Sensors and Process Control in Copper Smelters: a Review of Current Systems and Some Opportunities

Review Sensors and Process Control in Copper Smelters: A Review of Current Systems and Some Opportunities

Luis Arias 1 , Eduardo Balladares 2 , Roberto Parra 2,*, Daniel Sbarbaro 1 and Sergio Torres 1

1 Department of Electrical Engineering, University of Concepción, Concepción 4070371, Chile; [email protected] (L.A.); [email protected] (D.S.); [email protected] (S.T.) 2 Department of Metallurgical Engineering, University of Concepción, Concepción 4070371, Chile; [email protected] * Correspondence: [email protected]

Abstract: Despite the widespread and extended use of sensor systems in different industries, there is an important gap to reach equivalent conditions in pyrometallurgical processes for primary production. In the specific case of copper pyrometallurgy, the situation is particularly challenging to incorporate the Industry 4.0 concept for the optimization of their operations. Currently, only two instruments can be identified at the commercial level: the Noranda pyrometer and the Online Production Control (OPC) system. The iron-making and steelmaking industries, however, present an advanced level of control based on monitoring and sensing networks throughout the entire process. This reality has served as a basis for developing a series of solutions based on radiometric sensors for copper pyrometallurgy. We present two types of sensing concept. The first one is applied to smelting and converting reactors based on the measurements of the radiation of the oxidation of different copper and iron sulfides. The second one considers hyperspectral imaging of molten phases flow during operations. The idea of this proposal is to transfer some commercial sensing technologies already in use in the steelmaking industry. In this article, the fundamentals of the sensor design, proofs of concept, and the initial industrial validations are reviewed. Finally, a discussion on the contribution of this knowledge and development opportunities within the framework of Industry 4.0 are addressed.

Citation: Arias, L.; Balladares, E.; Parra, Keywords: copper pyrometallurgy; sensing; spectral measurement R.; Sbarbaro, D.; Torres, S. Sensors and Process Control in Copper Smelters: A Review of Current Systems and Some Opportunities. Minerals 2021, 11, 1. 1. Introduction https://dx.doi.org/min11010001 It is well accepted that modern industry requires a range of effective plant sensors and Received: 21 October 2020 appropriate process control systems leading to a level of automation and third-generation Accepted: 8 December 2020 process control for the highest efficiency [1]. This is the case of steelmaking plants, the oil Published: 22 December 2020 and gas industry, chemical processes and the food industry to mention some examples. In all of them, the 3rd industrial revolution has been adopted in their control systems. Publisher’s Note: MDPI stays neu- In this set of industries, the copper pyrometallurgical processes can be considered as the last tral with regard to jurisdictional claims step of the mining industry or the first one for manufacturing. Today, they face significant in published maps and institutional challenges for establishing sustainable operations due to higher complexity in the ores to affiliations. be treated and increasingly stringent environmental standards. On-line sensors of important operational variables play a key role in the improvements of operational standards. In copper pyrometallurgy, the degree of use of analytical instru-

Copyright: © 2020 by the authors. Li- mentation for the monitoring and control of the process is limited, so that the information censee MDPI, Basel, Switzerland. This available for the decision-making process during the operation is scarce. Currently, the article is an open access article distributed control of copper smelters operations is mainly based on macroscopic mass and energy under the terms and conditions of the balances, and the dynamics of the process depends, to a large extent, on the experience Creative Commons Attribution (CC BY) of operators. license (https://creativecommons.org/ This diagnosis is clearer if the iron and steel industry is taken as a reference, a process licenses/by/4.0/). industry with clear similarities and analogies to copper pyrometallurgy. The steel industry

Minerals 2021, 11, 1. https://dx.doi.org/10.3390/min11010001 https://www.mdpi.com/journal/minerals Minerals 2021, 11, 1 2 of 17

