Statistical Techniques to Assist Decision Makers in Selecting Complex Mixturesa

摘要

The relative proportions of the componentsrnof complex mixtures often control the propertiesrnof interest. Statistically designed experimentsrnreveal not only which ingredients control a givenrnproperty but also generate response surfacernmodels to predict performance as a function ofrncompositions. A mixture experiment is inrnprogress to predict properties of common interestrnfor four generic grout ingredients plus onernadmixture used to fill and close tanks. Twornexamples of response surface models arernpresented for a previously generated statisticalrndesign. The generic 30-mix design in progress isrnalso presented with preliminary results for therntank closure properties. A baseline formulationrncan be selected from the results and models fromrnsuch designs targeting the site-specific criteriarnfor a given tank closure can be generated.