The implementation of autonomous haulage systems in open-pit mines is a progressive step in the industry, but it brings potential safety risks that need careful evaluation. This study developed a discrete event simulation model to analyze and evaluate these risks in different operating scenarios -fully autonomous, hybrid (a mix of autonomous and human-operated vehicles), and non-autonomous operations.The simulation model was developed using the HAULSIM and Anylogic software, integrated mine layout, haulage operations, and various fleet compositions. It provided insights into collision risks, a significant concern in mining safety literature. Results showed that collisions were inversely proportional to the number of autonomous vehicles in operation, indicating the potential safety advantages of fully autonomous operations. However, certain high-traffic intersections were identified as high-risk areas, emphasizing the need for targeted risk mitigation strategies. Further, a profile risk matrix was developed to provide a comprehensive view of collision severity and likelihood in each scenario. This highlighted the impact of collisions on both human safety and project operations. Based on the results, risk mitigation strategies were proposed, with a focus on redesigning intersections, implementing strict rules for human-operated trucks in autonomous zones, and improving autonomous vehicle capabilities. The study, while insightful, was limited by assumptions and the generic nature of operational data used in the simulation. Therefore, future research should seek to incorporate more detailed, mine-specific data and empirical insights from projects that have implemented autonomous haulage systems. Continuous advancements in autonomous technology and simulation modeling will be key to ensuring a safe and productive mining environment.

Haul truck driving remains to be the most dangerous occupation in the mining industry (Kecojevic et al., 2007), leading in fatalities over other occupations in the mining industry. As the transition to autonomous haul trucks over the coming decades will not address thisproblem today, an immediate solution is needed that can be adopted at short notice to help aid in minimizing haul truck accidents. A gamified training simulator was developed that teaches haul truck drivers how to handle the hazardous situations that lead to the most injuries in the mine environment. The simulated nature of the training method allows truck drivers to gain knowledge on hazardous situations and immediately put them into practice by applying the gained knowledge during the different levels. Gamification aspects like a score system and progression through levels were added to increase engagement during the training simulation.
A statistical analysis was made on haul truck accidents from 1983 until 2022. The accidents that proved to be most relevant were included in the content of the training software. A training software was developed using Unreal Engine 5, where trucks could be driven using a Thrustmaster T150 Pro steering wheel and driving pedals. 18 participants from ages 19-65 underwent the training program and two tests of identical difficulty were taken before and after undergoing the training, in order to make a judgement on the progress of the participants by undergoing the training simulation. Participants were judged based on the number of crashes and traffic violations committed, attempts required to complete the test, as well as the time taken to finish the test. A score was calculated based on these factors providing an overall judgement of the participants’ level of skill.
Average skill of participants increased by undergoing the gamified training, and training showed particularly good progress for underperformers in the initial test. After completing the gamified training the average score of participants was notably higher and the standard deviation was notably lower. It was concluded that all participants were brought to a comparably high level after completing the training.

