Development and validation of a method to predict coal pillar life in South Africa

One of the most difficult aspects of mine closure procedures, is to predict the long term stability of pillars in the case of coal mining. In South African coal mining, pillars have been designed since the late 1960's with the aid of a pillar strength formula based on statistical analysis of failed pillar cases by the well known team of Salamon and Munro. They developed the widely used power formula for pillar strength. Since that time, however, the data base of failed pillars has effectively doubled in size and re-analysis of the new data indicated that the original pillar strength of small pillars may have been over estimated. The data was then extensively re-analysed and a more effective linear formula for pillar strength was found. The most important differences between the two formulae are that the predicted strength of small pillars are lower and the strength of larger pillars, higher with the new formula. However, neither of the two methods explicitly cater for the prediction of the expected time of stability of coal pillars. No direct correlation between the safety factors and the period of stability of pillars could be found. It was then determined that the most frequent mode of pillar failure was by progressive scaling. Using the new formula to determine a minimum value of pillar safety factor (i.e. the safety factor at which failure can be taken as guaranteed to occur), the final sizes - after scaling - at which the failed pillars had to be in order to fail, were determined. The differences between the original dimensions and the postulated final dimensions were then used to calculate a rate of pillar scaling. The rate was then re-applied to the original data bases of both failed and intact pillar cases and distinct differences were found. The projected lives of the failed pillars were substantially shorter than the projected lives of the intact pillars. While this inspired confidence in the procedure, it was still based on an assumed and unproven rate of pillar scaling. The next step was then to measure the actual amounts of scaling on un-failed pillars underground. This was done on a large number of pillars in a variety of conditions. The measurements showed sufficiently close correspondence with the assumed values, to confirm the assumptions and by consequence, the life prediction procedure. With the aid of the real measurements, it was also possible to statistically determine confidence and accuracy levels of the procedure. (author)