Search Documents

By M. Frenzel, K. Bachmann, M. Rudolph, R. Tolosana-Delgado, E. Schach, M. Buchmann, L. Pereiraa, K. G. van den Boogaart, N. Kupka

Geometallurgy - the science of predicting the behaviour of ores along the value chain based on their geological and mineralogical characteristics - has barely moved beyond process minerology studies. These studies typically link process behaviour to the mineralogical and microstructural characteristics of pristine or processed ores in a qualitative or semi-quantitative manner. Eventually, a trial-and-error optimisation of the operation is achieved. However, to deliver on its promise, geometallurgy should progress beyond this current approach to a fully quantitative understanding of process behaviour, including the various uncertainties involved, namely predictive geometallurgy. One of the barriers currently hindering this development is the neglect of most ore characterisation data obtained in geometallurgical studies. This is mostly due to the large extent of this information exceeding the capacity of conventional software. To address this challenge, we developed a data mining platform consisting of an SQL database, an R front-end in-house package, and back-ends to interpret data from several analytical instruments. This platform enables the use of the statistical and data mining power of the R open-source environment for any geometallurgical study. Potential applications range from ore characterisation to process understanding and forecasting, and from geostatistical prediction to operational control and optimisation. The scripting abilities of R allow for the: 1) processing of many streams in loops, thus freeing the user from repeating tedious click-and-drop tasks; 2) computation of virtually any derivate or aggregate quantity from the data; 3) distribution of the work among several processor clusters; and 4) keeping of a log file for the calculations done. This contribution presents the building blocks of this platform, and illustrates with several examples its potential to enable predictive geometallurgy. Keywords: Data mining, entropy, kernel methods, LASSO regression, predictive geometallurgy

Jan 1, 2020

By Ronald J. Roman

All mining and milling complexes can, within limits, operate at a variety of production rates. Factors of economic importance that are affected by production rates are costs per unit production, recovery and product grade. Once the relationship between the production rate and cost is established, dynamic programming techniques can be used to define the optimum production schedule (that is, the production schedule that maximizes the present worth of the operation) for the life of the deposit. In general, assuming that the capacity and efficiency of the physical plant remain unchanged throughout the life of the deposit, the optimum production rate is not constant for the life of the deposit, but declines gradually as the deposit is being exhausted.

Jan 1, 2014

By J. Yianatos, P. Vallejos

In the last decades, the exponential increase in the cells’ volume resulted in significant advantages such as investment costs, footprint and energy savings. However, in larger size cells, despite the increased diameter/height ratio, new constraints arose related to the need for decreasing the larger froth transport distances to enhance the froth recovery. Thus, the use of froth crowders and internal launders became compulsory to maintain the froth transport distances but decreasing the ratio between gas and collected mineral flowrates. Consequently, the bubble load will increase until reaching a critical bubble surface coverage. In this paper, the concentrate carrying capacity (tph/m2) is described in terms of the bubble surface coverage (bubble loading) at the pulp-froth interface and the froth recovery (froth transport) and is evaluated as a function of the cells’ volume and operating variables. A sensitivity analysis, based on industrial operating and design conditions, such as superficial gas rate, particles size, froth crosssectional area and cells’ volume, was developed for evaluating the metallurgical performance of cells from 100 to 630 m3. Results show that limiting carrying capacity can arise for the whole range of cells’ volume, depending on critical operating conditions, particularly in first rougher cells and cleaning stages of industrial flotation circuits. The increase in cells’ volume consistently increases the bubble loading (bubble surface coverage) under all conditions, approaching the limiting conditions more rapidly in the case of cells without internal launders. Otherwise, the limiting condition was achieved when the particle size DPS decreased from 50 to 20 (μm) and for superficial gas rates significantly less or larger than 1 (cm/s). The use of internal launders allows for increasing the range of the operating conditions needed to reach a certain carrying capacity. Keywords: Bubble carrying capacity, flotation cells, bubble surface coverage, cells’ volume

