7.1 Characteristics of turbulent flows

Turbulent vs Laminar Flow

Turbulent and laminar flows are the two main flow regimes in fluid dynamics. In laminar flow, fluid moves in smooth, parallel layers with no mixing between them. In turbulent flow, the motion becomes chaotic and irregular, with strong mixing across the flow. The transition between these regimes depends on fluid velocity, viscosity, and the geometry of the flow domain.

Understanding which regime you're dealing with matters enormously: turbulent flows produce more friction, more mixing, and more energy loss than laminar flows. Nearly every engineering system involving fluid motion, from pipe networks to jet engines, requires you to account for turbulence.

Characteristics of Turbulent Flows

Highly Irregular Velocity Fluctuations

The defining feature of turbulent flow is rapid, seemingly random fluctuations in velocity. These fluctuations happen in all three spatial dimensions and change with time. They're superimposed on top of the mean flow velocity, creating a complex, chaotic velocity field.

This is what separates turbulence from laminar flow at a fundamental level. In laminar flow, the velocity profile is smooth and predictable. In turbulent flow, if you placed a sensor at a fixed point, you'd see the velocity jumping around erratically even though the average flow remains steady.

Chaotic and Random Motion

Fluid particles in turbulent flow follow chaotic trajectories. The motion is highly sensitive to initial conditions: a tiny perturbation can cause two nearby particles to end up in completely different locations a short time later. This sensitivity makes it essentially impossible to predict the exact path of any individual particle.

Because of this unpredictability, turbulence is described using statistical tools rather than deterministic equations. Probability density functions, correlation functions, and spectral analysis are the standard ways to characterize what's happening in a turbulent flow.

Enhanced Mixing and Diffusion

Turbulent flows mix momentum, heat, and mass far more effectively than laminar flows. The chaotic motion of fluid particles carries material across the flow domain much faster than molecular diffusion alone could manage.

This enhanced mixing is why turbulence matters so much in practice. Combustion engines need turbulence to mix fuel and air quickly. Chemical reactors rely on it to bring reactants into contact. Heat exchangers use it to boost thermal transport. In each case, the turbulent mixing rates can be orders of magnitude higher than what you'd get from molecular diffusion in a laminar flow.

Increased Friction and Energy Dissipation

Turbulent flows generate higher friction and dissipate more energy than laminar flows. The irregular velocity fluctuations create additional shear stresses (called Reynolds stresses) beyond the viscous stresses present in laminar flow. This leads to higher pressure drops and energy losses in pipes, channels, and around bodies.

The energy dissipation follows a specific pathway: kinetic energy transfers from large-scale motions down to progressively smaller scales through what's called the energy cascade. At the smallest scales, viscosity converts that kinetic energy into heat. This cascade process is central to how turbulence works and will come up repeatedly.

Transition from Laminar to Turbulent

Critical Reynolds Number

The transition between laminar and turbulent flow is governed by the Reynolds number ($Re$), a dimensionless ratio of inertial forces to viscous forces:

$Re = \frac{\rho U L}{\mu}$

where $\rho$ is fluid density, $U$ is a characteristic velocity, $L$ is a characteristic length, and $\mu$ is dynamic viscosity.

Each flow geometry has a critical Reynolds number ($Re_{cr}$) above which the flow transitions to turbulence. For internal pipe flow, $Re_{cr} \approx 2300$. For flow over a flat plate, transition typically occurs around $Re_x \approx 5 \times 10^5$. These values aren't absolute thresholds; they depend on the level of disturbances present in the flow.

Factors Affecting Transition

Several factors push a flow toward turbulence or help it stay laminar:

• Increasing velocity or decreasing viscosity raises $Re$, promoting transition
• Surface roughness introduces disturbances that can trigger turbulence at lower $Re$ than a smooth surface would
• Freestream disturbances such as vibrations, acoustic noise, or upstream obstacles can trip the flow into turbulence earlier
• Adverse pressure gradients (pressure increasing in the flow direction) destabilize the boundary layer and accelerate transition
• Flow curvature can either stabilize or destabilize the flow depending on the geometry

Controlling these factors is how engineers manipulate the transition point in applications like aircraft wing design or drag reduction.

Turbulent Boundary Layers

Velocity Profile in Turbulent Flows

The mean velocity profile in a turbulent boundary layer looks quite different from a laminar one. Near the wall, the velocity gradient is much steeper, meaning the fluid accelerates from zero (at the wall, due to the no-slip condition) to near-freestream values over a shorter distance. Further from the wall, the profile is more uniform because turbulent mixing redistributes momentum effectively.

The logarithmic law of the wall describes the mean velocity in the overlap region of the boundary layer:

$u^+ = \frac{1}{\kappa} \ln(y^+) + B$

where $u^+$ is the non-dimensionalized velocity, $y^+$ is the non-dimensionalized wall distance, $\kappa \approx 0.41$ is the von Kármán constant, and $B \approx 5.0$ for smooth walls.

