An intelligent decision support system for bridge safety assessment based on data mining models

Abstract

Bridges are one of the primary infrastructures in our society. During the life cycle of a bridge structure, the service conditions should be evaluated on a regular basis in order to assure the necessary levels of strength and durability. Taking into account (i) the social-economic importance of bridges use, (ii) the necessary safety assurance and (iii) the high costs of any physical intervention, there is a need for continuous online bridge monitoring, for investment and use optimization. Recently, smart structures, which combine remote sensors (which send a stream of time series data) with intelligent information systems for real-time decision support through embedded Data Mining (DM) models, have been proposed to handle this task. Indeed, the application of DM techniques to analyse civil engineering data has gained an increasing interest in recent years, due to intrinsic characteristics such as the ability to deal with nonlinear relationships. In this study, Artificial Neural Networks have been used to predict the following ratios: global efficiency, structural adequacy and safety, serviceability, essentiality for public use, and special reductions, using a ratio-based framework, and data collected during inspections of bridges in the north of Portugal. In particular, the global efficiency ratio is very useful to identify intervention priorities and to schedule the repair, strengthening and rehabilitation needs. The obtained results are encouraging and the most accurate model for global efficiency presents a low error (Root Mean Squared Error of 0.149). This approach opens room for the development of intelligent decision support systems for Bridge Management Systems. These systems are being recognized as a good way to systematize all the management process and to minimize the ratio cost/benefit during the bridge lifetime.