Simulation Training and Route Optimization for Bridge Inspection

Presented:  December 4, 2019, 11:00 AM Central Time
Speaker: Dr. Sushil Louis, University of Nevada-Reno

View Webinar

Abstract

Since the 1970s simulation training has developed operational trainers for a variety of complex systems from pilot flight training simulators to cultural awareness training simulations. When connected to the real world, such simulation training interfaces can drive real vehicles and systems. We have been building a Simulation Training And Control System (STACS) for autonomous bridge inspection that uses a simulated world to train inspectors to control a heterogeneous group of robots. The objective is that, once trained, inspectors can use the same STACS interface used in training to control multiple real robots simultaneously during a bridge inspection task. We first built a multi-robot control interface and simulated environment so that a single operator may manage at least two types of robots. Second we developed a new optimization algorithm for automatically and quickly generating near-optimal routing for n robots to cooperatively cover every truss while minimizing inspection time. This webinar describes and demonstrates STACS, provides optimization results corresponding to time (and thus cost) saved in bridge inspection. Results show that we can significantly reduce bridge inspection time with inspection robots and that the time needed decreases  in inverse proportion to the number of robots available for inspection.

Speaker

Dr. Sushil Louis is a full professor of Computer Science at the University of Nevada, Reno and director of the Evolutionary Computing Systems Laboratory (ECSL). He works in the areas of evolutionary computation and its applications in search, optimization, and machine learning. His current interests are in evolutionary approaches to game AI, design optimization, and human behavior modeling. Dr. Louis was an associate editor of the IEEE transactions on Computational Intelligence and Artificial Intelligence in Games and was co-general chair of the 2006 IEEE Symposium on Computational Intelligence and Games.  Please visit his website at http://www.cse.unr.edu/~sushil for additional information.

Data to Risk-Informed decisions through bridge model updating

Presented:  September 25, 2019, 11:00 AM Central Time
Speaker: Dr. Iris Tien, Georgia Institute of Technology

View Webinar

Abstract

Across the country, bridge structures are aging with use commonly extending beyond their original design lives. Decisions to repair, retrofit, or rehabilitate these structures will support continued reliability and resilience of these structures. To better understand the states of the bridges at any point in time, there are increasing new technologies to inspect, monitor, and assess the conditions of these bridges. In order to be able to effectively use the results of these inspections and monitoring activities to support repair, retrofit, and rehabilitation decisions, we need an understanding of the relation between varying inspection parameters and the predicted performance of bridge structures. In this seminar, we will describe methods to use inspection data to update our assessment of bridge performance, focusing on corrosion inspection data. Through bridge model updating, we account for the effects of corrosion, including reduction of longitudinal and transverse reinforcement and bond deterioration between the steel and concrete through corrosion-induced cracking in reinforced concrete bridges. Corrosion is measured by percent mass loss. The impact of measured corrosion parameters on performance is assessed with results quantifying the increase in risk or vulnerability of these structures as corrosion levels increase. Comparing results across bridges supports risk-informed decisions in the management of bridges to protect these structures and ensure their reliability and resilience under future loading and hazard scenarios.

speaker

Dr. Iris Tien joined the faculty in the School of Civil and Environmental Engineering at the Georgia Institute of Technology in 2014 after receiving her Ph.D. in Civil Systems Engineering from the University of California, Berkeley. Dr. Tien’s research interests are in probabilistic methods for modeling and reliability assessment of civil infrastructure systems. She has a unique interdisciplinary background that encompasses traditional topics of civil engineering, sensing and data analytics, stochastic processes, probabilistic risk assessment, and decision making under uncertainty. Her work on interdependent infrastructure systems modeling and analysis has twice won best paper awards in resilient critical infrastructure. Dr. Tien’s research has been funded by the National Science Foundation, Department of Homeland Security, and U.S. Department of Transportation. Dr. Tien has been selected by the National Academy of Engineering to participate in three Frontiers of Engineering Symposia. She was also selected to organize the session on Resilient and Reliable Infrastructure at the U.S. Frontiers of Engineering Symposium.

A Performance-Based Approach for Loading Definition of Heavy Vehicle Impact Events

Presented: June 5, 2019, 11:00 AM Central Time
Speaker: Dr. Anil K. Agrawal, The City College of New York

VIEW Webinar

Abstract

Based on bridge failure data compiled by New York State Department of Transportation, collision, both caused by vessel and vehicles, is the second leading cause of bridge failures after hydraulic.  Current AASHTO-LRFD (2012) recommends designing a bridge pier vulnerable to vehicular impacts for an equivalent static force of 600 kips (2,670 kN) applied in a horizontal plane at a distance of 5.0 feet above the ground level.  This research presents a performance-based approach for designing a bridge pier subject to impacts by tractor-semi-trailer weighing up to 80,000 lb based on an extensive investigation using finite element model of a tractor-semitrailer in LS-DYNA.  In order to ensure the reliability of the proposed approach, parameters of concrete model were calibrated using small-scale impact test and were validated using a large scale test.  Mechanics and modes of failure of bridge pier bents during vehicular impacts were verified through pendulum impact test on a large scale model of three column pier-bent system.  A performance-based approach in terms of shear distortion, plastic rotation and demand / capacity (D/C) ratio has been proposed for the design of bridge piers vulnerable to heavy vehicle impacts.

