Research Papers

Dual Faceted Linearization of Nonlinear Dynamical Systems Based on Physical Modeling Theory

[+] Author and Article Information
H. Harry Asada

Department of Mechanical Engineering,
Massachusetts Institute of Technology,
77 Massachusetts Avenue,
MIT Room 3-346,
Cambridge, MA 02139
e-mail: asada@mit.edu

Filippos E. Sotiropoulos

Department of Mechanical Engineering,
Massachusetts Institute of Technology,
77 Massachusetts Avenue,
MIT Room 3-346,
Cambridge, MA 02139

1Corresponding author.

Contributed by the Dynamic Systems Division of ASME for publication in the JOURNAL OF DYNAMIC SYSTEMS, MEASUREMENT,AND CONTROL. Manuscript received December 24, 2017; final manuscript received August 28, 2018; published online October 5, 2018. Assoc. Editor: Dumitru I. Caruntu.

J. Dyn. Sys., Meas., Control 141(2), 021002 (Oct 05, 2018) (11 pages) Paper No: DS-17-1631; doi: 10.1115/1.4041448 History: Received December 24, 2017; Revised August 28, 2018

A new approach to modeling and linearization of nonlinear lumped-parameter systems based on physical modeling theory and a data-driven statistical method is presented. A nonlinear dynamical system is represented with two sets of differential equations in an augmented space consisting of independent state variables and auxiliary variables that are nonlinearly related to the state variables. It is shown that the state equation of a nonlinear dynamical system having a bond graph model of integral causality is linear, if the space is augmented by using the output variables of all the nonlinear elements as auxiliary variables. The dynamic transition of the auxiliary variables is investigated as the second set of differential equations, which is linearized by using statistical linearization. It is shown that the linear differential equations of the auxiliary variables inform behaviors of the original nonlinear system that the first set of state equations alone cannot represent. The linearization based on the two sets of linear state equations, termed dual faceted linearization (DFL), can capture diverse facets of the nonlinear dynamics and, thereby, provide a richer representation of the nonlinear system. The two state equations are also integrated into a single latent model consisting of all significant modes with no collinearity. Finally, numerical examples verify and demonstrate the effectiveness of the new methodology.

Copyright © 2019 by ASME
Your Session has timed out. Please sign back in to continue.


