Abstract
The problem of modeling a linear dynamic system is discussed and a novel approach to automatically combine black-box and white-box models is introduced. The solution proposed in this contribution is based on the usage of regularized finite-impulse-response (FIR) models. In contrast to classical gray-box modelling, which often only optimizes the parameters of a given model structure, our approach is able to handle the problem of undermodeling as well. Therefore, the amount of trust in the white-box or gray-box model is optimized based on a generalized cross-validation criterion. The feasibility of the approach is demonstrated with a pendulum example. It is furthermore investigated, which level of prior knowledge is best suited for the identification of the process.
Zusammenfassung
Als Problemstellung wird die Modellierung linearer dynamischer Systeme betrachtet. Hierzu wird ein neuer Ansatz zur automatischen Kombination von datenbasierten und physikalischen Modellen vorgestellt. Der in diesem Aufsatz vorgeschlagene Ansatz basiert auf der regularisierten Schätzung endlicher Impulsantwortmodelle. Im Gegensatz zur klassischen Kombination physikalischer und datenbasierter Modelle, bei denen einzelne Parameter des physikalischen Modells basierend auf Eingangsdaten geschätzt werden, ist unser Ansatz in der Lage, ebenso Fehler in der Modellstruktur des physikalischen Modells zu berücksichtigen. Dafür wird ein Strafterm für die Abweichung vom physikalischen Modell bei der Modellierung des datenbasierten Modells berücksichtigt und die Höhe des Strafterms auf Basis des verallgemeinerten Kreuzvalidierungskriteriums optimiert. Die Machbarkeit des Ansatzes wird an dem Beispiel eines an einer Feder befestigten beweglichen Pendels illustriert. Weiterhin wird untersucht, welcher Umfang an Vorwissen zur Identifikation des Prozesses am Besten geeignet ist.
About the authors
Tobias Münker received his B. Sc. from Universität Siegen in 2010 and his M. Sc. from TU Darmstadt in 2012. After 3 years of industry experience, he has been working towards his Ph. D. under the supervision of Prof. Nelles since 2015. His main research interests are new techniques for the identification of linear and nonlinear systems.
Timm J. Peter graduated with a Master of Science degree from Universität Siegen in 2018. After finishing his masters thesis about regularized FIR models he joined the working group Automatic Control – Mechatronics of Prof. Nelles as a research assistant. His research topics focus on new techniques for linear and nonlinear system identification.
Oliver Nelles is Professor at the University of Siegen in the Department of Mechanical Engineering and chair of Automatic Control – Mechatronics. He received his doctor’s degree in 1999 at the Technical University of Darmstadt. His key research topics are nonlinear system identification, dynamics representations, design of experiments, metamodeling and local model networks.
AppendixParameters of the physical model
For completeness, the parameters of the physical model are listed below. The values from Tab. 3 allow the calculation of the coefficients in Tab. 4.
Physical constants for the cart and pendulum.
| Constant | Value |
| Spring Rate | |
| Motor Armature Resistance | |
| Motor Armature inductance | |
| Motor Torque Constant | |
| Motor Efficiency | |
| Back-Electromotive-Force Constant | |
| Gear Ratio | |
| Planetary Gearbox Efficiency | |
| Rotor Moment of Inertia | |
| Cart Mass | |
| Cart Weight Mass | |
| Motor Pinion Radius | |
| Motor Pinion Number of Teeth | |
| Position Pinion Number of Teeth | |
| Position Pinion Radius | |
| Rack Pitch | |
| Cart Travel | |
| Cart Encoder Resolution | |
| Pendulum Encoder Resolution | |
| Damping Coefficient Engine | |
| Acceleration of Gravity | |
| Cart Mass Total | |
| Pendulum Mass | |
| Pendulum Full Length | |
| Pendulum Distance Center of Gravity | |
| Pendulum Moment of Inertia | |
| Damping Coefficient Pendulum |
Formulas for the computation of the white-box-model coefficients.
