Webinar: Optimization Strategies for Different Underground Deposit Types

Introduction and webinar overview

Hi everyone, and thank you for joining. My name is Anna, and I am part of the business development team at K-MINE. Before we start, I would like to invite you to follow our LinkedIn page where we share product updates, mining events, case studies and announcements about upcoming webinars and training sessions.

In this session we focus on how underground optimization principles come together in real projects. We will look at geometry as the foundation of optimization, the algorithms and methods that sit on top of that geometry, and how they connect to underground mine design and production planning. Finally, we will review case studies that highlight common mistakes and best practices for different underground deposit types.

About K-MINE and our focus in underground mining

K-MINE started in 1994 in Ukraine and has since expanded its footprint across Europe, the United States and Canada. Our team combines software developers, geologists and mining engineers, including Qualified Persons certified under NI 43-101, JORC and SEC-equivalent reporting standards.

We work with a wide range of mining companies from early exploration groups to large producers. Our goal is to deliver reliable, cost effective solutions that are tailored to each project rather than forcing a one size fits all workflow.

K-MINE provides an integrated mining software platform that supports the full value chain from early exploration through production. We cover geological data management, 3D modelling, mine design, open pit and underground scheduling, stability analysis, survey and reporting. The platform includes flexible tools for both open pit and underground operations.

One of our key strengths is adaptability. A site can start with a single module and then extend to the full suite as the project develops. If required, we can integrate IoT or dispatch systems to provide real time monitoring and analytics so that planning decisions stay aligned with actual underground conditions.

Our software is shaped by real consulting work. Alongside K-MINE Desktop we provide geological modelling, mineral resource estimation, underground design, scheduling and technical report preparation under NI 43-101, JORC and other standards. We have deep experience in hard rock commodities with a strong focus on critical minerals. If you would like to discuss your project, you can reach out via comments or the contact form on our website.

Geometry as the first layer of underground optimization

In underground optimization discussions, people often focus on economics and algorithms. In practice the real turning point comes earlier when we decide what geometry we are even willing to mine.

If we do not define a strict geometric representation of the orebody and connect it to a specific mining method at the start, then even very advanced IP, MIP or MILP formulations will optimize shapes that can never be mined at the face. This is why the first layer of optimization is geometric, and why it must be fully computable.

Different orebody types behave very differently when we discretize them for optimization:

• Vein deposits form elongated bodies in plan and down dip, with variable thickness and gaps. They need orientation aware subblocking and clustering so that stopes follow the vein without cutting into hanging wall or footwall.

• Stratiform deposits form layered volumes with relatively consistent thickness and clear roof and floor contacts. They work well with a regular 10 x 10 or 15 x 15 mining grid that maps directly to room and pillar layouts.

• Massive and stockwork deposits form compact but irregular 3D volumes with internal lenses. The boundary of the viable zone rarely aligns with a single plane, and not all parts are accessible from existing workings. For these we often need an explicit accessibility mask.

Once the orebody is formalized and discretized in a way that respects its geometry, the resulting blocks or clusters can feed a realistic optimization engine.

Linking geometry to stability and mining method

The next step is linking geometry to rock stability. Traditionally this is done using a modified stability number and hydraulic radius frameworks such as the Mathews Potvin approach. Instead of applying this manually during design, we integrate it directly into the optimization workflow.

Every candidate stope generated from the block model is evaluated in stability space. If it falls into supported or unstable zones, its dimensions are adjusted until it fits within acceptable limits, or required backfill is added. This means we do not build an economically perfect shape and then throw away half of it after geotechnical checks.

We tell the optimizer up front that along strike the allowable span is limited, and down dip the height is constrained by the mining method. All subsequent models then operate only on objects that are physically constructible.

This also allows us to introduce multi criteria analysis at the geometric level. We already have conflicting objectives before economics appear:

• Minimize dilution

• Maximize recovery

• Respect geotechnical spans

• Preserve workable stope orientation and access

Methods such as AHP or other MCDM approaches are well suited here. We can set mine method and stability at the top of the hierarchy, recovery and dilution on the next level, and routing or access at the third. For stratiform deposits we give more weight to grid conformity and pillar stability. For veins we prioritize mining width and sidewall stability. For massive bodies we emphasize recoverability and potential scale up to larger stopes.

Alternatively, we can normalize criteria and use a distance to ideal solution approach. The ideal stope is fully stable, non diluting and aligned with the chosen method. Candidate stopes are ranked by their distance from this ideal. The key point is that this geometric screening happens before financial evaluation. We first select shapes that make geometric and technological sense and then apply economics.

