Assessing the Consequences After an Incorrect Multiple-Component Constraint Was Used to Develop a Constrained Mixture Experiment Design

• Activity Number: 355
• Type: Contributed
• Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Greg Piepel* and Scott Cooley
• Companies: Pacific Northwest National Laboratory and Pacific Northwest National Laboratory
• Keywords: Mixture experiment ; Experimental design ; Multiple component constraint ; Waste glass

Abstract

This presentation uses a case study to discuss what was done to assess the consequences of performing a mixture experiment design when one of the multiple-component constraints used to specify the constrained mixture region (CMR) was later found to be incorrect. Specifically, the mixture experiment involved 15 waste-glass components. The CMR was specified by lower and upper bounds on the component proportions (single component constraints) and four multiple-component constraints (MCCs). One of the MCCs was developed from a linear mixture model for viscosity of the waste glass, and the model was subsequently found to be based on data for which a subset of the viscosity values were wrong by a factor of 10. The experimental work was completed by the time this problem was detected. The presentation covers the work performed to develop a corrected viscosity model, viscosity MCC, and CMR. The presentation also discusses the work to assess the effect of the incorrect MCC on the (1) incorrect CMR explored by the experimental design compared to the corrected CMR, and (2) properties of the experimental design for the incorrect CMR versus the corrected CMR.