Search Documents

By F Owusu-Mensah, F Nsarko

The Obuasi Mine has been in operation for the past 100 years. Within this period a host of mining methods have been employed during its various changing phases. The causes for the changes are multiple, ranging from orebody capability, dynamics, technology, safety, economics, production levels, geotechnical, etc. Mining methods applied have also ranged from timber square setts, conventional cut-and-fill, shrinkage stoping, sublevel caving, mechanised cut-and-fill, longitudinal and transverse open stoping, and lately longitudinal retreat mining method (LRMM).The Obuasi Mine has a world class underground resource potential of about 27 million ounces. A third of this is in reserve. The orebody is subvertical, with multiple lodes and with widths ranging from 1.5 - 40 m. It stretches along 8.5 km N/S strike. A cursory statistical analysis on the various resource ore widths show a higher percentage (60 per cent) of the resource being within the range of about 1.5 - 10 m. Thus a high proportion of the resource is amenable to narrow vein mining methods. This has enhanced the choice of LRMM as an alternative method. Also, with the increasing cost of mining over time, LRMM, which eliminates expensive and time consuming footwall drives, will become far more advantageous as compared to the transverse/longitudinal open stoping methods which currently constitute about 40 per cent of the total underground production.The LRMM currently delivers about 20 per cent of the Obuasi Mine production. Plans are underway to improve total underground production from the current average of 160 000 tpm to about >180 000 tpm in the next few years and about 20 - 40 per cent more in the foreseeable future. Recoverable ounces will also increase from 26 000 to about 34 000 ounces per month. The proportion of LRMM in the business plan is also expected to increase from the current modest level to about 50 per cent. Despite its tremendous advantages in terms of economics and productivity, the method is facing some challenges during its inception. The objective of this paper is to highlight some of the emerging challenges and their effect on production and offer possible remedies that could mitigate the current and possible future challenges.CITATION:Nsarko, F and Owusu-Mensah, F, 2012. The Obuasi mine - longitudinal retreat mining method application and challenges, in Proceedings Narrow Vein Mining 2012 , pp 59-70 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Mar 26, 2012

By A Lyon

The minerals industry, as many other industries, has a wealth of indicators and a thriving æindustryÆ in their development, measurement and reporting. It is natural for this to occur for a new endeavour. However, perhaps it is time for the community concerned with minerals industry indicators to ask whether we have achieved a certain maturation whereby proliferation of indicators may be replaced by a more disciplined and constrained approach. In many natural and engineered systems, thermodynamics provides well-understood and tested constraints. Indicators that are uninterpretable in this framework are not considered acceptable. This paper provides initial evidence that such a formalism may be possible for mine sustainability. The formalism developed does not claim to have the rigour of thermodynamics. However, it does provide a first step towards formalising selection of indicators on the basis of their true comprehensibility. A set of simple equations is developed that represent the conventional five capitals (actually we use six because we separate renewable and non-renewable natural capital) and how they are linked. This is coded into a simulation model. The base equations are then used to develop a system level indicator of sustainability performance, which is asserted to be a reflection of the mineÆs reputation. We define a least sustainable state, termed æmaximum revenue mining (MRM)Æ in the system and define the degree of system sustainability as how far the system is from the MRM. We then introduce a simple mechanism whereby different actors can express differing views on the emphasis for the distance from MRM. Sustainability thereby becomes a non-unique thing that is defined by a distance from MRM and a set of preferences across the capitals.