speaker

Dr. Anil Agrawal is currently a Herbert G. Kayser Professor of Structural / Bridge Engineering at the City College of New York and the Chief Editor of the ASCE Journal of Bridge Engineering.  He has been the past-chair of ASCE Committee on Bridge Inspection, Rehabilitation and Monitoring.  His research interests include inspection and deterioration of bridge elements, robotic inspection of bridge components, post-hazard assessment using drones, behavior of bridges during extreme hazards such as earthquakes, blast, fire, and vehicular impacts on highway bridges, redundancy of long span cable supported bridges and advanced geophysical methods on foundation characterization.  Dr. Agrawal has published more than 250 articles, including more than 100 peer preview journal articles and more than 20 reports.  A report published by FHWA on “A Performance-Based Approach for Loading Definition of Heavy Vehicle Impact Events” can be downloaded from https://rosap.ntl.bts.gov/view/dot/38226.

battery-free wireless strain measurement using an antenna sensor

Presented: March 6, 2019, 11:00 AM Central Time
Speaker: Dr. Yang Wang, Georgia Institute of Technology

VIEW WEBINAR

Abstract

A battery-free antenna sensor can wirelessly measure strain on a structure. Bonded to the surface of a base structure, the antenna sensor deforms when the structure is under strain, causing the antenna’s electromagnetic resonance frequency to change. This resonance frequency change can be wirelessly interrogated and recorded by a reader through electromagnetic backscattering.  A radio frequency identification (RFID) chip on the sensor harnesses a small amount of energy from the interrogation signal and responds to the reader.  The resonance frequency change identified by the reader is then used to determine the strain applied on the structure.  The latest antenna sensor prototype adopts a thermally stable substrate as demonstrated in outdoor tests.  Considering nonlinear constitutive relations, multi-physics simulation is performed to more accurately model the behaviors of the antenna sensor.  In both simulation and laboratory experiments, the antenna sensor is shown to be capable of wirelessly measuring small strain changes.  Finally, an emulated crack testing of the antenna sensor is presented, demonstrating the capability of measuring crack growth in application settings.

speaker

Dr. Yang Wang is an Associate Professor in Civil and Environmental Engineering at the Georgia Institute of Technology, where he also holds an adjunct position in Electrical and Computer Engineering.  Prior to joining Georgia Tech, Dr. Wang received a Ph.D. degree in Civil Engineering and an M.S. degree in Electrical Engineering at Stanford University in 2007.  Dr. Wang’s research interests include structural health monitoring, wireless and mobile sensors, structural system identification, model updating, and decentralized structural control.  He received an NSF Early Faculty Career Development (CAREER) Award in 2012 and a Young Investigator Award from the Air Force Office of Scientific Research (AFOSR) in 2013.  Dr. Wang is the author and coauthor of over 120 journal and conference papers.  He currently serves as an Associate Editor for the ASCE (American Society of Civil Engineers) Journal of Bridge Engineering and for the Structural Health Monitoring journal.  Dr. Wang is an Associate Director and Principal Investigator of the INSPIRE University Transportation Center.

Assistive Intelligence (AI):
Intelligent Data Analytics Algorithms to Assist Human Experts

Presented: January 30, 2019, 11:00 AM Central Time
Speaker: Dr. Zhaozheng Yin, Missouri University of Science and Technology

VIEW WEBINAR

Abstract

Artificial Intelligence, particularly deep learning, has recently received increasing attention in many applications, such as image classification, speech recognition, and computer games. The success of deep learning algorithms requires big annotated datasets for training, gradient-based optimization algorithms, and powerful computational resources. In the case of civil infrastructure inspection, we can collect big data from different imaging sensors such as color, thermal, and hyperspectral cameras. Three issues encounter in this application. First, it is tedious and expensive to let human experts annotate the datasets to train deep learning algorithms. Second, the offline trained deep learning algorithms may not be able to adapt to new civil infrastructures. Third and lastly, the trained deep learning algorithm works like a black box on new data, without the domain knowledge from human experts. In this project, we investigate intelligent data analytics algorithms with human experts in the loop, called Assistive Intelligence (AI). Using the bridge inspection as a case study, we aim to find regions-of-interest (e.g., joints with damages) over long video sequences. The data analytics algorithm is initially trained from a small set of data. Given the dataset of a new bridge, bridge experts only need to annotate a few region-of-interest examples as the seed; our algorithm will retrieve corresponding examples in the rest of videos. Human experts can also return some incorrectly retrieved samples to the data analytics algorithm for further refinement. Thus, while the data analytics algorithm can assist human in an efficient way, bridge experts can leverage their domain knowledge in the adaptation of the computational tool in different scenarios.

speaker

Dr. Zhaozheng Yin received his PhD in Computer Science and Engineering from Penn State in 2009 and received his Master and Bachelor degrees at the University of Wisconsin-Madison and Tsinghua University.  He is specialized in Computer Vision, Image Processing and Machine Learning, with broad applications in structure health monitoring.  Dr. Yin has been a faculty member in Computer Science at Missouri S&T since September 2011. He is a recipient of CVPR Best Doctoral Spotlight Award (2009), MICCAI Young Scientist Award (2012, finalist in 2015 and 2010), NSF CAREER Award (2014), Department Outstanding Junior Faculty Research Award (2014) and Missouri S&T Faculty Research Award (2014, 2017), Best Paper Award of CVPR Workshop on Understanding Hands in Actions (2015). He is a Daniel St. Clair Faculty Fellow in the Computer Science department since 2015, and a Dean’s Scholar in the College of Engineering and Computing since 2016.