Observer-Based Simultaneous States and Parameters Estimation Method with Application to System Heath Monitoring

Joint state and parameter estimation is paramount in many engineering and scientific fields, as it involves determining the internal states of a system and the estimation of its time-varying / unknown parameters simultaneously. This twin estimation aspect is crucial for real-time system monitoring, fault diagnosis and the optimization of system control strategies. Provided in this thesis is a comprehensive review of the existing simultaneous states and parameters estimation techniques from the literature, discussing the strengths and limitation of each approach. As an industrial application, an augmented extended Kalman filter is implemented to estimate the substrate temperature in the Advanced Metal Organic Chemical Vapor Deposition (AMOCVD) process, used for High-Temperature Superconductor tapes manufacturing. Precise temperature control in this process is crucial to minimize nano-scale defect growth during production. However, accurate substrate temperature measurement is hindered by sensor drift caused by the deposition of precursor gases over the crystal rod, obstructing the area between the pyrometers and the substrate. To overcome this challenge, a physics-based heating model for the substrate is developed and an augmented extended Kalman filter is implemented for the simultaneous estimation of the substrate temperature and heating model parameters in real-time. A particle swarm optimization algorithm is employed to systematically tune the filter’s noise covariance matrices and avoid the ad-hoc manual selection of those matrices. The estimated temperature is then fed into the control system to regulate the flow of current through the tape, driving the substrate temperature to the desired set point. Inspired by the latter work, a novel simultaneous state and parameter estimation algorithm based on Luenberger observer is proposed. In this approach, a modified Luenberger observer is employed for system states estimation and a recursive least squares estimation is used to estimate the unknown time-varying parameters of the model, in real-time. In the formulation of the novel approach, the unknown parameters of the model are represented as additive uncertainty matrices to the state and input matrices in the state space system representation. The utility of this formulation is that the model uncertainties are confined to the structure of the state matrices and thus allows the identification of the location(s) within the state matrices requiring parameter adaptation. Based on the location(s) and size of these adaptations, real-time health monitoring and health degradation isolation for a system can be realized thereby enabling prognostics, remaining useful life estimation, and forecasting.

Author dc:creator

• Chebbi, Amal

Contributors dc:contributor

• Franchek, Matthew A.
• Grigoriadis, Karolos M.
• Chen, Zheng
• Song, Gangbing
• Becker, Aaron T.
• Cescon, Marzia

• Engineering, Mechanical

Language dc:language

en

Identifier

Portions of this document appear in: Chebbi, Amal, Karolos Grigoriadis, Matthew Franchek, and Marzia Cescon. "Substrate temperature estimation and control in advanced MOCVD process for superconductor manufacturing." The International Journal of Advanced Manufacturing Technology (2024): 1-13. https://doi.org/10.1007/s00170-024-13699-1. Chebbi, Amal, M. A. Franchek, and K. Grigoriadis, "A Modified Observer Method for the Joint Estimation of States and Parameters for the Class of Linear Uncertain Discrete-Time Systems," International Journal of Engineering Research and Applications, vol. 14, no. 5, pp. 07-18, 2024.

OAI identifier oai:identifier

oai:tdl-ir.tdl.org:10657/18366