Jan 1, 2009

By L Soebari, W Sunyoto

The Wabu Ridge Gold Skarn deposit is located at elevations of up to 3100 metres above sea level, in the Central Highlands of Irian Jaya, Indonesia, 35 km NNW of the Grasberg porphyry deposit. Wabu is a grass-roots discovery in an area with no previous gold mining history or record of gold mineralisation. A geological resource of 117.26 million tonnes at an average grade of 2.16 g/t Au (1.0 g/t Au cut-off) has been defined to-date. Mineralisation is hosted within a skarn occurring along the southern boundaries of the Late-Miocene age monzonite-diorite Pagane intrusive complex. The intrusive complex occurs in the footwall of a major EW structure, the Derewo fault, which displays sinistral strike-slip and reverse fault movement. The Derewo fault separates Derewo metamorphic units to the north from platform sediments of the Australian craton to the south. The skarn has a strike length of 6 km and is up to 1.5 km wide. It is developed within calcareous shelf sediments of the Tertiary New Guinea limestone group and is dominated by prograde garnet development. Gold mineralisation was deposited during retrograde alteration and is directly associated with arsenophyrite and bismuth minerals. Retrograde alteration is distinguished by a sequence of clinopyroxene , magnetite, epidote and hornblende, biotite and chlorite and lastly carbonate. Gold is associated with the biotite/ chlorite phase together with sulphides and precipitated as the result of fluid mixing. The Wabu Ridge Gold Skarn deposit was identified in April 1990 during a regional program of helicopter assisted stream sediment sampling of -80 mesh (180 microns) fraction and panned concentrate sampling. The initial discovery sample site was situated 15 km down stream from the deposit and returned 1.49 ppm Au for the stream sediment sample and 138 ppm Au for the panned concentrate sample. A regional airborne helicopter magnetics survey, conducted in early-1992, helped defined the main magnetite associated skarn zone and the extent of associated igneous rocks. Detailed follow-up commenced in May 1992 and consisted of detailed mapping, soil sampling and ground magnetics prior to drilling the first hole in October 1992. Drilling to-date has delineated a contained geologic resource of 8.1 million ounces of gold and potential exists for a larger resource.

Jan 1, 1999

Quay Magnesium Limited (QMG) listed on the Australian Stock Exchange in late September 2004 having raised approximately A$30 million with the object of designing, constructing, commissioning and operating a 15 000 tpa magnesium alloy production facility in Nanjing, China. Site construction commenced in February 2005, with installation of all major components completed by the end of October 2005. Hot commissioning was simultaneously commenced with the final stages of installation, leading to the first stage of continuous ingot casting by mid November 2005. Since that time, ramp-up to the name plate capacity has been undertaken. This presentation highlights some of the more significant technical and logistical hurdles that have had to be overcome in order to meet the high quality product specifications required by a range of QMG customers.

Jan 1, 2006

By E Kabwe, M Karakus

Deep hard rock underground mines tend to experience high stress as such drives exhibit dynamic responses attributed to a rock burst variant (strainburst). This dynamic behaviour is difficult to predict, and every effort is made to devise the appropriate dynamic support system for its associated failure restraint. This paper outlines the back analysis and the support scheme design for an underground decline that exhibited damage from a dynamic strainburst occurrence. It employs a decline section that experienced a dynamically loaded strainburst that initiated rock fragment ejections. The procedure involves a preliminary structural analysis, a seismic intensity measure, and a rock mass simulation to determine the static stress concentration and ground support system response. The numerical simulation is conducted in FLAC3D which employs the advanced strain-softening IMASS constitutive model governed by two Hoek–Brown (HB) bounding yield surfaces. To account for the post-peak brittleness response associated with strainburst the plastic strain multiplier () is adjusted to 0.0 < < 1.0. The back-analysis shows that the occurrence was initiated by a low-intensity seismic wave that superimposed an extra stress component to the initial static stress which triggered rock fragment ejections. Additionally, the appropriate support scheme comprised of Weldmesh, Fibrecrete, static (split set) and yieldable (Kinloc) reinforcements is recommended and implemented. Thereafter, the bolt’s efficiency is evaluated based on the axial force and elongation capacities and then compared with the published bolt performance data. The evaluation results show that the maximum bolt axial force (0.1 per cent) and elongation (0.006 per cent) are below the designed capacity and rupture strain.

Nov 29, 2022

No paper available.

Jan 1, 1996

Underground mine ventilation is now com monly simulated on a digital computer. Such a programme, incorporating natural ventilation has been modified by others to account for methane emission and elevated local tempera- tures. At Wollongong University, the pro- gramme was further developed and extensive interactive graphics incorporated such that any view of the three dimensional mine could be viewed, and superimposed upon that view are areas where a nominated parameter has exceeded a particular value - generally at a nominated time after a particular event and usually a fire/explosion of one type or another. A further programme was developed such that up to 20 &apos;times&apos; could be examined at a particu- lar location after say, a fuel rich fire at some other place in the mine. The output from this programme includes a plot of methane 1, smoke 1, temperature, pressure and air flow versus time - a most useful graphic to observe the limits of safe attendance in a particular Airway.

