Date of Award
Doctor of Philosophy (PhD)
The Federal Highway Administration (FHWA) recognizes the necessity for cost-effective and practical system identification (SI) techniques within structural health monitoring (SHM) frameworks for asset management applications. Indirect health monitoring (IHM), a promising SHM approach, utilizes accelerometer-equipped vehicles to measure bridge modal properties (e.g., natural frequencies, damping ratios, mode shapes) through bridge vibration data to assess the bridge's condition. However, engineers and researchers often encounter noise from road roughness, environmental factors, and vehicular components in collected vehicle signals. This noise contaminates the vehicle signal with spurious modes corresponding to stochastic frequencies, impacting damage monitoring assessments. Thus, an efficient and reliable SI technique is required to process vehicle signals and extract bridge features effectively before practical deployment. To achieve this, vehicle-bridge interaction (VBI) models are often developed to simulate physical data for either initial verification of SI methodologies or for use in a model-updating algorithm to determine the bridge modal properties by tuning the model to the physical data. Common steps in the SI process include signal processing of the raw data, operational modal analysis (OMA), and leveraging machine learning (ML). This dissertation proposes a framework for efficient creation of VBI models using commercial code, develops an autonomous SI technique (APPVMD) to extract bridge frequencies from passing vehicles, provides guidelines for improving bridge frequency extraction with multi-vehicle scenarios via an extensive analytical study, demonstrates the need for improved methodologies for simulating road surface roughness effects in VBI models via comparison with physical data, and provides a substantial archive of test data and models that can be leveraged in future studies (road surface profiles using laser profilometers and vehicle acceleration data from four-post shaker testing with the associated vehicle model). The work encompasses four major studies aimed at achieving these research objectives.
The first study presents a computationally efficient VBI modeling framework in commercial finite element (FE) software (Abaqus) requiring minimal user coding, suitable for industrial and research communities. The framework's dynamic response is verified using literature data, and a damage modeling methodology is proposed to extend the framework to SHM applications with SI techniques in IHM.
In the second study, an autonomous peak-picking variational mode decomposition (APPVMD) framework is introduced to enhance scalability in SI techniques for the IHM of a bridge network. APPVMD leverages signal processing techniques and heuristic models to autonomously extract bridge frequencies from vehicle acceleration responses without prior information or model-informed training. The framework is tested on different vehicles and bridge classes to assess its feasibility, achieving successful bridge frequency extractions in many cases.
In the third study, an extensive parametric study is undertaken to determine if multiple vehicle scenarios would enhance bridge frequency identification and what vehicle types and driving speeds would be most effective. Four vehicles are considered representative of true vehicle properties found in the literature, and six bridges are taken from physical bridge data, including drawings obtained from both the literature and the South Carolina Department of Transportation (SCDOT). The study unveils interesting phenomena regarding the complex interaction between vehicles and bridges, performs brief case studies to improve bridge frequency extractions further, and proposes guidelines that researchers and engineers can follow when preparing to collect acceleration data from vehicles for bridge SI.
The fourth study presents preliminary work to experimentally show that current methodologies for representing road surface roughness effects are insufficient. First, a vehicle model is developed to include road surface roughness effects and compared with experimental data collected in a previous study at Clemson University. The results suggest that commonly referenced roughness factors in the literature underestimate road surface roughness effects while inputting the average values based on road class from the ISO-8608 standard tends to exaggerate road surface roughness effects. A moving-average filter (MAF) was found to help attenuate noise but requires appropriate parameter selection. Recommendations for improving road surface roughness modeling in VBI problems are provided. Further work is conducted on a BMW 535 Xi with enhanced ride quality, including verification exercises using a four-post shaker and extensive road tests for real-life road roughness measurements during driving. The study concludes with the suggested path forward for utilizing collected data. It suggests additional tests that can unveil the tire behavior during road tests to compute the transfer function between the road surface roughness and the unsprung masses in VBI models.
This dissertation concludes by summarizing the contributions made to the field IHM of bridges and outlines the next steps for future research.
Abuodeh, Omar, "Improved Vehicle-Bridge Interaction Modeling and Automation of Bridge System Identification Techniques" (2023). All Dissertations. 3364.
Author ORCID Identifier