Multilayer feedforward networks are universal approximators
Neural Networks
Neural networks for control systems: a survey
Automatica (Journal of IFAC)
Inverse model control using recurrent networks
Mathematics and Computers in Simulation - Special issue from the IMACS/IFAC international symposium on soft computing methods and applications: “SOFTCOM '99” (held in Athens, Greece)
Control System Design
ICCS'03 Proceedings of the 1st international conference on Computational science: PartI
Nonlinear control structures based on embedded neural system models
IEEE Transactions on Neural Networks
IEEE Transactions on Neural Networks
Fuzzy Sets and Systems
| Hi-index | 0.00 |
This paper shows the viability of implementing a control strategy based on the internal-model control paradigm, which is a useful synergy of a dynamic ANN trained from real-life data and used to predict process output and a fuzzy-logic control (FLC) that enhances the control system's overall performance. A force control problem involving a complex electromechanical system, represented here by the machining process, is considered as a case study. The main goal is to control a single-output variable, cutting force, by changing a single-input variable, feed rate. The proposed neurofuzzy-control (NFC) scheme consists of a dynamic model using ANNs to estimate process output, and a fuzzy-logic controller (FLC) with the same static gain as the inverse model to determine the control inputs (feed rate) necessary to keep the cutting force constant. Four approaches, the fuzzy-logic controller (FLC), the direct inverse controller based on ANNs (DIC-NN), the internal-model controller (IMC-NN) and a neurofuzzy controller (NFC), are simulated and their performances are assessed in terms of several performance measurements. The results demonstrate that the NFC strategy provides better disturbance rejection than the IMC-NN and the FLC for the cases analyzed.