Mining Publication: Development of Multiple Regression Functions for Performance Prediction of Gob Gas Ventholes for Sealed and Active Longwall Mines
Original creation date: February 2010
Authors: HN Dougherty, C� Karacan, GV Goodman
NIOSHTIC2 Number: 20036456
2010 SME Annual Meeting and Exhibit, February 28 - March 3, Phoenix, Arizona, preprint 10-201. Littleton, CO: Society for Mining, Metallurgy, and Exploration, Inc., 2010; :1-10
In the absence of values for each influencing parameter and a known relationship of input-output space, novel approaches should be implemented as proxy solutions to mathematically complicated problems. Proxy solutions can be in the form of linear or non-linear multiple regression equations. Predicting the performance of gob gas ventholes is one of these mathematically complicated problems as it is difficult to know the exact relationships between various parameters influencing gob gas venthole (GGV) performance from active and sealed (completed) longwall panels. In this paper, a discussion of the NIOSH-developed software suite for methane control and prediction (MCP) based on an artificial neural network (ANN) is presented. The application of the ANN-based software suite to develop algebraic relationships for gob gas venthole performance prediction in sealed and active longwall panels is sought. Output information obtained from the NIOSH-MCP software as a response to various input parameters were used to create an input-output database for total flow and methane percentage from GGVs operating in active mines during completed and sealed panel situations. Due to the multi-variable nature of the software, outputs generated using the software for various inputs were used to develop analytical equations first using linear and then second using non-linear multivariable regression techniques. The advantage of this method is its ability to better understand the significant parameters of the relationships between the inputs and outputs for gob gas venthole production in active and sealed panels and to be able to predict their performances using relatively simple analytical functions.
NIOSHTIC2 Number: 20036456
2010 SME Annual Meeting and Exhibit, February 28 - March 3, Phoenix, Arizona, preprint 10-201. Littleton, CO: Society for Mining, Metallurgy, and Exploration, Inc., 2010; :1-10