1、PDF原文：http:/ 中文3570字英文原文 Practical Neural Network Applications in the Mining Industry L. Miller-Tait, R. Pakalnis Department of Mining and Mineral Process Engineering, University of British Columbia, Vancouver, B.C., Canada ABSTRACT The mining industry relies heavily upon empirical analy
2、sis for design and prediction. Neural networks arecomputer programs that use parallel processing, similar to the human brain, to analyze data for trends andcorrelation. Two practical neural network applications in the mining industry would be rockburst predictionand stope dilution estimates. This pa
3、per summarizes neural network data analysis results for a 1995Goldcorp/Canmet study on rockbursting and a 1986 UBC/Canmet study on open stope dilution at theRuttan Mine. 1. INTRODUCTION Many aspects of mine design are based upon empirical data. Neural Networks analyze data and predictions based on p
4、revious results. Neural networks haveadvantages over conventional empirical designapproaches.These advantages include: Neural networks can easily use multiple inputs to analyze data. By using multiple hidden layers and nodes neural networks investigate the combined influence of inputs. Neural networ
5、ks can be easily retrained as new data becomes available making them a more dynamicand flexible empirical estimation approach. Neural network software is inexpensive and easy to use. Neural networks have demonstrated a more accurate empirical estimate over conventional methods. The advantages of usi
6、ng neural networks are illustrated in a rockburst prediction example and an open stope dilution example. 2. ROCKBURST PREDICTION The first example of a potential situation where neural networks could be useful in the mining industry isthe prediction of rockbursts through physical inputs. To quote di
7、rectly from the Ontario Ministry ofLabor“.we do not have the ability to predict when and where rockbursts will occur, and the experts in the fieldagree that we are not close to make such predictions” 1. Between 1984 and 1993 eight undergroundminers were killed in Ontario due to rockbursts. This acco
8、unted for approximately 10% of underground fatalities during this period. If neuralnetworks were to have success in predicting where rockburstsoccur, additional ground support, remote equipment, and/or design modifications could reduce or possibly eliminate fatalities due to rockburst. As saf
9、ety is the primary responsibility of mining engineers, thepotential for neural networks to assist in predicting rockburst inputs should be investigated. In 1995, a joint project was completed by Goldcorp Inc. and Canmet called “Development of EmpiricalDesign Techniques in Burst Prone Ground at A. W.
10、 White Mine” 2. Part of the study was to collect inputinformation on rockburst, caving, ground wedge, and roof fall failures at the A. W. White Mine between1992 and 1995. This resulted in a failure database consisting of 88 ground failures with correspondinginputs for each failure. The six inputs co
11、llected for each failure were RMR 3, Q 4, span 5, SRF2,RMR adjustment, and depth. These input factors were set up and run in a neural network with 73 examplesbeing used for training and 15 examples being used to test the network. The output factor, stability, can beone of four failures2 - PUN-RF (po
12、tentially unstable roof fall), PUN-GW (potentially unstable groundwedge), BUR (rockburst), and CAV (cave). A brief description of the input and output factors are listed below. 2.1. Input factors RMR - The RMR system, initially developed by Bieniawski in 19733, bases rock mass quality on fiveparamet
13、ers.These parameters are: Uniaxial compressive strength of the rock Rock quality designation (RQD) Spacing of discontinuities Condition of discontinuity Ground water conditions. These factors are given a numerical value and totalled together to get an RMR value. This value will be anumber between 0
14、and 100 with zero being very poor rock and 100 being extremely good rock. The groundwater conditions were assumed to be dry conditions. Q -The Q factor refers to the rock quality tunnelling index 4. Developed in 1974, by Barton, Lien and Lunde,from the Norwegian Geotechnical Institute, the Qfactor i
15、s based on six factors, which are: RQD - rock quality designation Jn -joint set number Jr -joint roughness number Ja -joint alteration number Jw - joint water reduction factor SRF - stress reduction factor. The actual Q formula is Q= RQD/Jn Jr/JaJw/SRF. The Jw/SRF factor was assumed to be 1.0
16、 for this study because dry conditions are assumed. Stress is factoredthrough modelling and strain measurements. The Q factor ranges on a logarithmic scaleranging from 0.001 to1,000 where 0.001 is extremely poor rock and 1,000 is virtually perfect rock. Span5 - the meaning of span refers to the widt
17、h of an underground opening in plan view. Span can bedetermined through the largest diameter of a circle within an underground excavation. SRF 2 - refers to the adjusting of RMR values relative to stress ratios and previous history of groundconditions. It does not refer directly to SRF used in the c
18、alculation of Q. Stress criteria is based upon the ratioof induced stress overunconfined compressive strength (UCS) of the rock. 2.2. Output Factors Burst refers to a stope in which a rockburst has occurred. A rockburst is an instantaneous rock failure in orabout an excavated area characterized/acco
19、mpanied by a shock or tremor in the surrounding rock. PUN-RFrefers to potentially unstable ground with respect to a roof fall. A stope is considered potentiallyunstable if any of the following conditions occur2: The opening may exhibit strong discontinuities having orientations that form potential w
20、edges in the back. Extra ground support may have been installed to prevent a potential fall of ground. Instrumentation installed in the stope has recorded continuing movement of the stope back. There may be an increased frequency of ground working or scaling. PUN-GW refers to a stope considered pote
21、ntially unstable due to the likelihood of a ground wedge failure.This is a subset of PUN-RF collected separately to identify areas where jointing may result in wedge failures. Cave refers to when uncontrolled ground failures result in caving. 3. NEURAL NETWORK ANALYSIS The above inputs and outputs w
22、ere run on a neural network to see if a neural network could predictresults from the input data and also to see which inputs had the greatest effect on output prediction. A two layernetwork consisting of 13 nodes was run for 10105 cycles reaching a 1.69 percent error. Seventy threeobservations were used to train the network. The remaining 15 observations were used to test the networks predicting ability. The results of the neural network showthat the network correctly predicted all outputs from the training. The reason that this is not surprising is that the network
