Tutorial 9

Design of Experiments

Run a fixed-budget formulation campaign as a rubber-film R&D team. Your team is assigned one of three classic strategies — OFAT, fractional factorial, or CCD — and has to defend its predicted-best formulation against the other two.

Alix Schmidt
90 min
Excel/Sheets · Google Colab
Back to schedule

Start here

Pick up your team's briefing doc and template below (under Choose your strategy). Design and submit experiments on paper, in chat, however you like — the facilitator runs them and reads your results back. Once you have data, open the analysis notebook: it's the same notebook for all three archetypes, you just set which one you ran.

Analysis notebook

The system

You are formulating a latex rubber film. You control five factors, each with a hard range:

factorrangenotes
latex_pct5 – 20 wt%hard constraint: latex ≤ 20 wt%
crosslinker0.5 – 3.0 phrphr = parts per hundred rubber
filler_loading0 – 40 phrphr = parts per hundred rubber
cure_temp140 – 180 °C
cure_time5 – 30 min

Every experiment reports back three responses: tensile_mpa (your target), elongation_pct, and hardness_shore_a. You don't get the underlying formulas — that's exactly what your design has to probe for.

Choose your strategy

Your team is assigned one of three classic designs. Each comes with its own team briefing (what your design lets you see, and how to run it) and its own spreadsheet template (blank for OFAT, a pre-computed run list for FRACTIONAL and CCD).

OFAT

One-Factor-At-A-Time. No pre-computed run list — pick a baseline, vary one factor at a time, freeze the best, move on. Full freedom over what you test.

Briefing Template

Fractional

A 2^(5-1) half-fraction: 16 corner runs, cleanly separating main effects and two-way interactions across 4 of your 5 factors (your call which one to drop).

Briefing Template

CCD

A trimmed, face-centered central composite design: corners + axial "star" points, enough to fit a full quadratic response surface in 4 factors — saturated, so no lack-of-fit test.

Briefing Template

Run your experiments

  1. Budget: 15 experiment-dollars. One submission = one dollar. Objective: maximize tensile_mpa.

  2. Submit to the facilitator — paper, chat, or verbally. They run each combination and read back your three responses, or tell you why a run is infeasible (no penalty for a caught typo; it costs a dollar if you hadn't checked the constraint yourself).

  3. No comparing notes with the other archetype teams. This is a competition — each team's answer is theirs to defend.

Analyze and predict

Open the analysis notebook (button above) once your data is in. It's the same notebook for every archetype — set TEAM_ARCHETYPE to 'OFAT', 'FRAC', or 'CCD' and it fits whichever model your design supports, then scans for the formulation it predicts will maximize tensile. Hand that predicted-best formulation to the facilitator — that's what you'll defend.


Discuss with your team once you have a predicted-best formulation — some of these have no single right answer.

  1. Warm-up

    Which archetype did your team run — OFAT, fractional factorial, or CCD? What appealed to (or worried) you about it before you ran a single experiment?

    Easy
  2. Core

    Walk through your factor-assignment decisions (FRACTIONAL/CCD) or your sweep order (OFAT). Where did your chemistry intuition get it right or wrong once the model was fit?

    Medium
  3. Challenge

    Compare your predicted-best formulation against the other two archetypes in class. Which team's design gives you the most confidence in its answer — and what specific feature (corners? axial points? center points?) earns that confidence?

    Challenge

Team materials

OFAT briefing OFAT template Fractional briefing Fractional template CCD briefing CCD template

Analysis

Analysis notebook (Colab) — same notebook for all three archetypes

General

Paired lecture notes