Engineering Design and Analysis
Engineering Design and Analysis tests the chain from need to proven design: measurable requirements, concept selection against explicit criteria, sizing for the governing failure mode, and a defined verification path.
Core rules
- Requirements are design inputs and must be measurable and traceable - 'withstand 12 bar for 10 000 cycles', not 'robust'. Outputs (drawings, calculations, specifications) map back to each input.
- Identify the governing failure mode before sizing: yield (σ ≤ S_y/n), fatigue (S–N curve, endurance limit, Goodman line for combined mean and alternating stress), buckling (Euler P_cr = π²EI/(KL)², K set by end conditions), creep at temperature, corrosion allowance, brittle fracture.
- Tolerance stack-up: worst-case arithmetic sum guarantees assembly; root-sum-square predicts scrap rate and must never be used where a guaranteed fit or safety clearance is required.
- DFMEA scores risk as RPN = Severity × Occurrence × Detection, but severity is acted on first; a low-RPN high-severity item still needs control.
- Verification (does it meet the specification - test, analysis, inspection) is distinct from validation (does it meet the user need).
Worked example
Four stacked parts have tolerances ±0.10, ±0.15, ±0.05 and ±0.10 mm on a 40 mm stack. Worst case: ±(0.10+0.15+0.05+0.10) = ±0.40 mm, giving 39.60–40.40 mm. RSS: √(0.10²+0.15²+0.05²+0.10²) = √0.045 = ±0.21 mm. Use ±0.40 to guarantee every assembly closes; the ±0.21 figure only describes the probable spread and predicts defect rate.
Traps
Treating RSS as a guaranteed limit; omitting the effective-length factor K in buckling; ignoring thermal growth ΔL = αLΔT in tolerance chains; reporting validation as verification; sizing on average load instead of the specified peak.