Speed Controller by Neural Networks Trained by Invasive Weeds for a DC Motor

Throughout history, the implementation of intelligent machines capable of performing activities that help humans has been a complicated task, then the need to seek new sources of inspiration to solve effective control solutions arises, giving rise to the emergence of bio-inspired algorithms which adopt phenomena present in nature. In this research, speed control by Artificial Neural Networks (ANN) or Neuro controller (NC) is developed for its application on an industrial machine such as a DC motor. For the training of the ANN, a novel and almost unexploited algorithm such as the Invasive Weed Optimization (IWO) is used, as a useful tool when training a neuro-controller for complex systems. The neuro controller has superior characteristics to a conventional controller, and if parameterized correctly it does not require a large computational effort. The MatLab/Simulink ANN toolbox is used to develop the basic structure of the ANN and a Control Plant Trainer (CPT) with a DC motor is used as a test plant. An ARDUINO board is used as an acquisition and control board. To validate the performance, the Wilcoxon test is used to compare the Time Weighted Error Integral (TWEI) of an NC trained by Back-propagation with the one trained by IWO and a conventional Proportional Integral and Derivative (PID) controller. The results obtained are good and interesting from the point of view of industrial automatic control.