Details:

Creators:
Burgueño, Rigoberto ; Zhe, Li
Burgueño, Rigoberto ; Zhe, Li Less -
Corporate Creators:
Michigan State University. Dept. of Civil and Environmental Engineering
Corporate Contributors:
Michigan. Dept. of Transportation. Construction and Technology Division
Subject/TRT Terms:
[+]
Bridge Abutments Bridge Design Bridges And Other Structures Highways I24: Design Of Bridges And Retaining Walls Neural Networks Structural Analysis Time Dependence
Publication/ Report Number:
RC-1508
Resource Type:
Tech Report
Geographical Coverage:
Michigan ; United States
Michigan ; United States Less -
TRIS Online Accession Number:
01115034
Corporate Publisher:
Michigan. Dept. of Transportation. Construction and Technology Division
Abstract:
Structural distress in the abutments walls of bridges in the form vertical cracks along the visible wall height and u-shape cracks under girder supports is problem that the Michigan Department of Transportation has been trying to understand and address. While possible causes of damage have been hypothesized, the specific mechanisms and their relative importance as a cause of the damage are not well understood. The objective of this research was thus to identify the causes behind abutment damage, propose solution strategies and develop prediction models to improve maintenance and future design. Given the time-dependent nature of the problem, the research was based on the use of field inspection data from the National Bridge Inventory record for Michigan’s bridges. However, data from manual inspections is qualitative, unbalanced, subjective, with errors and incomplete. The research approach was thus to use statistical methods, data mining techniques, and artificial intelligence models to interpret the information captured in this database. Statistical analyses were used to extract an information database from the general inspection record and thus identify parameters that could serve as explanatory variables in prediction models. A family of bridges sharing statistically significant parameters related to abutment damage was carefully inspected and four bridges were monitored for 1-year. Strains and displacements on the abutment walls of the monitored bridges were used as a dynamic database for the identification of damage sources. A large case-matrix of finite element simulations was used to develop a virtual database of abutment performance to support the evidential database in prediction models and help establish the relative importance of damage-causing mechanisms. Four different artificial neural networks (ANN) models were developed and validated to predict the structural condition of existing and new bridge abutments. The individual ANN models provided satisfactory performance but suffered from the unbiased and subjectivity of the inspection databases. ANN ensembles with novel data handling techniques and diverse voting in virtual committees were developed and proven to alleviate these problems and led to improved accuracy in the prediction models. An ANN ensemble model was implemented into a computer program (SbNet) that can predict bridge abutment condition and life-time degradation given design parameters or a bridge identification number. The findings indicate that the main causes behind abutment distress are the pressures from pavement approaches and temperature gradients. Strategies to relieve these effects are well known and they include: the use of flexible pavements, pavement pressure relief joints, improved expansion joint seals, smaller skew angles, use of expansion bearings at abutments, and the elimination of pin-and-hanger assemblies.
Format:
PDF
Collection(s):
US Transportation Collection
Main Document Checksum:
[+]
urn:sha256:92a2faedee0cfaf81d88f02675334644a39a41dacbeb036c7ba0fff46cc2f3cd
Download URL:
https://rosap.ntl.bts.gov/view/dot/39515/dot_39515_DS1.pdf
File Type:

Supporting Files

• No Additional Files

More +