Dimensional analysis and similitude

In short

Dimensional analysis groups the many physical variables of a flow problem into a few dimensionless numbers, so that an experiment on one model describes a whole family of real situations. Similitude is the set of conditions (geometric, kinematic and dynamic) that must hold for a scale model's readings to be transferable to the full-size prototype. It is the basis of every wind-tunnel, towing-tank and hydraulic-model test — spillways and ports at CWPRS Pune, ship hulls at NSTL Visakhapatnam, aerofoils at NAL Bengaluru, pump and turbine model acceptance tests at KBL, BHEL and Andritz. In exams it carries guaranteed marks in GATE/ESE (Buckingham pi, model laws, scale ratios) and in placement rounds it appears as "what is the Reynolds number and why does it matter".

Explained

Dimensions and dimensional homogeneity

• Every mechanical quantity reduces to the fundamental dimensions M (mass), L (length), T (time) — plus θ (temperature) in thermal work. The MLTθ system is standard; the FLT system uses force instead of mass.
• Dimensional homogeneity: every term of a physically correct equation has the same dimensions. Both sides of Bernoulli's equation, for example, are L (head form) or L²T⁻² (energy per unit mass).
• Uses: (i) check an equation or a derived result, (ii) convert units and change systems, (iii) build dimensionless groups, (iv) plan experiments — 4 independent variables needing 10 readings each means 10⁴ tests, but 2 pi terms need only ~10.
• Limitations: it tells you the *form* of the relation, never the constant or the function; the variables must be chosen correctly by physical insight; it cannot distinguish quantities with the same dimensions (work and torque, both ML²T⁻²).

Rayleigh's method and Buckingham's pi theorem

• Rayleigh's method: assume the dependent variable is a product of powers of the others, X = C·a^x·b^y·c^z, write the dimensional identity, equate the exponents of M, L and T and solve. Practical only up to about 4-5 variables.
• Buckingham's pi theorem: if a problem has n variables involving m fundamental dimensions, the relation can be written with (n − m) independent dimensionless pi terms.
• Steps: (1) list all n variables; (2) find m from the dimensional matrix (strictly, m = rank of the matrix); (3) choose m repeating variables that together contain all dimensions, are independent, and are not the dependent variable; (4) form each pi term as the repeating variables raised to unknown powers times one remaining variable; (5) solve exponents from homogeneity; (6) write φ(π₁, π₂, …) = 0.
• Choice of repeating variables: conventionally one geometric (L or D), one flow (V), one fluid property (ρ or μ). Avoid μ as repeating if viscosity is the effect under study, otherwise Reynolds number will not appear.
• Any pi term may be inverted, raised to a power or multiplied by another and remain valid — hence the same result can look different in two textbooks.

The important dimensionless numbers

• Each comes from the ratio of inertia force (ρL²V²) to one other force.
• Reynolds number Re = ρVL/μ = VL/ν — inertia/viscous. Pipe flow (laminar below ~2000, turbulent above ~4000 for pipes), submerged bodies, lubrication, low-speed aerodynamics.
• Froude number Fr = V/√(gL) — inertia/gravity. Free-surface flows: channels, spillways, weirs, ship waves, hydraulic jumps (Fr = 1 critical, <1 subcritical, >1 supercritical).
• Mach number Ma = V/c, c = √(K/ρ) — inertia/elastic. Compressible flow: above ~0.3 density changes matter; subsonic, sonic, supersonic, hypersonic.
• Euler number Eu = V/√(Δp/ρ) — inertia/pressure. Pressure-driven conduit flow, cavitation (its cousin is the cavitation number σ).
• Weber number We = ρV²L/σ — inertia/surface tension. Droplets, sprays, atomisation, capillary waves, very thin sheets over weirs.
• Others: Strouhal number St = fL/V (vortex shedding, flow-induced vibration), friction factor f, drag and lift coefficients, pressure coefficient, pump/turbine head and power coefficients, and in heat transfer Nu, Pr, Gr.

