Multivariate Regression Analysis Method for Two-Variance Model

M. A. Mondal, De. H, Stefaan, & J. K. Ladha

Department of Mechanical Engineering, RUET, Kazla-6204, Rajshahi, Bangladesh

ABSTRACT: The multivariate two-variance regression model and its analysis method are established. The multivariate regression and confidence limit curves (or surfaces) with high confidence level and high reliability are also given. The multivariate two-variance regression model involves two kinds of independent variable, one of which is associated with the process such as test time, and the other is associated with test state such as temperature, therefore the presented method extends traditional multivariate regression analysis to process data and can analysis stochastic processes under various test conditions simultaneously, such as aero engine thrust curves and performance degradation curves. The concept of polynomial two-variance process and regression analysis method for polynomial two-variance model are also put forward. The presented method exploits the lengthwise information among different test times, therefore it can solve the problem of reliability assessment with small sample and higher precision.