Development of a dynamic model for direct copper electrowinning operations

Date
2023-12
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Innovation is essential for fostering sustainable and environmentally conscious growth in copper production, particularly for operations employing resource-intensive direct copper electrowinning. A dynamic model can be coupled with advanced control strategies in an innovative approach to addressing the control and optimisation challenges associated with copper electrowinning. Previous studies have primarily focused on steady-state models, and limited research has been conducted on dynamic models for copper electrowinning. Consequently, this project aimed to develop a dynamic model for copper electrowinning, with a specific focus on the direct electrowinning process. The main original contribution of this project is the validated semi-empirical dynamic copper electrowinning model. The model can be calibrated for a specific tankhouse, including direct electrowinning operations. An offline parameter-fitting approach was developed for fitting initial model parameters, and for use when limited data are available. The project also introduced an accompanying online parameter-fitting approach that uses moving horizon estimation to continuously adjust the model parameters based on evolving input data. The approach ensures the parameters remain up to date as process conditions change. The least-squares error objective function was selected for use in the online approach, with two types of system models investigated: fundamental and surrogate. The surrogate models were investigated mainly as a future-orientated strategy for online parameter-fitting using computationally intensive datasets. The model incorporated a conceptual resistance network, mass conservation equations, and reactionrate and mass-transfer kinetics. Key performance indicators (copper yield, current efficiency, and specific energy consumption) were used to quantify electrowinning performance. The model included input variables such as current, and the concentrations of copper, iron, nickel, cobalt, and sulfuric acid. The effect of nickel and cobalt were accounted for through existing empirical density and conductivity correlations, and a newly regressed limiting-current density correlation. Validation using dynamic industrial tankhouse data showed the credibility of the model for representing real-life systems. The average normalised residual mean square errors over the five 14-day validation cycles investigated (with the online approach activated) were 10.0%, 29.3%, 79.2%, and 3.9%, for the current efficiency, copper plating rate, specific energy consumption, and potential, respectively. The quantifier values for the specific energy consumption were consistently above the threshold for acceptable model fit. Caution was, therefore, advised in interpreting the model-predicted specific energy consumption values. Overall, the model's performance, particularly with inclusion of the online parameter-fitting approach, however, exhibited satisfactory agreement with the industrial data. The developed model has the potential to make a meaningful contribution to the field. The model's versatility and accuracy make it a valuable tool for use in operator training, process monitoring, and early-fault detection. It also opens avenues for exploration of advanced control strategies. By everaging these potential benefits, operations can enhance productivity, reduce costs, and minimise environmental impact. It is recommended that future work should focus on developing online data validation strategies to further enhance model fidelity, as well as exploring advanced surrogate model structures.
AFRIKAANSE OPSOMMING: Innovasie is noodsaaklik om volhoubare en omgewingsbewuste groei in koperproduksie te kweek, veral vir bedrywe wat hulpbron-intensiewe direkte koperelektroherwinning gebruik. ’n Dinamiese model kan gepaar word met gevorderde beheerstrategieë in ’n innoverende oplossing tot die beheer en optimaliseringsuitdagings geassosieer met koperelektroherwinning. Vorige studies het hoofsaaklik op bestendige toestand modelle gefokus, en beperkte navorsing is uitgevoer op dinamiese modelle vir koperelektroherwinning. Vervolgens het hierdie projek beoog om ’n dinamiese model vir koperelektroherwinning te ontwikkel, met ’n spesifieke fokus op die direkte elektroherwinningproses. Die hoof oorspronklike bydrae van hierdie projek is die gevalideerde semi-empiriese dinamiese koperelektroherwinningsmodel. Die model kan gekalibreer word vir ’n spesifieke tenkhuis, insluitend direkte elektroherwinningsbedrywe. ’n Aflyn parameter-passing benadering is ontwikkel vir die passing van aanvanklike modelparameters, en vir gebruik wanneer beperkte data beskikbaar is. Die projek het ook ’n gepaardgaande aanlyn parameter-passing benadering bekendgestel wat bewegende horisonberaming gebruik om aanhoudend die modelparameters aan te pas gebaseer op dinamiese insetdata. Die benadering verseker dat die parameters op datum bly soos proseskondisies verander. Die minste kwadraatfout doelfunksie is gekies vir gebruik in die aanlynbenadering, met twee tipes sisteemmodelle: fundamenteel en surrogaat. Die surrogaatmodelle is hoofsaaklik ondersoek as ’n toekomsgeoriënteerdestrategie vir aanlyn parameter-passing met rekeningkundige intensiewe datastelle. Die model het ’n konseptuele weerstandnetwerk, massabehoudvergelykings, en reaksietempo en massaoordrag kinetika geïnkorporeer. Sleuteldoeltreffendheidindikators (koperopbrengs, stroomdoeltreffendheid, en spesifieke energiegebruik) is gebruik om die doeltreffendheid van die proses te kwantifiseer. Die model het insetveranderlikes soos stroom, en die konsentrasies van koper, yster, nikkel, kobalt, en swaelsuur, ingesluit. Die effek van nikkel en kobalt is in rekening gebring deur bestaande empiriese digtheid en konduktiwiteitskorrelasies, sowel as ’n oorspronklike beperktestroomdigtheidskorrelasie. Die geldigheid van die model is bevestig deur validasie wat, onder andere, dinamiese industriële tenkhuisdata gebruik het. Die gemiddelde genormaliseerde residu gemiddelde vierkantsfoute oor die valideringsiklusse (met die aanlynbenadering geaktiveer) was 10.0%, 29.3%, 79.2%, en 3.9%, vir die stroomdoeltreffendheid, koperplateringtempo, spesifieke energiegebruik, en potensiaal, onderskeidelik. Die waardes vir die spesifieke energiegebruik was konstant bo die drumpel vir aanvaarbare modelpassing. Oor die algemeen het die model, veral met insluiting van die aanlyn parameter-passing benadering, egter bevredigende ooreenstemming met die industriële data getoon. Die ontwikkelde model het die potensiaal om ’n betekenisvolle bydra tot die veld te maak. Die model se veelsydigheid en akkuraatheid maak dit ’n waardevolle hulpmiddel vir gebruik in operateurafrigting, prosesmonitering en vroeë foutopsporing. Die model kan verder ook gebruik word vir die ondersoek van gevorderde beheerstrategieë. Deur middel van vir hierdie potensiële modeltoepassings kan produktiwiteit verhoog word, kostes verminder word, en omgewingsimpak geminimeer word. Dit word voorgestel dat toekomstige werk moet fokus op die ontwikkeling van aanlyndatavalideringstrategieë om die modelgetrouheid te verbeter, sowel as op die ondersoek van gevorderde surrogaatmodelstrukture.
Description
Thesis (PhD)--Stellenbosch University, 2023.
Keywords
Citation