The importance of decarbonization in various industries is growing significantly, especially when it comes to LHD equipment in the minerals and metals industry. Conventional diesel load and haul (LHD) equipment used in quarrying and mining operations have drawbacks that result in the emissions of hydrocarbon and greenhouse gases that need to be reduced to reach the Net Zero Emissions scenario. Decarbonization options that could replace diesel equipment include battery electric vehicles, green hydrogen, and alternative fuels. This project aims to give an overview of current and near future available decarbonization methods and to analyse the environmental benefit, cost effectiveness and technological feasibility of these alternatives to diesel fuel. The case study used to simulate and compare the three different decarbonization options is a mountaintop marble quarry based in Ipoh, Malaysia. A comparison has been made with Ipoh located in Malaysia and a hypothetical situation of Ipoh located in France to show the true potential of decarbonization as Malaysia does not incentivize businesses to reduce emissions due to their subsidies on the use of fossil fuel. Malaysia’s energy mix consist of primarily unabated fossil fuels.
The proposed workflow involves an extensive literature review of current and near-future decarbonisation technologies to replace diesel equipment and to create a haulage network based on block model data using the software Xpac Quarry solutions and its Haulnet package. Diesel equipment is imported from the Quarry solutions database and used as a base case to compare BEV and trolley assist simulations. Hydrogen and HVO fuel consumption have been calculated manually to obtain NPVs and associated emissions. An investigation into required infrastructure and energy requirements has been done to accurately define costs and resulting total CO2eq emissions.
The results indicate that replacing diesel at Ipoh with any of the decarbonization methods will result in a negative NPV as the infrastructure requirements need extensive capital investment of which the payback period often exceeds the life of mine. The least expensive method to implement HVO fuels, followed by BEVs and green hydrogen. Furthermore, it has been found that using BEVs instead of diesel at Ipoh is more environmentally polluting because producing 1 kWh of electricity is more polluting than producing the same energy worth of diesel. Green hydrogen is the least polluting method with no greenhouse gas emissions associated with it, followed by alternative fuels. When investigating the technological feasibility of the three main decarbonization options, hydrogen is the furthest away of being technologically feasible as from now. No green hydrogen using water electrolysers has been produced at an industrial scale, and a lack of technical expertise and infrastructure is present regarding distribution, transportation, and storage. Currently, no hydrogen LHD equipment is commercially available currently, and the same applies to BEVs that could replace the current equipment used at Ipoh. To bridge this gap, alternative fuels are the best option, but the market is competitive, and supply is limited. This project has given a good overview of current possibilities and clearly shows today’s options for decarbonisation and where industries stand with respect to infrastructure and equipment.

The mining industry is a significant contributor to greenhouse gas emissions and operational costs. The transportation of materials within mining operations, especially using diesel-powered trucks, accounts for a substantial portion of both emissions and costs. Optimizing fuel consumption in this context is crucial for environmental sustainability and financial efficiency. While there is some existing research on minimizing fuel consumption in mining operations, most of these studies prioritize production over fuel efficiency. Additionally, there is a lack of solutions addressing real-time dispatch problems while also minimizing fuel consumption. This thesis proposes the development of a model to optimize truck speeds in mining operations, with the primary goal of reducing fuel consumption. The focus is on minimizing waiting times for trucks at loading stations by adjusting their speeds using a metaheuristic algorithm, specifically Simulated Annealing (SA). To achieve this, a discrete event simulation (DES) model made in HaulSim to replicate a simplified mining site in Nevada, USA. The simulation includes Caterpillar 793 series trucks. The SA algorithm is applied to find optimal truck speeds, with the aim of reducing fuel consumption with the same level of production. The hypothesis is that lowering truck speeds in mining environments with queueing trucks can significantly reduce fuel consumption. The study explores how speed optimization impacts efficiency across different mining conditions, identifies the most effective optimization techniques, and quantifies potential fuel savings. The research reveals a 27-32% reduction in fuel consumption across various truck scenarios. However, these findings should be interpreted cautiously due to limitations in data constraints and modelling simplifications. In conclusion, the study highlights the effectiveness of SA in optimizing truck speeds and the potential for substantial fuel savings in mining operations. We recommend the integration of Artificial Neural Networks (ANNs) for a more nuanced approach to fuel consumption estimation and optimization, considering factors like road maintenance intervals and payload. This synergistic approach, combining metaheuristics and machine learning techniques, aligns with sustainable practices in the mining industry and offers promising avenues for further research and application.

The goal of a mining operation is to extract the maximum value from exploiting the orebody. The equipment used by these mining operations has a nominally rated performance to achieve needed annual production. However, this nominally rated performance is not achieved during operations. This study assesses the underlying causes of reduced shovel production in a mining operation.

The following research questions will be answered in this study: What factors contribute to deviations from the predicted performance? What factors are the most significant? What is the effect of automated trucks on shovel productivity? Can a data-driven model be developed to predict the actual productivity of the shovel in the real world?

The underlying causes of reduced shovel production are assessed through a literature review and five case studies. The case studies consisted of three data analytical desk studies, interviews with industry professionals, and a mine visit to north Africa.