Turbulent Boundary Layer Structure

A turbulent boundary layer has distinct regions, each with different physics:

• Viscous sublayer ($y^+ < 5$): Viscous stresses dominate. The velocity profile is nearly linear ($u^+ = y^+$).
• Buffer layer ($5 < y^+ < 30$): A transition zone where both viscous and turbulent stresses are significant. This is where turbulence production peaks.
• Log-law region ($30 < y^+ < 0.2\delta^+$): The overlap region where the logarithmic velocity profile holds. Turbulent stresses dominate.
• Wake region (outer layer): The outermost part of the boundary layer, where the velocity approaches the freestream value. The profile deviates from the log law and is described by Coles' wake function.

Boundary Layer Separation in Turbulence

Boundary layer separation occurs when the flow detaches from a surface, typically due to an adverse pressure gradient. Turbulent boundary layers resist separation much better than laminar ones. The reason is straightforward: turbulent mixing brings high-momentum fluid from the outer region down toward the wall, giving the near-wall fluid enough energy to push against the rising pressure.

This is why a golf ball has dimples. The dimples trip the boundary layer into turbulence, which delays separation and reduces the size of the wake behind the ball, cutting drag significantly. The same principle applies in flow control techniques like vortex generators on aircraft wings, which energize the boundary layer to prevent or delay separation.

Turbulence Scales and Energy Cascade

Energy-Containing Eddies

Turbulent flows contain eddies spanning a huge range of sizes. The largest eddies, called energy-containing eddies, hold most of the turbulent kinetic energy. Their size is characterized by the integral length scale ($L$), which is typically comparable to the dimensions of the flow domain (pipe diameter, boundary layer thickness, etc.).

These large eddies extract energy from the mean flow through a process called vortex stretching. They're anisotropic, meaning their shape and orientation depend on the geometry and boundary conditions of the specific flow.

Inertial Subrange and Kolmogorov Scale

Between the largest and smallest eddies lies the inertial subrange, where energy transfers from larger eddies to smaller ones without significant viscous dissipation. In this range, the energy spectrum follows Kolmogorov's five-thirds law:

$E(\kappa) \propto \varepsilon^{2/3} \kappa^{-5/3}$

where $E(\kappa)$ is the energy spectrum, $\varepsilon$ is the dissipation rate, and $\kappa$ is the wavenumber. This is one of the most well-established results in turbulence theory.

At the bottom of the cascade sit the smallest eddies, characterized by the Kolmogorov length scale:

$\eta = \left(\frac{\nu^3}{\varepsilon}\right)^{1/4}$

where $\nu$ is the kinematic viscosity and $\varepsilon$ is the dissipation rate. These are the scales at which viscosity finally acts to convert kinetic energy into heat.

Energy Dissipation at Small Scales

The energy cascade is a one-way street: energy flows from large scales to small scales, never the reverse (in the mean). At the Kolmogorov scales, the velocity gradients are steep enough for viscosity to dissipate the kinetic energy as heat.

A key result is that the dissipation rate $\varepsilon$ is set by the large scales, not the small ones. The large eddies determine how much energy enters the cascade; the small eddies simply dissipate whatever arrives. In equilibrium turbulence, the production rate of turbulent kinetic energy equals the dissipation rate, which is a common assumption in many turbulence models.

The ratio of the largest to smallest scales grows with Reynolds number: $L/\eta \sim Re^{3/4}$. This is why high-Reynolds-number turbulence is so computationally expensive to simulate directly.

Statistical Description of Turbulence

Turbulent Velocity Fluctuations

Since turbulence is inherently random, it's described statistically. The standard approach is Reynolds decomposition, which splits any instantaneous quantity into a time-averaged (mean) component and a fluctuating component:

$u(x,t) = \bar{u}(x) + u'(x,t)$

where $\bar{u}$ is the mean velocity and $u'$ is the fluctuation. By definition, the time average of the fluctuation is zero: $\overline{u'} = 0$. But the average of $u'^2$ is not zero, and this quantity relates directly to the turbulent kinetic energy.

The root-mean-square (RMS) velocity $u'_{rms} = \sqrt{\overline{u'^2}}$ quantifies the typical magnitude of the fluctuations.

Probability Density Functions

Probability density functions (PDFs) describe the statistical distribution of velocity fluctuations at a point. For many turbulent flows, the single-point velocity PDF is approximately Gaussian (normal distribution), but deviations from Gaussian behavior reveal important physics.

• Intermittency shows up as heavy tails in the PDF, meaning extreme fluctuations occur more often than a Gaussian distribution would predict.
• Skewness of the PDF indicates asymmetry in the fluctuations, which can signal preferred directions of momentum transport.
• Joint PDFs describe correlations between different velocity components or between velocity and other quantities like temperature or concentration.