Jan 1, 2020

[Consider a square block W in an ore body S2, to be valued from one centre sample w, and one corner sample w2 (Fig. 9.1). The sample set is ws = Iw,;w21. We can estimate the block average µw by the average of the sample values x, and x2: µw = 'z (x, + xz). (9.1) The error made is measured by the estimation variance: o' (ws to W) = 2 - (ws; W) - - (ws;ws) - - (W; W). (9.2) However, we know that the centre sample w, gives more information about the block value than the corner sample w2. In the example studied in § 8.4, where w, and w2 were considered to be points, we found that of (w, to W) was nearly one third smaller than o2E (w2 to W), therefore when estimating W one should give more weight to w, than to w2. In general, when evaluating a block W from a set of n samples ws, where WS = l w;w2; ... wn1, a better estimator of µw can be obtained if we give different weights to the sample values, depending on the size and the positions of the samples. Some desirable properties of such an estimator µw are: ? It should be a linear function of the sample values xi: n µw = bi xi (9.3) =t where bi = weight given to sample x . ? It should be unbiased, i.e. its expected value should be equal to the true block value: E [µw - µw] = 0. (9.4) ? The mean squared error of estimation of µw should be a minimum E [(µw - µw)2] = a minimum. (9.5) The estimator which satisfies these properties is a minimum variance, linear and unbiased estimator, and is known as the kriging estimator. The corresponding error of estimation is the kriging error. This estimator is also known as the BLUE (for Best Linear Unbiased Estimator) estimator. This was named by Matheron in acknowledgement of the early work of Krige's in the optimization of the evaluation of the South African gold mines. The following notations will be used: µK kriging estimator. o2k = o2K (w,, to W) kriging error.]

Jan 1, 1978

"This study examines the problems faced by open pit mine superintendents who make choices about how to direct their materials, either to the waste dump or to the mill. The paper explores the effects of introducing a 10% sampling error and a 0.9-times to 1.1-times sampling bias on positively skewed distributions for precious and base metals, negatively skewed distributions in the case of bulk commodities, and normal distributions as is the case for coal deposits. Parent distributions for each commodity were created on a 25 x 25 m grid using transformations of gold, iron ore, and coal data-sets, spatially based on a nonconditional Gaussian simulation. Ordinary kriging of grades for the three commodities into a 10 x 10 m grid provided the reference case against which the distributions with the sampling error and sampling bias for the commodities were compared. Imposing cut-off grades on the actual-versus-estimated scatterplots of the three commodities allowed the distributions to be classified into components of waste, dilution, ore, and lost ore. Ordinary kriging of values for each deposit type acted as the reference data-set against which the effects and influence of 10% sampling error and 0.9-times to 1.1-times sampling bias are measured in each deposit type. Indications are that the influence of error and bias is not as significant in gold deposits as it is in iron ore and coal deposits, where the introduction of small amounts of error and bias can severely affect the deposit value.IntroductionThe effects of poor sampling and the financial implications for mining companies, traders in mineral assets, and sellers of metal as dore or commodities is documented in a number of studies, the most notable of which is that by Carasco (2004). He examined the financial impact of poor sampling practices in the Chilean copper industry and found that finacial losses due to poor sampling amounted to hundreds of millions of dollars over the life of a mining operation. Holmes (2004) examined the effects of correct sampling and measurement as the foundations of metallurgical balances, and found that revenues from the sales of large iron ore shipments may be profoundly affected by poor sampling practices. In South Africa the way in which sampling of in situ gold-bearing reefs affects the mine call factor has been an ongoing study since the disparity between the estimation of gold in the reef and the actual gold bullion produced was noted by early investigators such as Beringer (1938), Jackson (1946), Sichel (1947), Harrison (1952), and a number of others."

Jan 1, 2017

By C. O. Aksoy, G. G. U. Aksoy, A. Fisne, I. Alagoz, E. Kaya

Coal produced in underground mines is transported to the surface by means of conveyor belts throughout long roadways. The combustion of belts puts a large number of employees at great risk. Underground collieries differ from other underground mines -due to the combustible nature of the product. The most prevalent cause of belt fires in underground coal mine is spontaneous combustion due to oxidation of the coal, which also enables the conveyor belt to burn over time. In 2014, a belt fire in an underground coal mine in Manisa-Soma, Turkey caused 301 fatalities. A study has been conducted on this accident for approximately 3 years, consisting of combustion tests in a purpose-built research gallery, comparison of the test results with mine records of the accident, and CFD modelling of the mine environment. The intensity of the fire was sufficient to redirect the air flow underground, causin large amounts of toxic gases to fill almost the entire mine in approximately 15 minutes. It is recommended that CFD analysis be used in planning emergency action strategies in underground mines.

