Hydraulic Actuator Identification Using Interval Type-2Fuzzy Neural Networks

Mohsen Vatani
2012 International Journal of Information and Electronics Engineering  
In recent years, intelligent based approaches have been introduced as one of the best potential methods for solving many problems in control literature. Neural Networks (NN) and Fuzzy Logic are widely used in nonlinear system modeling and identification. These approaches require a high number of model parameters, which impose more complex computation. Using Interval Type-2 Fuzzy Neural Network (IT2FNN) method, one needs considerably fewer numbers of required parameters. It can also model
more » ... inty and nonlinearity of the system much more effectively. In this paper, we suggest to use this neuro -fuzzy based network for nonlinear modeling of a hydraulic actuator. Simulation studies of this challenging benchmark confirm the excellent nonlinear modeling properties of the IT2FNN. Index Terms-Nonlinear systems, identification, type-2 fuzzy neural network, hydraulic actuator. I. INTRODUCTION System identification is often the first and the critical step in process control, prediction of behavior, fault detection, estimation of immeasurable variables and improving understanding of system behavior. In the control literature, there exist a large number of active researches and various well developed algorithms for linear system identification [1]. However, nonlinear system identification is far from mature both in theory and in practice [2], [3] . When linear models are not appropriate for accurately describing the system dynamic, nonlinear identification should be employed. Because the structure of nonlinear systems is so rich and complicated, it is not expected that a single method could be effectively applied to all nonlinear systems. Indeed, in many cases, knowledge of the physical phenomena involved is often incomplete, some critical variables may not have been measured, and some physical parameters may be unknown. The model must then be determined from observed data. Several methods have been developed for non-linear system identification. Early works begin by considering the functional series of Volterra. For linear systems, convolution integral models the input-output data set. However, for nonlinear systems, Volterra series serves a generalization of the convolution integral. In [4] wiener series have been used for identification of nonlinear systems. However, by using this method, identification of even a simple system that contains second-order nonlinearity would require the evaluation of, typically, coefficients [3] .
doi:10.7763/ijiee.2012.v2.160 fatcat:amnbbqe7jbgn3cedl7nnefvt2i