| Constant | Formula |
| Denominator | |
| Denominator | |
| Engine Coefficient | |
| Engine Coefficient | |
References
1. Hirotugu Akaike. A new look at the statistical model identification. Automatic Control, IEEE Transactions on, 19(6):716–723, 1974.10.1109/TAC.1974.1100705Search in Google Scholar
2. Torsten Bohlin. A case study of grey box identification. Automatica, 30(2):307–318, 1994.10.1016/0005-1098(94)90032-9Search in Google Scholar
3. Torsten Bohlin and Stefan F Graebe. Issues in nonlinear stochastic grey box identification. International Journal of Adaptive Control and Signal Processing, 9(6):465–490, 1995.10.1002/acs.4480090603Search in Google Scholar
4. Tianshi Chen and Lennart Ljung. Implementation of algorithms for tuning parameters in regularized least squares problems in system identification. Automatica, 49(7):2213–2220, 2013.10.1016/j.automatica.2013.03.030Search in Google Scholar
5. Tianshi Chen and Lennart Ljung. Regularized system identification using orthonormal basis functions. arXiv preprint arXiv:1504.02872, 2015.Search in Google Scholar
6. Tianshi Chen and Lennart Ljung. On kernel design for regularized LΤΙ system identification. arXiv preprint arXiv:1612.03542, 2016.Search in Google Scholar
7. Tianshi Chen, Henrik Ohlsson, and Lennart Ljung. On the estimation of transfer functions, regularizations and Gaussian processes-revisited. Automatica, 48(8):1525–1535, 2012.10.1016/j.automatica.2012.05.026Search in Google Scholar
8. Francesco Dinuzzo. Kernels for linear time invariant system identification. SIAM Journal on Control and Optimization, 53(5):3299–3317, 2015.10.1137/130920319Search in Google Scholar
9. Torkel Glad and Lennart Ljung. Control theory. CRC press, 2000.Search in Google Scholar
10. Rolf Isermann. Mechatronic systems: fundamentals. Springer Science & Business Media, 2007.Search in Google Scholar
11. Lennart Ljung. System identification, volume 2. Prentice-Hall, 1999.10.1002/047134608X.W1046Search in Google Scholar
12. Anna Marconato, Maarten Schoukens, and Johan Schoukens. Filter-based regularisation for impulse response modelling. IET Control Theory & Applications, 11(2):194–204, 2016.10.1049/iet-cta.2016.0908Search in Google Scholar
13. Tobias Münker, Julian Belz, and Oliver Nelles. Improved incorporation of prior knowledge for regularized FΙR model identification. In American Control Conference (ACC), 2018 (accepted for publication).10.23919/ACC.2018.8431661Search in Google Scholar
14. Tobias Münker and Oliver Nelles. Sensitive order selection via identification of regularized fir models with impulse response preservation. In IFAC Symposium on System Identification (SYSID18), 2018 (submitted for publication).10.1016/j.ifacol.2018.09.129Search in Google Scholar
15. Oliver Nelles. Nonlinear system identification: from classical approaches to neural networks and fuzzy models. Springer Science & Business Media, 2001.10.1007/978-3-662-04323-3Search in Google Scholar
16. Gianluigi Pillonetto, Tianshi Chen, Alessandro Chiuso, Giuseppe De Nicolao, and Lennart Ljung. Regularized linear system identification using atomic, nuclear and kernel-based norms: The role of the stability constraint. Automatica, 69:137–149, 2016.10.1016/j.automatica.2016.02.012Search in Google Scholar
17. Gianluigi Pillonetto and Giuseppe De Nicolao. A new kernel-based approach for linear system identification. Automatica, 46(1):81–93, 2010.10.1016/j.automatica.2009.10.031Search in Google Scholar
18. Gianluigi Pillonetto, Francesco Dinuzzo, Tianshi Chen, Giuseppe De Nicolao, and Lennart Ljung. Kernel methods in system identification, machine learning and function estimation: A survey. Automatica, 50(3):657–682, 2014.10.1016/j.automatica.2014.01.001Search in Google Scholar
© 2018 Walter de Gruyter GmbH, Berlin/Boston
