STATISTICS SEMINAR SERIES :: ADAPTIVE EXPERIMENTATION, EXPECTED VALUE OF IMPROVEMENT, AND ROBUST DESIGN

GUEST LECTURER
Daniel D. Frey

AFFILIATION
Mechanical Engineering and Engineering Systems, MIT

ABSTRACT
This seminar will present research into adaptive experimentation as a means for making improvements in design of engineering systems. A simple method for experimentation is described entitled “adaptive one-factor-at-a-time.” A mathematical model is proposed and theorems are presented concerning the expected value of the improvement provided and the probability that factor effects will be exploited. It is shown that adaptive one-factor-at-a-time provides a large fraction of the potential improvements if experimental error is not large compared to the main effects and that this degree of improvement is more than that provided by resolution III fractional factorial designs if interactions are not small compared to main effects. The theorems also establish that the method exploits two-factor interactions when they are large and exploits main effects if interactions are small. A case study on design of electric powered aircraft supports these results. These results are then extended to robust parameter design. A method is proposed for evaluating the effectiveness of robust parameter design methods. The method employs a hierarchical probability model to express assumptions about effect sparsity, hierarchy, and inheritance. Using this approach, it is shown that adaptive one-factor-a-a-time crossed with a resolution III array generally outperforms other robust parameter design methods in terms of improvements attained and the costs incurred.

DATE & TIME
Thursday, October 27, 2005 -- 11:00 AM

DURATION
1 hour

LOCATION
Exectutive Room #228

CONTACT PERSON
Ming Yuan