Dec 13, 2023

By S. G. Mgenge, J. D. Steenkamp

Tronox Namakwa Sands operates two ilmenite smelters for the production of high-titania slag and low-manganese pig iron. Slag or metal is tapped from one of two product-specific, bi-level tap-holes. In this paper the refractory design specific to the tap-hole refractory layout is described, as well as practices for slag and metal tapping and refractory maintenance. Further developments planned are briefly described.

Jan 1, 2014

By R Linares

For centuries, diamond's superlative properties have earned it a reputation as the 'dream' material for mankind in its search to create advanced applications. Until recently, this dream has been realized only in the use of diamond for gemstones and abrasive applications where several tons of diamonds are now used on an annual basis. The vision of having large, high quality diamond as part of a broad engineering tool-kit for many diverse high tech industries is now becoming a reality however, with the advent of early stage commercial availability of single crystal and poly crystalline CVD diamond.

Jan 1, 2010

By M. B. C. Brandt

The near-surface wave attenuation factor k (kappa), which describes the attenuation of seismic waves with distance in the upper 1–3 km of the Earth, was determined for the Far West Rand gold mining area using seismic data recorded by station Parys (PRYS) of the South African National Seismograph Network. Twenty micro-events in the magnitude range 0.7 ? ML ? 1.8 for the period 1 July to 15 November 2015 were analysed. For the analysis a 10-second window for the S phase portion of the vertical component seismogram was selected. The result was an average k = 0.048 ± 0.014 s, which is much higher than for a stable continental region where k = 0.006 s. This is because the assumed Brune source model is inappropriate for describing mining-related events that often have moment tensors with a volumetric (implosive) component. The average k was calculated to be 0.098 ± 0.038 s for explosions of a similar magnitude. This higher value derived for mining-related k may be employed as a means of correcting the Brune source spectrum to calculate seismic moment as well as corner frequency for events in the mining areas of South Africa.

Jan 1, 2017

By M. Lange

The postulation that primary platinum group metal (PGM) matte will chemically react with magnesia-chrome bricks when temperatures exceed 1500°C was tested. Magnesia-chrome brick samples were heated in contact with matte at 1300°C to 1750°C for 30 minutes, after which the refractory samples were analysed using reflected light microscopy and scanning electron microscopy. The samples were all completely penetrated by matte. As the temperature increased the matte also penetrated the fused aggregate grains and disintegrated them. The chromium concentration of the matte inside the refractory samples was found to be slightly higher than that of the bulk matte. At temperatures of 1500°C and higher, MgO, FeO, and magnesium-rich silicate crystals could be identified in the matte directly adjacent to the refractorymatte interface. Phase relations clearly indicated that chemical reactions take place between primary PGM matte and the magnesiachrome refractory material at temperatures above 1500°C, but that these reactions are more complex than expected from FactSAGE® calculations.

Jan 1, 2014

By M. I. Mabala, B. P. Watson, M. J. Kanda

A failed and stable Upper Group 2 chromitite pillar database was collected in 2007 as part of the PlatMine 1.2 pillar strength project. The database consisted of 177 pillars, of which 33 had failed. Most of the data were collected from various intermediate-depth mines near the town of Thabazimbi. Spans between rows of chain pillars were about 30 m. Pillar loads were estimated using pseudo-3D, elastic, boundary element modelling software. The pillar-load data were evaluated using a maximum likelihood regression analysis to determine strength parameters, following a similar approach to historically acceptable practice. The original formula was named the PlatMine Formula. This paper describes re-evaluations of the database, using a linear formula and several variations of the power formula. The formula providing the lowest standard deviation to the data in the database was considered the best representation of the pillars in the database. Through this rigorous process, a new formula for peak pillar strength was developed for Upper Group 2 chromitite pillars, which is unaffected by volume. The small standard deviation provided by this formula suggests a reliable relationship between strength and w/h ratio, in the w/h ratio range between 1.5 and 4.7. Accuracy, precision, and recall analysis confirmed the efficacy of this equation to the data in the database. Importantly, the new formula provided in the paper does not predict a significantly different strength to the current PlatMine formula, within the confines of the database.