Similitude: the three similarities

• Geometric similarity: all corresponding lengths in the same ratio — the scale ratio Lr = Lm/Lp. Then areas scale as Lr², volumes as Lr³. Roughness must also scale, which is where real models struggle.
• Kinematic similarity: velocities and accelerations at corresponding points in the same ratio and in the same direction — so streamline patterns match. Requires geometric similarity first.
• Dynamic similarity: corresponding forces in the same ratio. Achieved by matching the governing dimensionless number; then force, pressure and power ratios follow automatically.
• Complete similarity needs *all* numbers matched, which is usually impossible (matching Re and Fr together in the same fluid requires Lr = 1). So the engineer identifies the dominant force and matches only that number — the other is accepted as a scale effect.
• Model = the scaled replica, prototype = the real machine or structure. Models may be smaller (ships, dams), equal or larger (a watch mechanism, a small valve).

Model laws and when each is used

• Reynolds model law — viscous force dominant, no free surface: pipes and ducts, fully submerged submarines, aerofoils at low speed, bearings, flow meters. Condition Re_m = Re_p.
• Froude model law — gravity with a free surface: spillways, stilling basins, weirs, notches, open channels, ship resistance (wave-making), hydraulic jump. Condition Fr_m = Fr_p.
• Mach model law — compressibility dominant: aircraft and missiles above Ma 0.3, compressors, water-hammer and gas pipelines.
• Weber model law — surface tension dominant: capillary rise, atomisers, very low heads over weirs, thin films.
• Euler model law — pressure dominant, friction and free surface negligible: flow through closed pipes under pressure, cavitation studies in pumps and turbines.
• Ship testing (Froude's hypothesis): total resistance = frictional + wave-making. The model is run at equal Fr; frictional resistance is calculated separately from Re-based friction formulae for both model and prototype and corrected — because both numbers cannot be matched at once.

Distorted models, scale effects and Indian practice

• A distorted model has different horizontal and vertical scales (e.g. 1:600 horizontal, 1:60 vertical) — used for rivers, estuaries, harbours and tidal models where an undistorted model's depth would be a few millimetres, making the flow laminar and surface tension significant.
• Advantages: measurable depths, turbulent flow retained, sediment movement possible, cheaper and smaller. Disadvantages: pressure and velocity distributions are not similar, slopes and waves are exaggerated, results need judgement and verification against field data.
• Scale effect: the error from the un-matched dimensionless numbers and from unscalable features — relative roughness, surface tension, air entrainment, cavitation. It shrinks as the model gets larger, hence "the biggest model the laboratory can afford".
• Indian practice: CWPRS, Khadakwasla (Pune) runs physical models for spillways, ports and river training for CWC, NHPC and state irrigation departments; NSTL, Visakhapatnam and IIT Madras/Kharagpur towing tanks for hulls; NAL/CSIR Bengaluru and ISRO wind tunnels for aerospace. Turbine and pump model acceptance tests follow IEC 60193 and the BIS/ISO pump test codes (editions are revised — verify the current one); model studies for dams are routinely required under CWC/Dam Safety Act 2021 review.
• Specific speed, the practical outcome of dimensional analysis for turbomachines — Ns = N√Q/H^3/4 for pumps, N√P/H^5/4 for turbines — lets one family of machines be designed and compared from a single model test.

Key points

• Fundamental dimensions - M, L, T (plus θ); all mechanical quantities are combinations of them.
• Dimensional homogeneity - every term in a valid equation has identical dimensions; the fastest exam check.
• Buckingham pi theorem - number of independent dimensionless groups = n − m (m = rank of the dimensional matrix).
• Repeating variables - m of them, must cover all dimensions, be independent and exclude the dependent variable.
• Rayleigh's method - power-product assumption; quick for few variables, clumsy beyond five.
• Reynolds number - inertia/viscous; governs pipes, submerged bodies and lubrication.
• Froude number - inertia/gravity; governs every free-surface flow and ship waves.
• Mach number - inertia/elastic; matters above about 0.3.
• Weber number - inertia/surface tension; drops, sprays, very thin sheets.
• Euler number - inertia/pressure; conduit flow and cavitation.
• Geometric similarity - same shape, one length scale, including roughness.
• Kinematic similarity - same velocity ratio everywhere, so identical streamline patterns.
• Dynamic similarity - same force ratios; obtained by matching the dominant dimensionless number.
• Scale ratio Lr - model length / prototype length; areas Lr², volumes Lr³.
• Froude scaling - Vr = √Lr, Tr = √Lr, Qr = Lr^2.5, force ratio = Lr³ (same fluid).
• Distorted model - unequal vertical and horizontal scales, for rivers and estuaries.
• Scale effect - unavoidable error from the dimensionless numbers that were not matched.
• Limitation - dimensional analysis gives the form of a relation, never the numerical constant.