uses different types of sensor and well-developed control systems. This condition is a key element in the operational versatility that these plants have, producing a wide range of products with the same reactors and the same operators, and changing the characteristics of the steel produced from one charge to another. Similarly, the consumption of refractory materials in the steel industry is orders of magnitude lower than those of copper pyrometallurgy, expressed per ton of metal produced. It should also be borne in mind that the steel industry operates at a temperature higher than the temperature of the copper pyrometallurgy, at least 300 ◦C and with more aggressive chemical systems. Even with the tremendous advance in the development of passive and active photonic sensors such as hyperspectral imaging (HI) and different optic sensors, as simple as digital optic pyrometers, there are still only two real-time analytical sensors specially designed for the unique condition of the pyrometallurgical copper processes. The direct study of the encrypted information in the spectral radiation produced by combustion reactions facilitates the analysis of the process, due to the non-invasive and wireless nature of the technologies used. This has enabled, for example, the estimation of reaction temperatures and the observation of the main species of the reaction and its concentrations. This concept has been successfully applied to the combustion of liquids or gases [2] and pyrotechnic mixtures [3–5]. In these applications, the information provided by online spectra has allowed investigation of morphological structures of flames, types of reactions in mixtures pyrotechnics, the movement and burning of individual particles, the processes of igniting powders, propagation of gas explosions, and in situ observations of reactions in closed combustion chambers through optical fibers with optomechanics adapted for these processes. The application of optical spectroscopy to copper pyrometallurgy was discussed during the Annual Meeting of the Minerals, Metals & Materials Society in Denver 1993 where professor Queneau defined the advantage of the information that the operator can have, where up to that time (until today) there had been no option to have the same information “without sticking the head into the vessel”, this information was then developed in [6]. The maturity of the development of this type of sensor for the steelmaking industry is clear today. The first development at the industrial scale was done early in the 1970s [7,8] and gained real maturity in the 1980s [9,10]. This led to very robust process control of the steelmaking processes and reactors. The use of spectral characteristics has also been validated in the spatial recognition of phases of interest in the iron and steel making process, as will be discussed in detail in Section 4.2. Within copper smelters, the reality is that even the 3.0 industrial revolution (information technology (IT) and automation) still has room for continued development. This current state limits the ability to integrate Industry 4.0 developments for process optimization. Indeed, the availability of analytical instrumentation for the monitoring and control of the smelter processes is limited. Thus, the information available for operational decision- making is mainly based on static mass and energy balances. The operational dynamics continue to depend largely on the experience of ground operators. This condition is also re- flected in the literature review (both on the eastern and western world) where the scientific and technical information is related only with the two instruments that have gained industry acceptance for operational monitoring and control in copper pyrometallurgical reactors. The first one is the Semtech Online Production Control (OPC) system. The OPC system was developed in collaboration with the Swedish company Boliden Mineral AB to provide continuous online information on the status of the process in the Pierce Smith converter (Equations (3) and (4)). The company started as a spin-off from the Physics De- partment, Lund Institute of Technology at Lund University in 1987, and its first commercial system was installed in 1994 at Boliden, already working [11]. The instrument is based on spectrometric measurements to identify the presence of PbS and PbO in the off-gas of the converter. The dynamic of these trace elements in the gas, increasing concentration of PbO and decreasing concentration of PbS, signal when the process must be stopped (by stopping the O2(g) injection in Equation (3)) [12,13]. Precise determination of the endpoint of O2(g) Minerals 2021, 11, 1 3 of 17

injection is of great value to control the operation. It increases the operating time while stabilizing the operation of the whole smelter due to the optimization of slag recirculation, which at the same time diminishes its copper content increasing copper recovery. The second is the Noranda Pyrometer [14], configured solely for bath furnaces using blowing tuyeres. It has been on the market since the 1980s without any modifications to the original design. By measuring radiation through an analog array, this instrument monitors the temperature of the molten bath by applying Planck’s law for two fixed wavelengths. In this case, more than the application of a radiometric measurement to determine a temperature, which is quite trivial, the value of the instrument is in the design and operation of an optical system that allows a direct and permanent field of vision of the molten bath inside the reactor. One option for the advancement of process monitoring and control systems in copper smelters is to modernize the use of radiometric measurements since this approach has been previously validated with both the OPC system and the Noranda Pyrometer. The analysis presented in this article is based on the physicochemical characteristics of the reactions that occur in the different reactors of the pyrometallurgical process for copper concentrates to relate them with similar sensing processes present in other industries. The concept for the development of new sensors for copper smelters is based then, in adopting and adapting principles already validated to the very particular conditions of the chemical transformations of the primary production of copper from concentrates.

2. Pyrometallurgical Process Description 2.1. Basic Chemical Scheme for a Cu Smelter and Energy Released The Cu pyrometallurgy process treats Cu concentrates to produce Cu anodes that are then electro-refined by an electrochemical step in aqueous acid media. The Cu concentrates are copper and iron sulfide minerals in the form of particles with sizes under 150 µm. The elemental composition is around 25 to 30 wt% of Cu, 25 to 35 wt% of S, iron from 20 to 40 wt% and gangue (oxides). The smelter operation allows the selective oxidation of Fe and S keeping the Cu in the final product. This oxidation is thermodynamically favored by the high exothermic reactions that allow to melt the system. The overall chemical transformations that occur in a Cu smelter can be represented by the following scheme:

(Cu,Fe,S) + O2(g) + (SiO2-CaO)fluxes = Slag (2FeO*SiO2 + Fe3O4 + CaO + others) + SO2(g) + Cu (1) The fluxes diminish the liquidus temperature in the Fe-O system allowing the separation of the Cu metallic phases from the slag by density. The Cu smelter follows a two-step process to achieve this overall transformation. The first stage is aimed at melting the concentrate by the exothermic reactions controlling the oxygen mass balance to produce an intermediate phase called matte that is within the Cu-Fe-S system and can be reasonable represented by the pseudo binary Cu2S-FeS:

(Cu,Fe,S) + O2(g) + (SiO2-CaO)ﬂuxes = (2FeO*SiO2 + Fe3O4 + CaO + others)slag + SO2(g) + (Cu2S-FeS) (2) This matte phase follows the oxidation in the same way in the second step:

(Cu2S-FeS) + O2(g) + (SiO2-CaO)ﬂuxes = (2FeO*SiO2 + Fe3O4 + CaO + others)slag + SO2(g) + Cu2S (3)

The Cu2S produced, also known as white metal, follows the oxidation, in this case to eliminate the last remaining sulfur:

Cu2S + O2(g) = 2Cu + SO2(g) (4)

