Prediction of hazards in coal mines through directional change-point analyses

Abstract

Change in the mean trend or preferred direction of cleat measurements have been thought of as good predictors of potentially hazardous mining conditions lying ahead. The detection of such a change is thus of great importance and usefulness. This poses the change-point problem for the mean direction of circular data. Usual (arithmetic) mean for linear data is nonsensical for the circular case. Appropriate statistics and techniques for such circular data need to be developed. First some new graphical tools arc presented as quick diagnostic or exploratory data analytic techniques. These are followed up by formal parametric statistical tests. Data from the Wallsend Borehole Colliery is analyzed using these new tools and methods, which seem to be quite promising. DDSTAP-Statistical Analysis Package for Directional Data, recently developed by this author has been utilised for such analyses.