Formulas and examples

Example 1 — Buckingham pi for drag. Drag F on a sphere depends on D, V, ρ, μ. n = 5, m = 3 → 2 pi terms. Repeating: ρ, V, D.

• π₁ = ρ^a V^b D^c F. Dimensions: (ML⁻³)^a(LT⁻¹)^b(L)^c(MLT⁻²) = M⁰L⁰T⁰ → M: a+1 = 0 → a = −1; T: −b−2 = 0 → b = −2; L: −3a+b+c+1 = 0 → 3−2+c+1 = 0 → c = −2. So π₁ = F/(ρV²D²).
• π₂ similarly with μ gives μ/(ρVD) = 1/Re.
• Result: F = ρV²D²·φ(Re), i.e. C_D is a function of Re alone — exactly what the standard C_D vs Re curve plots.

Example 2 — Rayleigh's method. Pressure-wave speed c = f(K, ρ): c = C·K^x·ρ^y → LT⁻¹ = (ML⁻¹T⁻²)^x(ML⁻³)^y. M: x + y = 0; T: −2x = −1 → x = ½, y = −½. So c = C√(K/ρ); experiment gives C = 1. For water (K = 2.07 GPa, ρ = 1000): c = √(2.07×10⁹/1000) ≈ 1,439 m/s.

Example 3 — Reynolds model law (oil pipeline). Prototype: D = 600 mm, V = 2 m/s, oil ν = 3.0×10⁻⁵ m²/s, ρ = 900 kg/m³. Model: D = 60 mm with water, ν = 1.0×10⁻⁶ m²/s, ρ = 1000 kg/m³.

• Re_p = 2 × 0.6 / 3.0×10⁻⁵ = 40,000.
• Equal Re: V_m = Re·ν_m/D_m = 40,000 × 1.0×10⁻⁶ / 0.06 = 0.667 m/s.
• Pressure coefficient equal: Δp_p = Δp_m (ρ_p/ρ_m)(V_p/V_m)². If the model reads Δp_m = 1.8 kPa over the test length, V_p/V_m = 2/0.667 = 3.0 → Δp_p = 1.8 × 0.9 × 9 = 14.58 kPa.

Example 4 — Froude model law (spillway). A 1:36 model of a dam spillway; prototype design discharge 1,800 m³/s, design head 7.2 m. Lr = 1/36.

• Head: H_m = 7.2/36 = 0.20 m.
• Discharge: Lr^2.5 = 36^2.5 = 36² × 6 = 7,776 → Q_m = 1,800/7,776 = 0.2315 m³/s ≈ 232 L/s.
• Velocity: V_m/V_p = 1/6; time: a surge taking 10 s on the model represents 60 s on the prototype.
• Force: F_m/F_p = Lr³ = 1/46,656, so a model gate thrust of 25 N implies 1,166 kN on the prototype.

Example 5 — Ship resistance. Prototype ship 150 m long at 10 m/s in sea water (ρ = 1,025 kg/m³); 1:25 model in fresh water (ρ = 1,000).

• Equal Fr: V_m = 10 × √(1/25) = 10/5 = 2 m/s.
• Wave-making resistance measured on the model = 15 N → F_p = 15 × (1,025/1,000) × 25³ = 15 × 1.025 × 15,625 = 240,234 N ≈ 240 kN.
• Re_m = 2×6/1.0×10⁻⁶ = 1.2×10⁷ against Re_p = 10×150/1.19×10⁻⁶ ≈ 1.26×10⁹ — far apart, so the frictional part must be computed separately (Froude's hypothesis).

