Robust Central Composite Design
Robust Central Composite Design for Response Surface Methodology · Also known as: Robust CCD, CCD with robust optimization, robust RSM with CCD, robust response surface CCD
Robust Central Composite Design (Robust CCD) combines the efficient quadratic fitting capability of the central composite design with robust optimization principles to find factor settings that simultaneously achieve a target mean response and minimize the effect of uncontrollable noise factors on response variability. It is widely applied in manufacturing, chemical engineering, and product development when both performance and consistency under real-world variation are critical.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use Robust CCD when you need to fit a full quadratic response surface (requiring second-order terms) and simultaneously reduce sensitivity to noise or uncontrollable variation. It is appropriate when the number of controllable factors is between 2 and 6, a continuous response variable is available, and both the mean performance and the variability of the response matter. Prefer it over a standard CCD when field conditions introduce significant noise that cannot be controlled during production. Do not use it if only linear effects are expected (a factorial design suffices), if the experimental region has constraints that preclude axial points (use Box-Behnken instead), if budget limits rule out the center-point replicates needed to estimate variance, or if noise factors cannot be identified or quantified.
Strengths & limitations
- Simultaneously optimizes mean performance and minimizes variability in a single experimental framework.
- The CCD structure supports full second-order modeling with relatively few runs compared to a full factorial at three levels.
- Rotatable or near-rotatable designs provide uniform prediction quality throughout the factor space.
- Widely supported by statistical software (JMP, Minitab, Design-Expert, R rsm package), making implementation practical.
- Applicable across a broad range of engineering and scientific domains wherever both performance and robustness matter.
- Requires identification and quantification of noise factors; if these cannot be measured or controlled in the experiment, the robust variance model cannot be built.
- With more than five or six controllable factors the number of CCD runs grows rapidly; fractional CCD variants reduce runs but sacrifice resolution.
- The quadratic model assumes a smooth, continuous response surface; if the true surface has sharp discontinuities or multiple distinct optima, the model may be misleading.
- Center-point replicates are essential for pure-error estimation; omitting them prevents reliable variance modeling.
Frequently asked
How does Robust CCD differ from the standard CCD?
A standard CCD fits a quadratic surface and finds the factor settings that optimize the mean response. Robust CCD adds an explicit noise model: it fits both a mean model and a variance model over the CCD factor space and then optimizes so that the mean hits the target while the variance is minimized. The run structure (factorial points, axial points, center points) is the same; the analysis and objective are extended.
How many runs does a Robust CCD require?
For k controllable factors the CCD core requires 2^k factorial points (or a fraction), 2k axial points, and typically 3–6 center-point replicates. For k=3 a full CCD needs approximately 20 runs; for k=4 approximately 30 runs. If noise factors are also crossed into the design (combined array), the run count multiplies further. Fractional factorial cores can cut the run count for k >= 5.
When should I choose Robust CCD over Box-Behnken design for robust optimization?
Choose CCD when axial points are physically feasible and you want rotatability or near-rotatability, or when you already have a two-level factorial dataset that can be augmented. Choose Box-Behnken when experimental boundaries prevent running axial points (which exceed the +/-1 coded range), or when you have three factors and want a more economical design without extreme vertex combinations.
Can I conduct a Robust CCD with only controllable factors and no explicit noise factors in the design?
Yes, using a propagation-of-error (POE) approach: fit the quadratic mean model from the CCD, then analytically propagate assumed noise distributions through the fitted surface using the gradient at each point to estimate transmitted variance. This avoids explicitly crossing noise factors into the design but relies on assumptions about the noise distribution and requires the mean model to be accurate.
What software supports Robust CCD analysis?
Design-Expert (Stat-Ease) has dedicated dual-response and robust optimization modules. Minitab and JMP support response surface designs and variance modeling. In R, the rsm package fits response surface models and custom scripts can implement dual-response optimization. SAS PROC RSREG can also be used.
Sources
- Myers, R. H., Montgomery, D. C., & Anderson-Cook, C. M. (2009). Response Surface Methodology: Process and Product Optimization Using Designed Experiments (3rd ed.). Wiley. ISBN: 978-0470174463
- Khuri, A. I., & Mukhopadhyay, S. (2010). Response surface methodology. WIREs Computational Statistics, 2(2), 128–149. DOI: 10.1002/wics.73 ↗
How to cite this page
ScholarGate. (2026, June 3). Robust Central Composite Design for Response Surface Methodology. ScholarGate. https://scholargate.app/en/experimental-design/robust-central-composite-design
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Box-Behnken DesignExperimental design↔ compare
- Central Composite DesignExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare
- Robust Response Surface MethodologyExperimental design↔ compare