Nov 27, 2024

By M. C. van Eeden, R. Vermeulen, K. B. Le Bron

This paper presents the lessons learnt during the transition from an open pit method to a sub-level open stoping method at the Koidu Limited kimberlite pipe. These learnings may assist other mines with planning, both from a practical implementation perspective (layout, charging up challenges), and a method for estimating potential dilution and stability challenges associated with the interaction between the surface and underground excavations. Predictive geotechnical stability modelling was successfully applied to highlight potential failure zones, which allowed modification and enhancement of the mining layouts and quantification of the potential dilution of the kimberlite from indicative instabilities.

Sep 1, 2025

The theory of geostatistics covers a branch of .applied statistics aimed at a mathematical description and analysis of geological observations. Geostatistics can be used in pure geology (for example, for the analysis of trace elements in a metamorphic rock), in mineral exploration (for example, for the analysis of geochemical exploration data), as well as in mine valuation. This book is intended to provide a practical introduction to the theory of geostatistical methods of mineral evaluation. Over the years, various mathematical models have been developed to represent the distribution of values in mineral deposits. The simpler models are based on the assumption that the values are randomly distributed. Classical statistical methods, based on this assumption of the random distribution of values, are used to analyse mineral deposits to which these models apply or are assumed to apply. In all mineral deposits, however, one recognizes the presence of areas where the values are higher or lower than elsewhere. Also, the values of two samples in a mineral deposit are more likely to be similar if the samples are taken close together than if they are taken far apart. This indicates that there exists a degree of correlation between sample values, and that this correlation is a function of the distance between the samples. Models have been developed which take this correlation into consideration, with the degree of correlation between sample values being usually measured by the semivariogram function. In these models the fact that two samples taken next to each other will most probably not have the same value, must also be considered; even for very short distances the correlations are usually not perfect and a purely random component is present in the value distribution. The mathematical models will therefore assume the presence of two sources of variability in the values: a correlated component and a random component. Finally, one must consider the particular and very common case of mineral deposits in which the values present a systematic variation in space. This variation is usually referred to as a drift, or a trend. For example, the grade of an ore body may increase with depth of the ore, or it may decrease when one moves away from a central volcanic pipe. The earlier models did not give a satisfactory representation of drifts, and more complex models have been developed, in which three sources of variation are represented. These models are made up of: a deterministic component, a correlated component, and a random component. The deterministic component is used as a model of the drift. The correlated component explains regular changes in values which are not represented by the drift. The random component represents variations which cannot be explained by any of the above factors. The simpler models, based on the assumption of a single random component, will be described first (Chapter 2). The models based on the hypothesis of the superimposition of a correlated component on a random component, will then be analysed in detail. These models are most commonly used in the analysis of mineral deposits (Chapters 3-1 I). Finally, how to deal with the presence of a drift will be briefly described (Chapter 12). This book has been written essentially for students in mining engineering and for mining engineers who are interested in the background to the theory of geostatistics as well as its practical applications. The assumption is made that the reader has an elementary knowledge of statistics. Some knowledge of linear algebra is useful in part of Chapter 9, and is necessary to read Chapter 12. A proof is given of all the equations related to geostatistics, and which are not usually found in elementary textbooks on statistics. Understanding of these proofs is not necessary for practical application of the theory, and the reader may wish to skip them on a first reading, concentrating attention on the numerous simple practical examples given. Although the theoretical geostatistician will not find much new material in this publication, it is anticipated that he will develop some interest in the practical approach chosen to prove the various geostatistical equations. Many people and institutions contributed to the preparation and completion of this work. I am much indebted to Dr D. G. Krige and the Anglo Transvaal Consolidated In- vestment Company Limited, who gave me the opportunity to spend a considerable amount of time working on geostatistical problems, both theoretical and practical, during the time that I was in their employment. Dr Krige contributed greatly in developing my interest in studying both the theory and practice of geostatistics, always insisting that a correct balance be kept between theory and practice. I am grateful to Professor H. M. Wells and the Mining Department of the University of the Witwatersrand for inviting me to give lectures in a post- graduate course on geostatistics. The notes 1 prepared for that course became the foundation of the present work. I am also indebted to the Centre de Geostatistique of the École des Mines de Paris, where I received my first formal education in geostatistics during a summer course given by Charles Huijbregts, after many years of my lonely plodding through the published literature. Many graphs in the present volume are reproduced with the permission of the Centre de Geostatistique. The Department of Metallurgical and Mineral Engineering of the University of Wisconsin-Madison has also contributed in making this work possible, by allowing me to spend a considerable amount of time and resources in the writing, typing, and correcting of successive drafts. I am thankful to Lynn D. Kendall, who typed the entire manuscript under constant pressure.