The mining method link becomes explicit:

• Narrow steep veins naturally point to mechanized cut and fill or narrow long hole stoping. Optimization must generate stopes that follow the vein and respect minimum mining width.

• Gently dipping stratiform deposits lead to room and pillar variants. We start from a grid of potential rooms and later decide which cells are economic in the first phase.

• Massive ore bodies with typical rock strength often point to sublevel stoping or similar methods where large stopes are acceptable and outline smoothing improves productivity.

If we do not fix orebody type and allowed orientations at geometric stage, optimization will produce shapes that designers later have to slice apart to match the chosen mining method.

Accessibility, routability and realistic search space

In underground mining there is no value in optimizing stopes that cannot be reached with acceptable gradient, turning radius or ventilation path.

To handle this, we add an access graph to the geometric model. It represents levels, ramps, blind drifts, raises and ventilation crosscuts. From this we derive reachability of each potential stope. In the simplest case this is a binary matrix between access nodes and stope centers. In more advanced cases it becomes a development cost function integrated into economic evaluation.

Only after we align clusters with orebody geometry, apply geometric trimming, run multi criteria ranking and confirm reachability does it make sense to move to economics in linear or integer optimization models. At that stage we are no longer working with idealised polygons but with stopes that can realistically be drilled, blasted, ventilated and backfilled.

From block model to stope clusters in K-MINE

Let us look at how this is implemented in K-MINE. Our approach starts from the observation that the key unit in underground mining is the stope, not an individual block. A stope already carries orientation, practical dimensions, geotechnical meaning and alignment with a mining method.

The first step is to build a geometric framework for future stopes. The block model is grouped into clusters according to orebody strike and dip while respecting angular limits defined by the method.

• In narrow veins optimized for mechanized cut and fill, we aim for slender, vein parallel stopes with minimal deviation to control dilution.

• In sublevel stoping of massive bodies, stopes can be larger and orientation tolerances wider.

• Minimum and maximum cluster sizes are set by method: small for narrow veins, fixed to room dimensions for stratiform deposits so room and pillar layouts do not need a different discretization.

Once this spatial structure is defined, optimization can start. In K-MINE the objective can be grade driven or economics driven.

• Grade driven mode is valuable for highly variable vein deposits where secured metal content is critical.

• Economic mode is often preferred for more homogeneous massive bodies. Clusters are evaluated against cutoff, metal prices and costs. Only those with positive contribution remain in the stope, avoiding marginal volumes that would later need manual trimming.

Constraints are applied in sequence:

1. Geometric constraints – height range, maximum width, avoidance of impractical tapers and bottlenecks. In stratiform room and pillar layouts, the optimizer constructs stopes that match room width and height directly. In stockwork deposits under sublevel stoping, constraints reflect level height and room length so the stope can go straight into detailed design.

2. Geotechnical constraints – exclusion of weak zones, limiting spans, enforcing sidewall stability. Sections of poor rock quality along strike or down dip are marked inadmissible even if grade is high. This is critical in faulted underground gold mines.

3. Economic constraints – applied only after geometric and geotechnical checks pass. In veins the focus is dilution control. Waste blocks that add little metal but a lot of tonnage are left out. In massive bodies a small reduction in average grade may be acceptable if it produces a cleaner, more productive shape.

After that, K-MINE refines the stope. Sidewalls can be adjusted, incidence relative to the orebody improved and adherence to the payable zone tightened. For narrow gold veins this preserves selectivity and avoids barren rock. For sublevel stoping, refinement smooths edges and removes block model artefacts, simplifying drilling, blasting and backfilling.

The result is a set of stopes that are ready for use in underground design.

K-MINE also allows restriction of the optimization domain. Engineers can define the current working volume, exclude protective pillars or inaccessible areas and prevent stope generation below a certain elevation or outside a ventilation window. The optimizer remains aware of the intended mining method through all of this, adjusting cluster size, angular limits and dilution rules accordingly.

Integrating optimized stopes into underground design

Once optimized stopes are generated, they must be integrated into the real mine layout. The optimizer works in the coordinate system of the deposit, while designers work in the coordinate system of the mine: levels, ramps, crosscuts and ventilation drifts. Integration is about transforming an already workable shape into a set of actual workings.

The first manual pass is a site specific validation. The optimizer enforce general method constraints, but each mine has its own rules that seldom appear in public documentation:

• Protective pillars

• Zones of reduced rock mass quality

• Levels where drilling direction is constrained

• Areas that must be backfilled

Each stope receives a local status: accepted as is, accepted with backfill, accepted only with reduced height or to be rerun under stricter constraints. This avoids repeatedly correcting the same issues later in detailed design.

