Real-time hardware ANN-QFT robust controller for reconfigurable micro-machine tool

Abstract

This paper shows a reconfigurable micro-machine tool (RmMT) controlled by an artificial neural network based on a robust controller with quantitative feedback theory (QFT). In order to improve the performance of the controller, a field programmable gate array (FPGA) was applied. Since micro-machines present parametric uncertainties under different points of operation, linear controllers cannot deal with those uncertainties. The parametric uncertainties of a micro-machine could be described by a set of linear transfer functions in frequency domain to generate a complete model of the micro-machine; this family of transfer functions can be used for designing a robust controller based on QFT. Although robust control based on QFT is an attractive control methodology for dealing with parametric uncertainties in CNC micro-machines, the real-time FPGA implementation is difficult because robust controllers require a complex discrete representation. In contrast, artificial neural networks (ANNs) work with basic elements (neurons) and run using a parallel topology. Moreover, they are described by simple representation, so the real-time FPGA implementation of ANN controller is a better alternative than the QFT controller. The proposed ANN-QFT controller gives excellent results for controlling the CNC micro-machine tool during the transitory response.