Ph.D. Thesis Defense Announcement

State, Parameter, and Input Estimation Through MMSE Estimation for Structural Systems

By 

Peter Lander

Advisor:

Dr. Yang Wang (CEE) and Dr. Jacob Dodson (AFRL)

Committee Members: 

Dr. Lauren Stewart (CEE), Dr. Ryan Sherman (CEE),

Dr. Xi Liu (Intuitive Surgical, Inc.), Dr. Ying Zhang (ECE)

Date and Time:  June 8, 2022 at 1:00 PM EST

Location: SEB122 and Online: Zoom

Complete announcement, with abstract, is attached.

The research areas of bridge weigh-in-motion and high-rate structural health monitoring seek to use sensor data to gain real-time insights into monitored structures and their operating conditions. This thesis contributes to both research areas by applying minimum mean square error estimation (MMSE) algorithms. Specifically, bridge weigh-in-motion estimates the weights of vehicles as they cross a bridge by measuring the bridge’s dynamic response. This thesis details a full-scale experimental validation of the recently developed finite input covariance estimator for bridge weigh-in-motion in tandem with the development of a low-cost autonomous bridge weigh-in-motion system. Experimental results using acceleration and strain measurements demonstrated that the finite input covariance estimator achieves axle weight estimates with less than 10% error. When combined with the extended Kalman filter, the same finite input covariance estimator can be applied to nonlinear systems. High-rate structural health monitoring seeks to track the state of structures experiencing nonlinearities while being subjected to high-rate events, like impacts. This research presents the development and application of the extended version of the finite input covariance estimator towards performing joint state-input-parameter estimation. The proposed estimator is validated through simulations and experiments based on a nonlinear high-rate testbed. The estimator is found to be capable of performing accurate joint state-input-parameter estimation in the presence of impacts and high-rate parameter changes. Additionally, the proposed estimator maintains the stability of the original finite input covariance estimator in scenarios where only acceleration measurements are available.