resource/reserve model
Update reported in October 2021 NI-43101 Technical report (effective July 26)
Jan 2020-June 2021 data
185Km drilling/336 holes
Update to Saddle zone
Expect 100km by year end 2021
Litho-Chemical Domains defined are a good discriminant of gold distribution and should be added to resource model.
Confirmed by statistical validation (Mario Rossi Dec 2021)
GRADE CONTROL MODEL
Machine Learning:
Defining domains is slow.
Re-assaying older drill holes that didn't include data.
Machin learning expected to speed up domain definition process-can then share this with Medium term and GC model
Calibration against in-situ grade Control Model:
will include dilution/ore loss and information effect accounting of impact of modelling from DDH to RC holes
Dilution:
Inconsistencies in modelled dilution.
Resource SMU- 10x10x7.25 vs Reserve SMU-40x30x7.25
(MSO process where diluted average grade is assigned to all blocks inside a panel)
calibration dilution based on conditional simulation models and re-blocking to appropriate SMU size of in-situ gc model
Three additional options to to calibrate model and compare to practical smu:
grade-tonnage curve at assumed smu from theoretical change of support models (discrete gaussian model.
In-situ gc model which can provide dilution estimates by simply re-blocking to assumed smu size.
conditional simulation models (as used in Information Effect study) that can provide dilution for a number of SMUs including theoretical and practical SMUs currently assumed by DL. :
Top Cuts:
Capping calibrated against reconciliation but could inherit uncertainties embedded in overall recon process.
consider re-defining values based on first principles (geological) such as continuity of high-grades including directional continuity for each domain
traditional capping increases smoothing. Replace procedure by using original grade but restricting spatial influence at time of estimation.
North Pit Resource is Risky:
Use of IMOK or Pack-type model that defines with a 40% probability can easily overestimate T&G above cut-off
BLAST HOLE SAMPLING
Single sample per bench or half bench in some areas
Estimate economic model/Loss function:
actual pit costs (flitch vs. full bench)
processing
G&A
Recoveries
Long-term or budget gold prices
Conditional Simulation/Indicator Krigging for daily ore/waste delineation.
Comparative study - get 1 year of production and re-model blast by blast.
Add auxiliary variables:
Moisture Model
Additional density data values may allow estimating locally in-situ density in gc model
Optimize block sizes. Currently too big. Opportunity to improve resolution.
Geomet Variables:
Add as they become available
introduce TPH estimates based on f80/p80
Alteration types can be proxies for litho-chem domains
Local structural information can be included in gc model
SAMPLING IMPROVEMENTS
Trays unlikely to result in good sample
Autosamplers not feasible
Use shovel with improved protocols (or hollow pipes)
Avoid sub-drill material
Sample on 4 sides of cone
Specialist consultant to review sampling methodology
ORE PRO 3D
Use dynamic update of Litho-chem domains in GC model
Optimize data gathering process for updating database (including rc drilling, face and BH chip logging)
Ore/Waste polygons follow block edges creating high dilution corners.
Mining direction and flitching not applied consistently.
Single CoG used. Optimize by sequentially defining HG blocks, followed by MG then LG
Data Preparation for Machine Learning Model:
Update of Database and Data preparation for feed into the machine learning model needs to be expedient and efficient, Need to develop a optimal process for this.
RC samples better than BH but can be compensated by higher density of blast holes if neither is systematically biased.
RC Bias vs DDH:
may be due to drilling in water combined with presence of coarse gold.
prioritize investigation as it may indicate high-grade bias
RC Bias vs DDH:
may be due to drilling in water combined with presence of coarse gold.
prioritize investigation as it may indicate high-grade bias
Density & Moisture Values:
Only 8,691 density values (Wood Audit Oct 2020)
Insufficient to provide robust in-situ density modelAdd denisty values to database by testing more core intervals.
Add moisture values from RC samples on regular basis (assuming original wet and then dry sample weight is recorded in sample drying process)
Information Effect Study
Simulated drill holes are vertical. Optimize to simulate inclined drill holes
SMU in operation is larger than GC block size (10 x 5 x14.5) used in comparison which implicitly assumes more selectivity.
If less selectivity is assumed then more widespread drilling would result in more precise ore/waste definition
Instead of krigging from tightest drill hole spacing (10 x 5) average original simulated nodes to the reference SMU block size to avoid smoothing implicit in krigging.
misclassification has been done on block by block basis which assumes perfect selectivity. More representative to apply ore/waste selection criteria used in defining mineable polygons.
