Mineral Engineers: Statistical Methods For
Most ore grades (especially precious metals) follow a lognormal rather than normal distribution. This means:
Finally, a sobering reality for the mineral engineer is the nature of sampling. Pierre Gy’s Theory of Sampling (TOS) is a statistical framework that dominates this area. Gy demonstrated that the fundamental sampling error is inversely proportional to the number of particles in a sample. For a coarse, high-grade gold ore, a single 5 kg sample might contain only a few gold particles. The variance in the assay result from replicate samples of this material is enormous—a false sense of precision is created by finely grinding the sample before assaying, which does not correct the initial sampling error. Statistical thinking forces the engineer to design sampling protocols (correct cutters, appropriate sample masses, proper splitting techniques) that ensure a sample is truly representative, because no statistical test can validate an incorrectly taken sample.
1. Data Characterization and Exploratory Data Analysis (EDA)
Developing customized water quality monitoring and mineral sampling procedures to minimize variance. Process Optimization: Statistical Methods For Mineral Engineers
Testing new reagents or grind sizes with minimal trials. 2. Fundamental Statistical Techniques Descriptive Statistics
[Screening: Fractional Factorial] ➔ [Optimization: Central Composite Design] ➔ [Result: Response Surface Map] 7. Mass Balancing and Data Reconciliation
is the standard deviation (uncertainty) of the measurement instrument. Most ore grades (especially precious metals) follow a
Before applying advanced modeling techniques, a mineral engineer must understand the baseline characteristics of the operational data. Mineral processing data is notoriously noisy due to sensor errors, changing ore mineralogy, and process disturbances. Central Tendency and Variability
Determining the average grade (e.g., %Cu) or throughput.
Once the ore is delivered to the processing plant, the challenge shifts from estimation to efficiency. The comminution circuit (crushing and grinding) and the separation circuit (flotation, leaching, magnetic separation) are complex systems with multiple interacting variables: feed rate, solids density, pH, reagent additions, and particle size. Here, classical statistical methods take center stage. is particularly powerful. Instead of the traditional "one-factor-at-a-time" approach, DOE allows engineers to vary multiple factors simultaneously, revealing not just their individual effects but critical interactions. For example, the effect of a collector reagent in flotation might depend entirely on the pulp pH. DOE, through factorial designs and response surface methodology, can map this interaction and identify the optimal operating region with a minimum of expensive plant trials. Gy demonstrated that the fundamental sampling error is
) quantify a plant's ability to produce within designated target specifications. A Cpkcap C sub p k end-sub
A copper porphyry deposit. Inverse distance weighting might over-weight a single high-grade assay near a fault. Kriging detects the anisotropy (directionality) and assigns weights based on the continuity along the ore body vs. across it.
Variance and standard deviation quantify process stability. A high standard deviation in flotation feed grade indicates significant ore blending challenges.
Engineers use ANOVA (Analysis of Variance) to determine if a change in production—such as a new chemical collector—actually improved recovery or if the gain was just random noise. 🛠️ Essential Statistical Toolkit According to the definitive guide Statistical Methods for Mineral Engineers by Tim Napier-Munn , the core toolkit includes: Statistical Methods for Mineral Engineers - Google Books
Modern plants generate thousands of data points every second via distributed control systems (DCS). Univariate statistics cannot handle this complexity, necessitating multivariate statistical methods. Principal Component Analysis (PCA)