Adaptive SVSF-KF estimation strategies based on the normalized innovation square metric and IMM strategy

• Published in: Results in engineering Vol. 16; p. 100785
• Main Authors: Goodman, Jacob, Hilal, Waleed, Gadsden, S. Andrew, Eggleton, Charles D.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2022; Elsevier
• ISSN: 2590-1230; 2590-1230
• DOI: 10.1016/j.rineng.2022.100785

The smooth variable structure filter (SVSF) is highly robust to modeling uncertainty and unknown disturbances. Recent developments to the SVSF have allowed for the creation of an adaptive estimation scheme termed the SVSF-KF, which balances the optimality of the Kalman filter (KF) with the robustness of the SVSF. The approach utilizes the KF estimate during normal operation, and utilizes the robust SVSF gain to estimate the states during the presence of a fault. However, the gain adaptation involved in detecting a fault and switching between the KF and SVSF estimates suffers from several limitations, including unwanted chattering. In this work, we review the original SVSF-KF approach and present two novel SVSF-KF strategies based on the normalized innovation square metric and the interacting multiple model strategy to address these limitations. Experimental simulations involving a simple harmonic oscillator subject to a fault condition are conducted, which verify the effectiveness of our proposed approaches. •Existing SVSF-KF approaches suffer from chattering due to high-frequency gain switching.•Significant degradation in estimation performance is encountered due to this chattering.•Two new approaches to control the SVSF-KF gain are proposed, the NIS SVSF-KF and IMM SVSF-KF.•Experimental simulations on a system subject to a fault prove that both approaches eliminate the effects of chattering.•Furthermore, the NIS and IMM demonstrate improved estimation performance over the traditional SVSF-KF.