00763136

Component

The National Bridge Inventory Database (NBI) is the most extensive repository of data on highway bridges in the United States. Initiated in 1972, this database now contains detailed historical data on over 600,000 bridges for a span of 26 years. The archived and current NBI files contain more than 6.2 billion bytes of data. To efficiently utilize this information for research and analysis, the authors investigated and developed several different relational database management approaches and data warehousing techniques. From the resulting system, data mining methods were used to efficiently access this data and extract information on the Nation's highway bridges. This has resulted in a significantly better understanding of the bridge inventory. Key descriptive statistical summaries of the NBI resulting from this work are presented. Although the NBI, in itself, is a tremendous resource, the true power of data mining methods were not realized until the data inherent in the NBI was expanded by implementing a spatial relationship capability utilizing a geographic information system (GIS). This facilitated extensive visualization of geographic patterns in the NBI data, several examples of which are included in the paper. More importantly, it enabled a study of relationships between bridge behavior and other factors, such as climate. Advanced analyses have been performed using the GIS capabilities coupled with statistical modeling and analysis methods. One research study, which is summarized in this paper, focused on the development of a new model of bridge deterioration. Using the expanded data sets available from the combined NBI and GIS databases, three different regression methods were applied to model the relationship between condition state and plausible factors causing deterioration. The variables included in the study were age, average daily traffic, precipitation, frequency of deicing, temperature range, freeze thaw cycles and type of bridge construction. Different models were developed for deck, superstructure and substructure deterioration. Generalized linear models, generalized additive models and a combination of the two were applied. The generalized linear model gave the best prediction. This new model is presented.

C-6, IBMC-047

English

16p

1999

Eighth Transportation Research Board Conference on Bridge Management

Figures (10) ; References (8) ; Tables (10)

Data mining; Databases; Deterioration; Geographic information systems; Highway bridges; Mathematical models; Regression analysis; Relational databases; Statistical analysis; Visual presentations

Condition (Bridges); Data warehousing; National bridge inventory

United States

Bridges and other structures; Data and Information Technology; Design; Highways; I24: Design of Bridges and Retaining Walls

TRIS, TRB

Apr 27 2000 12:00AM