Factors
Controlled or deliberately varied inputs such as temperature, time, current density, concentration, pressure, speed, preparation method or cure profile.
Special Processes Institute
sp-i.org
Design of Experiments Report
Design of Experiments Project
Learn and apply Design of Experiments to special-process development, optimisation, troubleshooting and validation. Define factors, levels, responses, controls, risks and measurement methods; generate a randomised two-level factorial run matrix; enter results; calculate main effects and two-factor interactions; identify the best observed settings; record conclusions and confirmation evidence; save or load projects as JSON; load a comprehensive zinc-nickel plating example; and produce a professional PDF-ready report.
Process knowledge Factorial experimentation Controlled change · objective evidence
2-level
Selected design
0
Factors
0
Experimental runs
0%
Results complete
Draft
Study status
What Design of Experiments Means
A planned, statistical method for learning how controlled inputs affect measured outputs.
Design of Experiments (DoE) changes multiple factors systematically so their individual effects, interactions and useful operating regions can be understood efficiently. It is more powerful than one-factor-at-a-time experimentation because the effect of one parameter may depend on the setting of another.
DefineProblem, scope and response
PlanFactors, levels and controls
RunRandomised controlled trials
AnalyseEffects, interactions and variation
ConfirmRepeat predicted optimum
ControlValidate and sustain the window
Levels
The chosen settings for each factor. A two-level study commonly uses coded low (−1) and high (+1) settings selected within safe and approved limits.
Responses
Measured outputs such as thickness, hardness, adhesion, strength, porosity, roughness, corrosion resistance, defect rate, yield or cycle time.
Main effects
The average change in response when one factor moves from low to high across the other conditions in the design.
Interactions
An interaction exists when the effect of one factor changes depending on another factor. This is a key reason to use factorial designs.
Noise and robustness
Uncontrolled or deliberately varied conditions such as operator, batch, rack position, ambient condition or material lot. Robust settings reduce sensitivity to noise.
Common DoE Designs
| Design | Best used for | Strength | Important limitation |
|---|---|---|---|
| Full factorial | Small number of factors where interactions matter. | Estimates all main effects and interactions in the chosen model. | Runs increase rapidly: 2ᵏ for a two-level design. |
| Fractional factorial | Screening many factors with fewer runs. | Efficient early learning. | Some effects are aliased and cannot be separated without assumptions or follow-up runs. |
| Plackett–Burman / screening | Finding influential factors from a long candidate list. | Very low run count. | Primarily estimates main effects; interactions may distort conclusions. |
| Response surface | Optimising known important continuous factors. | Models curvature and supports an optimum operating region. | Requires more runs and stronger statistical analysis. |
| Taguchi / robust design | Reducing sensitivity to noise factors. | Focuses on robust performance. | Interaction treatment and interpretation require care. |
| Mixture design | Formulations where ingredient proportions sum to a fixed total. | Appropriate geometry for blends and chemical formulations. | Ordinary factorial designs are unsuitable when proportions are constrained. |
Applying DoE to Special Processes
Control requirement: A DoE does not authorise processing outside approved specifications, qualified ranges or customer requirements. Trials must be risk assessed, technically approved, segregated where necessary and supported by suitable measurement capability. Production implementation may require revalidation, engineering approval or customer approval.
Electroplating / anodising
Factors may include current density, temperature, time, chemistry, pH, agitation, load area, rack position and pre-treatment. Responses may include thickness, distribution, adhesion, appearance, porosity, corrosion resistance and internal stress.
Heat treatment
Study furnace temperature, soak time, transfer time, atmosphere, quench condition, loading pattern or temper cycle against hardness, tensile properties, case depth, distortion, grain size and microstructure.
Welding / brazing
Study current, voltage, travel speed, heat input, shielding gas, joint gap, filler, preheat or post-weld treatment against penetration, strength, hardness, porosity, distortion and defect rate.
Painting / thermal spray
Study surface preparation, material viscosity, gun distance, pressure, passes, flash time, cure temperature and humidity against thickness, adhesion, roughness, appearance and corrosion performance.
Adhesive bonding / composites
Study preparation, contamination, bond-line thickness, pressure, out-time, cure temperature and cure time against lap shear, peel, void content, glass transition, failure mode and dimensional stability.
NDT and inspection processes
Use carefully designed trials to understand sensitivity, probability of detection contributors, scan speed, gain, concentration, dwell, lighting or developer time while preserving approved techniques and qualified personnel requirements.
Good Practice Before the First Run
Prepare: define a clear response; confirm the measurement system; select technically meaningful factor ranges; identify noise, nuisance and blocking variables; randomise run order; plan replicates and centre points where needed; define stop criteria; and retain complete traceability.
Avoid: changing too many uncontrolled conditions; using ranges wider than approval permits; treating correlated settings as independent; relying on a single sample; ignoring failed or inconvenient runs; selecting an optimum without confirmation; and confusing statistical significance with practical importance.
DoE Project Definition
Define the study0% complete
Factors
Up to five factors for automatic full-factorial generation.Responses
Design Controls
Automatic generation creates a two-level full factorial design. Fractional, response-surface, mixture and custom designs require a design selected and justified by a competent practitioner.
Experimental Runs and Results
Enter actual settings, response results, status and observations.
Define factors and generate a design.
Analysis Summary
Effect calculations are descriptive and intended to support, not replace, competent statistical review.
Complete run results and select Analyse Results.
Interpretation and Confirmation
Design of Experiments Report
Generated from the project, factors, runs, results and conclusions.
Complete the project and select Generate Report.