When a longwall geomechanics research program at Kestrel first began to deliver some unexpected results the research team could have rejected their findings. It would have been tantamount to throwing out the baby with the bathwater, but those involved were not to know that at the time.

CSIRO was conducting a study of geomechanics occurring with a retreating longwall face. Conventional longwall geomechanics theory at the time said most rock failure took place near the face line and was of a tensile nature. Therefore, if face retreat rates were fast enough roof failure would occur well back in the goaf and the rock over and in front of the face would be unbroken.

What was most baffling about the work at Kestrel and other mines such as Appin was that it seemed to challenge these ideas. CSIRO's coal mining coordinator, Michael Kelly, recalled that initial results were treated with skepticism.

"The original results from Kestrel were contrary to our understanding of what should be happening. Failure was well in front of the face, failure was shear and not tensile in nature, and there was an undrained component in the failure. Conventional rock mechanics says that failure should be immediately in front or behind the face and that it would be tensile ... undrained behaviour could not happen."

CSIRO has been investigating longwall geomechanics with consultant groups and mine sites through several ACARP funded projects for the past six years. Only recently, with the total suite of studies completed, analysed and cross referenced has the full significance of the results started to emerge.

Said Kelly: "Results at Appin, which showed failure down to 120m below coal pillars and events over 300m in front of the face on the block-side were quite remarkable. The Appin results were almost dismissed until the full picture emerged."

One of the main determinations of the work is that the extensive forward abutments can activate faults and joints ahead of the face. This can result in increased loads on gateroads and very poor conditions when the faults are intercepted. Another finding was that increasing roof strengths moved the area of initial breakage to over the face resulting in "periodic weightings" or further back into the goaf resulting in convergence and face spall. Other factors which were found to influence failure are layout geometry and pore water pressure. Another key discovery is that strata failure activity under adjacent goafs can be re-initiated as a subsequent longwall passes by.

Kelly said one of the strengths of the investigations was the combination of microseismic monitoring and computational modelling, supplemented with measurements from a range of other tools including surface extensometers and geological mapping.

"Although the seismic monitoring has been the facilitating technology, it has been a real team effort by groups inside and out of CSIRO," Kelly said.

What is yet to be determined is whether the microseismic noises being detected can be physically matched to changes in rock mass strength. If this can be done it will provide a bridge of accurate calibration of numerical models of strata behaviour to predict rock failure.

And if microseismics can be used to accurately predict falls or periodic weighting events perhaps something can be done to better prepare for them. Similarly, if predicted floor failure can be correlated to gas inflows from lower seams a useful predictive tool will have been gained.

The numerical modelling methods for strata behaviour around longwalls are currently being modified to recognise the CSIRO findings. If the model is wrong it has no worth as a tool for designing ground reinforcing techniques. The models, therefore, have to be matched with what is being detected by microseismics.

The next step is the development of a cost-effective predictive tool that could be successfully incorporated into a functioning longwall operation.

"Three-D modelling as a predictive tool is in its infancy," Kelly said. "It is clear that longwall geomechanics is a 3D issue and requires 3D assessment."

Kelly said current models did not have enough flexibility which meant compromises had to be made when dealing with different issues such as fluid content and faulting.

"In addition, even with the power of modern computers, element size is compromised to enable computation within reasonable time frames."

Further CSIRO work is attempting to develop composite codes to alleviate some of these problems.

So significant is this work that the Australian Coal Association Research Program (ACARP) awarded $298,000 in this year's round of funding for ongoing research. ACARP is also ensuring that project results will be provided to all Australian geomechanics consultants.