Cover, T. , 1965, “ Geometrical and Statistical Properties of Systems of Linear Inequalities With Applications in Pattern Recognition,” IEEE Trans. Electron. Comput., EC-14(3), pp. 326–334. [CrossRef]
van Overschee, P. , and de Moore, B. L. , 1996, “ Subspace Identification for Linear Systems—Theory, Implementation, and Applications,” Kluwer Academic Publishers, Boston, MA.
Tohru, K. , 2010, Subspace Methods for System Identification, Springer, London.
Korenberg, M. J. , Bruder, S. B. , and McIlroy, P. J. , 1988, “ Exact Orthogonal Kernel Estimation From Finite Data Records: Extending Wiener's Identification of Nonlinear Systems,” Ann. Biomed. Eng., 16(2), pp. 201–214. [CrossRef] [PubMed]
Simone, O. , 2014, “ Improving the Approximation Ability of Volterra Series Identified With a Cross-Correlation Method,” Nonlinear Dyn., 78(4), pp. 2861–2869. https://link.springer.com/article/10.1007/s11071-014-1631-7
Cleveland, W. , and Devlin, S. , 1988, “ Locally-Weighted Regression: An Approach to Regression Analysis by Local Fitting,” J. Am. Stat. Assoc., 83(403), pp. 596–610. [CrossRef]
Chen, S. , Cowan, C. F. N. , and Grant, P. M. , 1991, “ Orthogonal Least Squares Learning for Radial Basis Function Networks,” IEEE Trans. Neural Networks, 2(2), pp. 302–309. [CrossRef]
Daubechies, I. , 1992, “ Ten Lectures on Wavelets,” Society for Industrial and Applied Mathematics (SIAM), Society for Industrial and Applied Mathematics, Philadelphia, PA.
Krzysztof, K. , and Willi-Hans, S. , 1991, Nonlinear Dynamical Systems and Carleman Linearization, World Scientific, Teaneck, NJ.
Shankar, S. , 1999, Nonlinear Systems: Analysis, Stability, and Control, Springer, New York.
Mayalu, M. , Kim, M.-C. , and Asada, H. , 2017, “ Latent Space Superposition of Multiple Solutions to Predict Emergent Behaviors of Nonlinear Cellular Systems,” American Control Conference (ACC), Seattle, WA, pp. 2146–2151.
Karnopp, D. C. , Marcolis, D. L. , and Rosenberg, R. C. , 2000, System Dynamics – Modeling and Simulation of Mechatronic Systems, 3rd ed., Wiley Inter-Science, Hoboken, NJ.
Forbes, B. , 2007, Engineering System Dynamics – a Unified Graph-Centered Approach, Taylor and Francis CRC Press, Boca Raton, FL.
Asada, H. , and Bhattacharyya, S. , 2017, “ Precise Linearization of Nonlinear, Non-Autonomous Systems Based on Physical System Modeling Theory,” American Control Conference, Seattle, WA, pp. 2850–2856.
Bernard, P. , and Wu, L. , 1998, “ Stochastic Linearization: The Theory,” J. Appl. Probab., 35(3), pp. 718–730. [CrossRef]
Roberts, J. B. , and Spanos, P. D. , 1990, Random Vibration and Statistical Linearization, Wiley, Chichester, UK.
Oyama, H. , Yamakita, M. , and Asada, H. , 2016, “ Approximated Stochastic Model Predictive Control Using Statistical Linearization of Nonlinear Dynamical System in Latent Space,” 55th IEEE Conference on Decision and Control, Las Vegas, NV, pp. 7153–7158.


Grahic Jump Location
Fig. 1

Conceptual diagram of DFL. A nonlinear dynamical system is viewed from an augmented space consisting of independent state variables and auxiliary variables, which are nonlinearly related to each other.

Grahic Jump Location
Fig. 2

Example of bond graph

Grahic Jump Location
Fig. 3

Examples of constitutive laws of elements: (a) resistor R1, (b) capacitor C, and (c) inertia I1

Grahic Jump Location
Fig. 4

Algebraic loop involved in the bond graph in Fig. 2

Grahic Jump Location
Fig. 5

Example of causal auxiliary variables where input u(t) is not involved in each causal path

Grahic Jump Location
Fig. 6

Example of an anti-causal auxiliary variable where input u(t) is involved in the causal path. Backtracking the causal path terminated at auxiliary variable f leads to input u(t).

Grahic Jump Location
Fig. 7

Bond graph of a dynamical system inspired by an excavator powered by a hydraulic system

Grahic Jump Location
Fig. 8

Constitutive laws of the system in Fig. 7

Grahic Jump Location
Fig. 9

Comparison of three linearization methods, Taylor expansion, statistical linearization, and DFL, in terms of the root mean square of predicting η˙ . The bar chart is normalized by the root-mean-square (RMS) of DFL: σDFL .

Grahic Jump Location
Fig. 10

Schematic of deviated trajectories of linearized models from the exact nonlinear model

Grahic Jump Location
Fig. 11

Deviation envelopes of different linearization models. The maximum deviation from the exact nonlinear model over the time horizon of 0.3 s is shown by the envelope.

Grahic Jump Location
Fig. 12

Sum of mean squared errors of all the state variables. Comparison of the three linearization models against time horizon.

Grahic Jump Location
Fig. 13

Bond graph of a system with anti-causal auxiliary variable

Grahic Jump Location
Fig. 14

Deviation envelopes of simulations using the causal auxiliary variables η*



Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In