Prior Knowledge for System Identification with Linear and Nonlinear FIR Models

• Applicant: Professor Dr.-Ing. Oliver Nelles
• Subject Area: Automation, Mechatronics, Control Systems, Intelligent Technical Systems, Robotics
• Term: from 2020 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 439767479

• Final Report Year: 2023

Final Report Abstract

The goal of this project was to develop a novel method for system identification for both linear and nonlinear finite impulse response (FIR) models. These models are flexible and inherently stable output error models, which also are linear in their parameters. The key challenge is the high number of parameters and subsequently the high variance error. Incorporation of prior process knowledge through regularization enables the adjustment of the number of effective parameters. For this purpose, a novel regularization technique with the impulse response preserving (IRP) matrix is investigated in detail and improvements of the regularization scheme are presented. These findings are transferred to nonlinear processes by using local model networks (LMNs). The focus lies on the interpretability of the local models and their applicability to real-world processes.

Publications

• Gray-box identification with regularized FIR models. at -Automatisierungstechnik, 66(9), 704-713.
  Münker, Tobias; Peter, Timm J. & Nelles, Oliver
  
• System Identification and Control of a Polymer Reactor. IFAC-PapersOnLine, 53(2), 437-442.
  Münker, Tobias; Kampmann, Geritt; Schüssler, Max & Nelles, Oliver
  
• Adaptive Model Predictive Control with Finite Impulse Response Models. In Proceedings - 31th Workshop Computational Intelligence, pages 149– 168, 2021.
  Christopher Illg; Tim Decker; Jonas Thielmann & Oliver Nelles
  
• Adaptive Model Predictive Control with Regularized Finite Impulse Response Models. ATHENA Research Book, Volume 1, 327-331. University of Maribor Press.
  Illg, Christopher & Nelles, Oliver
  
• Adaptive System Identification with Regularized FIR Models. IFAC-PapersOnLine, 55(12), 1-6.
  Illg, Christopher; Kösters, Tarek; Kiroriwal, Saksham & Nelles, Oliver
  
• Excitation Signal Design and Modeling Benchmark of NOx Emissions of a Diesel Engine. 2022 IEEE Conference on Control Technology and Applications (CCTA), 907-912. IEEE.
  Smits, Volker; Schussler, Max; Kampmann, Geritt; Illg, Christopher; Decker, Tim & Nelles, Oliver
  
• Regularized Local FIR Model Networks for Enhanced Physical Interpretability. In Workshop on Nonlinear System Identification Benchmarks, 2022.
  Christopher Illg & Oliver Nelles
  
• Handling of time delays in system identification with regularized FIR models. at -Automatisierungstechnik, 71(10), 833-844.
  Kösters, Tarek; Illg, Christopher & Nelles, Oliver