A modern bridge is such a complicated system that is difficult to analyze by conventional mathematic tools. A rational bridge monitoring requires a good knowledge of the actual condition of various structural components. Fatigue analysis of concrete bridges is one of the most important problems. Concrete bridges are often undergoing a fatigue deterioration, starting with cracking and ending with large holes through the web. There is a need for the development of efficient health assessment system for fatigue evaluation and prediction of the remaining life. This information has clear economical consequences, as deficient bridges must be repaired or closed. The goal of this research is to provide a practical expert system in bridge health evaluation and improve the understanding of bridge behavior during their service. Efforts to develop a functional bridge monitoring system have mainly been concentrated upon successful implementation of experienced-based machine learning. The reliability of the techniques adopted for damage assessment is also important for bridge monitoring systems. By applying the system to an in-service PC bridge, it has been verified that this fuzzy logic expert system is effective and reliable for the bridge health evaluation.
Smart Health Monitoring System (SHMS) is a set of integrated system of hardware and software designed to automatically collect and analyze the data from a faraway bridge. The real-time data can be preprocessed in the sub-workstation on the bridge then transferred to the main server with a wired or wireless internet access. SHMS is based on the statistical analysis of the static and dynamic characteristics of structures. In order to automate the procedure of processing and analyzing all the raw data, a rule-based expert system was developed for the monitoring system with Bootstrap Method. In general, the estimation of parameters from measurement always contains systematic perturbations and random fluctuations. The systematic perturbations mainly come from periodic environmental factors, especially temperature. Random fluctuations result from irregular disturbance including instrumentation sources and numerical processing algorithms. The former can be identified and characterized. Based on the historical data, a set of correction models have been built to remove the influence from systematic perturbations. Random fluctuations are difficult to be characterized by traditional statistical methods. But with Bootstrap Method, we can minimize the random error.
The Kishwaukee bridge (circa 1979) is a five-span pre-cast post-tensioned segmental concrete box girder bridge. The structure has been under increasingly stringent inspection since extensive cracking adjacent to the piers in the webs was observed. There is limited observational evidence that continued propagation of the cracks has been occurring, and static load testing of the structure in 2000 provided direct indications of locally excessive stresses in the shear reinforcement. Since that time, the authors have developed
instrumentation which continuously monitors the bridge. Our previous thermal analyses and modeling based upon the data collected
demonstrates that thermal equilibrium is rare for this structure, a situation that undoubtably applies to most bridges of this type. Partly due to this constant disequilibrium, basing alarm criteria
on real time measurements is untenable. In this paper we develop a two-step monitoring strategy which can be applied to both global and local deformation data. The strategy is based on preliminary modeling to identify the thermomechanical influences, followed by a second step using a bootstrap comparator.
A sizeable number of efforts have sought to instrument bridges
for the purpose of structural monitoring and assessment. The outcomes
of these efforts, as gaged by advances in the understanding of the
definition of structural damage and their role in sensor selection as well as in the design of cost and data-effective monitoring systems, has itself been difficult to assess. The authors' experience
with the design, construction, and operation of a monitoring system for the south-bound Kishwaukee Bridge has provided several lessons that bear upon these concerns. In this paper we describe certain aspects of the design of our Unix-based monitoring system. The system, patterned after similar systems developed for kick detection and well-control in the oil industry, has performed well in providing a continuous, low-cost monitoring platform for bridge engineers with immediately relevant information.