Design of Experiments terminology

Learning objectives

• Explain the purpose of design of experiments terminology in a garment-factory improvement project.
• Apply the described method to a representative shop-floor situation.
• Recognize the common mistakes and how to avoid them.

Concept

DOE changes multiple inputs through a planned design to estimate effects efficiently and discover interactions.

Randomization protects against time bias; replication estimates error; blocking controls known nuisance variation; center points reveal curvature.

DOE differs from trial-and-error by protecting inference through structure, not intuition.

Variables, units, and assumptions

Factors (X, engineering units), response (Y, e.g. shade ΔE), run order (integer).

Independent runs, stable measurement system, factor levels feasible and safe.

Free-spreadsheet workflow

In a free spreadsheet, list all combinations, randomize with =RAND(), execute in that order, record Y, compute effects as difference of means at high/low levels.

Interpretation and limitations

An effect is meaningful when it exceeds noise (2× SD of replicates) AND is practically important.

Conclusions apply only within tested factor ranges and observed noise conditions.

Garment-factory example

Experimental units are fabric panels processed under planned combinations of laser intensity, speed, and wash time; shade delta is the response.

Method

1. State problem and response.
2. Identify factors and noise.
3. Choose design.
4. Randomize.
5. Execute consistently.
6. Analyze effects.
7. Confirm with holdout runs.

Common mistakes

• Changing one factor at a time and calling it a DOE.
• Confusing repeated readings on one panel with independent replication.