Pareto Select

Pareto Select shows the trade-off front that Desirability found and lets you pick one recipe from it by clicking. The chosen recipe - its settings, its score per goal and its overall D - comes out of the component and stays with the document, even when the DoE components upstream are switched off or their front changes.

Panel: 6 Calibrate. Nickname: Pareto. Compute button: no (runs on every change).

Inputs

Name	Type	Default	Description
front	Number (tree)		The pareto output of Desirability, one branch per recipe: factor values, one desirability per active goal, D. Optional once a selection is stored
factorNames	Text (list)		Factor names from DoE Design in model order, exactly one per factor value
objectiveNames	Text (list)		Names of the active goals in model order, the objectives output of Desirability. The axes are desirabilities from 0 to 1, not response units
selectionKey	Text		Recipe key from this component’s key output for selection without the picker, valid only for its original front
selectionIndex	Integer	-1	Select a recipe by current branch index from 0 (-1 = none), reapplied only when this input changes
optimumSettings	Number (list)		Desirability settings output to mark the recommended recipe on the plot (unwired = highest D)
weights	Number (list)		Importance per active objective in objectiveNames order, 0 or more each. Marks the recipe with the highest weighted geometric mean of the front’s own desirabilities, response goals only, without re-running Desirability

All inputs are optional. Without a front the component still emits a recipe it stored earlier.

Optional group: Confirm selected recipe

Right-click the component to switch this group on. The inputs appear only while the group is enabled.

Name	Type	Description
confirmValues	Number (list)	Measured values for the selected recipe in objectiveNames order

Outputs

Name	Type	Description
settings	Number (list)	Selected factor values in actual units, factorNames order. Empty without a selection
di	Number (list)	Selected recipe’s desirability per active objective, 0 to 1
D	Number	Selected recipe’s overall desirability from 0 to 1
key	Text	Recipe key for selectionKey, valid only for the front it was issued from
status	Text	Selection status: current, stale (front changed), restored, ambiguous, or why a request was refused
current	Boolean	True only when the stored recipe is verified on the live front under the same identity
confirmReport	Text	Confirmation measurements checked against the selected recipe’s prediction intervals

How to use it

1. Enable the pareto group on Desirability and compute; wire its pareto output to front, the factor names to factorNames, and its objectives output to objectiveNames.
2. Click a point on the strip under the component (two goals: a scatter; three or more: one line per recipe across one axis per goal; one goal: a strip). Press Pick… for a larger window; there the arrow keys walk the front and Enter confirms. Hovering shows the full recipe.
3. Read settings for the print. Reset forgets the choice; both are undoable.
4. Optional: wire a list of weights (one per goal) to see which recipe a given importance would pick. Move the numbers and watch the green square walk along the front. This is a probe for your preference, not a selection: click the recipe to choose it.

The axes are desirabilities from 0 to 1, not measured units: a value of 0.8 on the width axis means the recipe meets the width goal to 80 %.

Why the front is drawn over desirabilities

A Pareto front and a desirability optimum are often described as two different methods, and they are - but they share one ingredient. The goal you type for each result (min, max, target=7 in 6..8) does two things: it turns a measured value into a score from 0 to 1 where bigger is always better, and, together with the weights, it folds all the scores into the single number D. Only the first step is needed to draw the front. Without it a target goal could not be put on a trade-off plot at all, because “close to 7 mm” has no direction in millimetres. So the front uses your goals and their bounds, but not your weights. The weights are a preference, and the front is there precisely so that you can see the trade-offs before committing to one. Pareto Select is where that preference is applied afterwards: by clicking, or by wiring weights to see where it would land.

Two things follow. Recipes outside a goal’s bounds all score 0 on that axis and are shown as one point, so the bounds you type decide what counts as acceptable. And factor priors (prefer a low layer height) are not on the axes, which is why Desirability’s recommendation can sit off the front while still carrying the highest D.

What happens when the front changes

The choice is stored by its numbers, not by its position in the list. If Desirability recomputes and the same recipe is still on the front, the choice stays current, even when the list is in a different order. If the front is different - a goal, a weight or a bound changed - the component keeps the stored numbers, sets current to false, and says front changed on the component; it never quietly switches to another recipe. A recipe that still matches is shown dashed as a candidate; click it to accept the new front. Two identical recipes on one front are reported as ambiguous.

Back to Components.