The ﬁrst step (Equation (2)) is called smelting and the second converting (Equations (3) and (4)). The different smelting technologies induce similar chemical transformations but differ in how the oxygen contact the concentrate. Until today, the best available technology has been the ﬂash smelting reactor where the concentrate is oxidized through a burner, with clear similarities with coal combustion. A second important family for the smelting stage is the bath smelting, where the concentrate is injected with air in a molten bath through tuyeres. The bath smelting technology family is gaining increasing relevance due to the Minerals 2021, 11, 1 4 of 17

development of new reactors in China, the so-called bottom blowing reactors that simplify the injection of concentrates through tuyeres charging the concentrate from the top of the reactor and blowing the enriched air from the bottom at high speed. For the converting stage, the Peirce Smith converter is the most widely used in the world despite being a reactor that has not undergone substantial improvements since its introduction in 1906. Equation (2) takes place with different mineral sulﬁdes species, the main ones being Chalcopyrite (CuFeS2), Pyrite (FeS2) and Bornite (Cu5FeS4). The oxidations of these species are strongly exothermic as shown in Figure1, where a mass and energy balance were undertaken for two types of classic concentrate (Table1) The adiabatic temperature of the oxidative system is presented as a function of the Cu content in the matte as a target of the control process. The energy and mass balance was developed using HSC 6.0 software and database. Silica was the ﬂux to form the slag assuming a target of Fe/SiO2 = 1.38.

Minerals 2021, 11, x The chemical representation of the slag considers 2FeO*SiO5 of 19 2, CaO*Al2O3*SiO2 and free oxides to close the balance if needed.

2 000

1 900

1 800

1 700

1 600 Adiabatic temperature (K) temperature Adiabatic S/Cu=1; Cu=30.97% S/Cu=1.15; Cu=29.247% 1 500

1 400 60 65 70 75 80 % Cu in matte Figure 1. Adiabatic temperature of the reactive system during smelting (oxidation process) of con- centratesFigure 1 and 2 1. (TableAdiabatic 1). The enriched temperature air for the process of thehas 65% reactive O2(g) (pO2(g) system = 0.65). during smelting (oxidation process) of concentrates 1 and 2 (Table1). The enriched air for the process has 65% O 2(g) (pO2(g) = 0.65). 2.2. Dynamic of the Oxidation of Sulfides The temperatures obtained (Figure 1) are a useful reference to describe the thermal processTable of sulfides 1. Chemical minerals oxidation. and mineralogical The dynamics of characterization the oxidation reactions of of two those concentrates used for calculation speciespresented are extremely in Figurefast, being1. completed in few milliseconds for particles of the size of the ones from a concentrate. In this way, the generation of heat is also very fast which could be a faster process than the heat transfer dynamic. So, if the temperature at the reaction interfaceSpecies/Elements could be measured, it would be verified that Concentrate the reached temperature 1 is Concentrate 2 effectively close to the adiabaticFeS temperature. 21.77 30.00 The mathematical modelling2 of the oxidation dynamics for different minerals particles has been developedCuFeS by different2 researchers [15,16] and is a classical40.00 element for anal- 35.00 ysis of the oxidation/combustionCu2S process for sulfide species. We have 15.00 applied those mod- 15.00 els for chalcopyrite oxidationCuS and other sulfide minerals as an example 3.93 to have a reference 3.93 for the thermal evolution that a particle of concentrate follows during the oxidation pro- Cu5FeS4 4.00 4.00 cess. The model takes into account the O2(g) partial pressure in the gas, the temperature of the environment, the sizeAl 2ofO the3 particle and the thermal properties 3.00of the different mineral 3.00 species. CaO 4.00 4.00 2.3. Fundamentals of the SiOPhysical2 Chemistry Processes for Designing Optical7.30 and Radiometric 5.07 Sensors Cu 30.972 29.241 In the case of high-temperatureFe processes, the generation of 22.752 electromagnetic radia- 25.062 tion in the visible and near-infraredS spectrum from the involved 30.972phases has served to de- 33.624 velop and commercializeS/Cu instruments that obtain information on 1.000 the reactive process at 1.150 distance. In other words, the dynamic changes in the characteristics of spectral emission have been related to the chemical process that is monitored. Examples directly related to the copper smelting process are the combustion of fossil fuels (char, coal, fuel oil, natural gas, methane,2.2. Dynamic etc.) and the of characterization the Oxidation of molten of Sulfides phases in the iron and steel industry. The characteristicsThe temperatures of pyrometallurgical obtained oxidation (Figure process1) for are sulfide a useful minerals reference are to describe the thermal depicted in Figures 1–3. One can conclude that the kinetics of the reactions, as well as the temperatureprocess rise ofof the sulfides process, is minerals on the same oxidation.order of magnitude The as dynamics that of the oxidation of the oxidation reactions of those and combustionspecies areof fossil extremely fuels [17–19]. fast, Therefore, being it is completed evident to analyze in fewwhich millisecondssolutions for particles of the size have ofbeen the developed ones fromto monitor a concentrate. and follow online In real-time this way, reactions the ofgeneration fossil fuel com- of heat is also very fast which bustion, which precisely correspond to the structure of Equation (2). could be a faster process than the heat transfer dynamic. So, if the temperature at the reaction interface could be measured, it would be verified that the reached temperature is effectively close to the adiabatic temperature.

