Graphical Optimization
With multiple responses, regions where requirements simultaneously meet the critical properties are sought. By overlaying critical response contours on a contour plot the best compromise can be picked (the sweet spot).
Graphical optimization displays the area of feasible response values in the factor space. Regions that do not fit the optimization criteria are shaded gray. Any “window” that is NOT gray shaded satisfies the goals for every response. Right-click on the graph to plant flags showing predictions for all responses at that location in space.
Graphical Optimization Criteria
Click on a response to specify the response limits, lower and/or upper, for each response to be included in the optimization. If the goal is to minimize a response, specify the highest acceptable upper limit. If maximizing, specify the lowest acceptable lower limit. If trying to achieve a target, specify both a lower and upper limit. Leave the low and high limits blank to remove a response from consideration. Interval estimates can be added to the optimization plot to account for uncertainty in the point predictions.
Graphical Optimization Intervals
Interval estimates can be added to the graphical optimization plots to help understand the impact of uncertainty on achieving process goals. Interval estimates are turned on by selecting the Show Interval box in the criteria.Once selected, choices for type of interval and confidence level can be specified.
On the plot, the bright yellow (default) now shows where the entire range of all intervals meet the specified criteria. The dark gold corresponds to where the point estimate meets the criteria requirements, but part of an interval estimate does not.
Intervals can be applied independently to each response.
User Tip: Consider starting with loose restrictions on the responses and gradually tighten them up. This can be done by typing tighter limits and observing the change in the graph, or (more intuitively) by “dragging” the limit contours closer together. You can drag a contour on any 2D contour graph simply by clicking on it and moving the mouse (while keeping the mouse button pressed). Moving contours this way is not as precise as typing in numeric limits, but gives you a good feel for how the surface behaves.