APPARATUS FOR CONTROLLING POSITION OF STRIP USING NEURAL NETWORKS IN TWIN ROLL STRIP CASTER

摘要

PURPOSE: An apparatus for controlling a position of a strip using neural networks in a twin roll strip caster is provided which accurately controls a target position of the strip by estimating the position of the strip, correcting a position of the actually controlled position and controlling a cylinder using neural network theory. CONSTITUTION: The apparatus for controlling the position of a strip using neural networks in a twin roll strip caster comprises a controlled error detector(31) detecting errors of a target position of the strip and a corrected position of the strip; a neural controller(32) controlling a cylinder(21) of a plant(20) with a cylinder reference which is an optimum output variable according to neural network theory in which the strip position errors between the controlled error detector(31) and cylinder parameters are input variables; a neural identifier(33) sending the obtained cylinder parameters to the neural controller(32) by estimating strip positions based on the cylinder reference of the neural controller(32) and obtaining optimum cylinder parameters using errors between the corrected strip positions and the estimated strip positions; an image processor(34) finding an actual position(measured position) of the strip by processing images from a camera(22) taking a photograph of images corresponding to positions of the strip of the plant(20); a strip position corrector(35) correcting the strip positions estimated from the neural identifier(33) based on the actual strip position of the image processor(34); and a model error detector(36) sending the detected errors to the neural identifier(33) by detecting errors between the strip positions corrected in the strip position corrector(35) and strip positions estimated in the neural identifier(33).