Minerals 2021, 11, 1 5 of 17

The mathematical modelling of the oxidation dynamics for different minerals particles has been developed by different researchers [15,16] and is a classical element for analysis of the oxidation/combustion process for sulﬁde species. We have applied those models for chalcopyrite oxidation and other sulﬁde minerals as an example to have a reference for the thermal evolution that a particle of concentrate follows during the oxidation process. The model takes into account the O2(g) partial pressure in the gas, the temperature of the environment, the size of the particle and the thermal properties of the different mineral species.

2.3. Fundamentals of the Physical Chemistry Processes for Designing Optical and Radiometric Sensors In the case of high-temperature processes, the generation of electromagnetic radiation in the visible and near-infrared spectrum from the involved phases has served to develop and commercialize instruments that obtain information on the reactive process at distance. In other words, the dynamic changes in the characteristics of spectral emission have been related to the chemical process that is monitored. Examples directly related to the copper smelting process are the combustion of fossil fuels (char, coal, fuel oil, natural gas, methane, etc.) and the characterization of molten phases in the iron and steel industry. The characteristics of pyrometallurgical oxidation process for sulﬁde minerals are depicted in Figures1–3. One can conclude that the kinetics of the reactions, as well as the temperature rise of the process, is on the same order of magnitude as that of the Minerals 2021, 11, x 6 of 19 oxidation and combustion of fossil fuels [17–19]. Therefore, it is evident to analyze which solutions have been developed to monitor and follow online real-time reactions of fossil fuel combustion, which precisely correspond to the structure of Equation (2).

(a)

(b)

FigureFigure 2. 2.TemperatureTemperature evolution evolution for different for different sizes of CuFeS sizes2 of(chalcopyrite) CuFeS2 (chalcopyrite) particles with particlesdiffer- with dif- ◦ entferent pO2(g) pO in 2(g)the processin the processgas. The gas.temperature The temperature of the environment of the was environment set at 973 K was (700 set°C). at(a) 973 K (700 C). pO2(g) = 0.8; (b) size of particle = 27 μm. (a) pO2(g) = 0.8; (b) size of particle = 27 µm.

(a)

Minerals 2021, 11, x 6 of 19

(a)

(b)

Figure 2. Temperature evolution for different sizes of CuFeS2 (chalcopyrite) particles with different pO2(g) in the process gas. The temperature of the environment was set at 973 K (700 °C). (a) Minerals 2021, 11, 1 6 of 17 pO2(g) = 0.8; (b) size of particle = 27 μm.

Minerals 2021, 11, x 7 of 19

(a)

(b)

Figure 3. Temperature evolution for CuFeS2 (chalcopyrite) and different minerals species for particle Figure 3. Temperature evolution for CuFeS2 (chalcopyrite) and different minerals species for parti- size of 27 µm and pO = 0.8. (a) Chalcopyrite at different Temperature of the environment (b)T cle size of 27 μm and pO2(g)2(g) = 0.8. (a) Chalcopyrite at different Temperature of the environment (b) Tenvironment environment 973 973 K K and and size size of of particle particle = = 27 27µ μm.m. 3. Sensors and Control in Cu Pyrometallurgy 3. Sensors and Control in Cu Pyrometallurgy The accurate understanding of the physical chemistry of reactive and non-reactive The accurate understanding of the physical chemistry of reactive and non-reactive systems is of great value for designing sensors and has been the basis for development systems is of great value for designing sensors and has been the basis for development in in other industries. Despite of a good level of understanding on the physical-chemical other industries. Despite of a good level of understanding on the physical-chemical fun- fundamentals of oxidation and combustion of sulfides, until today the spectral analysis damentals of oxidation and combustion of sulfides, until today the spectral analysis in- involved in copper pyrometallurgy has been studied only within the research activities volvedof the in group copper preparing pyrometallurgy this paper. has Asbeen already studied noted, only within the lack the of research instrumentation activities forof themonitoring group preparing the dynamics this paper. of the processAs already leads no toted, almost the lack non-existent of instrumentation or poorly performing for moni- toringcontrol the strategies dynamics of of the the smelting process leads and converting to almost non-existent processes, where or poorly the performing human factor con- of troloperators strategies play of an the important smelting roleand andconverting imposes processes, big instabilities. where Itthe is human important factor to highlightof oper- atorsthat play the variability an important on role the mineralogicaland imposes big characteristics instabilities. It of is the important copper to concentrate highlight that is a thenatural variability perturbation on the mineralogical to the system wherecharacteristics the characterization of the copper of concentrate this stream is is a the natural basis perturbationfor the definition to the ofsystem the macroscopic where the characterization mass and energy of balancethis stream for theis the smelting basis for rector. the definitionIf an online of the sensor macroscopic can provide mass any and information energy balance on thefor the chemical smelting evolution rector. If during an online the sensoroxidation/combustion can provide any information process, it would on the provide chemical information evolution during for fine-tuning the oxidation/com- online and bustionreal-time process, the smelting it would furnaces. provide Thus, informat the productivityion for fine-tuning indexes willonline be improvedand real-time by mean the smeltingof higher furnaces. stability ofThus, the operation,the productivity decreasing indexes copper will losses, be improved and diminishing by mean thermal of higher and stabilitychemical of deviationthe operation, for their decreasing set points. copper A serieslosses,of and proposed diminishing concepts thermal and and results chemical from deviationinitial industrial for their trials set points. are discussed A series hereafter.of proposed concepts and results from initial industrial trials are discussed hereafter.

