ANN Based Improved Dynamic Set Point Weighted PI Controller in Real Time Experimentation

Abstract

Artificial Neural Network or ANN, modeled by the neuron structure of the human brain, have the ability to process and interpret complex, non-linear patterns through interconnected layers and activation functions. In this work, a novel control strategy for DC servo positioning systems is proposed by integrating neural network concepts into classical control. Traditional PID controllers, especially without derivative action, face challenges in tuning and exhibit sluggish behavior and noise sensitivity. While dynamic set point weighting (DSW) techniques improve PI controller performance, their design is complex and often unsatisfactory for systems with high open-loop gain. To overcome these limitations, neuron-based improved dynamic set point weighted PI controller (IDSWPI) is introduced. In this approach, the weighting factor is adaptively computed online using a neuron with a sigmoidal activation function based on instantaneous error changes. Simulation and real-time experimental results demonstrate that the proposed IDSWPI controller offers superior performance compared to conventional fixed and dynamic set point weighting methods.