Jan 1, 1983

By Z Lianbi, Z Yibin, Z Youhong, P Bin

Without adequate soil available, it is difficult to rehabilitate a large area of acid mine waste dump with sustainable plant communities, which has a strong potential acid generation. In the present project, paste tailings capping method has been put forward to reconstruct plant ecosystem on acid mine waste rock dump. This rehabilitation method uses paste tailings as capillary break and plant growth media. The tailings used contained 26.3 per cent of particles of less than 20 ¦m which satisfy the requirement of grain of paste tailings preparation. The tailings are not acid- producing materials. In addition to Cu, total contents of Zn, Pb, Cr, Cd, As and Ni in tailings are lower than those in local soil. Paste tailings concentration 70 per cent ~ 72 per cent was prepared by concentrating ten per cent tailings by using polypropylenecrylic sodium as flocculation. The potential pollution risks are low by using paste tailings as capillary break layer. Plants directly grown in the tailings did not particularly accumulate Cu, Zn, Pb, Cr, Cd, As, Ni in roots, stem and leaves, which were 0.01 ~ 1.6 times of those grown in soil.CITATION:Yibin, Z, Lianbi, Z, Bin, P and Youhong, Z, 2012. Study on paste tailings capping ecology reclamation on acid mine waste rock dump, in Proceedings Life-of-Mine 2012 , pp 243-246 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Jul 10, 2012

By P Hagan, S Campbell

A study was undertaken using various methodologies to determine the optimum loading and hauling configuration for different haul routes at Callide Mine, a coal mine located in central Queensland. The study involved development of an effective flat haul model that was used to estimate production rates in truck and shovel circuits that employ two excavators of differing capacities then Queuing Theory to construct a capacity constraint model for scheduling purposes to assist in truck allocation and finally a cost model to determine the most cost effective fleet configuration.The model was calibrated using haul road design and performance data of trucks and excavators currently in use at the mine. The study examined the effects of cycle time for different truck and excavator fleet configurations on production rate and costs.The study found haulage cost was directly related to cycle time while production rate reduced with cycle time. The study indicates the approach taken in combining a cost model with a capacity constraint model offers a simple and effective method of scheduling an optimum number of trucks in a haulage circuit.CITATION:Campbell, S and Hagan, P, 2012. Selection of an optimum truck and shovel fleet based on effective flat haul and capacity constraint modelling, in Proceedings Eighth Open Pit OperatorsÆ Conference , pp 19-26 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Sep 18, 2012

By V Glenn, P McKenna, H Vickers, P Audet, M J. Gillespie, A Gravina, D R. Mulligan

A requisite corollary of mineral sand mining on North Stradbroake Island (Queensland (Qld), Australia) has been the progressive rehabilitation of post-mined lands for the revegetation of native subtropical Eucalypt communities. Since 2005, rigorous environmental monitoring has been undertaken to verify the annual development of post-rehabilitated sites. From these long-term analyses, a chronosequence of land revegetation (ranging from three to 26 years postrehabilitation) was developed to examine the rehabilitation trajectories of various post-mined sites in relation to proximal undisturbed locations. Here, we identified that some sites may be deviating from the intended natural analogues as evidenced by the opportunistic colonisationand otherwise æinvasiveÆ behaviour of a regionally native species (Black Sheoak - Allocasuarina littoralis), which appear to be displacing the intended Eucalypt varieties. Although it is recognised that the broader data trends correspond to distinct periods of adaptive modifications to the implementation of rehabilitation (ie pre- and post-1995), this analysis focuses on further identifying potential above- and belowground feedback mechanisms underlying these divergent ecological outcomes.CITATION:Audet, P, Gravina, A, Glenn, V, McKenna, P, Vickers, H, Gillespie, M J and Mulligan, D R, 2012. Above/belowground feedback may facilitate and sustain mono-dominance on rehabilitated North Stradbroke Island, in Proceedings Life-of-Mine 2012 , pp 15-24 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Jul 10, 2012

By E M. Jones

On 26 June 2000 a hydraulic fill retaining barricade failed at the Bronzewing mine in Western Australia. The barricade failure resulted in the release of 19 000 m3 of saturated hydraulic fill into the mine, which caused the deaths of three miners. Following this tragic accident, the practices and procedures of mines using hydraulic fills came under intense scrutiny by regulatory bodies and mining companies. At this time, the use of hydraulic fill at NormandyÆs zinc/copper Scuddles mine was suspended and waste rock became the sole backfill material in use underground. By 2003, development of the mine adjacent to Scuddles, Gossan Hill had progressed to a point where a self-supporting fill material was required to enable full recovery of high-grade ore. Cemented hydraulic fill (CHF) was identified as the optimal fill type and work began on designing appropriate management procedures and controls to ensure the safe implementation. This work included definition of the fill properties, plant design and construction, reticulation system design, design of stope barricades and drainage systems and fill placement control and monitoring practices. CHF placement began on 16 July 2004 and has continued to date. Significant operational experience has since been gained in areas including plant performance, fill product quality assurance, barricade construction, stope drainage and in situ performance.