Jan 1, 1978

By G. C. S. Burns

"A system has been developed to simulate the mining of the Mufulira orebody. The main output from the system is a picture-gram from which the start date and effective life of all proposed mining faces can be measured. This information provides schedule dates for the completion of major capital development projects and stope preparation activities.The orebody must be divided horizontally into main levels and vertically into blocks of equal interval. Ore reserves are defined by specifying the tonnage contained in each block and main level. Mining faces are defined by their co-ordinates, and may themselves be split into sublevels. The simulation maintains an echelon between contiguous sublevels. Other face parameters control the speed of mining, the availability and relative priority of the face.The required production from each crusher/shaft system is specified. The technique of simulating the mining process in order to achieve the required production is described."

Jan 1, 2014

It is suggested that an increase in the production from thin coal seams in South Africa is necessary, and that a substantial amount of any increase will be mined by the bord-and-pillar method. For the purpose of this study, a thin coal seam was regarded as any seam 1,5 m thick or less to the minimum thickness at which mechanized equipment can operate in bord-and-pillar workings. The equipment being utilized in thin-seam bord-and-pillar layouts includes conventional equipment, load-hauldump cars (scoops), and drum-type and Fairchild-Wilcox continuous miners. Comparisons are made of these types of equipment for bord-and-pillar mining at various seam thicknesses within the thin-seam range. The evaluations involved production rates and the cost of extraction for South African conditions, use being made of an existing computer modelling technique in the generation of production rates. The indications are that, for seam thicknesses greater than 0,9 m, drum-type continuous miners utilizing continuous haulage are the most productive and cost-effective equipment at present operating in thin-seam bord-and- pillar workings. Where seams thinner than 0,9 m are to be exploited, Fairchild-Wilcox continuous miners, particularly the Mark 22, produce the most coal at the lowest cost.

Jan 1, 1984

By E. Lwamba

The aim of this study was to examine the effect of grain size distribution control of the raw material mixture on the permeability of the green sinter bed and the properties of the produced sinter. This was achieved by evaluating the granulation characteristics of the sinter mixture (moisture content, granulation time and mean granule diameter) in terms of its green bed permeability, and evaluating the productivity of the sinter bed, the coke rate, tumble index (TI), abrasion index (AI), reduction disintegration index(RDI) and reducibility (RI) of the produced sinter. The raw sinter mixtures contained combinations of Thabazimbi iron ore, Sishen iron ore, coke, lime and return fines. The grain size distributions were controlled by removing the ?0.5 mm size fractions of the return fines and coke, and the ?1 mm size fraction of lime. Of the examined mixtures, a mixture of 20 mass% Thabazimbi iron ore, 80 mass% Sishen iron ore and fluxes where the coke, lime and return fines were all sized had the highest granulation effectiveness and permeability. The sintering properties of the mixtures in which the grain size distributions were controlled, were very similar for all the mixtures, but superior to the base case mixture in which the grain size distribution was not controlled. Keywords: Thabazimbi iron ore; Sishen iron ore; granulation, permeability, grain size distribution; sinter.

Jan 1, 2008

By M. D. G. Salamon

In view of an urgent need for a simple means of simulating the behaviour of stratified rock masses, a long-dormant linear model is revived, generalized, and discussed in some depth. The model is piecewise homogeneous, consisting of a pile of homogeneous isotropic laminae in which the interfaces between beds are parallel and free of shear stresses and cohesion. The continuity of stresses and normal displacements is assured across the bedding planes and, to maintain simplicity, the theory of thin plates is used as the kernel of the model. The paper covers the case where the rock mass consists of discrete beds in which the layers are arranged either heterogeneously or systematically. The stratification is regarded as uniform when all the layers are of the same thickness and their Young's modulus changes exponentially with depth; and as homogeneous when all the laminae have identical thickness and properties. The case of uniform stratification is also treated on the basis of an equivalent quasi-continuum derived from the basic model. Attention is focused on the derivation of influence functions. It is demonstrated that the quasi-continuous models, when they are available, can replace their distinct-layer counterparts. The homogeneous quasi-continuum, which permits the derivation of its influence functions in closed form, facilitates the determination of the roof and floor convergence. In the case of a significant class of problems, the convergence distribution is defined by a second-order differential equation in the xy plane, which simplifies to a Poisson's equation if the mining excavations are unsupported. With this result, the paper prepares for the solution of a host of practical problems. These solutions are presented in a series of papers that are in the course of publication.

