engleski

Design of artificial neural network based PID controller

This work deals with design of Artificial Neural Network based PID controller which provides better transient characteristics and better control properties than conventional PID controllers. The controller uses one of the most common ANN structures (multi-layer perceptron) and training algorithm (back-propagation), also briefly described in this work. PID Neural Network controller model was designed using Matlab/Simulink. Performance of the proposed controller was verified through the MATLAB simulation. To show effectiveness of the PID Neural Network controller in system control it was compared to a common PID controller tuned by Ziegler-Nichols ultimate cycling method. PID controller that uses Artificial Neural Network for optimization of its parameters can be of great benefit to the process of complex system controller design and development, for system that can not be easily described via traditional mathematical models or systems whose models are too complex, such as Marine diesel engine.

Artifical Neural Networks; PID Neural Network controller; back-propagation; controller parameters optimization

nije evidentirano