Actions for New Developments in Design of Experiments

Email RIS file
Bookmark
Report an Issue

New Developments in Design of Experiments

Author
Peng, Jiayu
Published
[University Park, Pennsylvania] : Pennsylvania State University, 2019.
Physical Description
1 electronic document
Additional Creators
Lin, Dennis Kon-Jin
Access Online

Availability

I Want It
I Want It

Finding items...

Graduate Program
Restrictions on Access
Open Access.
Summary
This dissertation consists of three projects on the design of experiments.The first project is on order-of-addition experiments. In an order-of-addition problem, the output of a process depends on the order of adding different components into the system. The term ``order-of-addition'' originated from chemical experiments, in which the order of adding different reagents often affects the properties of final product. The importance of order-of-addition problem goes well beyond the area of chemistry though, for example, in the job scheduling field. An order-of-addition design is a set of pre-designed orders to test, so that the order effects can be most efficiently learned (in terms of some criteria), under certain experimental budget. In this project, we investigate the optimal design under different models of order effects, with the main results being: (1) a theory of D-optimal design under any order-of-addition linear model, and more specific theories under the pairwise-order model; (2) closed-form construction for a class of optimal and robust designs; (3) a fast and flexible algorithm for the construction of efficient designs with large number of components.The second project is on the optimal design for estimating error variance. In screening experiments, the ``saturated design with center points'' is a conventional choice (in Response Surface Methodology). We propose an alternative design -- the projection of a larger Hadamard matrix. In this project, we show that with certain assumption, the proposed design achieves the most efficient estimate of not only the factor effects but also the error variance, and is thus preferable over the conventional choice. Under various common error distributions, theories and simulation results are given to demonstrate that the proposed design achieves a more reliable estimate of error variance, as well as a reliable significance testing. For more practical use cases, the optimal follow-up design problem is studied. Relevant optimality theories are established and a convenient construction method for optimal follow-up design is proposed.The third work is about optimal run orders. In many industrial experiments, systematic run orders are desirable for reducing the costs from resetting factors, i.e., level changes. For special designs such as fractional factorials, minimum-level-change run orders are available in the literature. We consider the run order optimization for any arbitrary design, and also for more flexible costs criteria. We apply Traveling Salesman Problem (TSP) devices to solve the optimal run order. Specifically, the run order problem is converted and solved using a ``Concorde'' solver of TSP. Many new optimal run orders for a broad class of designs have been thus obtained.
Other Subject(s)
Genre(s)
Dissertation Note
Ph.D. Pennsylvania State University 2019.
Reproduction Note
Microfilm (positive). 1 reel ; 35 mm. (University Microfilms 28097190)
Technical Details
The full text of the dissertation is available as an Adobe Acrobat .pdf file ; Adobe Acrobat Reader required to view the file.
View MARC record | catkey: 28948202