Jan 1, 2008

By J Johansson, B Johansson

The mining industry needs to attract and recruit young talented people. An important task is, therefore, to create an attractive mining workplace that engages with, and motivates youngsters who are not particularly interested in working in the industry. This paper aims to illustrate a way forward for the mining industry to form healthy and safe workplaces honouring key components such as knowledge, skill, co-operation, continual development and acceptance of diversity. This is a way to create a vision of healthy and safe mining work. The mining industry must work towards zero vision; a future without accidents and hazardous environments. In this paper the authors discuss different aspects of an ‘attractive workplace’ and summarises lessons learnt in 26 statements dealing with safety, both physical and social work environments, and social responsibility. The authors hope these recommendations will provide some guidance for how attractive workplaces should be designed for future deep mining.The paper is based on lessons from two large European projects, a Swedish-Polish project: The Smart Mine of the Future; and a European Union project: Sustainable Intelligent Mining Systems (SIMS). One of the authors was the WP leader of health and safety in both projects.CITATION:Johansson, J and Johansson, B, 2017. Attractive workplaces in future underground mining, in Proceedings 13th AusIMM Underground Operators’ Conference 2017, pp 99–102 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Oct 16, 2017

By M Z. Abzalov

Drilling mineral sand deposits has specific challenges, which are different from drilling hard rock deposits: Drilled sequences are represented by non-consolidated free-flowing sands to semi-consolidated clayey sands and indurated layers. Valuable heavy minerals (ilmenite, rutile, zircon, leucoxene) have an average density of 4.6&nbsp;g/cm3, which is significantly heavier than the hosting sand matrix; mainly represented by quartz (density = 2.6 g/cm3). The large difference in the mineral densities can result in sample segregation in the drilling rods and capturing devices causing biased composition of the recovered samples. The drilled strata can contain beds of consolidated sediments, hard pan lenses and concretions. Therefore, chosen drilling equipment should be flexible and permit a switch from drilling the free-flowing sands to hard rock drilling. Thickness of the sand hosted mineralisation varies from several metres to more than 170 m. Mineralisation is distributed above and below the water table. Evaluation of the mineral sands deposits is based on hundreds of drill holes and many thousands of samples, therefore drilling methods should be fast and cost effective. The drilling methods commonly used for mineral sand deposit evaluation, include aircore, triple tube diamond and sonic; including its variants, such as vibrocore drilling technique. The ability of the studied drilling techniques to properly address the above listed challenges in mineral sand deposits, whilst maximising the quality of the obtained samples has been compared in three deposits: Corridor Sands (Mozambique), Richards Bay (Republic of South Africa) and Fort-Dauphin (Madagascar).This study has shown that aircore samples tend to be negatively grade biased, underestimating the heavy minerals content, in particular when drilled intervals are located below the water table. Triple tube diamond core method can produce non-biased samples. However, results can be marred by incorrectly chosen drilling mud. Sonic drilling is a relatively new technique based on a high frequency vibratory technology. It is the only production drilling technique that allows a continuous relatively undisturbed sample of unlithified sands to be obtained. The main limitations of this technology preventing its wider use for developing mineral sand resources are the relatively high drilling costs, in particular when the depth of drilling exceeds 100 m, and speed of drilling advance.