Then stopes are positioned into the existing network:

• In vein deposits mined by long hole or mechanized cut and fill, integration means generating short access drifts and drilling bays at correct elevation and azimuth while keeping distance to the face minimal.

• In stratiform deposits with room and pillar layouts, integration is mostly a snapping exercise. Stopes are aligned to the mine grid, pillars are preserved and established ventilation branches are used.

• In massive ore bodies under sublevel stoping, stope geometry is often sound but needs ramps or sublevel access, drill drifts and ventilation connections.

Design does not invent geometry at this point. It materialises the geometry created by the optimizer.

Connecting optimization with production planning

After spatial placement, the time dimension is added. Optimization indicates what is worth mining. Design and scheduling decide when and in what order it can be mined.

Sequences vary by method:

• In narrow vein cut and fill with backfill, adjacent stopes rarely run in parallel due to backfill, stability and ventilation constraints.

• In room and pillar mining more parallel work is possible, but only if pillar spacing and air flow remain safe.

• In sublevel stoping, level by level directionality and slot sequencing dominate. Mining a lower stope before its upstream partner may look profitable on paper but can break access and disrupt the production flow.

During integration we attach precedence links and mutual exclusion rules to each stope. Examples include:

• A stope can only be mined after the neighbour is backfilled and cured.

• Two stopes sharing the same ventilation branch cannot be mined in parallel.

• A stope requiring a long blind drift must wait until development is complete.

With these relationships embedded, the scheduling system receives a coherent network of tasks, not just a flat list of profitable shapes.

Reoptimization is also critical. Prices change, cutoffs move, new drilling updates the block model. The whole point of using an optimizer is the ability to rerun it. But reruns are only useful if stope identity and connections are preserved.

Each stope in K-MINE has a stable identifier. Development created for that stope knows which object it belongs to. If stope geometry changes slightly, only attached crosscuts are regenerated. If geometry changes significantly, the stope is flagged for redesign. If only economics change, geometry stays intact and updates flow into scheduling.

Tolerance for automatic updates depends on method. Vein mining has low tolerance, so updates are mostly manual. Stratiform room and pillar deposits are more forgiving. Sublevel stoping is the most flexible as long as level spacing and stope height remain valid.

Method specific planning: cut and fill, room and pillar, sublevel stoping and sublevel caving

In cut and fill mining an optimized stope represents a full cycle: drilling fan patterns, blasting, mucking, cleaning, placing backfill, waiting for strength gain and moving equipment to the next slice.

Planning maps each stage into a calendar with durations derived from optimized geometry. Longer stopes and narrower widths imply denser drill patterns and longer draw cycles.

Dependencies capture:

• Prohibition on opening neighbouring stopes simultaneously in the same ventilation wing

• Waiting periods while backfill cures before mining above or adjacent

• Resource loading for drill rigs, LHDs and paste plants to avoid peaks in demand

Economic attributes help choose which of the accessible stopes move first based on margin and contribution to plant feed quality. Actual paste cure times are tracked and used to fine tune later plans.

In room and pillar mining optimized stopes correspond to economic cells in a regular grid. Planning is driven by roof stability and pillar preservation rather than pure economics. Panels and access headings are laid out to support ventilation and haulage efficiency.

Dependency rules include:

• Adjacent rooms cannot be mined without preserving a pillar between them

• A cell cannot be mined if it breaks a ventilation circuit

• Rooms linked to the same narrow drift cannot be worked simultaneously

Scheduling assigns time windows panel by panel within geomechanically permissible options. Backfill strategies are inserted where optimizer marked the need for pillar unloading. Large panels are often split into smaller sub panels to reduce equipment deadheading and stabilise production.

In sublevel stoping, optimized stopes already align with the sublevel design. Planning focuses on synchronising drilling, blasting and backfilling across many levels. Stopes are grouped into panels to concentrate work and reduce travel.

A checkerboard pattern is used: drill the first stope, blast and draw, drill the second in parallel, backfill the first while blasting the second and so on. Dependency rules prevent blasting next to fresh paste, drilling in high gas zones or simultaneous drawdowns that overload shared conveyors.

Economic attributes guide priority when drilling or paste capacity is limited. Access development is integrated into the schedule so distant stopes start later while closer stopes fill early production windows. Level by level strategies help maintain vertical stability and safe working conditions.

In sublevel caving, optimized stopes are interpreted as acceptable draw zones along a caving front rather than standalone stopes. Planning is built around continuous cave propagation, draw control and uniform advance of the front.

Panel maps define rows of drawpoints, sublevel sequences and target advance rate along strike. Optimization highlights priority draw zones, but only areas on the active front under good ventilation and draw conditions are scheduled.

