THE USE OF STATISTICS IN MINERAL PROCESS ENGINEERING

Sound metallurgical decisions rest on data, yet data is rarely as trustworthy as the confidence placed in it. A recovery figure falls and a meeting must decide whether the plant has a fault or the number is merely noise. A reagent trial reports a three percent uplift. A testwork report recommends a circuit costing tens of millions to build. A metallurgical account fails to reconcile with the smelters, and the discrepancy is measured in money. Each of these turns on a single quiet question: can these numbers be trusted, and what do they truly indicate? Answering that question, reliably and defensibly, is the purpose of this book.

The Problem This Book Addresses

Most engineers first encounter statistics as an abstract university course built on symmetric distributions and independent observations. Plant data seldom obeys those assumptions. Grades are skewed rather than normally distributed; readings taken minutes apart are correlated rather than independent; precious-metal assays scatter in ways governed by the nugget effect. Most significantly, the largest source of error in a typical result is committed before any calculation is performed — at the sampling point, the cutter, or the splitter — where no statistical treatment can recover what a poor sample has lost. A handbook for practitioners must begin from these realities rather than from textbook idealisations.

Who Benefits

This handbook is written for the professionals who must apply statistics while operating and improving a plant, rather than for specialists in the subject. It will serve:

Plant metallurgists and process engineers interpreting daily results and judging the significance of changes.

Laboratory and QA/QC managers responsible for the precision, bias, and defensibility of analytical data.

Metallurgical accountants producing balances that must withstand audit and reconciliation.

Testwork chemists and research metallurgists designing experiments and fitting models to laboratory data.

Students and graduates in extractive metallurgy seeking the applied statistical grounding that formal courses often omit.

The reader is assumed to be numerate and practically minded, but not a statistician, nor obliged to become one.

Practical Benefits

On completing the relevant chapters, the reader will be able to:

Design a representative sampling protocol and quantify the error it carries.

Distinguish a real process change from measurement noise, and report results with defensible confidence limits.

Plan an experimental programme that extracts the most information from the fewest, costliest tests.

Fit and critically assess kinetic, isotherm, and grade–recovery models without over-interpreting them.

Construct and reconcile mass balances and metallurgical accounts and propagate the uncertainty through to recovery.

Apply control charts and process-capability methods to plant data that is drifting and autocorrelated.

Why This Book Is an Invaluable Tool

Three features distinguish this handbook from a conventional statistics text. First, it is organised around the metallurgist's workflow — from the sample taken, through measurement, inference, experimentation, and modelling, to the balances and control charts of daily operation — so that each method is introduced by the problem it solves rather than by its mathematical category. Second, every technique is supported by a worked example using realistic metallurgical data, together with spreadsheet templates that allow the reader to apply the method to their own figures without rebuilding it. Third, and most importantly, the book treats statistics not as a means of manufacturing certainty but as a discipline of honesty about uncertainty. Its highest value lies in telling the practitioner, truthfully, how much confidence a number deserves — and, frequently, in establishing that the data are insufficient to support a decision at all. A metallurgist who can place a defensible interval on a recovery or demonstrate that a trial result is indistinguishable from noise, carries more authority than one armed with a single impressive average.

How to Use This Book

The chapters may be read in sequence, but each is largely self-contained and may be consulted as the need arises. Every chapter opens with the metallurgical problem it addresses and closes with a fully worked example. Call-out boxes throughout flag the recurring and costly errors of practice — treating skewed grade data as normal, trialling with non-independent samples, reading significance into differences smaller than the measurement error, and over-fitting models to inadequate data. The accompanying templates are intended for immediate use in the laboratory and on the plant.

Scope and Boundaries

This handbook addresses the statistics of operating and improving a processing plant: sampling, measurement quality, experimental design, modelling, mass balancing, metallurgical accounting, and process control. It deliberately leaves geostatistics and ore-reserve estimation to the mining and resource literature, engaging with them only where the grade and variability of plant feed are concerned.

A practitioner who, having read these pages, questions a number once accepted without thought, or designs a test once run on intuition alone, will have drawn from the book its full intended value.