The literature review shows that the mining industry uses formulas to determine the theoretical production rates. The factors in the shovel production formula should be predicted with frequency distributions. Also, truck automation will increase trucking hours, impacting shovel productivity through truck exchange time. The first case study shows a sensitivity analysis of the shovel formula. This sensitivity analysis shows that the swell factor, density, bucket fill, efficiency, and cycle time have a more significant impact than truck exchange time, dumping of the first bucket, and the number of cycles. The second case study shows the underperformance of different electric rope shovel models in different mining operations. The third case study compares the theoretical shovel capacity with the total material moved for 16 years. This shovel capacity and total material ratio should be between 1-2.5 for a mining operation to be classified as above-average-in-class. The fourth case study shows the different factors influencing shovel productivity based on the interviewees' responses. The last case study shows the effects of operational decisions on a mining operation. The literature review and case studies are used to develop a flowchart regarding the factors influencing shovel productivity.

This flowchart was used to synthesise the results. The difference between theoretical and real-world production rates can be decreased when the frequency distributions are known for all the shovel production formula factors. This shovel production formula for annual production rate can be divided into three main pillars. These three pillars are the hourly production rate, use of availability, and mechanical availability. The pillars allow OEMs or mining companies to implement the correct improvement measures to improve shovel productivity. However, one solution for every mining operation will be impossible due to the uncertainty in the data and variability of each mining operation.

To conclude, the factors that contribute to deviations from the predicted performance are categorised as uncontrollable (weather and geographical location), direct (density and cycle time), and indirect (fragmentation and face dimensions), which all impact shovel productivity. The most significant factors are not found during the study, but solving underperformance in the direct factors will solve most of the problems. The effect of automated trucks on shovel productivity will result in additional trucking capacity, which will need to be absorbed by the shovel. Lastly, a data-driven model can be developed with access to all the data from a mining operation. However, this data is often not available to an OEM. Therefore, it is not advisable to develop such a model.

Digital Soil Mapping (DSM) of soil types in geotechnical project areas is a top priority. These maps are often used in decision making and can have significant consequences related to costs and risks. Usually, these maps are generated by digital soil models that interpolate soil types at known locations. In practice, conventional spatial interpolation techniques are still often used for DSM of soil types, such as inverse distance weighting and kriging. However, conventional models are not well suited for predicting or interpolating soil types because of their inability to deal with categorical data properly. Besides, the design of the conventional models does not allow for incorporating the abundance of meaningful covariate information that is available nowadays. The flexibility of machine learning algorithms vanquish both problems and has become increasingly popular for DSM of soil properties in recent years. The results of machine learning techniques for DSM of soil properties are promising and generally outperform conventional models. However, few studies have used machine learning for DSM of soil types and is therefore still a relatively unknown field. Moreover, at the time of writing, there are no studies that use sequence models for DSM of soil properties or types. Hence, the author proposes to introduce a new method for DSM of soil types, namely a Long Short-Term Memory (LSTM) network. The intuition behind this introduction is that the spatial correlation can be captured in sequences and can improve soil type prediction.

Real project data from a cable burial project is used to evaluate and compare the performance of the conventional interpolation methods triangulation and kriging, the machine learning models random forest and XGBoost, and the newly proposed deep learning model LSTM. The project data consist of 757 vibro cores (VC), 718 cone penetration test (CPT), bathymetry data and sub-bottom profilers. The geotechnical data, i.e. VCs and CPTs, is received on separate PDF pages that require to be digitized first. This thesis describes a simple yet precise manner to extract this data from the PDFs. The VCs and CPTs are provided with a soil type interpretation and can be used directly for developing the models. The data is split into a training set to develop/train the models and a test set for evaluation. Ultimately, the best performing model is used to build a 3D stratigraphic soil model for the project area with associated prediction accuracies.

All state-of-the-art techniques outperform the conventional models and especially in predicting minority classes. The best performing model is random forest with an overall accuracy of 85.44\% and is comparable to the performance of XGBoost of 85.11\%. LSTM network achieved a slightly lower accuracy of 84.27\%. The results show that LSTM is suitable for DSM of soil types and has considerable potential for improvement as only a few possibilities of the model have been examined.