Jan 1, 1991

By V. Moodley

Strict implementation of International Standards for representative sampling of coal is difficult for routine monitoring, due to time, technical, and economic constraints. A large percentage of the total sampling variance arises from errors accumulated during the earliest stages of sample collection. Modified procedures on the precision of the final result leads to controversies between the sellers and customers for coal. Manual collection of coal samples violates the principle of correct sampling, that all fragments must have an equiprobable chance of being in the sample. In addition, manual size reduction of large numbers of samples is time consuming and involves human errors. A coal quality management procedure in alignment with the relevant ISO standards that describes the sampling protocol has been implemented, and current sampling and laboratory analysis of coal from base load mines (mines situated at the mouth of the power station) are well established. For import coal, sampling is performed during unloading by auto-mechanical sampling (AMS). Due to logistical and safety issues associated with multiple suppliers, Eskom opted for the coal to be sampled at source on a pre-certification basis. Online analysers provide a fast, accurate, real-time method of determining the total moisture and elemental composition of coal and is important to the coal industry for pricing, quality control, and reduction of SO2 emissions. Elemental analyses include S, Si, Al, Fe, Ca, Na, Mg, Mn, Ti, K and Cl. Measurement of S is dictated by the control of SO2 emissions from coal-fired power plants. Elements such as Na and Cl have adverse effects on boilers, causing fouling and slagging. Available online elemental analyser technologies include prompt gamma neutron activation analyses (PGNAA), recommended for major element determination, and pulsed fast and thermal neutron analysis(PFTNA), that is capable of measuring the major and minor chemical elements contained in coal. These systems utilize nuclear reactions to produce characteristic gamma rays that are used for identifying various chemical elements. By acquiring the gamma rays in three different time windows, there is a significant reduction of the background as compared with the spectra taken with a radio isotopic source.

Jan 1, 2009

By C. W. Siyasiya

Process and steel chemistry differences in hot rolled Al-killed low carbon strip steel between the hot charge route (HCR) as practised in a compact strip production (CSP) plant and the conventional cold charge route (CCR) have been studied. This was done with reference to differences in reheating practice, hot rolling practice and steel chemistry through emphasis on the effects of solubility and reprecipitation of AlN in the two process/product combinations and its effect on ferrite grain size development. An earlier modelling study on the evolution of the austenite grain size in both of the above hot rolling processes has also now been extended to include an estimate of the possible effects of retained strain after the last finishing pass on the ferrite grain size. As part of this study the equilibrium solubility trends of AlN in a number of boron-free low carbon Al-killed strip steels with low and high sulphur content from the two respective processes was investigated by the thermoelectric power (TEP) technique and differences in the effects through the derived solubility equation of Log[Al][N]=2.6?9710/T at typical slab reheating temperatures was demonstrated. Ferrite grain size modelling using the significantly lower Ar3 transformation temperatures generally found in hot rolled low carbon strip steel produced in the CSP plant using the HCR compared well to the larger grain size produced in the conventional CCR strip rolling plant. Keywords: austenite grain size modelling; thermoelectric power(TEP); compact strip production (CSP); cold charge route (CCR);ferrite grain size

Jan 1, 2008

By X. Ndlovu

Underground observations in a coal mine have indicated failure oft he immediate roof above the bords. Careful observations and photographic recording of occurrences of these roof failures, termed ?roof guttering?, have been made. The occurrence of guttering results directly in roof instability, and indirectly, due to the interaction of the guttering with natural planes of weaknesses in the immediate roof rock. In this paper, relevant observations of roof guttering, possible mechanisms of formation of roof gutters, laboratory testing for determination of material strength and deformation properties, and numerical analyses to explain underground observations are described. Mapping of roof failures showed that these took place mainly towards the centre of the roadways. Roof guttering was observed often to occur violently and with little warning. Occurrence of roof guttering has a negative impact on production?some panels are abandoned, production times are increased and safety of workers is compromised. The numerical stress analyses showed that, although none of the constitutive behaviour criteria used predicted the observed underground failures correctly, the extension strain criterion gave the best agreement.

Jan 1, 2007