Identification of Mechatronic Systems with Dynamic Neural Networks using Prior Knowledge

This paper introduces a new approach for the identification of mechatronic systems with dy- namic neural networks. Based on a given continu- ous signal flow chart a discrete chart is constructed. This discrete chart is implemented in a general dy- namic neural network (GDNN). Certain weights of the neural network model correspond with physical parameters of the system and can be trained by an optimization algorithm. Moreover, nonlinear parts of the signal flow chart can be identified by nonlin- ear subparts of the neural network. The parameters are trained with the Levenberg-Marquardt (LM) opti- mization algorithm. Therefore the Jacobian matrix is required. The Jacobian is calculated by real time re- current learning (RTRL). The proposed identification method is tested with a nonlinear two mass system.

[1]  Christian Endisch,et al.  System Identification with General Dynamic Neural Networks and Network Pruning , 2008 .

[2]  M.T. Hagan,et al.  Backpropagation through time for a general class of recurrent network , 2001, IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222).

[3]  O. De Jesus,et al.  Forward perturbation algorithm for a general class of recurrent network , 2001, IJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222).

[4]  Dierk Schröder,et al.  Online identification of a nonlinear mechatronic system , 2002 .

[5]  Martin T. Hagan,et al.  Backpropagation Algorithms for a Broad Class of Dynamic Networks , 2007, IEEE Transactions on Neural Networks.

[6]  O. Nelles.  Nonlinear System Identification , 2001 .

[7]  Paul J. Werbos,et al.  Backpropagation Through Time: What It Does and How to Do It , 1990, Proc. IEEE.

[8]  Christian Hintz.  Identifikation nichtlinearer mechatronischer Systeme mit strukturierten rekurrenten Netzen , 2003 .

[9]  Orlando De Jesus,et al.  Training general dynamic neural networks , 2002 .

[10]  Christian Endisch,et al.  Optimal Brain Surgeon for General Dynamic Neural Networks , 2007, EPIA Workshops.

[11]  Dierk Schröder,et al.  Comments on "Backpropagation Algorithms for a Broad Class of Dynamic Networks" , 2009, IEEE Trans. Neural Networks.

[12]  P J Webros.  BACKPROPAGATION THROUGH TIME: WHAT IT DOES AND HOW TO DO IT , 1990 .

[13]  Ronald J. Williams,et al.  A Learning Algorithm for Continually Running Fully Recurrent Neural Networks , 1989, Neural Computation.

[14]  Mohammad Bagher Menhaj,et al.  Training feedforward networks with the Marquardt algorithm , 1994, IEEE Trans. Neural Networks.

[15]  Dierk Schröder,et al.  Elektrische Antriebe - Regelung von Antriebssystemen , 2001 .

[16]  Christoph Hintz,et al.  Online identification of a mechatronic system with structured recurrent neural networks , 2002, Industrial Electronics, 2002. ISIE 2002. Proceedings of the 2002 IEEE International Symposium on.