As a consequence of the difficulty to obtain complete direct information about the subsurface, and the importance of heterogeneity of soil properties in geotechnical design, a desire exists in the geotechnical community for a smarter use of the available data in common practice. The objective of this thesis is to provide a potential solution for this, by implementing the Random Finite Element Method (RFEM) in PLAXIS, a well-known geotechnical calculation software package, allowing users to stochastically model engineering problems while accounting for the spatial variability of soil properties.

As a result of this project, practitioners can use the upside of stochastic analysis using RFEM to derive new insights about the behavior of their system, such as potential soil-related asymmetries, all while obtaining a better picture of the risks associated with a certain design. As with any technological advancement, responsible use of the framework is required to ensure the upside is attained and computational concerns do not hinder the use of the model, feature which still requires further research and improvement.

Mineral Resource Modeling (MRM) is used to predict the properties of an orebody, however it does not come without uncertainties. Multiple approaches can be used to reduce the latter. Due to limited knowledge about the subsurface, predictions are difficult to be made. In this thesis the application of Bayesian Evidential Learning (BEL), in order to reduce uncertainties on MRM, will be researched. The uncertainties present in MRM will be linked to the knowledge obtained from case studies in different geological domains where BEL has been applied successfully and reduced certain parameter uncertainties. The gap in todays industry is the knowledge and proof that using BEL for MRM will reduce uncertainties and risks, works effectively and will consume less money. The aim of this study is to show that BEL is a useful approach to reduce uncertainty of the predictions in MRM. The success of a project is supported by the accuracy of the model utilized and the geological interpretation. BEL is a framework based on statistical relationships between data and prediction variables. It will predict the posterior distribution of the prediction variable. The various case studies that have been discussed are (Hermans et. al, 2019), (Hermans et. al, 2018), (Thibaut et. al, 2021) and (Tadjer and Bratvold, 2021). BEL has successfully been able to reduce uncertainties related to geological problems of the following:
• The temperature in an alluvial aquifer
• The efficiency of the thermal energy storage capacity in an alluvial aquifer
• The wellhead protection areas surrounding the pumping well using tracing experiments as predictors
• The prediction of leakages of CO2 and the storage of CO2
Thus, BEL can be seen as a potential approach to also reduce uncertainties in a mineral resource domain; the research of this thesis. In attempt to prove this, a descriptive case study on Tropicana Gold Mine has been executed. The aim is to show that the use of BEL will, based on the geological and geochemical properties such as lithology, grade and mineral type, reduce the uncertainty in the prediction of a geometallurgy property, namely the hardness of a rock. The six steps of BEL’s framework will be followed consisting of Monte Carlo simulations, Principal Component Analysis (PCA) and Canonical Correlation Analysis (CCA) in order to obtain relations between the data and predictor variables. Where the data variables are from exploration drillhole data and prediction variable the hardness of the rock. It shows that BEL is able to reduce the uncertainty of a geometallurgy property by using geological and geochemical properties. Meaning BEL can be applied to MRM to reduce uncertainties.

