Prediction of key flotation indicators using a CEEMDAN-PSO-optimized transformer-BiLSTM model

Qing Shi

doi:10.1016/j.mineng.2026.110677

Prediction of key flotation indicators using a CEEMDAN-PSO-optimized transformer-BiLSTM model

2026-07-27

0

PRE

AI

Q

Qing Shi *

S

Shuxun Fan

S

Shixin Wang

G

Guofan Zhang

DOI：10.1016/j.mineng.2026.110677

Original

Original request for help

Save

Abstract

En 中文

• A dual-output framework predicts zinc concentrate and tailing grades in oxidized zinc flotation. • CEEMDAN improves time-series quality through signal decomposition and reconstruction. • A PSO-optimized Transformer-BiLSTM captures global dependencies and local dynamic features. • The proposed model outperforms benchmark models in MAE, RMSE, R² and MAPE.

Keywords:

Oxidized zinc ore flotation

Complete ensemble empirical mode decomposition with adaptive noise

Particle swarm optimization

Transformer-BiLSTM

Flotation index prediction

Prediction of key flotation indicators using a CEEMDAN-PSO-optimized transformer-BiLSTM model

Abstract

Journal

Organization

Cited Papers

Citing Papers