🔧 Quality Security Quality Simulation

Watch a production line, catch the drift before parts go out of tolerance.

Controls

Live readout

Independent checks

The science

Model. Part diameters are normal with mean μ = 10.000 mm and σ = 0.05 mm (spec 10.00 ± 0.15 mm, so Cp = 1.00). Every second an inspector measures a subgroup of n = 5 parts. Control limits come from the first 25 in-control subgroups: UCL/LCL = X̄̄ ± A2R̄ (A2 = 0.577) and R-chart UCL = D4R̄ (D4 = 2.114).

Quality maths. Capability: Cp = (USL − LSL)/6σ, Cpk = min(USL − μ, μ − LSL)/3σ; scrap rate = Φ((LSL − μ)/σ) + 1 − Φ((USL − μ)/σ). Average run length of an X̄ chart with 3-sigma limits after a shift of δ sigmas: ARL = 1/p, p = Φ(−3 − δ√n) + 1 − Φ(3 − δ√n). The button runs a Monte Carlo test of exactly that formula.

Why it matters for security-critical parts. Locks, seals and fasteners must be inside tolerance every time; a chart catches a drifting process long before the scrap appears. Shifts of 1.5σ take about 1.6 subgroups to be caught, 1σ about 4.

Honest simplification. Normal, independent measurements, no autocorrelation, no gauge error, only the simplest alarm rules (a point beyond a limit); real plants also use run rules and CUSUM/EWMA charts.