Loading...
Master design of experiments with practical DOE engineering examples including factorial, fractional factorial, Plackett-Burman, and response surface methodology. Optimize processes and ace ASQ exams.
Free practice questions for every ASQ certification.
Create a free account and start practicing in under 60 seconds. No credit card required.
Injection molding engineers constantly face the challenge of balancing multiple process parameters — temperature, injection pressure, cooling time, hold time — to minimize defects like warpage. Trial-and-error approaches are costly and time-consuming, often missing interaction effects that significantly impact quality.
Design of Experiments (DOE) offers a structured, efficient way to explore these parameters simultaneously, revealing main effects and interactions with fewer runs than a full trial matrix. Yet, engineers often struggle to choose the right DOE approach: full factorial, fractional factorial, Plackett-Burman, or escalate to response surface methodology (RSM).
This article walks through these DOE types with a practical injection molding example, illustrating key concepts, trade-offs, and analysis techniques engineers need to optimize their processes and prepare for ASQ certifications like CSSBB and CQE.
Ignoring DOE principles can lead to:
For engineers targeting ASQ certifications, DOE forms a significant portion of the body of knowledge (BoK) for CSSBB and CQE exams. Proficiency in DOE not only drives process excellence but also signals mastery in quality engineering disciplines.
A full factorial design tests all possible combinations of k factors each at two levels (commonly coded as -1 and +1). This design captures all main effects and interactions but grows exponentially with factors.
For example, with 4 factors (temperature, pressure, cooling time, hold time), the runs are:
Runs = 2^4 = 16
Factors = {T, P, Ct, Ht}
Levels = {low (-1), high (+1)}
This design enables estimation of:
A fractional factorial design uses a fraction of the full factorial runs to save time and cost, accepting some confounding (aliasing) of effects. For example, a half-fraction (2^(4-1) = 8 runs) can estimate main effects and some interactions but aliases certain effects together.
Alias structure depends on design resolution:
Selecting resolution depends on prior knowledge and resource constraints.
Primarily used for screening many factors to identify the few that significantly impact the response. They are highly fractional with minimal runs (multiples of 4), but only estimate main effects assuming interactions are negligible.
Once significant factors are identified, RSM explores the relationship between variables and response in more detail to find optimal settings.
Common RSM designs:
These allow modeling curvature and interaction effects to optimize process parameters precisely.
Minimize warpage in a plastic part molded with these factors:
| Factor | Low (-1) | High (+1) |
|---|---|---|
| Temperature (°C) | 220 | 260 |
| Pressure (MPa) | 50 | 70 |
| Cooling Time (s) | 15 | 30 |
| Hold Time (s) | 5 | 15 |
| Run | T | P | Ct | Ht | Warpage (mm) |
|---|---|---|---|---|---|
| 1 | -1 | -1 | -1 | -1 | 0.45 |
| 2 | +1 | -1 | -1 | -1 | 0.38 |
| ... | |||||
| 16 | +1 | +1 | +1 | +1 | 0.29 |
Note: Hypothetical warpage values from experimental runs.
Using ANOVA, we quantify the significance of each factor and interaction:
| Source | DF | SS | MS | F | p-value |
|---|---|---|---|---|---|
| Temperature | 1 | 0.032 | 0.032 | 18.4 | 0.002 |
| Pressure | 1 | 0.015 | 0.015 | 8.6 | 0.015 |
| Cooling Time | 1 | 0.005 | 0.005 | 2.9 | 0.11 |
| Hold Time | 1 | 0.003 | 0.003 | 1.2 | 0.3 |
| T*P | 1 | 0.008 | 0.008 | 4.6 | 0.05 |
| Residual | 8 | 0.014 | 0.002 |
Temperature and pressure are significant main effects; their interaction is borderline significant.
Plotting main effects shows warpage decreases with increasing temperature and pressure. Interaction plots reveal combined influence of T and P is non-additive.
Residuals appear randomly scattered without patterns, confirming model adequacy.
To reduce runs, a half-fraction design aliasing some 2-factor interactions is used. The engineer must accept that estimation of some interactions is confounded.
If 8+ factors existed, a 12-run Plackett-Burman design would screen for critical factors before detailed study.
Using CCD with center and axial points, engineer models quadratic effects to find the optimal temperature and pressure settings that minimize warpage.
The RSM model might be:
Optimization algorithms then identify the settings minimizing warpage.
Tip: Always start with screening or fractional factorial designs to identify critical factors before investing in complex RSM.
DOE is a cornerstone topic in both the CSSBB and CQE exam BoKs:
Mastering DOE concepts and practical application through examples like injection molding warpage prepares candidates to tackle exam questions and real-world problems with confidence.
If you're ready to formalize this expertise into a credential employers respect, our CSSBB and CQE courses cover DOE and the rest of the body of knowledge — see our certification programs to get started today.
Written by
John Lee
Founder & Lead Instructor, Alpha Training & Consulting
John Lee is the founder of Alpha Training & Consulting, holds 19 ASQ certifications, an MBA in Quality Systems, and a B.S. in Mechanical Engineering. He is a Shingo Prize-winning author and has trained over 2,500 engineers and quality professionals across 25+ years, with students achieving a 93% pass rate on ASQ certification exams.
Create your free ASQ Exam Prep account to access sample questions, BOK domain tracking, and your personalized study dashboard. Upgrade only when you're ready for the full 15,000-question library and John Lee's certification courses.