The magnetorheological (MR) damper is a type of controllable fluid damper that can be applied to modify
the response of structures for improved performance and disaster resilience under extreme loading events.
The MR fluids inside the damper have the ability to reversibly change from a liquid to a semi-solid upon application of a
magnetic field, increasing the yield stress in the fluid, and generating forces in these devices that
are up to 50 times larger than without the field (see Blog
for more information on MR fluid).
Both small- and large-scale MR damper testing platforms (see figures above) are currently available in the lab. We are designing advanced MR damping systems with novel controllers to provide optimal response reduction effects. The developed MR damping systems can be widely applied in structures subjected to random dynamic loading contions.
We are interested in developing novel damping devices and systems. The figures shown above is the shake table tests conducted on a special typed mass damper--disruptive mass damper (DTM).
The figures shown above is the shake table tests conducted on a multi-unit rolling isolation system (RIS). This type of isolation systems is used to isolate mission-critical equipment (server cabinet in this case) from earthquake-induced ground and floor motions.
With recent advances in sensing technologies, materials science and damage identification algorithms,
many innovative structural health monitoring (SHM) systems have been developed. We are intersted in developing advanced SHM
systems that are capable to provide more rapid structural condition assessment, and more robust to
environmental influences and noise impacts.
A campus bridge (river-side pedestrian bridge, shown in the above figure) will be instrumented to serve as a test bed for this effort. Different types of sensors will be installed on the bridge to collect vibration, deformation and environmental data. A numerical model of the bridge will be updated using the collected data, and then applied to damage diagnosis and prognosis study.
We are conducting research on a steel box-girder bridge on interstate I20/59. Various types of sensors have been installed on the bridge to measure stress and deflection features of bridge as indications of its health condition.
Real-time hybrid simulation (RTHS) is a novel, powerful and cost-effective experimental technique
for examining the behavior of complex, full-scale structural systems under realistic loading conditions. It leverages the
basic framework of pseudodynamic testing with both physical and simulated components, but applying advanced algorithms
to reconcile these two components to provide real-time loading rate as the experiment progresses (see the figure above).
We are intersted in developing novel real-time hybrid simulation platform to investigate rate-dependent behavior of large- or full-scale structural components. The platform needs to be accurately compensated for delay and lags caused by servo-hydraulic systems. It also needs to have good robustness in response to the complex conditions during large-scale structural testing.
Learning-based methods can potentially alleviate the above subjectivities by directly learning from data,
making predictions via self-learned pattern recognition and extraction. In literature, some common machine learning
methods such as k-nearest neighbors (k-NN) , support vector machine (SVM), and neural network (NN) ,
have been explored by researchers for structural condition assessment, such as defect identification.
As computer algorithms and high-performance computing devices rapidly evolve in recent years, deep convolutional neural network (DCNN), a type of deep learning-based methods, has been extensively explored for crack detection tasks. Compared with regular machine learning methods, DCNN has the advantage of achieving multi-level feature abstraction through a deep architecture layout, and thus enabling the ability of self-adaptation and adjustment to reflect a highly complex model.
We are intersted in developing new DCNN based algorithms to evaluate structural defects. Currently, we have been conducting DCNN-based research on crack identification problems (see figure above). The goal is to provide robust and accurate crack identification results using images with real-world complexities.
The image data used in the research are collected in the roadway surveys by a mobile automated 3D laser system. Our DCNN algorithms have shown robust and accurate performance in crack classification and segmentation with these real-world data.