Statistical Methods For Mineral Engineers -
Many flotation recovery curves follow a sigmoidal shape. The Hill equation (borrowed from biochemistry) models recovery as a function of residence time:
$$ R(t) = R_max \cdot \fract^nK^n + t^n $$
Where $K$ is the time to 50% recovery and $n$ is the slope (kinetics). Fitting this using non-linear least squares allows engineers to optimize residence time for maximum throughput.
Once significant factors are identified, RSM (e.g., Central Composite Design, Box-Behnken) models curvature. This is essential for finding true maxima (recovery) or minima (cost, reagent consumption).
Output: A contour plot showing predicted recovery vs. two continuous variables, with a clear stationary point.
PCA reduces dozens of variables (e.g., particle size bins, mineral abundance, XRD peaks) into a few uncorrelated “principal components.”
Application: A plant processing a complex sulfide ore used PCA on 25 QA/QC variables. Two components explained 78% of variance: PC1 (sulfide content) and PC2 (clay content). Monitoring just these two components instead of 25 separate charts simplified control.
Scenario: A copper concentrator sees recovery drop from 92% to 88% on the Monday morning shift. The shift supervisor blames the operators. Operations blames the ore.
Statistical Investigation (Hypothesis Testing):
Two-sample t-test result: p-value = 0.003 (<0.05). Reject H₀.
Conclusion: The drop is statistically significant. It is not random. Statistical Methods For Mineral Engineers
Root cause found: The primary crusher gap had drifted open over the weekend, producing coarser feed. Flotation kinetics slowed.
Lesson: Without statistics, you’d blame people. With statistics, you fix the crusher.
Pierre Gy’s theory of sampling is the bedrock of statistical mineral engineering. The fundamental sampling error (FSE) is given by:
[ s^2 = K \cdot d^3 \cdot \left( \frac1M_L - \frac1M_T \right) ]
Where:
Key insight: To reduce sampling variance by half, you must either:
Rule of thumb for mineral engineers: Never trust an assay result without knowing how the sample was collected, crushed, and split. A statistically invalid sample is worse than no sample—it leads to wrong decisions with false confidence.
The primary resource for this topic is the book Statistical Methods for Mineral Engineers: How to Design Experiments and Analyse Data Professor Tim Napier-Munn
. It is widely regarded as an essential text for plant metallurgists and assay chemists to manage experimental uncertainty and make data-driven decisions.
Below is a draft of the key features and statistical methods used by mineral engineers to optimize plant performance and minimize risk. 1. Essential Statistical Tools Many flotation recovery curves follow a sigmoidal shape
Mineral engineers use specific statistical tests to compare data sets and validate results from plant trials: t-tests, F-tests, and Chi-square tests
: Used for comparing quantities and determining if differences in performance (e.g., between two circuit configurations) are statistically significant. Analysis of Variance (ANOVA)
: Critical for analyzing the impact of multiple variables simultaneously on a process output. Regression Analysis
: Essential for establishing relationships between measurements, such as modeling how reagent dosage affects recovery rates. 2. Experimental Design (DoE)
Properly designed experiments are necessary to ensure that trial results are definitive and cost-effective: Factorial Experiments
: Used to study the effects of several factors on a process and identify interactions between them. Randomized Block Designs
: A method to reduce the influence of known but uncontrollable variables (like ore hardness variations over time) on trial results. Response Surface Methodology (RSM)
: A collection of mathematical and statistical techniques used to model and optimize processes, such as finding the temperature and pressure that maximize yield. 3. Monitoring Plant Trials
Specialized methods are used to track performance changes in real-time or over long durations: Cumulative Sum (CUSUM) Charts
: A powerful tool for detecting small, persistent shifts in process performance that might be missed by standard control charts. Paired Testing Once significant factors are identified, RSM (e
: Used to compare a "new" versus "old" approach under similar operating conditions to isolate the effect of the change. Time Series Modeling
: Helps analyze data collected over time to account for cycles or trends in ore quality and plant performance. 4. Uncertainty and Measurement Error
Statistical methods help quantify the inherent "noise" in mineral processing: Error Propagation
: Calculating how measurement errors in individual instruments (like flow meters or belt scales) affect the overall calculated recovery or mass balance. Confidence Limits
: Establishing ranges within which the "true" value of a parameter likely falls, allowing engineers to report results with a defined level of certainty. 5. Advanced & Emerging Methods
Modern mineral engineering increasingly incorporates data-driven and machine learning techniques:
Statistical Methods for Mineral Engineers heads for third reprint
Scenario: A lead-zinc plant sees erratic recovery (70–85%).
Statistical approach:
Result: $2.5M/year additional metal value.
Mass balance and metal balance reconciliation is where statistics meets accounting.