Comparative Analysis of ANN and PID Controller of Aero-pendulum on Simscape

Abstract

This paper presents the analysis and control of driven pendulum or aero-pendulum simulated on simscape, a classic example of a mechatronic system by using both, conventional Proportional Integral Derivative (PID) and Artificial Neural Network (ANN) Controller. Aero-pendulum has highly non-linear dynamics, therefore its angular position control by using conventional control techniques is very difficult, and the PID controller parameters can only be tuned for single operating point. The work carried out mainly focuses on developing an ANN controller which can overcome the shortcomings of conventional PID controller. An ANN controller has been trained by data acquired from the system when PID controller was incorporated for different operating points. Finally, the angular position control of driven pendulum has been achieved using the ANN controller, having comparable performance and stability with the PID controller. Also instead of using a linearized mathematical model of aero-pendulum for simulation, a 3D multibody model has been developed using simscape, which incorporated all the nonlinearity related to the physical system.