School of Civil and Environmental Engineering

Ph.D. Thesis Defense Announcement

Finite Element Model Updating through Nonconvex Optimization

By

Xinjun Dong

Co-Advisors:

Dr. Yang Wang (CEE), Dr. Rafi L. Muhanna (CEE)

Committee Members:

Dr. Santanu S. Dey (ISYE), Dr. Glaucio Paulino (CEE), Dr. Phanish Suryanarayana (CEE)

Date & Time: Friday, December 7, 10am

Location: Mason 1133

During the past few decades, great efforts have been devoted towards finite element (FE) modeling of structures, in order to simulate the structural behavior under various loading conditions. Due to the complexity of large-scale civil structures, the simulation results generated by an FE model are usually different from these of the as-built structure. To reduce the difference, selected structural parameters can be updated by utilizing the data collected from the actual structure. This process is known as FE model updating.

This research explores FE model updating utilizing the measured frequency-domain modal properties, i.e. resonance frequencies and mode shapes. Naturally, frequency-domain FE model updating is formulated as optimization problem aiming to minimize the difference between simulated and experimentally-measured modal properties. This research focuses on three frequency-domain model updating formulations, i.e. MAC value, eigenvector difference and modal dynamic residual formulations. Local search optimization algorithms are first studied for comparison. The performance of model updating formulations and local search optimization algorithms are validated through numerical simulations, laboratory and field experiments. To collect structural vibration data from the actual structure, a new wireless sensing node, named Martlet, is developed.

To overcome the limitation of local search optimization algorithms, this research also investigates two global optimization algorithms, i.e. branch-and-bound and primal-relaxed dual algorithms to solve the optimization problems in FE model updating. Again, the performance of the two global optimization algorithms are validated through both numerical simulation and laboratory experiment.