More Conditional base simulations done internally:
Resource classification: provide unceratinity measure (confidence limits) to exisiting resource categories
Provide grade-tonnage uncertainty curves and compare to resource model grade tonnage curve by domain and gloabbaly to evaluate dilution/oreloss at various cog's
Evaluation of GC model. GC model provide best result when they reproduce grade variablity.
Evaluate long and short term min plan risks. Based on expected confidence for different planning perios.
Validation of sample intervals and effect of compositing to 14.5m can be evaluated
Medium-term Model:
Improve short-term mine plans using medium term model obtained on a monthly basis.
RC drill can be done on 40 x 10m grid. Optimized with experience and Information Effect Study.
Litho-chemical domains should be updated for 3 month volume then complete grade estimation include density, moisture and geomet variables.
Short-term schedule can then be adjusted as necessary resulting in recommendation for flitch mining.
Geometallurgical Model
JK Tech do testing and modelling. Then implement in Resource/Reserve, medium-term and GC models.
Currently focused on throughput variables. Future development to consider ore and gangue minerology assuming it can be useful.
Current litho-chem domains may provide info on certain gangue minerals impact recoveries.
Consider available info from exploration and other data sets:
Point Load tests, RQD, Drill Pen rates, in-situ, density and others.
May give prelim indication while actual tests (DWi and BWi) are being developed.
Consider proper QAQC procedures for these geomet tests. Treat as regular exploration drilling campaign.
When estimating, some don't average linearly as in TPH. Instead of using distance or krigging on tph values directly, use kWh/t and overall intended power draw.
Alternate comminution models (practical or heuristic) be looked at applied at similar plants to test applicability at DL. Actual tph numbers from mill can be used to compare, calibrate and reconcile at different TPH.
Reconciliation:
Include on tonnes and grades separately not just metal content
Reconcilier package being implemented
Add additional comparisons on a routine basis to improve understanding of material flow and impact on value chain of specific aspects, sources of dilution and impact of changes in modelling
Reconciliation can be used as optimization and calibration tool. Mapping uncertainties in process to better understand sources of errors provided opportunities for improvement and calibration of LT and ST models
Resource Model vs GC model in-situ polygons (not as marked in pit-D4). This may be comparison that may aid in defining top cuts.
GC in-situ vs GC mark-out (D5). Checks consistency of mark-up in pit to blast heave
Reserve Model vs Mill Feed (R5) - although mill feed is back calculated grade, the tonnage can provide a key data point.
Mine Production vs Mill Feed (R6). More reliable indication of true dilution and ore loss assuming data used is reliable
After adding density and moisture to gc model, this can be correlated to truck and belt tonnages reconciliation depending on assumptions about stockpile feed.
Routinely keep track of specific comparison to allow for review of the degree of uncertainty of different options to report specific T&G
Tonnages expit- truck count vs volumetrics + density assumptions; consider statistical variance of on-board scales in short term comparisons
Analyze data from dispatch; trends in misallocations or sporadic
Follow up assumed moisture content in trucks vs determined on belts vs moisture values in gc model; look for seasonal effects
Stockpile movement estimates
Input tonnages to mill. 510 vs 410, thickener underflow
Mill should provide t&g sampled at different points in mill; particularly tails and au produced; but not only in gold content
As part of uncertainty mapping - quantify operational dilution and ore loss. Unplanned dilution added from observations in pit and dispatch.
Flitch Mining
Decision making process starts by defining criteria (example high grade ore zones and also spatial distribution of ore zone) and then find areas where flitch mining can provide economic advantage because of increased selectivity.
cost-benefit analysis performed on medium term model and confirmed in GC model. Consider that reserve model is diluted to bulk mining scenario with MSO process. Use grade used prior to MSO dilution.
Issue potential flitch mining zones to short term planners for practicality analysis
Operational Issues to consider:
Trade off between 7495 bench development vs dilution/ore loss
Impact on drill fleet capacity and 795 tire life
Evaluate FS vs Backhoe config
Frost blasting
additional support equipment
metrics to track progress against plan
Reserve model should include flitched and full bench mining dilution.
Full impacts of 14.5 vs 7.25m bench have not been considered in LOM. Emphasis of plan is on operational deployment of equipment than economic trade of of dilution/ore loss.
Uncertainty Mapping
Mapping key processes and potential areas of uncertainty within each process. Subjective flagging of importance of each uncertainty within each process is performed followed by mitigation measures of main uncertainties identified.