Multi-Network-Feedback-Error-Learning in pelletizing plant control

This work is devoted to present a control application in an industrial process of iron pellet cooking in an important mining company in Brazil. This work uses an adaptive control in order to improve the performance of the conventional controller already installed in the plant. The main strategy approached here is known Multi-Network-Feedback-Error-Learning (MNFEL), it uses multiple neural networks in the strategy Feedback-Error-Learning (FEL). EEL is a strategy control which a neural network (NN) learns to improve the control actuation of a Conventional Feedback Controller (CFC), in this case a Proportional-Integral-Derivative (PID) controller. The advantage of the FEL strategy is to provide cooperation between the adaptive controller and the conventional controller. The NN learns not only the actuation necessary for the control, but new actions can be acquired as consequence of changes in the process. The approach of MNFEL is to add a new neural network whenever the network's error stops decreasing, so that avoid the conventional approach, restart the learning of the NN. It is emphasized MNFEL can be used when wish to improve the results obtained with FEL strategy, just adding many neural networks in the system. That is good option because FEL improves the performance of the CFC and MNFEL is a improvement of FEL, so MNFEL improves more than FEL, the performance of the control system. In this work, due to the unknown mathematic model of the plant and, in order to simulate the control of the process, a neural model of the plant is also presented. In a simulation environment, PID, FEL and MNFEL strategies are compared and the results are discussed.