Turbulence Intensity and Length Scales

Turbulence intensity ($I$) measures how strong the fluctuations are relative to the mean flow:

$I = \frac{u'_{rms}}{\bar{U}}$

Typical values range from about 1% in a clean wind tunnel to 10-20% in atmospheric flows. Higher turbulence intensity means stronger fluctuations relative to the mean.

Turbulent length scales characterize the size of eddies at different points in the cascade:

• Integral length scale ($L$): Size of the largest, energy-containing eddies. Determined from the autocorrelation of velocity fluctuations.
• Taylor microscale ($\lambda$): An intermediate scale related to the mean strain rate of the turbulence. It's not a dissipation scale, but it's useful for estimating the Reynolds number of the turbulence itself ($Re_\lambda$).
• Kolmogorov microscale ($\eta$): The smallest scale, where dissipation occurs.

Turbulent Flow Modeling Approaches

Direct Numerical Simulation (DNS)

Direct Numerical Simulation resolves every scale of turbulence by solving the Navier-Stokes equations on a grid fine enough to capture the Kolmogorov scales, with time steps small enough to resolve the fastest fluctuations. No turbulence model is needed.

The cost is enormous. The number of grid points scales as $Re^{9/4}$ in three dimensions, which limits DNS to relatively low Reynolds numbers (typically $Re \sim 10^3$ to $10^4$). DNS is primarily a research tool used to study turbulence physics and to generate benchmark data for validating other models.

Large Eddy Simulation (LES)

Large Eddy Simulation directly resolves the large, energy-containing eddies and models only the small-scale eddies using a subgrid-scale (SGS) model. The Navier-Stokes equations are spatially filtered to remove scales below the grid resolution.

LES captures the unsteady, three-dimensional structure of turbulence much better than RANS models, making it more accurate for flows with separation, strong curvature, or other complex features. The computational cost is significantly less than DNS but still substantial, especially near walls where the turbulent scales become very small.

Reynolds-Averaged Navier-Stokes (RANS) Models

RANS models solve the time-averaged Navier-Stokes equations and model all turbulent fluctuations using closure models. This makes them by far the cheapest approach computationally, and they're the workhorse of industrial CFD.

The averaging process introduces unknown terms called Reynolds stresses ($\overline{u'_i u'_j}$), which must be modeled. The most common approaches are:

• $k$-$\varepsilon$ model: Solves transport equations for turbulent kinetic energy ($k$) and its dissipation rate ($\varepsilon$). Good for free-shear flows; less accurate near walls without modifications.
• $k$-$\omega$ model: Solves for $k$ and the specific dissipation rate ($\omega$). Better near-wall behavior than standard $k$-$\varepsilon$.
• SST $k$-$\omega$ model: Blends $k$-$\omega$ near walls with $k$-$\varepsilon$ in the freestream. Widely used for aerodynamic flows.

RANS models work well for many attached, steady flows but struggle with massively separated flows, strong streamline curvature, and unsteady phenomena.

Turbulence in Engineering Applications

Turbulent Flow in Pipes and Channels

Most pipe and channel flows in engineering operate in the turbulent regime. Water flowing through a typical household pipe at moderate velocity already exceeds $Re_{cr} \approx 2300$.

Turbulence increases friction losses, quantified by the Darcy-Weisbach equation:

$\Delta p = f \frac{L}{D} \frac{\rho U^2}{2}$

where $f$ is the friction factor, $L$ is pipe length, $D$ is diameter, and $U$ is mean velocity. The friction factor depends on $Re$ and relative wall roughness, and is commonly read from the Moody chart. Wall roughness, flow obstructions, and changes in cross-section all influence the turbulent friction losses.

Turbulence in Aerodynamics and Wind Engineering

Turbulence has a major impact on aerodynamic performance. On aircraft wings, the state of the boundary layer (laminar vs. turbulent) directly affects drag, and turbulent separation can lead to stall. Tripping the boundary layer to turbulence is sometimes deliberately done to delay separation, trading a small increase in skin friction for a large reduction in pressure drag.

Wind turbines operate in the highly turbulent atmospheric boundary layer, where turbulence intensity can reach 15-20%. This turbulence affects both power output and fatigue loads on the blades. In wind engineering for buildings and structures, turbulent fluctuations determine peak wind loads, and turbulent dispersion governs how pollutants and heat spread in urban environments.

Turbulent Mixing in Combustion and Chemical Processes

Efficient combustion depends on rapid mixing of fuel and oxidizer, which turbulence provides. In gas turbine combustors, for example, turbulent mixing controls flame stability, combustion efficiency, and pollutant formation. Without turbulence, the mixing would rely on molecular diffusion alone, which is far too slow for practical combustion rates.

In chemical reactors, turbulent mixing increases contact between reactants and catalyst surfaces, boosting reaction rates and product yields. The design of these systems often centers on controlling the turbulence level and mixing patterns to achieve the desired performance while avoiding problems like incomplete mixing or hot spots.