A hybrid nonlinear autoregressive neural network for permanent-magnet linear synchronous motor identification
ABSTRACT The modeling of permanent-magnet linear synchronous motor is very important to the control and the static and dynamic characters analysis of the system. In this paper, the model of permanent-magnet linear synchronous motor is presented by using neural networks of the nonlinear autoregressive with exogenous inputs. Based on the same cost function, residual signal analysis is mixed into the networks, and then the networks can identify motor's ranks automatically. First, the nonlinear autoregressive with exogenous inputs model is expanded into the polynomial function, then the condition which true ranks satisfy is presented by using residual signal analysis. Some shortages of BP (back-propagation algorithm) are considered, so NDEKF (node-decoupled extend Kalman filter) is applied to train networks. The experiment results show that the hybrid neural networks of the nonlinear autoregressive with exogenous inputs can identify object's (a vertical transport system driven by permanent-magnet linear synchronous motor) ranks precisely, and the output of networks is very close to the experimental result. In the experiments, the performance of NDEKF is often superior to that of BP, while requiring significantly fewer presentations of training data than BP and less over training time than that of BP.
Conference Proceeding: Optimized feedforward neural networks for on-line identification of nonlinear models[show abstract] [hide abstract]
ABSTRACT: Optimization of a class of nonlinear approximators corresponding to feedforward neural networks is investigated for on-line identification of nonlinear models in high-dimensional settings. The parameters are optimized by minimizing a cost function, which consists of the summation of two terms: a fitting penalty term and a term related to changes in the parameters. The relative influence of the two terms on the overall minimization can be tuned, according to a proper scalar. The resulting algorithm has properties of convergence and robustness. Simulation results are performed to compare its performance with classical algorithms, such as back-propagation and learning based on the extended Kalman filter, used for adjusting parameters in neural-network identification of nonlinear models. The advantages of the proposed approach are shown.Decision and Control, 2002, Proceedings of the 41st IEEE Conference on; 01/2003
- [show abstract] [hide abstract]
ABSTRACT: The paper presents simulation and experimental studies of identification of civil engineering structures using neural networks. The identification of structural models by using measured data is an important issue in engineering. Although static function mapping may be achieved using neural networks without knowing the fundamental physics of the system, dynamic model identification is still a challenging topic in neural network applications. A generalized neural network-based technique for structural dynamic model identification is developed based on the dynamics of structure. During the simulation study, structural response records from a 10-storey San Jose apartment building subjected to three different earthquakes are adopted for the dynamic model identification. For the experimental study, a series of experiments were conducted in which a designed scaled model structure, mounted on a shake table, was tested. The neural network is trained and examined using the measured structural responses under different earthquake loading conditions. It is shown that the trained neural network is capable of providing sensible outputs when presented with input data that has never been used during its trainingIntelligent Information Systems, 1997. IIS '97. Proceedings; 01/1998
- [show abstract] [hide abstract]
ABSTRACT: This paper presents a fuzzy modeling approach for identification of dynamic systems. In particular, a new fuzzy model, the Dynamic Fuzzy Neural Network (DFNN), consisting of recurrent TSK rules, is developed. The premise and defuzzification parts are static while the consequent parts of the fuzzy rules are recurrent neural networks with internal feedback and time delay synapses. The network is trained by means of a novel learning algorithm, named Dynamic-Fuzzy Neural Constrained Optimization Method (D-FUNCOM), based on the concept of constrained optimization. The proposed algorithm is general since it can be applied to locally as well as fully recurrent networks, regardless of their structures. An adaptation mechanism of the maximum parameter change is presented as well. The proposed dynamic model, equipped with the learning algorithm, is applied to several temporal problems, including modeling of a NARMA process and the noise cancellation problem. Performance comparisons are conducted with a series of static and dynamic systems and some existing recurrent fuzzy models. Simulation results show that DFNN compares favorably with its competing rivals and thus it can be considered for efficient system identification.IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics) 02/2002; 32(2):176-90. · 3.24 Impact Factor