Where it is used

• Hydraulic structures: CWPRS Pune and state irrigation research stations build Froude-scaled models of spillways, energy dissipators, canal falls and harbour layouts before construction of dams and ports.
• Turbomachinery: pump and turbine manufacturers (Kirloskar Brothers, BHEL, Andritz, Voith) test homologous model runners and scale head, discharge, power and efficiency using dimensionless coefficients and specific speed under IEC 60193.
• Automotive and aerospace: wind-tunnel scale models at NAL Bengaluru, ARAI Pune and OEM tunnels give C_D and C_L at matched Re or Ma; drag coefficient directly drives vehicle fuel economy and CAFE compliance.
• Ship and submarine design: NSTL Visakhapatnam, IIT Kharagpur and Cochin Shipyard towing tanks; hull resistance and propeller thrust predicted from Froude-scaled models.
• Process and chemical plants: scale-up of stirred vessels, mixers, cyclones, cooling-tower and chimney plume models (densimetric Froude based), and heat-exchanger correlations in Nu-Re-Pr form.
• Job roles: CFD/validation engineer, hydraulic design engineer, test and instrumentation engineer, R&D engineer — all read and generate dimensionless performance curves.

Interview questions

• Q: What is dimensional homogeneity? A: Every term of a physically valid equation must have the same dimensions; it is used to check derived formulas and to build dimensionless groups.
• Q: State Buckingham's pi theorem. A: A relation among n variables containing m fundamental dimensions can be reduced to a relation among (n − m) independent dimensionless pi terms.
• Q: How do you pick repeating variables? A: Take m of them that between them contain all the dimensions, are dimensionally independent, exclude the dependent variable, and conventionally are one length, one velocity and one fluid property.
• Q: What does the Reynolds number physically mean? A: The ratio of inertia force to viscous force; it decides laminar or turbulent flow and governs friction factor and drag coefficient.
• Q: Why is the Froude number used for spillway and ship models instead of Reynolds? A: In free-surface flows gravity waves dominate the behaviour, so matching gravity-to-inertia ratio reproduces the flow pattern; viscosity is secondary.
• Q: Can Reynolds and Froude similarity be satisfied together? A: Only if the scale ratio is 1 or an impractical fluid is available, because Re needs Vr = νr/Lr while Fr needs Vr = √Lr.
• Q: What are the three types of similarity? A: Geometric (shape and lengths), kinematic (velocities and streamlines) and dynamic (forces); dynamic similarity presupposes the other two.
• Q: What is a distorted model and why use one? A: A model with different horizontal and vertical scales, used for rivers, estuaries and harbours so that depths stay large enough to keep the flow turbulent and surface tension negligible.
• Q: What is a scale effect? A: The discrepancy between model and prototype caused by dimensionless numbers and features (roughness, surface tension, air entrainment) that could not be matched; it reduces with larger models.
• Q: Above what Mach number must compressibility be considered? A: About 0.3, where density changes exceed roughly 5% and incompressible relations start to err noticeably.
• Q: What are the limitations of dimensional analysis? A: It gives only the functional form, not the constants; it needs the right variables chosen by physical understanding; and it cannot separate quantities with identical dimensions.
• Q: Why is drag plotted as C_D versus Re? A: Dimensional analysis shows F/(ρV²D²) is a function of Re alone, so one curve covers all sizes, speeds and fluids.

Common mistakes

• Taking m as the count of dimensions appearing rather than the rank of the dimensional matrix — occasionally gives one pi term too few or too many.
• Choosing the dependent variable (or two variables of the same dimensions) as repeating variables, so the pi terms come out dependent or meaningless.
• Including μ as a repeating variable when viscous effects are the subject, which hides the Reynolds number.
• Applying Froude scaling to a submerged body or Reynolds scaling to a spillway — picking the wrong dominant force.
• Forgetting that force scales as Lr³ (Froude, same fluid) and velocity as √Lr, and simply scaling everything linearly.
• Mixing units — mm with m, poise with Pa·s, stokes with m²/s — so a "dimensionless" number comes out with units.
• Claiming a model test is exact; omitting to mention scale effect and the need for field verification.

Related topics

Viscous flow and pipe losses · Boundary layer, drag and lift · Fluid flow: Bernoulli and flow measurement · Hydraulic turbines · Centrifugal and reciprocating pumps · Compressible flow and gas dynamics