Main content

Sliding Innovation Filtering: Theory and Applications

Show full item record

Title: Sliding Innovation Filtering: Theory and Applications
Author: Lee, Andrew Sanghyun
Department: School of Engineering
Program: Engineering
Advisor: Gadsden, Stephen Andrew
Abstract: Estimation strategies aim to extract useful state or parameter information from systems with uncertainties and noisy measurements. The sliding innovation filter (SIF) is an estimator based on the predictor-corrector methodology that achieves robustness to modeling uncertainties and disturbances. The SIF, as originally presented, does not produce the optimal solution to the linear estimation problem. Instead, the SIF implements a corrective gain using the innovation and a fixed sliding boundary layer based on variable structure techniques. The research presented in this thesis expands upon the original SIF formulation to improve its performance and increase the number of useful applications. Unlike classical model-based filters, the original SIF does not incorporate the state error covariance in the calculation of the corrective gain. The first contribution of this research involves the derivation of a corrective gain for the SIF based on minimizing the state error covariance. This results in a time-varying sliding boundary layer that adapts to changes in the system model. The adaptive formulation is applied to an electrohydrostatic actuator (EHA) and demonstrates increased accuracy over the original SIF. The second contribution of this research combines the SIF with other estimators such as the Kalman filter (KF) and its variants. The time-varying sliding boundary layer is used as a criterion for switching between the two filters. This new strategy is applied to a magnetorheological (MR) damper setup, which was designed, built, and modelled to support these research activities. When the system operates in the presence of faults, the new filters demonstrate improved accuracy and robustness when compared with standard Kalman filters for nonlinear systems. The final contribution of this research is the formulation of the new IMM-ESIF which combines the interacting multiple model (IMM) strategy with the SIF (extended for nonlinear systems) for fault detection and diagnosis. The IMM-ESIF was applied to the MR experimental setup and demonstrates improved estimation accuracy and classification confidence over the well-known IMM-EKF and IMM-UKF strategies. The research presented in this thesis expands upon the SIF and creates a foundation for further research and advancement.
URI: https://hdl.handle.net/10214/26361
Date: 2021-09
Terms of Use: All items in the Atrium are protected by copyright with all rights reserved unless otherwise indicated.
Related Publications: A. S. Lee, S. A. Gadsden and M. Al-Shabi, "An Adaptive Formulation of the Sliding Innovation Filter," in IEEE Signal Processing Letters, vol. 28, pp. 1295-1299, 2021, doi: 10.1109/LSP.2021.3089918


Files in this item

Files Size Format View Description
Lee_Andrew_202109_PhD.pdf 3.593Mb PDF View/Open PhD Thesis

This item appears in the following Collection(s)

Show full item record