ENGGEN 115 Chap.7 Prototyping, Testing and Iteration
Prototyping, Testing and Iteration
A prototype is a learning instrument selected to answer a design risk, not an unfinished product judged by polish. Fidelity should match the question: paper can test sequence, rough geometry can test reach or assembly, and material samples can investigate joints or strength. A fair comparison controls inputs, records observable responses, repeats where variation matters, and distinguishes observation from explanation.
Failure becomes useful when its mode is diagnosed and the next version changes the feature that plausibly controls it.
What this chapter covers
- 01
Risk-driven prototypes
- 02
Fidelity selection
- 03
Controlled test design
- 04
Observable acceptance conditions
- 05
Failure-mode diagnosis
- 06
Iteration records and evidence limits
Worked example · free
Compare tower bracing layouts
- +1Define the risk question as which bracing layout better limits lateral top displacement under a common action.
- +1Apply the same measured horizontal force at the same height to both towers using one test setup.
- +1Place a fixed scale beside the top and measure displacement from a consistent viewing position.
- +1Repeat each trial and record joint damage or abnormal loading instead of keeping only the preferred value.
- +1Compare representative displacement while treating connection failure as a separate observation.
- +1Retain the better bracing principle if supported, then revise the controlling joint or member weakness for the next test.
Key terms
- Risk question
- A focused uncertainty whose answer could change a design decision or development priority.
- Prototype fidelity
- The degree of realism included in geometry, material, appearance or interaction for a specific test.
- Controlled comparison
- A test in which relevant conditions are held consistent so one intended difference can be interpreted.
- Acceptance condition
- An observable threshold or outcome used to judge whether a requirement is met in a test.
- Failure mode
- The particular mechanism through which a design loses function, stability, strength or usability.
Prototyping, Testing and Iteration FAQ
How polished should a prototype be?
Include only the fidelity required to answer the risk question. A polished surface may help investigate appearance or recognition but adds little to a structural load-path test. Excess fidelity consumes time and can make failure causes harder to isolate. State what the prototype represents and what it deliberately omits.
Why repeat a prototype test?
Small builds and manual loading vary because of joints, material imperfections, alignment and measurement. Repetition helps distinguish a consistent response from one accident and exposes abnormal trials. It does not automatically make the experiment complete, so preserve raw observations and remain modest about what the sample supports.
What should an iteration record contain?
Record the version, risk question, controlled conditions, observable result, failure mode, interpretation, decision and next change. Include a photograph only when its scale and viewpoint make the claimed observation visible. The record should let another reader understand why the next version differs from the previous one.
How do I choose prototype fidelity?
List the uncertainties separately for geometry, material, appearance and interaction, then include only the fidelity required for the current question. A paper model can answer sequence or reach questions, while joint strength needs representative material and connection behaviour. State any omitted realism so the result is not overextended.
What should happen after a prototype fails?
Locate the failure, classify its mode and identify the feature that plausibly controls it. Change that feature while preserving comparable test conditions, then repeat the observation. Avoid adding material or complexity everywhere, because broad changes hide the mechanism and make it impossible to learn which revision improved performance.
Exam move
Choose one design uncertainty and list three possible prototypes at different fidelities. Select the cheapest representation that can answer it, define an observable response and write the acceptance condition before building. Run a controlled comparison, keep raw results, and classify any failure by member, joint, base, interface or setup.
End by changing one controlling feature and explaining why that modification should affect the measured response. Practise writing observation and interpretation in separate columns so measured displacement, visible damage or task time cannot quietly become an untested causal claim. Design a simple repeatability check with a fixed load point, camera position, scale and version label.
Alternate the test order when settling or operator learning could bias the later trial. For scaled models, list which geometric, force, stiffness and material relationships are not preserved, then write the narrowest conclusion the setup supports. A useful revision session ends with a new risk question, not merely a cleaner or stronger-looking object.
Revisit the first acceptance condition after testing and ask whether it measured the requirement that mattered. If not, repair the test question before rebuilding the prototype. Archive one failed version because its failure mode can guide later concepts more clearly than a final photograph.