Open AccessA Split Plot Design for an Optimal Mixture Process Variable Design of a Baking Experiment
Author(s) -
Mika Alvionita Sitinjak,
Utami Dyah Syafitri,
_ Erfiani
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1417/1/012018
Subject(s) - restricted randomization , design of experiments , variable (mathematics) , optimal design , mathematics , main effect , process (computing) , factorial experiment , mathematical optimization , engineering design process , variance (accounting) , design process , statistics , split plot , computer science , work in process , randomization , engineering , medicine , mechanical engineering , mathematical analysis , randomized block design , operations management , accounting , business , randomized controlled trial , operating system , surgery
A mixture process variable (MPV) design consists of mixture design and process variable(s). The problem in MPV experiment is the number of experiment runs will be larger if the process variable increases. An optimal design can be a solution to produce a good design with a certain criterion and a limited number of runs. In practice, the compositions of the mixture design are running on each level of the process variable(s). It has a consequence that the randomization is restricted. A split-plot design can be an alternative to overcome the problem. In this research, the whole plots of the split-plot design were the levels of the process variable(s) and the subplots were the compositions of the mixture experiments. In addition, two optimality criteria were used: D-optimality and I-optimality criterion. The D-optimal design is searching a design by minimizing covariance of model parameters meanwhile the I-optimal design is seeking a design by minimizes average of the prediction variance. The study case was a baking experiment in which consisted of three flours and a process variable. It is surprised that the D-optimal design out performed compared to the I-optimal design in terms of the variance prediction in this case.