Hybrid Design of Experiments — Combined Experimental Strategy
Hybrid Design of Experiments · Also known as: hybrid DOE, combined experimental design, mixed experimental design, hybrid experimental strategy
Hybrid design of experiments (hybrid DOE) combines two or more experimental design strategies within a single study to exploit the complementary strengths of each. Common combinations include factorial or fractional-factorial arrays paired with computer simulation runs, space-filling Latin hypercube designs merged with response surface augmentations, or Taguchi orthogonal arrays integrated with response surface methodology. The approach is widely used when a single design type cannot efficiently cover all phases of an engineering investigation — from screening through to optimization.
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 hybrid DOE when a single design type is insufficient — for instance, when screening must cover a wide factor space but optimization requires high-resolution local modeling, or when physical experiments are expensive and must be supplemented by simulation. It is appropriate in advanced engineering development, process optimization with both controllable and noise factors, and studies combining physical prototypes with computational models. Do not use it when the experimental budget is tight enough that the overhead of combining and analyzing multiple design types outweighs the benefit; in those cases a single well-chosen design (e.g., a definitive screening design or a central composite design) is more practical. Avoid hybrid DOE if the team lacks the statistical expertise to fit and validate a joint analysis model.
Strengths & limitations
- Exploits complementary strengths of different designs — breadth from space-filling or screening arrays, depth from response surface or confirmation designs.
- Can reduce the total number of expensive physical runs by using cheaper simulation or screening data to guide where physical resources are spent.
- Flexible: the combination can be tailored to the specific constraints (budget, time, precision) of each phase of the investigation.
- Enables simultaneous study of robustness and performance by integrating Taguchi-style noise arrays with optimization-focused response surface components.
- Produces richer data that support both mechanistic understanding and predictive modeling.
- Requires expertise in multiple design types and in joint analysis methods such as co-kriging or discrepancy-corrected regression.
- Integrating data from heterogeneous sources (physical versus simulation) can introduce systematic discrepancies that are difficult to model correctly.
- Planning and coordination overhead is higher than for a single-design study, and deviations from the planned design in one phase can propagate to the second.
- Software support for joint analysis of hybrid designs is less mature than for standard factorial or response surface designs.
- Results and recommendations can be harder to communicate to non-specialist stakeholders than those from a single, familiar design.
Frequently asked
What is the most common type of hybrid DOE in engineering practice?
The most common hybrid combines a computer-experiment design (typically a Latin hypercube or space-filling design) with a small set of physical confirmation or augmentation runs. This combination is especially prevalent in aerospace, automotive, and chemical engineering, where simulation is mature but physical validation is still required.
How do I handle the different error variances from simulation versus physical runs?
Simulation runs are usually deterministic (zero pure-error variance), while physical runs carry measurement noise. In the joint analysis model, assign a nugget or error variance term only to the physical-run data, or use a co-kriging or Kennedy-O'Hagan calibration framework that explicitly models the discrepancy between the simulator and reality.
Can I build a hybrid DOE sequentially, or must all runs be planned upfront?
Sequential (adaptive) hybrid DOE is common and often preferred. The typical approach is to run the cheaper or broader design component first, analyze the results, and then construct the second-phase design conditioned on what was learned. This requires a pre-specified stopping or augmentation rule to avoid data dredging.
Does hybrid DOE require specialized software?
Standard DOE software (JMP, Minitab, Design-Expert) handles most component designs and can analyze the phases separately. For joint analysis of physical and computer-experiment data, specialized packages such as R's DiceKriging, GPfit, or the Python GPy library are typically needed. Some platforms (e.g., JMP Pro) have built-in Gaussian process modules that support combined analysis.
When is a hybrid DOE NOT worth the extra complexity?
If all your runs are of the same type (all physical or all simulation), a standard single-design approach is simpler and equally effective. Hybrid DOE is also hard to justify when the team lacks experience with joint modeling, when the total run count is small enough for a single response surface design to cover the space adequately, or when the project timeline does not permit sequential execution.
Sources
- Santner, T. J., Williams, B. J., & Notz, W. I. (2003). The Design and Analysis of Computer Experiments. Springer. ISBN: 978-1441929921
- Loeppky, J. L., Sacks, J., & Welch, W. J. (2009). Choosing the sample size of a computer experiment: A practical guide. Technometrics, 51(4), 366–376. DOI: 10.1198/TECH.2009.08040 ↗
How to cite this page
ScholarGate. (2026, June 3). Hybrid Design of Experiments. ScholarGate. https://scholargate.app/en/experimental-design/hybrid-design-of-experiments
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.
- Central Composite DesignExperimental design↔ compare
- Design of experimentsExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare
- Simulation-assisted design of experimentsExperimental design↔ compare