4. Radiometric Proposals for New Instruments 4.1. Radiometric Measurements of Chemical Oxidation The development of this family of “new instruments” has started by reproducing the Equations (2)–(4) at laboratory level, adding online measurement of the spectral signal through an optic fiber connected to a radiometer with further treatment using mathematic tools. As was previously pointed out, there are two families of smelting reactors (Equation (2)): bath smelting and flash smelting. For the sake of the length of this article, we will refer specifically to developments, proof of concept, and preliminary tests at the industrial level for flash smelting reactions and reactors. 4.1.1. Experimental Set-Up A vertical laminar flow laboratory reactor was used to simulate the oxidation conditions of the copper concentrates. The combustion of concentrates was carried out under controlled conditions of feed rate and combustion gas mixture to simulate the physical and chemical transformations that the concentrate particles have when they burn in the

Minerals 2021, 11, 1 7 of 17

4.1.1. Experimental Set-Up

burnerA of vertical a flash smelting laminar furnace. flow In laboratory this way, the reactor high rate was of oxidation used to for simulate sulfide min- the oxidation condi- tionserals was of thereproduced copper at concentrates. laboratory scale. The The combustiontemperature and of oxygen concentrates potential wascondi- carried out under controlledtions in the gas conditions inside the furnace of feed were rate simila andr combustion to those that typically gas mixture occur in to the simulate concen- the physical and chemicaltrates burner transformations at the industrial level, that so the the concentrate burning material particles was assumed have to when have theycharac- burn in the burner teristics equivalent to the material that reaches the settler of the industrial furnace. The ofsystem a flash to feed smelting the concentrate furnace. was In equipped this way, with the a high-temperature high rate of oxidation optic fiber that for sulfidewas minerals was reproducedconnected to a at VIS-NIR laboratory spectrophotometer scale. The temperature USB4000 (Ocean and Optis oxygen Inc., potential Dunedin, conditionsFL, in the gas insideUSA) was the used furnace to acquire were the similarspectral data to those in the range that typicallyfrom 400–900 occur nm with in thean average concentrates burner at thespectral industrial resolution level, of ~0.22 so thenm; burningalso, the sp materialectrometer was and assumedoptical circuit to havewere calibrated characteristics equivalent to measure the emitted radiation in absolute irradiance units (μW × cm−2 × nm−1). With tosuch the a set-up, material the reaction that reaches zone of the the reactor settler is within of the the industrial field of view furnace. of the fiber The allow- system to feed the concentrateing to follow the was spectral equipped emission with of the aoxidation/combustion high-temperature of opticthe copper fiber concentrate that was connected to a VIS-NIR(Figure 4). spectrophotometerA detailed description of USB4000 the set-up (Oceanand its operation Optis Inc., is presented Dunedin, elsewhere FL, USA) was used to acquire[1–23]. the spectral data in the range from 400–900 nm with an average spectral resolution The spectrum collected during the oxidation process was analyzed to obtain valuable of ~0.22 nm; also, the spectrometer and optical circuit were calibrated to measure the information to correlate the spectral information with the chemical process. The measured −2 −1 emittedspectra have radiation a continuous in absolute part that follows irradiance Planck’s units radiation (µW behavior× cm and× discontinuousnm ). With such a set-up, thepattern reaction that are zone related of to the the reactoremission isfrom within atoms theand fieldmolecules of view formed of during the fiber the ox- allowing to follow theidation spectral process. emission The temperature of the of oxidation/combustion the flame produced was also of calculated the copper using concentratea clas- (Figure4). Asical detailed Planck’s description Law. of the set-up and its operation is presented elsewhere [1–23].

Figure 4. Experimental set-up [23]. Figure 4. Experimental set-up [23]. The results presented hereafter are the summary of different experimental campaigns and haveThe been spectrum reported collected individually during in other the articles oxidation or documents process for the was public analyzed domain. to obtain valuable informationIt is the first time, to correlatehowever, th theat they spectral have been information integrated to with give thea global chemical vision of process. what The measured spectrahas been havedeveloped a continuous at the laboratory part level. that This follows comprehensive Planck’s view radiation seeks to behavior present the and discontinuous patternconcept that that laboratory-level are related studies to the which emission reproduce from the atomssame chemical and molecules reactions of informed during the dustrial processes on a small scale are representative for studying the optical signals that processes have.

Minerals 2021, 11, 1 8 of 17

oxidation process. The temperature of the ﬂame produced was also calculated using a classical Planck’s Law. The results presented hereafter are the summary of different experimental campaigns and have been reported individually in other articles or documents for the public domain. It is the ﬁrst time, however, that they have been integrated to give a global vision of what has been developed at the laboratory level. This comprehensive view seeks to present the concept that laboratory-level studies which reproduce the same chemical reactions of industrial processes on a small scale are representative for studying the optical signals that processes have.

