Improvement of the Data Processing Algorithm for Diagnostic System of the State of Continuous Casting Machine Mold

Abstract

The paper is devoted to solving the task of creating the prototype of effective diagnostic system of the state of continuous casting machine mold used in iron and steel works. The relevance of the task is explained by the high cost of known similar systems of foreign production and of their maintenance as well. In addition, there are some technical complications of those systems integration into the equipment of local metallurgy enterprises. The authors have experience of designing the diagnostic systems of such kind, which were tested in several Ukrainian iron and steel works and showed satisfactory results. This work is aimed to improving the technical characteristics of the prototype of diagnostic system of the state of continuous casting machine mold designed by the authors, in order to increase the accuracy of diagnosing malfunctions of the oscillating mechanism of the mold. The state of the mold oscillating mechanism is evaluated by measuring the deviation of its movement trajectory from the reference signal. In the paper, the structure and operation principle of the diagnostic system prototype are shown and substantiated. The mathematical basics for processing data from the sensors used in the system, MEMS accelerometers, are represented. The procedures for improving the accuracy of measurements the mold movement parameters are described. The noise in the output signals of the sensors is defined as an undesirable factor to eliminate because of its effect on the accuracy of measurements. The Kalman filtering is suggested for decreasing the noise in the output signals of MEMS accelerometers. The filtering implementation is described mathematically.The improved algorithm including Kalman filtering for processing data from MEMS accelerometers, that reduces the measurement errors of the parameters of the mold movement under conditions of noise in the output signals of the sensors is represented. The simulation of the suggested algorithm is performed; the diagram of the sensor output noisy signal of acceleration and the diagram of filtered and integrated signal of displacement illustrating the processes of measuring the parameters of the mold movement and processing the measured data are obtained. The results of the simulation confirmed the effectiveness of the proposed solution. The application of Kalman filtering makes it possible to improve the signal-to-noise ratio from 13.9 dB to 38.7 dB, what is about 25 dB increase, that respectively gives the opportunities to increase the accuracy of mold movement trajectory measurements. In the future works, it is planned to develop this approach by combining it with machine learning, which is able to make the system as effective as possible.Ref. 15, fig. 4.