DOE Experiment Design Generator

Problem

When engineers test multi-factor processes, they need a clear run sheet before touching the system. Full factorials grow quickly as factors and levels increase, while limited budgets force a tradeoff between main effects, interactions, and resolution. Manual tables are easy to misorder and hard to feed into pandas, Minitab, or custom analysis scripts.

How it works

The skill reduces plan generation to a small set of explicit parameters. Full-factorial mode accepts --factors, --levels, and --output, producing the complete combination matrix. Fractional mode targets 2^(k-p) designs and accepts --base-levels, --num-factors, and --resolution, so a constrained budget can still preserve useful main-effect estimates. The script emits a JSON summary and writes a CSV with columns such as Run, Factor_A, and Factor_B. Levels are typically numeric codes like 0, 1, and 2, which makes downstream encoding and regression workflows straightforward.

Limits

This is useful when the task is to generate a standard DOE sheet, check the run count, and hand the result to an analysis pipeline. Fractional designs currently support two levels only. Resolution III designs impose stronger restrictions on main-effect estimation, so they should be chosen with factor importance and interaction expectations in mind. Output paths are relative to the working directory, and custom factor names must be maintained in the CSV or downstream analysis.