Clustering in optimization of RBF-based neural estimators for the drive system with elastic joint

In this paper the application of Radial Basis Function Neural Networks (RBF-NN) as neural estimators of state variables of electrical drive with elastic joint is presented. RBF network are used for estimation of the load speed and shaft torque of the two-mass drive system, in the control structure with the state controller. One of the most important stages of neural estimators design is correct selection of an internal structure of such models, which has a strong influence on the generalization properties of neural networks. In described application of RBF-NN the clustering is chosen for adjustment the number and distribution of radial function centers. Based on the literature review subtractive clustering method is chosen. High accuracy of the reconstructed signals is obtained without the necessity of the electrical drive system parameters identification and modeling. Simulation results show high quality of the estimation for wide range of changes of the load speed, load torque and inertia moment.