A single strain gauge measures strain along one axis. That's fine when you know the load direction — a tension rod, a beam in pure bending. But real parts see combined loading: a pressure vessel nozzle, a bracket weld, a crankshaft fillet. You don't know where the maximum stress points, and a gauge rotated 30° off the principal axis reads a fraction of the true value. The fix is a strain gauge rosette: three gauges bonded at known angles on the same backing, letting you solve for the full 2D strain state.

Two standard geometries dominate:

• Rectangular (0°/45°/90°) — three gauges at right-angle increments. Easy math, most common for general stress analysis.
• Delta (0°/60°/120°) — equally spaced. Better numerical conditioning when principal directions are truly unknown; preferred for fatigue work where small errors compound.

The math (rectangular rosette). Given measured strains ε₀, ε₄₅, ε₉₀:

• Principal strains: ε₁,₂ = (ε₀ + ε₉₀)/2 ± (1/√2)·√[(ε₀ − ε₄₅)² + (ε₄₅ − ε₉₀)²]
• Principal angle: tan(2θ) = (2ε₄₅ − ε₀ − ε₉₀) / (ε₀ − ε₉₀)
• Convert to stress via Hooke's law for plane stress: σ₁ = E/(1−ν²) · (ε₁ + ν·ε₂)

Real-world example. A pipeline company investigating a cracked elbow bonds a rectangular rosette to the outer surface near the crotch. Readings under operating pressure: ε₀ = 480 µε, ε₄₅ = 720 µε, ε₉₀ = 240 µε. Plugging in gives ε₁ ≈ 820 µε at roughly 34° from the pipe axis — nowhere near the hoop or axial direction anyone would have instrumented with a single gauge. With E = 200 GPa and ν = 0.3, σ₁ works out to about 190 MPa, well above what a naive hoop-stress calculation predicted. The crack orientation matched the principal direction the rosette identified.

Practical rules:

• Bond quality dominates accuracy — a poor adhesive joint reads low and drifts with temperature. Surface prep (degrease, abrade, neutralize, condition) is non-negotiable.
• Use a dummy gauge on unstressed material of the same alloy for temperature compensation, or specify self-temperature-compensated (STC) gauges matched to your substrate.
• Rosettes come stacked (three grids on top of each other, small footprint but poor heat dissipation) or planar (side-by-side, larger area but better thermal behavior and no strain gradient between grids). Planar is preferred unless space forces stacked.
• Wheatstone bridge completion: each grid gets its own quarter-bridge circuit; you cannot share a bridge across a rosette.