ABSTRACT
Prediction of deterioration of infrastructures is very important for rational maintenance planning and one of hot topics these days. Deterioration database on bridges generally contains various types of structures and deterioration mechanism. If single deterioration curve is used to predict future condition, it would involve very large uncertainty.
This study proposes a method to perform the grouping and regression analysis of existing bridges at the same time statistically by using EM (Expectation and Maximization) algorithm. Database on health index of 341 bridges in Gifu prefecture, Japan is used to estimate a group of deterioration curves. Based on the result of the principal component analysis of visual inspection data and some engineering considerations, it is judged that the deterioration degree of the RC deck can be evaluated by the first principal component (Health Index), whereas deterioration characteristics (mechanism) can be expressed by the second principal component, which is referred as Characteristic Index here.
Objective variable z in regression analysis is Health Index. The regression curve is made up of second order polynomial equation as to age y 1. () z = 5 + a y 1 2 + b y 1 https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/eq138.tif"/>
Bridges are separated into three groups by EM algorithm. Estimated deterioration curves for three groups are shown in Figure 1. Deterioration curve as to age for three groups of bridges, 2D (health index, age). https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig11_1.tif"/>
In addition to age y 1, Characteristic Index y 2 is considered as explanatory variable for the regression. Regression equation is expressed by, () z = 5 + a y 1 2 + b y 1 + c y 2 https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/eq139.tif"/>
Bridges are separated into four groups. The regression curves obtained for the four groups are shown in Figure 2. Deterioration curve as to age for three groups of bridges, 3D (health index, age, characteristic index). https://s3-euw1-ap-pe-df-pch-content-public-p.s3.eu-west-1.amazonaws.com/9781315207681/cd556cd4-4dcf-4efe-8e29-56fc67b8bfbd/content/fig11_2.tif"/>
It is noted that the number of group is increased by the additional information y2. It suggests that we can predict the deterioration with the more detailed models, when we have the more information.
