Design of Experiments (DOE): Designing the 'Optimal' Paper Airplane
Summary.
Course: ISE 202  Design/Analysis of Engineering Experiments
Team: 2 HFE (M.S.) and 2 ISE (M.S.) Graduate Students
Timeline: 2Weeks
Contribution: Experimental Design \ Prototype \ Data Collection and Analysis \ Preparation and Presentation of Deliverables
In the Spring 2019 semester, 4 HFE (M.S.) and ISE (M.S.) students at SJSU formed a team to develop the 'optimal' paper airplane (measured in flight distance) using a twofactor, 1/8 fractional factorial experimental approach. The experiment considered six paper airplane design parameters, with each design parameter containing two possible factors (+/):

Body Length

Body Width

Paper Clip

Paper Weight

Wing Length

Wing Type
Problem.

Flight distance depends on many design parameters, each with multiple levels to consider

Running a full factorial experiment can result in a large number of runs, becoming lengthy and expensive (with no practical significance)

The experimental design does not account for confounding variables or human factors (e.g., height of paper airplane operator)
Objectives.

Design the 'optimal' paper airplane (defined as flight distance) using a twofactor, 1/8 fractionalfactorial experiment

Consider six design parameters, with each design parameter containing two possible factors
Hypothesis.
H0: All treatment means are equal
H1: At least one treatment mean is different
alpha: 0.05
Methods.
Factor Screening \ DOE \ Experiment
Data Analysis.
ANOVA \ Posthoc Analyses
Tools.
Paper, Clipboard and Pencil \ Scissors \ Paperclips \ Measuring Tape \ A/V Equipment \ MiniTab DOE & Randomization \ SPSS Statistics \ Google Docs & Sheets
Process.
Procedure.
Phase I: Developing and Testing Initial Paper Airplane Designs
Factor Screening:

Estimating factors with large effects (main effects)

Estimating loworder interactions
Initial Airplane Development:
A total of eight paper airplane designs were developed using a randomization scheme, generated using MiniTab.
Experimental Trials:
Using a twofactor, 1/8 fractional factorial experimental design with six design parameters and each design parameter containing two levels each, each airplane design was tested through four experimental trials each. Researcher (human factors) and confounding variables were controlledfor by assigning consistent paper airplane developer, operator, data measurer, and data logger roles across the research team.
Statistical Analyses:

ANOVA

Posthoc analyses
Phase II: Developing the 'Optimal' Paper Airplane
Optimal Airplane Development and Testing:
Results obtained from Phase I of the experiment were used to inform and test the 'Optimal' paper airplane design.
Statistical Analyses:

ANOVA

Posthoc analyses
Results.
Phase I: Developing and Testing Initial Airplane Designs
ANOVA:
At an alpha level of 0.05, no significant differences were found between the treatment means, failing to reject the null hypothesis.
Posthoc Analyses:
Main Effects and Interaction Plots were analyzed to identify the parameters and levels that significantly affect flight distance.
Phase II: Developing the 'Optimal' Paper Airplane
ANOVA:
Data from the 'optimal' paper airplane experiment was compared to the data from the initial eight designs. At an alpha level of 0.05, at least one significantly different treatment mean was found.
Posthoc Analyses:
Analysis of Fisher Pairwise Comparisons revealed the final design significantly flies a further distance than the initial eight designs.
Deliverables.
The following deliverable was presented to the ISE 202 Design/Analysis of Engineering Experiments course at SJSU:
Report and Presentation (Slide Deck).
Next Steps.
This experiment provided insights into the main and interaction effects of six paper airplane design parameters (each with two levels (+/)) on total flight distance. Experimentation with additional factors, including human factors is necessary to improve the usability for various users.