Oct 5, 2011

By T Napier-Munn, D J. Bradshaw, M Chauhan, L Keeney

An important aspect of geometallurgy is quantifying mineral processing attributes, which leads to improved orebody understanding. One of these attributes is flotation performance. Several methods have been developed as indicators for flotation performance, including batch testing leading to kinetic information and the standard grade-recovery curve. This paper makes a preliminary statistical comparison between a newly proposed geometallurgical batch flotation procedure and the æstandardÆ batch flotation test commonly used to measure flotation performance in the laboratory. The new test uses much less material than the standard test, and is much faster. It is concluded that the new method follows the final recovery of the standard method well, suggesting that it will be effective in reflecting ore variability for geometallurgical purposes. Comparison of the methods based on the full grade-recovery curve shows less agreement, but this may be due to the fact that the curve is sensitive to the particular experimental set-up and conditions, which make such comparisons quantitatively challenging. Further development work is therefore planned.CITATION:Chauhan, M, Napier-Munn, T, Keeney, L and Bradshaw, D J, 2013. Progress in developing a geometallurgy flotation indicator, in Proceedings The Second AusIMM International Geometallurgy Conference (GeoMet) 2013 , pp 201-206 (The Australasian Institute of Mining and Metallurgy: Melbourne).

Sep 29, 2013

By Rayner J. G, Leckie J. F, Moorehead C. F

The Coronation Hill gold-platinum-palladium prospect is located in the valley of the South Alligator River (latitude 13&apos;35&apos;S, longitude 132&apos;36&apos;E), about 80 kilometres east of Pine Creek in the Northern Territory. Coronation Hill was one of thirteen sites in the South Alligator Valley at which uranium was mined between 1955 and 1964 after initial discovery by the Bureau of Mineral Resources in 1953. Most of the mines were operated by United Uranium No Liability. The Sleisbeck mine 25 kilometres south of Coronation Hill was operated by North Australian Uranium Corporation and the group of mines at Rockhole about 20 kilometres to the north-west of Coronation Hill was operated by South Alligator Uranium No Liability. The uranium mine at Coronation Hill produced about 26000 tonnes of ore averaging 2.6 kg/t U308 with an estimated average gold grade of 10.4 g/ gold. Gold was commonly associated with the uranium mineralisation and was recovered at the gravity treatment plant at El Sherana. Gold was also recovered in 1986 by Pacific Gold Mines Limited by treating the tailings from the Moline and Rockhole uranium leach plants at a CIP plant at Moline. With the completion of uranium contracts in 1964 mining in the South Alligator Valley ceased. Exploration for uranium continued in the area during the 1960s and 1970s with no success. The Coronation Hill Joint Venture (BHP Gold Mines Limited 45 per cent, Pioneer Minerals (Gold) Pry Ltd 45 per cent, Notgold Limited 10 per cent)carried out the first exploration specifically targetted at gold in 1984 with the drilling of five vertical drill holes which indicated high potential for a substantial gold resource at Coronation Hill with no associated uranium. The Joint Venture has sub- sequently drilled about 25000 metres of diamond drill holes at Coronation Hill alone. Analytical results from 1985 drilling established that platinum and palladium were associated with gold. An indicated resource of 3.49Mt at 5.12g/t Au, 0.21g/t Pt and 0.56g/t Pd has been calculated using a lg/t Au cut off.

Jan 1, 1991

The hoisting system at the Newcrest Telfer mine is designed for a final duty of 7.35 million tonnes per annum at a hoisting distance of 1450 m. In terms of hoisting capacity (t/h) times hoisting distance it is one of the largest friction hoist installations. Initially, the hoist will be equipped for hoisting 6.5 Mtpa from 1100 m depth. This new drive technology is now well proven by ABB for mine hoist applications with the first installation in commercial operation since July 2001. To date 17 systems for medium voltage drives and another five for low voltage drives have been purchased from ABB for mine hoists. This paper will also deal with skip design and powerful data loggers with diagnostics for local and remote monitoring that are installed in the Telfer mine hoist.

Jan 1, 2005

The reduction of titanium tetrachloride to trichloride can be achieved by employing hydrogen and electric arcs. It is possible to produce melts containing up to 50% TiCk in the eutectic mixture of LiCl/KCl by striking a D.C. arc at the surface in the presence of hydrogen and TiCl. vapour. If the electrode IS made to just penetrate the surface of the salt mixture, a hot zone results from the&apos; high resistance contact with both lower chlorides and sintered titanium may be produced. The sintered metal forms just below the melt surface while the trichloride forms at the gaslIiquid junction.The &quot;hotspot cathode&quot; appears to offer a solution to the problem of producing coherent deposits as the spong&apos;e metal can &apos;be continuously withdrawn in a similar manner to the &quot;calcium process&quot;.INTRODUCTIONConsiderable interest has been shown for some years in the possible development of a cheap electrolytic process for the extraction of titanium metal, since a method using simple materials and cheap electric power would materially reduce the cost of production. Even when cheap starting materials are used, however, there exist several general difficulties in the operation of plant which both limit the speed of production and increase the cost of the product. It happens that most of the techniques so far tried produce a powder as cathodic deposit, and when this is the case the following difficulties arise:1. The whole contents of the bath must be treated with solvent in order to extract the &apos;metal powder which may itself tend to oxidise, and thus become contaminated.2. Fine metal powders, which often contain residual entrained salts, are less easy to consolidate by arcmelting than metal sponge.

