SCIOVANTA

DoE and numerical optimization — worked example

DRAFT - requires author and legal review.

The synthetic response is 50+10A+5B−4A²−3B²+2AB on a two-factor, three-level grid. Fit a full quadratic in coded factors, then maximize the response with desirability limits 0 and 100. This noiseless fixture verifies calculation flow; it cannot provide a realistic estimate of experimental error. A numerical candidate must be confirmed experimentally. Only continuous factors are supported; no categorical or mixture interpretation is implied.

02468100204060Measured responseRun orderResponse (synthetic units)
Measured response (solid). Synthetic example, not experimental evidence.
Figure data
SeriesRun orderResponse (synthetic units)
Measured response130
Measured response236
Measured response336
Measured response442
Measured response550
Measured response652
Measured response746
Measured response856
Measured response960

Reproduce the example

Open the tool, enter the input below or copy it from the downloadable JSON, and select the stated options. These are synthetic measurements.

{
  "tool": "doe",
  "csv": "A,B,Y1\n-1,-1,30\n-1,0,36\n-1,1,36\n0,-1,42\n0,0,50\n0,1,52\n1,-1,46\n1,0,56\n1,1,60",
  "factors": [
    {
      "name": "A",
      "unit": "coded",
      "low": -1,
      "high": 1
    },
    {
      "name": "B",
      "unit": "coded",
      "low": -1,
      "high": 1
    }
  ],
  "model": "quadratic",
  "goals": [
    {
      "goal": "max",
      "low": 0,
      "high": 100,
      "target": 100
    }
  ]
}

Computed output

R21
codedCandidate[1, 1]
desirability0.6

Open DoE and numerical optimization Download exact input and output

Generated by scripts/build_content.py using the same scientific functions as the application. Independent numerical validation is documented in the source package; no commercial reference program was executed.