Skip to content
DSML.KZ
Publications

Article · Minerals · November 7, 2026

Control Selection and Prediction of Gold Recovery Gains in Multicomponent Microflotation

Serik Aibagarov, Aksultan Mukhanbet, Timur Imankulov, Adilbek Rustemov, Bakyt Daulet, Bagdaulet Kenzhaliyev

Aksultan Mukhanbet
Open via DOI

Abstract

Estimates of gold recovery gains from multicomponent microflotation (MMF) depend on the conventional test chosen for comparison. We analyzed 132 laboratory tests from 13 daily series, comprising 31 conventional controls and 101 MMF tests, to assess control sensitivity and the predictive value of reagent descriptors. Mean final-concentrate recovery was 5.037 percentage points higher than in the assigned conventional controls (95% grouped bootstrap interval 0.545–8.240). For the 88 MMF tests eligible for within-day matching, reassigning controls increased the mean difference from 4.318 to 5.979 points. Mean recovery differences remained positive across sensitivity analyses, although several intervals included zero. Recovery increased, and dump-tailings grade decreased in 59 of 101 reference comparisons; the average tailings-grade change was uncertain. In date-aware five-fold validation, direct-difference Ridge regression achieved a mean absolute error of 6.209 points, compared with 7.202 for Ridge using control recovery alone. Adding reagent descriptors to feed and control measurements improved direct-Ridge prediction on the fixed reference folds, but the benefit varied across models and validation schemes. Refitted resampling produced error-difference ranges that included zero, and individual attribution rankings varied across groups. The analysis supports retrospective screening of combined reagent and delivery regimes for confirmatory experiments. Confirmation requires traceable samples, independent experimental replicates, quantified assay precision, and independent validation.

Minerals Flotation and Separation TechniquesMineral Processing and GrindingMetal Extraction and Bioleachinggold recoveryrecovery predictioncontrol sensitivity

More by Aksultan Mukhanbet