Date of Award

Fall 2023

Document Type

Open Access Dissertation

Department

Civil and Environmental Engineering

First Advisor

Paul Ziehl

Abstract

The South Carolina Department of Transportation (SCDOT) is currently engaged in a multi-year effort to assess the structural integrity of its inventory of over 9,000 bridges. This assessment process is expected to result in an increase in the number of bridges that require load postings, repairs, or replacements across South Carolina. This could potentially lead to adverse economic repercussions due to restricted truck routes, bridge closures, repair work, and the need for bridge replacements. To alleviate the escalating costs associated with these challenges, it is imperative to explore methods aimed at reducing the need for load postings and bridge closures or replacements. This study is a part of an ongoing multi-year research investigation, which is being funded and supported by SCDOT. The study has three main objectives: (a) to investigate methods to strengthen the skinny leg channel girder bridges to reduce the number of load postings, (2) utilize acoustic emission (AE) parameters for condition assessment of in-service bridges and determine the condition factor for load ratings of the bridges, and (3) extend the use of AE to determine the vehicle loads on the bridges while monitoring the deterioration. Three studies were performed to fulfill the objectives. The first study addresses the existing gap in research by investigating different methods for strengthening prestressed skinny-leg girder bridges. Aluminum alloys possess desirable properties that make them attractive as external reinforcement materials. However, while previous studies have investigated the use of aluminum alloys on reinforced concrete members, there is a lack of investigation on their use on full-scale prestressed concrete bridge girders. To address this gap, this study investigates the feasibility of utilizing aluminum alloy channels as an external reinforcement material on full-scale prestressed concrete bridge girders. Nine girders obtained from decommissioned bridges in South Carolina were tested under monotonic loading to failure. The test program consisted of six unstrengthened girders, one strengthened with bonded aluminum channels (SE), one strengthened with bonded and bolted aluminum channels (SEB), and one strengthened with bolted aluminum channels (SB). The results indicate that externally anchoring aluminum alloy channels with bolts was the most effective strengthening method in terms of practicality and higher increase in the moment capacity. The second study proposes a data-driven condition assessment for in-service bridges using acoustic emission. The study aims to utilize AE parameters to assess the condition of the girders and to determine the condition factor used in the load rating of the bridges. Six prestressed concrete channel bridge girders, which were originally used in 30-ft span bridges constructed in the 1960s, were subjected to flexural tests at the University of South Carolina (USC). Acoustic emission (AE) was used to monitor the girders during the tests. The girders were visually inspected prior to testing and each girder was assigned a condition rating based on the Specifications for the National Bridge Inventory (SNBI). Intensity analysis charts were developed based on the collected AE data. The charts may detect if the girders are operating within the specified design criteria and are calibrated to theoretical cracking load and findings of cumulative signal strength analysis. In addition, the charts may assess the deterioration regardless of the initial girder condition, which can be utilized to determine the condition factor of the girders for load rating purposes. The innovation of the third study lies in presenting a potential approach for predicting the vehicle loads on prestressed concrete girder bridges from the collected AE data. To achieve this goal, three improved machine learning algorithms based on artificial neural networks (ANN), AdaBoost, and random forest were adopted to analyze the AE data. An ensemble training strategy was employed to eliminate the imbalance issue while training machine learning models. The AE data was collected by conducting a flexural test on a full-scale prestressed concrete girder. In this study, load determination is considered a classification problem. The loading procedure was divided into load steps and the AE signals were classified to their corresponding load steps. The results show that the improved random forest algorithm outperformed the improved ANN and AdaBoost algorithms in classifying AE hits to their corresponding load steps.

Rights

© 2024, Elhussien Khaled Elbatanouny

Download

Share

COinS