Mathematics · Statistics and Probability
A Multivariate Specification Development Approach Using Principal Component Analysis
Colin M. Haynes, Sean T. Kelly
See on the mapAbstract
The concept of a conservative, bounding environment for multiple-input, multiple-output (MIMO) vibration test-ing is inherently multivariate. However, some multivariate statistical methods proposed for MIMO specification development are very computationally expensive. This computational burden has motivated work to reduce the computational complexity of a specification with many control degrees of freedom (or variates). In this paper, a new method based on principal compo-nent analysis is presented that decomposes the multivariate specification problem into a univariate one while still accounting for the underlying correlation structure of the data. Numerical studies are used to illustrate the benefits of the new approach.