Jan 1, 1956

By Jaczkowski B

The lower orebody (LOB) at Chingola Divi- sion of Nchanga Consolidated Copper Mines Limited (NCCM) accounts for approximately 30 per cent of the total copper production for the company. The uniformity of the geology of the orebody lends itself readily to a computer based ore reserve calculation and accounting system. The mining units are physically defined such that a hierarchical type of com- puter file organisation can reflect the existing physical interdependencies. In 1972, a project team consisting of one representative from each of the mine planning, operational research, and computer services departments was formed to investigate and to define the technical mining requirements of the dynamic ore reserve system (DORS). The system was finally implemented in 1975ò

Jan 1, 1977

The Goonumbla district of central-west New South Wales, Australia, hosts four significant porphyry Cu-Au deposits that contain a combined probable ore reserve of 71.7 million tonnes at 1.2 per cent Cu and 0.5 g/t Au. Mineralisation in the Goonumbla deposits is associated with narrow, pipe-like intrusions of quartz monzonite porphyry, in which several intrusive phases have been recognised. Five textural characteristics have been used to distinguish the various intrusive phases recognised in the Goonumbla porphyry deposits: (1) phenocryst abundance; (2) nature of the groundmass; (3 and 4) proportion and mineralogy of the mafic and non-mafic phenocrysts; and (5) the presence or absence of anhedral quartz grains in the groundmass. In addition, each intrusive phase is classified according to its estimated original quartz, alkali feldspar and plagioclase contents. Features common to all of the intrusions are that they contain two types of plagioclase and two or three types of K-feldspar phenocrysts in variable proportions. Most of the intrusions also contain apatite micro-phenocrysts. The groundmass textures range from aphanitic through granular to euhedral K-feldspar in a quartz matrix. With only one exception, all of the intrusive phases recognised are either quartz monzonite porphyry or mafic (>5 per cent mafic minerals) quartz monzonite porphyry. The exception is a biotite - phyric monzodiorite, which occurs in the basal portions of one of the deposits. Textural variants of intrusive phases at Goonumbla have been investigated to establish whether specific combinations of characteristics indicate mineralised intrusions, while others suggest barren ones. Some correlations between intrusion types and Cu and Au grade have been recognised. The most important textural characteristic that relates to high grade mineralisation appears to be the low proportion of mafic minerals, because the quartz monzonite porphyry intrusions hosting most of the mineralisation in the two deposits studied contain

Jan 1, 1999

By M J. Nimmo

extremely powerful tools for building predictive and generative models. ML can be used for building highly accurate regression and classification models. But without careful data science and statistical analysis, a regression or classification model could be highly biased and completely wrong – and we may not even realise. An example rock density estimation task will be used to illustrate the potential gains and possible pitfalls in using ML and AI in Mineral Resource estimation. What could possibly go wrong? Nothing! For most simple regression or classification problems, very little. For density estimation, using the global mean may be all that is possible, but the resource classification would need to reflect the uncertainty in global estimation of tonnage. Everything! From data collection and measurement, from selecting ML and AI techniques and designing the analysis workflow to our own cognitive bias. The things that could go wrong are vast and may include but not limited to asking the wrong question, not asking questions, data bias, measurement errors, insufficient data coverage, not collecting the right variables, not collecting enough observations, collecting clustered samples, unbalanced data, filtering out data without reason (outliers), clipping data (outliers), relying on automatic feature selection, including irrelevant variables, excluding relevant variables, geological interpretation, changing context, blindly following best practice, splitting small data sets, subsetting large data sets, using the wrong tool for the task, insufficient budget to complete the analysis, not allowing enough time for testing and experimentation (the science), focusing too much on building the model (the engineering), not learning from the data, our assumptions, our preconceptions, and our skill. This paper will explore the question of how training regression models to predict rock density are affected by gaps in the data (missing observations and missing variables). What could possibly go wrong?

May 24, 2023