Title :
Pre-alarm system of coal mine based on ICA and Kalman filter
Author_Institution :
Sch. of Electron. & Control, Chang´´an Univ., Xi´´an, China
Abstract :
A method based on independent component analysis (ICA) and Kalman predictor is proposed to predict the methane and carbon monoxide in coal mine. Two disturbances of hydrogen and ethane are mixed into the gas samples to simulate the real world situation Features extracted with ICA are respectively predicted by Kalman filter step by step to 476 points, compared with BP neural network, the mean square errors(MSE) are decreased by 72 times and 94 times.
Keywords :
Kalman filters; air pollution; alarm systems; coal; feature extraction; hydrogen; independent component analysis; mean square error methods; mining; ICA; Kalman filter; Kalman predictor; carbon monoxide; coal mine; features extraction; hydrogen; independent component analysis; mean square errors; methane; pre-alarm system; Artificial neural networks; Carbon; Equations; Feature extraction; Kalman filters; Mathematical model; Safety; coal mine; independent component analysis; kalman filter; neural network;
Conference_Titel :
Natural Computation (ICNC), 2010 Sixth International Conference on
Conference_Location :
Yantai, Shandong
Print_ISBN :
978-1-4244-5958-2
DOI :
10.1109/ICNC.2010.5583137
