Abstract
Deterioration or health data base of infrastructure contains various types of structures and deterioration mechanism. It is preferable to separate the data according to detailed information before estimation of regression curves. It is, however, difficult to collect all relevant data to separate them. This study proposes the method to separate data and perform regression analysis at the same time statistically by EM algorithm and information criteria. We verify the usefulness of the proposed method by numerical experiments in which hypothetical data is generated from three lines. Four regression curves are estimated from an actual bridge deterioration data base by the proposed method. The physical meaning of separated groups is interpreted to some extent. It is shown that the proposed method can be useful to separate and perform regression analysis of large scale data base.
