

• Identification of Amplification Mechanisms:

  • Low Frequencies ($St < 0.01$): LSA revealed an unstable, stationary eigenmode ($m=1$) associated with a weak Coanda-type wall attachment. Resolvent analysis confirmed this mode drives low-frequency amplification, acting as a low-pass filter.
  • Intermediate Frequencies ($St \in [0.1, 1]$): RA identified a second, dominant amplification region within the shear layer. Contrary to findings at lower Reynolds numbers where sinuous waves ($m=1$) dominate, the most amplified fluctuations at $Re=4000$ are axisymmetric ($m=0$). The peak amplification occurs at $St \approx 0.59$.
  • Nature of Instability: Pseudospectral analysis demonstrated that the broadband amplification in the shear layer arises from strong pseudoresonance driven by non-normal interactions among multiple damped eigenvalues, rather than a single isolated modal resonance. This confirms the convective nature of the instability.

• Validation and Alignment:

  • The linear model was validated against SPOD results. In the shear-layer region ($St \in [0.1, 1]$), the optimal resolvent response modes demonstrated high gain separation and strong spatial alignment with the leading SPOD modes for $m \leq 3$.
  • The study noted that while the RA model assumes white-noise forcing, the SPOD spectrum exhibited sharp peaks at the vortex shedding frequency ($St=0.48$) and its harmonics. This discrepancy suggests that in transitional flows, correlated forcing couples with the natural receptivity of the damped shear-layer eigenmode to sharpen the output spectrum.

• Low-Order Reconstruction of Turbulent Statistics:

  • Turbulent Kinetic Energy (TKE): The study successfully reconstructed TKE using a rank-1 assumption based on the leading resolvent mode. In the shear layer, the optimally scaled RA model closely matched SPOD results. Even under a white-noise forcing assumption (without time-resolved data), the model qualitatively captured the spatial distribution of high TKE regions.
  • Turbulent Wall Shear Stress (tWSS): The relationship between near-wall TKE and tWSS was leveraged to reconstruct tWSS. In the immediate post-stenotic zone ($x/d_t < 4.3$), axisymmetric fluctuations ($m=0$) dominated the tWSS, with the leading SPOD mode capturing over 50% of the $m=0$ contribution. Downstream, sinuous fluctuations ($m=\pm 1$) became dominant. The RA model successfully captured the tWSS peaks in the shear-layer region for both $m=0$ and $m=1$, even with white-noise forcing assumptions. However, reconstructing the total tWSS magnitude required summing contributions from the first eleven azimuthal modes to reach 90% of the total energy.

Significance
The paper claims that linear mechanisms effectively capture the complex post-stenotic dynamics, providing physically interpretable insights into flow physics that data-driven methods alone cannot offer. A primary significance lies in the successful reconstruction of turbulent quantities (TKE and tWSS) using only the time-averaged mean flow data. This capability suggests that resolvent analysis could serve as a predictive tool for key turbulent quantities in clinical settings where only mean flow data is available, such as 4D-flow MRI, thereby opening new possibilities for physics-informed diagnostic tools without the need for high-resolution, time-resolved simulations or measurements.