Underground » Strata Control and Windblasts
Most Australian coal mines have shield leg pressure monitoring in real time to aid the identification and remediation of roof control problems and to assist in the identification of faulty legs and yield valves. Newer longwalls are now equipped with tilt sensors that allow convergence measurements in real time, albeit at a resolution that is sub-optimal. During ACARP project C21013, shield convergence data was obtained from 2 sites. Tilt sensors developed by CSIRO with a convergence resolution of about 0.8 mm were installed on 6 shields at the Austar Mine and data obtained. At this time Caterpillar installed tilt sensors with a resolution of about 8 mm on all shields at the Narrabri mine and data was also obtained from this site. From a back-analysis of this data, C21013 demonstrated that it is not possible to obtain a reliable understanding of how shield supports interact with the strata from leg pressures alone. It was demonstrated that a much greater understanding of shield-strata interaction is possible from shield convergence measurements. With the limited amount of back-analysis obtained from these two sites, it was not possible to develop reliable convergence triggers that could predict the on-set of roof control problems.
An extension project was approved to enable data to be obtained from additional sites. Additional leg pressure and convergence data was obtained from Broadmeadow Mine and the Dendrobium Mine, who have Joy tilt sensors installed on all shields. The Joy tilt sensors have a similar resolution to the Caterpillar sensors. Additional convergence and leg pressure data were obtained from Narrabri. CSIRO tilt sensors were installed on 6 shields at Newlands North and convergence and leg pressure data obtained. The leg pressure and convergence data from all these sites were back-analysed with the aim of developing reliable triggers of impending roof control problems in advance of mining using the LVA program.
The following convergence triggers have been found to correlate well with the formation of cavities, on at least 10 out of 20 adjacent shields:
· A convergence in a single cycle of ≥ 80 mm;
· An average convergence rate of ≥ 50 mm/hr; and
· A cumulative convergence over 3 consecutive cycles of ≥ 200 mm.
It was found that these were triggered mainly from 1 to 3 shears in advance of any roof control problems occurring although on occasions they are only triggered within the shear that the roof control problems occur. A load cycle is the time from setting of the support to releasing the support.
The only leg pressure signal that has been found to give any indication of impending roof control problems is the initial pre-yield loading rate. A pre-yield loading rate for at least 10 out of 20 adjacent shields of 10 to 15 bar per minute has been found to be the trigger level for impending roof control problems. This signal can be very unreliable, particularly where high set pumps are employed. It is often very difficult to judge when the pressure rise in shield legs is a result of the strata loading the shields or is a result of the pumps. High set pumps are now employed on the majority of longwalls in Australia. It is noted that the majority of longwalls in Australia do not currently have reliable convergence monitoring installed. An improvement in determining the time that the support is set and the pumps have ceased operating is required to make the pre-yield loading rate more reliable in the LVA program when high set pumps are enabled. This is not a simple task.
A rigorous analysis of the trigger levels noted above was initiated. It was found that cycle delineation within the LVA program was not accurate enough for this purpose. This is not a criticism of the program. To achieve better cycle delineation required research versions of the LVA program to be written. This was time consuming, as it needed to be site-specific and the programs so developed are not reliable enough yet to be a commercial product.
The research programs were then used to rigorously analyse the trigger levels. In general it was found that the trigger levels were about 80% accurate; i.e. the trigger levels accurately gave a warning of roof control issues in about 80% of the time. In the remaining about 20% of the time, the majority were false positives; i.e. the TARPs were triggered, but no cavities occurred. In the last two thirds of LW9 at Broadmeadow the amount of false positives significantly increased to about 60%, when using TWAP and low set pressures to indicate that cavities were occurring.
Time consuming analyses were carried out to understand the dramatic decrease in the accuracy of the trigger levels at Broadmeadow over this period, which necessitated both the use of the research version of the LVA program noted above and a great deal of manual analyses of the LVA trending data. The analyses concluded that the most likely reason was that operators were setting adequately when roof cavities and/or guttering were occurring and that cavities were not able to be detected by the LVA program. Changes in shield height between load cycles was used to determine if cavities had likely occurred and the accuracy of the triggers were determined to be about 80% using this method of cavity detection. It is noted that the use of changes in shield height between load cycles to determine the existence of roof cavities is not 100% certain. However, if the researchers' conclusions are indeed correct this is a very important observation. The researchers have been expounding the hypothesis that roof control problems can be significantly ameliorated on many occasions if adequate set pressures are attained when roof conditions deteriorate to the extent that guttering and cavities have occurred. This entails a trade-off between canopy attitudes and set pressures, which most operators are reluctant to engage in. In some situations (e.g. when cavities are high) such a trade-off is of course not possible. Further work is warranted to determine in the LVA program when cavities are occurring, which may need additional longwall monitoring signals to be captured by the program.
In the majority (>70%) of occasions, the analyses indicated that the TARPs were triggered ≥1 shear in advance of roof control problems occurring and impending roof control problems were detected up to 5 shears in advance. That does mean in about a quarter of occasions the TARPs are only able to determine the onset of roof control problems in the shear that they occur.
Please note, the stage one report and movie file are found within the attachment link, when you download this report.
An 
e-newsletter has also been published for stage one of this project, highlighting its significance for the industry.