4.1.2. Experimental Results: Temperature Measurements Considering the continuous part of the spectrum, it is very clear that a calculation for the flame temperature is possible by applying Planck radiation law. The mass balance requires a model of reactions based on the mineralogical species that are oxidized in the flame of the laboratory burner. This was achieved by performing a characterization using automated mineralogy techniques (QEMSCAN®) for the products generated. From this information, a set of k chemical reactions were proposed and a calculation of their degree of conversion allows us to close the mass balance to fit with the characterization of the l species produced. Knowing the degree of conversion for each reaction we can apply Equation (5) that corresponds to an energy balance and so on, estimating the temperature of the flame. k l+ 3 Z TCalc n H + Q = m × Cp (T)dT + σ × ε × A × T4 − T4 ∑ i∆ R,i,@TRef Furnace ∑ j j Calc wall , (5) i=1 j=1 TRef

where ni is the moles of a specie that reacts according to reaction i from the m reactions deﬁned; mj is the mole of one mineralogical specie from the l identiﬁed; TRef is the reference

temperature for the calculation of the enthalpy of the chemical reactions:∆HR,i,@TRef is the enthalpy of reaction i at the the temperature of reference; QFurnace is the heat released by the furnace to the environment; ε is the emissivity of the flame (understood as a reactive system which considers gases and condensed particles, both Cu concentrate and liquid droplets of sulfide and oxides, and as well solid oxides produced or coming from the gangue); A is the area of the flame; σ is Stefan–Boltzmann constant; Twall is the temperature of the furnace wall. Please note that j varies from 1 to l + 3, where 3 represents the 3 gases species: O2(g), N2(g) and SO2(g). In this case, the heat losses by the convective mechanism were neglected. Equation (5) requires an estimate of the flame area and an estimate of the emissivity. The emissivity is the emissivity of the whole reactive system, which is a mix of gases and condensed/liquid phases with a space density much lower than the case of an industrial burner. The emissivity was taken from Marin et al. [22] with a value of 0.28 × 10−3. This work presented a detailed model for the calculation of the emissivity with the same experimental set-up used in this work. The area of the flame was estimated through a set of photographs taken in a special configuration of the set-up where the steel tube of the furnace was replaced by a quartz one, and thanks to the characteristic of the furnace that can be open longitudinally (Figure5). The flame was approximated to a prolate spheroid, as described in Cartesian coordinates by Equation (6). Its area, A is given by Equation (6) [24].

x2+y2 z2 c 2 + 2 = 1 ⇒ A = 2πa a + arcsen(α) a cq α (6) = − c2 where α 1 a2 and A is the superﬁcial area. Minerals 2021, 11, x 10 of 19 Minerals 2021, 11, 1 9 of 17

z a a x y

Figure 5. Estimation of the ﬂame volume; a = 3.4 cm and c = 5.86 cm and A = 174 cm2 [24].

The adiabatic temperatures calculated from the mass and energy balance were com- paredFigure with 5. Estimation the temperature of the flame measured volume; using a = 3.4 Planck’s cm and lawc = 5.86 with cm two and wavelengths A = 174 cm2 [24]. (Figure 6). This calculation takes the online spectrum from the radiometric measurement and through a mathematical treatment of the signal eliminates the discontinuities of the spec- xy22+ zc2  trum. The continuous part+= was then used=π to apply + the two()α wavelengths method. A second 221 A2 a a arcsen algorithm, usingac an optimization method, choosesα the best pair of wavelengths for the calculation. The reported and calculated temperatures are on the same order of magni- (6) c2 tudes as those obtainedα= by modeling − the flash combustion for different mineral particles where 12 and A is the superficial area. (Figures2 and3 ). The results showa a very good agreement, which was expected due to the accurate characterization of the products. Nevertheless, it seems that the error is depending The adiabatic temperatures calculated from the mass and energy balance were com- on the temperature itself. One source of error could be the evaluation of the size of the pared with the temperature measured using Planck’s law with two wavelengths (Figure flame, which is kept the same for different temperatures. At higher temperatures, the flame 6). This calculation takes the online spectrum from the radiometric measurement and Minerals 2021, 11, x must expand, and its transfer area increases, as well as the temperature, so radiation losses11 of 19 through a mathematical treatment of the signal eliminates the discontinuities of the spec- are greater. Those two points go into the explanation for the error evolution. trum. The continuous part was then used to apply the two wavelengths method. A second algorithm, using an optimization method, chooses the best pair of wavelengths for the 2 100 calculation. The reported and calculated temperatures are on the same order of magni- tudes as those obtained by modeling the flash combustion for different mineral particles 2 000 (Figures 2 and 3). The results show a very good agreement, which was expected due to 1 900 the accurate characterization of the products. Nevertheless, it seems that the error is depending on the temperature itself. One source of error could be the evaluation of the size 1 800 of the flame, which is kept the same for different temperatures. At higher temperatures,

T measured (K) 1 700 the flame must expand, and its transfer area increases, as well as the temperature, so radiation losses are greater. Those two points go into the explanation for the error evolution. 1 600

1 500 1 500 1 600 1 700 1 800 1 900 2 000 2 100 T calculated (K)

FigureFigure 6.6.Comparison Comparison betweenbetween measured measured (radiometric (radiometric experiments) experiments) and and calculated calculated temperature temperature for laboratoryfor laboratory combustion combustion of concentrates of concentrates by means by means of a massof a mass and energyand energy balance balance (Equation (Equation (5)) [(5))25]. [25].

