Simulation-Assisted Full Factorial Design — Exploring All Factor Combinations via Computer Simulation
Simulation-Assisted Full Factorial Design of Experiments · Also known as: SA-FFD, computer simulation full factorial, virtual full factorial design, simulation-based full factorial DOE
Simulation-assisted full factorial design integrates full factorial design of experiments (DOE) with computer simulation models — such as discrete-event simulation, finite element analysis, or Monte Carlo methods — to systematically explore every combination of factor levels and quantify their effects on system responses. It enables comprehensive experimentation in contexts where physical trials would be costly, dangerous, or infeasible.
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 simulation-assisted full factorial design when: (1) physical experimentation is too expensive, slow, or hazardous to run all factor combinations; (2) a validated simulation model of the system exists or can be built; (3) you need to detect all main effects and interactions without confounding (the key advantage of full factorial over fractional designs); and (4) the number of factors is moderate (typically 2–5 factors), keeping the run count manageable. Do NOT use it when: no validated simulation model is available; the number of factors is large (prefer fractional factorial or response surface designs to limit runs); or when the simulation model has not been validated against real system behaviour — unvalidated surrogates produce misleading results.
Strengths & limitations
- Detects all main effects and all interaction effects without aliasing or confounding.
- Eliminates physical material, equipment, and safety costs by replacing real runs with simulation runs.
- Enables testing of extreme, hazardous, or infeasible conditions that cannot be run physically.
- Replications are inexpensive in simulation, allowing precise estimation of within-cell variance.
- Produces a complete factorial data set that supports ANOVA, regression, and response surface follow-up.
- Requires a validated simulation model — building and validating this model can itself be resource-intensive.
- Run count grows exponentially with the number of factors and levels (L^k), becoming impractical for many factors.
- Results are only as reliable as the simulation model; model assumptions and simplifications propagate into conclusions.
- Stochastic simulators require multiple replications per cell, multiplying total computation time.
Frequently asked
When should I use a full factorial simulation design rather than a fractional factorial simulation design?
Prefer full factorial when you suspect or need to detect interaction effects between all pairs of factors, and when the total number of runs is still computationally feasible. If you have more than 4–5 factors, a fractional factorial design is usually more efficient; it sacrifices some interaction estimates (aliasing) in exchange for a much smaller run count.
How many replications do I need per cell in a stochastic simulation?
Replications should be sufficient to estimate within-cell variance reliably. A common approach is a pilot study of 5–10 replications to estimate variance, then use power analysis to determine the final replication count needed to detect a meaningful effect size at a chosen significance level. For deterministic simulators, a single run per cell is typically adequate.
Does the simulation model need to be a perfect replica of the real system?
No, but it must be a sufficiently accurate approximation for the purposes of the study. Validation — comparing model outputs to real system data under known conditions — is mandatory before trusting factorial results. Document all model assumptions and verify that conclusions are robust to plausible variations in those assumptions.
Can I use simulation-assisted full factorial design with machine learning surrogate models?
Yes. When high-fidelity simulation runs are very expensive, a small initial set of simulation runs can be used to train a surrogate (e.g., Gaussian process, neural network), and the full factorial grid is then evaluated on the surrogate. This approach, sometimes called surrogate-assisted DOE or metamodeling, trades some accuracy for computational savings.
How do I handle factors with more than two levels?
A full factorial design accommodates any number of levels per factor: the total run count is the product of all level counts across factors. For quantitative factors at three or more levels, you gain the ability to detect nonlinear (curved) effects; for two-level designs only linear effects are estimable. When nonlinearity is suspected, at least three levels per factor are recommended.
Sources
- Montgomery, D. C. (2017). Design and Analysis of Experiments (9th ed.). Wiley. ISBN: 978-1119113478
- Kleijnen, J. P. C. (2015). Design and Analysis of Simulation Experiments (2nd ed.). Springer. ISBN: 978-3319185668
How to cite this page
ScholarGate. (2026, June 3). Simulation-Assisted Full Factorial Design of Experiments. ScholarGate. https://scholargate.app/en/experimental-design/simulation-assisted-full-factorial-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.
- Central Composite DesignExperimental design↔ compare
- Design of experimentsExperimental design↔ compare
- Simulation-assisted response surface methodologyExperimental design↔ compare
- Simulation-assisted Taguchi methodExperimental design↔ compare