Open-pit mining blast patterns are designed to safely break rock mass at desired boundaries and to produce uniformly sized fragments that lead to the optimal performance of subsequent mining operations. An issue that is often overlooked, however, is the deviation of drillhole locations with respect to their coordinates as designed in the blasting plan. This causes changes in the spatial distribution of explosive energy; some areas of the blast may receive a surplus of explosive energy, while others see a reduction. This results in less uniform fragmentation, producing more fines and increasing chances of oversized boulders, which reduces the efficiency of loading, hauling, and crushing operations. Additionally, deviated drillholes can lead to slope stability issues, flyrock, and uneven benches. To mitigate these problems, this study attempts to reduce differences in the explosive energy distribution (EED) through an optimization approach that adjusts the height of explosive columns in blastholes. A case study combines planned drillhole coordinates of two blast patterns with both simulated deviations and actual deviated drillhole locations. The corresponding benches are discretized into block models, for which the explosive energy distribution is calculated from the planned drillhole configuration as well as the deviated holes. The optimization objective is then to minimize block-to-block differences in explosive energy by making discrete adjustments to the height of individual column charges. Since mathematical programming is not possible due to the complexity of the evaluated formula, solutions are generated by metaheuristic algorithms. Among the Genetic Algorithm (GA) and Tabu Search (TS), the latter is preferred for its local search neighborhood that gives superior computation times. An inconvenience of the block-to-block comparison is the introduction of very large differences in explosive energy in blocks nearest to deviated blastholes, which cannot be accounted for by any charge adjustments except reductions in the deviated hole itself. Because optimization should focus on manageable differences in lower-valued blocks, these high-valued blocks are discarded by applying a cap based on data percentiles. Obtained results show that optimized charging adjustments vary with the chosen percentile, which makes it challenging to define comprehensive truly optimal solutions. A bias for charge reductions at high percentiles gradually shifts towards a bias for charge increases at lower values. Because a bias-induced imbalance in the total amount of explosives (compared to the original charging) is undesirable, the setting closest to equilibrium is selected, which is the use of only the lower 50\% of data. In this configuration, optimization of the case study blast patterns results in improvements of the objective value of 0.53-1.54\%, or 2.14-3.94\% when only the blocks affected by the recommended charge adjustments are considered. The block-to-block comparison handles any interchangement of block values as unwanted differences, which should not always be the case. Therefore, the true improvements to the explosive energy distribution might be better expressed using an alternative optimization objective. Computation times could be further reduced by expanding on the developed algorithm, and combination with prior research would be helpful to include any GPS-related inaccuracies in the actual drillhole locations.

The case study consist of a small underground mine with a small mining crew. The vehicle park is relatively large, and therefore it is necessary to establish the added value of additional miners or equipment for short-term production planning purposes, assuming that staff size currently limits production capacity to find out if staff size is indeed the bottleneck in the production capacity of the mine operation. When the bottlenecks of the mining system are known, it will be easier to focus on necessary areas and further implementations to improve the system.

The truck numbers used in the simulation study were ranging from 4 to 7, and the operator pool size was ranging from 10 to 15 people. Significant findings of this study are that with the current mine setup of 4 trucks, there would be no increase in production when adding operators. For the 24 scenarios the production increase was determined, the revenue change and the mining cost. By adding trucks and operators, a production increase of 19.38 % could be reached with 15 operator and 7 trucks.

Current evacuation plans for underground mines are inefficient and outdated. Simulations have proven that smart evacuation can significantly increase the efficiency of escape times in case of an emergency. In a smart evacuation, miners are given real time instructions on a smart device (such as a smart-watch) about their route to a safe haven if an emergency occurs. In order to make smart evacuation possible, an optimization model is needed that determines the most efficient route to safety. Currently, shortest path algorithms, such as Dijkstra’s algorithm, Floyd-Warshall’s algorithm, or ant colony optimization are used for solving the optimization models. These algorithms, however, are computationally inefficient (Dijkstra and Floyd-Warshall) or may not provide the optimal solution (ant colony optimization). In this thesis, a mathematical programming method will be used to calculate the most efficient escape plan. Mathematical programming translates the goal of finding the most efficient escape solution to an optimization problem. The goal of this optimization problem is to minimize the total distance travelled by all individuals that are underground in case of an emergency. This is done by setting an objective function, which sums up all the distances travelled by the miners, and a set of constraints, which are used to localize the miners and safe havens. The ensuing problem statement is solved using the network simplex algorithm. This principle was used to calculate escape solutions in four different scenarios: with or without blocked pathways and with or without correcting the distances for the stamina of the individual miner. Four different types of results were generated: the division of miners among the safe havens, the total distance travelled by all individuals taking part in the evacuation, the path of an individual miner (called Miner X), and the running times of the different scenarios. It was found that blocked paths can have a large impact on the division of miners among safe havens and can significantly increase the total distance travelled. Using stamina categories increases the path length of Miner X (who has a high stamina) and also the running times of the algorithm. It can be concluded that the algorithm works for all four scenarios. As including blocked paths does not give a time penalty, but will locate trapped miners and send their colleagues on safer escape routes, using this feature has no downsides. The usefulness of adding stamina categories to the algorithm can be debated. The running time of the algorithm increases, while the solutions for realistic numbers of miners do not change. Furthermore, there are philosophical and social questions about the ethics of using these types of categories.