4.1.3. Experimental Results: Molecular Identification of Oxides in the Flame The discontinuous part of the spectrum is related to some species, atoms or molecules, that emit at a very specific wavelength. The discontinuities observed in the laboratory and industrial measurements were analyzed to establish their relationship with species typically obtained from the concentrated oxidation process. One goal is to develop the capacity to correlate these discontinuities with the presence of species that are the ones to be formed (Equation (2)) and others that would indicate an anomalous operation. The increase of copper oxides (CuO and Cu2O) and magnetite (Fe3O4) is an unquestionable indicator of the bad quality of the combustion-oxidation that occurs in the flame. Indeed, Cu oxide should not be formed according to the scheme of Equation (2), and Fe3O4 must stay under a certain level. In industrial practice, the results of this anomalous condition can be identified only after several hours according to the chemical analysis of the products, as it was shown the oxidation process is on the time scale of milliseconds (Figures 2 and 3) so having the information or not after several hours is of questionable utility. If there is an increase in oxidized Cu, it represents a higher Cu content in the slag, and consequently losses in the value for the process. Magnetite affects the viscosity of the slag and the proper separation of matte and slag in the settler. Early identification of the formation of these molecules in the flame would provide information to relate poor oxygen distribution during the oxidation process and establish actions in the process control Toro et al. [26] have presented evidence of the FeO pick at laboratory scale by measurements with a built-in developed spectrometer with a high sensitivity and a high spectral resolution (Figure 7). Then, a second experiment was conducted measuring the spectra with a low spectral resolution and a low-sensitivity commercial spectrometer, but enhanced and analyzed with post-signal processing and multivariate data analysis such as principal component analysis (PCA) and a multivariate curve resolution–alternating least squares (MCR-ALS) method. In the same article, one can find the results from an experimental plan to propose for the first time in the scientific literature the characteristic wavelength for Fe3O4 molecular emissions. The experiment was undertaken by the combustion of pure pyrite (FeS2). This combustion test is very representative due to the possibility to produce Fe3O4 in the laboratory furnace according to the control of oxygen potential in the combustion chamber.

Minerals 2021, 11, 1 10 of 17

4.1.3. Experimental Results: Molecular Identification of Oxides in the Flame The discontinuous part of the spectrum is related to some species, atoms or molecules, that emit at a very specific wavelength. The discontinuities observed in the laboratory and industrial measurements were analyzed to establish their relationship with species typically obtained from the concentrated oxidation process. One goal is to develop the capacity to correlate these discontinuities with the presence of species that are the ones to be formed (Equation (2)) and others that would indicate an anomalous operation. The increase of copper oxides (CuO and Cu2O) and magnetite (Fe3O4) is an unquestionable indicator of the bad quality of the combustion-oxidation that occurs in the flame. Indeed, Cu oxide should not be formed according to the scheme of Equation (2), and Fe3O4 must stay under a certain level. In industrial practice, the results of this anomalous condition can be identified only after several hours according to the chemical analysis of the products, as it was shown the oxidation process is on the time scale of milliseconds (Figures2 and3) so having the information or not after several hours is of questionable utility. If there is an increase in oxidized Cu, it represents a higher Cu content in the slag, and consequently losses in the value for the process. Magnetite affects the viscosity of the slag and the proper separation of matte and slag in the settler. Early identification of the formation of these molecules in the flame would provide information to relate poor oxygen distribution during the oxidation process and establish actions in the process control. Toro et al. [26] have presented evidence of the FeO pick at laboratory scale by measurements with a built-in developed spectrometer with a high sensitivity and a high spectral resolution (Figure7). Then, a second experiment was conducted measuring the spectra with a low spectral resolution and a low-sensitivity commercial spectrometer, but enhanced and analyzed with post-signal processing and multivariate data analysis such as principal component analysis (PCA) and a multivariate curve resolution–alternating least squares (MCR-ALS) method. In the same article, one can find the results from an experimental plan to propose for the first time in the scientific literature the characteristic wavelength for Fe3O4 molecular emissions. The experiment was undertaken by the combustion of Minerals 2021, 11, x pure pyrite (FeS2). This combustion test is very representative due to the possibility12 of 19 to

produce Fe3O4 in the laboratory furnace according to the control of oxygen potential in the combustion chamber.

FigureFigure 7. Spectral 7. Spectral FeO FeO emission emission measur measurementsements and and reference reference spectral spectral features features taken taken from from [26]. [26 ].

