If successful, the project will deliver a step-change to data integration, decision support, gas monitoring, geo-technical monitoring, worker and equipment location, emergency response and personnel safety.
CSIRO are currently undertaking the jointly funded Japan-Australia safety project at Anglo Coal’s new Grasstree Colliery in central Queensland.
Entitled, Mine Communication and Information for Real-time Risk Analysis, the project will see the development of a system ultimately capable of providing pre-emptive alerts in advance of potential incidents and stoppages. An added benefit will see a marked reduction in the endless stream of false and non-critical alarms mines cope with on a daily basis.
The issues of false alarms, the need to constantly upgrade graphic interfaces and difficulties in exercising control over the computer interfaces themselves, were some issues raised by Control Room operators in a survey of underground mines in Queensland.
According to project coordinator, Greg Rowan, analysis of disaster events in underground coal mines shows that in many instances predictive data was available but was difficult to source, incomplete and often ambiguous.
This is partly because many existing mine communication systems are made up of multiple systems that cannot communicate with each other. Rowan estimates that a typical longwall mine may generate between 27,000 – 35,000 discreet Input/Output data streams. The sheer volume is a problem, as is the fact that the data sets are rarely integrated and the background noise makes it difficult to thoroughly scrutinise the “critical few”. Bottom line: mines are nowhere near using their available data to the best effect.
So for example in a typical system, equipment monitoring is handled by SCADA systems; video is provided by separate CCTV systems; a mixture of DAC intercom, analogue telephone and leaky feeder systems support voice communications; while tube bundle and instantaneous point sensors are usually employed for mine-wide gas monitoring – with each system speaking a different language.
Ethernet LANs are open systems which can, in principle, support this wide range of applications, from equipment monitoring, personal communications to video streaming. Developing a mine wide Ethernet network capable of doing all this with a common protocol will form part of the initial development strategy. (For those not au fait with computer language a common protocol is a common language that allows the various bits of monitoring equipment to ‘talk’ to each other.)
Specifically, the research will tackle developments in communication systems, gas and ventilation monitoring, strata monitoring and emergency response systems. A key objective with the communication systems backbone will be to provide an intrinsically safe Ethernet switch and the design criteria for this switch have been mapped out.
The idea is to install one infrastructure instead of several, built on a broad-band, high speed Ethernet fibre-optic data highways. Importantly, the system will integrate all the crucial data sets constantly flowing around a mine into one user interface and can be installed with new as well as with existing systems.
“It was critical to us that this new system be capable of integrating currently existing infrastructure in place at today’s mines – otherwise we would simply be creating another system to complicate things further,” Rowan said.
The system under development uses a self-learning, predictive architecture for decision support, which, Rowan emphasises, is not a science-fiction ‘neural net.’
“When we first drew up the requirements for decision support, we realised that terms such as ‘artificial intelligence’ and ‘predictive architecture’ would make the blood run out of your face,” Rowan recalled.
“No-one was going to trust their critical decisions to a support system they do not understand – this is quite different in that the design of the system is specific to each operation and the processes to be put in place are determined by the people at each mine”
In essence, the system aims to develop what is known as ‘decision support software,’ following on from one of the recommendations out of Moura.
“We’ve looked at 3 or 4 packages developed since Moura but no one appears to have cracked it yet. The decision support we’re designing will provide information to experienced, trained, qualified people in a way that is useable and valuable to them.
“It gives you some predictive ability in your control room so that pre-emptive corrective actions can be taken before an “event” can escalate beyond control - all done by something that doesn’t get tired,” Rowan said.
The predictive architecture will be individually built at each mine, driven from the face level with input from the mine operators. This rule based inference engine will be capable of analysing separate data sets, seeking new or required data, comparing the results of initial analysis with historical events and making recommendations based on the logic strings determined by the mine.
“We explain to people that computers need an algorithms to do certain things, measure this, or react in this way. The mine will determine what it wants - then we get the code crackers to write the code that builds the algorithm,” Rowan said.
“The designs, techniques, the rules and all the self learning abilities that make up the algorithm are built by the people at the mine, peculiar to the circumstances they have to face day in and day out.”
According to Rowan, this key analysis tool will enable this project to provide something not yet available to any mine – the ability to predict future events based on past history, current data and personal and corporate memory.
“All data streams, no matter how real-time, are measures of the past. The ability to use that data to alert operators to a current change in the risk profile of the operation and then predict a range of likely outcomes from that data - even to suggest a range of possible corrective and pre-emptive actions, presents a fundamental shift in the ability of mines to manage risk,” he said.
