DOE Optimization Technique

FOLLOWING TOPIC PASTED FROM YAHOO GROUPS

ORIGINATOR: Ambika Sahoo

DATE: Fri Sep 19, 2008

Dear Viswadeep sir,

we have DOE regression equation.From DOE analysis, we found no interaction effect.In DOE we've 2 levels & 3 factors.

I'm stuck how to get optimum point(Lowewst response value).

I think..we've 3 surface plots & 3 contour plots. Can we get minimum points from 3 surface plots? We can get minimum points by heat & trial from each surface plot.Then we put it in DOE regression equation & to

find out minimum point....Is it right method? Can we statistically prove this point as minimum point?

Project: Reduction of High current in Starter Motor Factor:Concentricity(0.1-1.0 mm),Ovality(0.1-1.0 mm),Flatness(0.1-0.5 mm)...from main plot we find ..lower the value, better the response fo all the factors)

Please help me.

Best Regards.

A.P.Sahoo

______________

FOLLOWING RESPONSE WAS POSTED TO YAHOO GROUPS

Dear Sahoo,

If you have no interaction effects in your model and only the main effects are significant, then you can interpret each main effect independently. If you are looking for a low value of output (Y), then if the effect value is positive, then the low value of X will provide a low value of Y, if the effect value is negative, then a high value

of X will provide a low value of Y. Hence, your optimal points will lie on the extreme points of your design space. You will not need to use the contour or surface plots to figure out the optimal point.

The main effect plots will help you identify your optimal values.

Hope this helps,

SJ.

_____________