Neural network generalized inverse of two-motor synchronous system working on constant volts per hertz control mode

Considering that the two-motor synchronous control is a nonlinear, strong-coupling, multi-input and multi-output (MIMO) system, a novel generalized inverse system (NNGI) is proposed in this paper. The NNGI system is based on a single hidden layer feed-forward neural network system (SLFNs). Here a novel learning algorithm called extreme learning machines (ELM) is used. This method has better generalization performance and much faster convergence rate for function approximation than traditional back-propagation (BP) and support vector machines (SVM). Being rightly designed, the NNGI system can transform the MIMO nonlinear system into a number of single-input single-output (SISO) linear subsystems with open-loop stability. The experimental results proved that the NNGI system has strong robustness with the changes of model parameters, pattern recognition capability of complex nonlinear system and good real-time capability. Furthermore, it can apply to general invertible nonlinear MIMO system.