Description: Performed synthesis of the neural network controller with prediction NN Predictive Controller to solve the problem of electromechanical system control taking into account the elasticity of the mechanical coupling. Analyzed neuro-controllers in the Control Systems section of the MATLAB Neural Network Blockset: controller based on the auto-regressive moving average model NARMA-L2 Controller, controller based on the reference model Model Reference Controller and the NN Predictive Controller. As a result of the research it was established that the NARMA-L2 Controller and Model Reference Controller neural network controllers do not provide satisfactory performance indicators of the system. NN Predictive Controller is found to be effective. Studied the structure and synthesis of the NN Predictive Controller neuro-regulator. Scheme of the neural network control system model is developed in MATLAB Simulink package with the NN Predictive Controller. It includes a control unit of the neural network controller (electromechanical system with elastic mechanical couplings) and a regulator block NN Predictive Controller. Developed a model of the electromechanical system in the form of a twolayer feed-forward neural network. Optimal number of neurons in the hidden layer is determined and the network is trained to provide high accuracy using the Levenberg-Marquardt algorithm. By a wide range variation of the neural network controller parameters, it is found which parameters significantly affect the control quality. Determined parameter values that satisfy the quality indicators of the transient response of the system state variables. Simulated a neural network system with a random amplitude step input. It is found that the use of a neural network model for an electromechanical system provided high quality of identification and optimal values of controller parameters. It allowed the NN Predictive Controller synthesis, that provided high dynamic characteristics of an electro-mechanical system with elastic coupling in kinematic transmissions.
Keywords: Neural network technology, neural network control system, two-mass electro-mechanical system, system with summing amplifier, neural network predictive controller