The dependency network prevents the front from running too far ahead and leaving undrawn rows behind. Geomechanical monitoring adds safety windows. If subsidence or microseismic data shows accelerating deformation, draw is reduced in that area even if modelled economics suggest otherwise.

Economic weighting is used only when choosing between equally ready rows, favouring higher grade and better selectivity. Loading charts by level and drawpoint row help avoid bottlenecks. Periodic reevaluation incorporates actual grades and dilution from draw control back into future plans.

Common pitfalls in underground optimization

When optimization is treated purely as an algorithmic exercise, the workflow looks simple: take a block model, apply economic parameters, generate stopes. In real deposits several recurring problems appear.

1. Vein deposits with coarse discretization

In narrow veins one of the most common issues is a block model built on a grid that works for geology but not for mining. Blocks are too large relative to vein thickness. Even with clustering and angular control, the optimizer is forced to pull in waste to get continuous stopes.

On paper, dilution looks acceptable and economics positive. But once the design enforces minimum mining width for equipment, the actual width exceeds the limit by tens of centimetres or more. The designer trims the stope back and the economics deteriorate.

The practical fix is to adjust model scale in the vein zone. Use finer discretization, subblocking at contacts or a separate grid for the vein. That allows the optimizer to shape stopes that follow the vein without pulling in unnecessary barren rock.

2. Stratiform deposits and misaligned grids

In stratiform deposits using room and pillar mining, a different problem appears. Optimization runs on a standard cubic grid that does not match the mine room spacing or pillar pattern.

The result is an irregular stope outline that:

• Pushes into future rooms

• Cuts across planned pillars

• Produces thin slivers that cannot be mined in practice

From the optimizer’s point of view, this is just a set of positive value blocks. From the designer’s point of view, it is unworkable. When the outline is snapped back to the mine grid, much of the apparent economic advantage disappears.

The correct workflow is to fix the mine grid first, evaluate cell economics on that discretization and then run optimization on those cells. K-MINE follows this principle, clustering aligned with the mine grid so that outputs are immediately compatible with real layouts.

3. Massive deposits and geotechnical boundaries

In massive and stockwork deposits, the orebody is large and sublevel stoping seems straightforward. Here errors often appear at the border between geotechnical limits and economic expansion.

Optimization tends to expand shells by adding marginal blocks that appear to improve metal output with limited dilution. On the block model this looks beneficial. In reality expansion may push the stope into areas where ground control requires extra pillars, reduced height or additional support. Designers then split the stope or trim it back, changing backfill volumes, access lengths and schedule logic.

To avoid this, geotechnical limits must be coded explicitly. Zones of reduced stability should be excluded from economic shells instead of being left for late stage correction. Slightly more conservative optimization usually produces a smoother workflow and a cleaner final project.

Best practices and real mine example

Several best practices help avoid these pitfalls:

• Separate zones by data quality. Use finer grids and more detailed stope shapes where drilling is dense. In poorly drilled areas restrict geometry to robust, easily designed forms.

• In selective narrow vein mining, consider geometric smoothing of vein contacts. Instead of oscillating widths every few metres, build smoother surfaces that lead to more regular stope dimensions and better operational consistency.

• Close the loop between optimization and scheduling. If the first scheduling pass reveals ventilation or access conflicts, feed those constraints back into the optimizer and rerun. It is better to generate shapes that already respect real constraints than to trim them manually afterwards.

A real mine example illustrates this. After introducing full 3D optimization, the team initially generated 15 to 20 percent more acceptable stopes than manual design. But when those stopes were placed into the mine layout, about one third conflicted with ventilation or needed development that was too long for the expected return.

Instead of fixing these one by one in design, the team added two constraints to the optimizer: a maximum allowed drift length to access a stope and an assignment to specific ventilation wings. With these in place more than 80 percent of optimized stopes went directly into planning with no modification, and the time needed to align them with the schedule was almost halved.

The key lesson is that the goal is not to perfect geometry manually but to push real world constraints back into the optimization engine so it produces shapes that already fit the mine.

Closing remarks

In this webinar we looked at underground optimization from the ground up:

• Geometry and orebody type as the first layer of optimization

• Direct integration of geotechnical constraints and mining methods

• Stope based clustering from the block model in K-MINE

• Integration of optimized stopes into design and scheduling for cut and fill, room and pillar, sublevel stoping and sublevel caving

• Common pitfalls and best practices across different underground deposit types

If you have questions or would like to discuss how K-MINE can support your underground project, please leave a comment under the video or contact us through our website. Thank you for watching, and we look forward to seeing you at our next session.