In theIn same the same scope scope and andanalyzing analyzing data data from from the the same same set-up set-up for for combustion combustion of of con- concen- trates, Arias at all [21] reported exploratory results depicted some spectral features related centrates, Arias at all [21] reported exploratory results depicted some spectral features to the presence of CuxO emissions at 606 nm and 616 nm. related to the presence of CuxO emissions at 606 nm and 616 nm. 4.1.4.4.1.4. Experimental Experimental Results: Results: Identification Identiﬁcation Characteristics Characteristics of the of Charge the Charge The lastThe result last result is related is related to the to possibility the possibility to identify to identify the characteristics the characteristics of the of charge the charge basedbased on the on analysis the analysis of the of thespectrum spectrum collect collected.ed. In Inthis this topic, topic, Marin Marin [27] [27 ]developed developed the ﬁrst first algorithms applied to identify threethree typestypes ofof concentratesconcentrates characterized characterized by by their their S/Cu S/Cu ratio ratio (Figure 8). Then Díaz et al. [23] used the same approach to extend the identification target to mineralogical species (Figure 9). The algorithms are based on principal component analysis (PCA) applied on the spectral data showing high correlation features among species with similar elemental compositions or the same type of concentrate. Classifica- tion algorithms were tested on the spectral data, and a classification accuracy of over 95% was shown. Detailed mathematical treatment of the spectrum and the mathematical development of identification procedures can be followed in [23,27]

Figure 8. PC1–PC2 scores from principal component analysis (PCA) applied to the identification of different types of concentrates [27]. Blend 1: Cu = 28.0%, Fe = 26.2% and, S = 27.1 with S/Cu = 0.96. Blend 2: Cu = 28.3%, Fe = 25.5% and, S = 26.6 with S/Cu = 0.94. Blend 3: Cu = 26.7%, Fe = 25.3% and, S = 26.6 with S/Cu = 1.00. The composition is given in weight percent.

Minerals 2021, 11, x 12 of 19

Figure 7. Spectral FeO emission measurements and reference spectral features taken from [26].

In the same scope and analyzing data from the same set-up for combustion of concentrates, Arias at all [21] reported exploratory results depicted some spectral features related to the presence of CuxO emissions at 606 nm and 616 nm. 4.1.4. Experimental Results: Identification Characteristics of the Charge Minerals 2021, 11, 1 The last result is related to the possibility to identify the characteristics of the charge11 of 17 based on the analysis of the spectrum collected. In this topic, Marin [27] developed the first algorithms applied to identify three types of concentrates characterized by their S/Cu ratio(Figure (Figure8). Then 8). DTheníaz etDíaz al. [et23 al.] used [23] the used same the approach same approach to extend to theextend identification the identification target to targetmineralogical to mineralogical species (Figure species9). (Figure The algorithms 9). The algorithms are based on are principal based on component principal analysiscompo- nent(PCA) analysis applied (PCA) on the applied spectral on datathe spectral showing data high showing correlation high features correlation among features species among with speciessimilar with elemental similar compositions elemental compositions or the same type or the of concentrate.same type of Classification concentrate. algorithmsClassifica- tionwere algorithms tested on were the spectral tested on data, the spectral and a classification data, and a classification accuracy of overaccuracy 95% of was over shown. 95% wasDetailed shown. mathematical Detailed mathematical treatment of treatment the spectrum of the and spectrum the mathematical and the mathematical development de- of velopmentidentification of identification procedures can procedures be followed can in be [23 followed,27]. in [23,27]

FigureFigure 8. 8. PC1–PC2PC1–PC2 scores scores from from principal principal component component analysis analysis (PCA) (PCA) applie appliedd to to the the identification identiﬁcation of of differentdifferent types types of of concentrates concentrates [27]. [27]. Blend Blend 1: 1: Cu Cu = = 28.0%, 28.0%, Fe Fe = =26.2% 26.2% and, and, S S= =27.1 27.1 with with S/Cu S/Cu = 0.96. = 0.96. Minerals 2021, 11, x Blend 2: Cu = 28.3%, Fe = 25.5% and, S = 26.6 with S/Cu = 0.94. Blend 3: Cu = 26.7%, Fe = 25.3%13 of and, 19 Blend 2: Cu = 28.3%, Fe = 25.5% and, S = 26.6 with S/Cu = 0.94. Blend 3: Cu = 26.7%, Fe = 25.3% and, S = 26.6 with S/Cu = 1.00. The composition is given in weight percent. S = 26.6 with S/Cu = 1.00. The composition is given in weight percent.

FigureFigure 9. 9.PC1–PC2PC1–PC2 scores scores from from PCA PCA applied applied to toexperi experimentsments with with mixtures mixtures of ofpure pure mineral mineral spe- species, cies,taken taken from from [23 [23].].

TheseThese results results provide provide a very a very interesting interesting opportunity opportunity to tointegrate integrate spectral spectral measure- measure- mentsments of of a asmelting smelting furnace furnace flame ﬂame into into operational operational control. As previously stated,stated, thethe control con- trolof of a smeltinga smelting furnace furnace (ﬂash (flash smelting smelting or or bath bath smelting) smelting) is basedis based on on a macroscopica macroscopic mass mass and energy balance. These balances are carried out by charges that represent several hours and energy balance. These balances are carried out by charges that represent several hours of furnace operation. These charges are statistically characterized and, therefore, there of furnace operation. These charges are statistically characterized and, therefore, there is is always a deviation from the values that have been established to control the process. always a deviation from the values that have been established to control the process. The main variables of this control are the amount of oxygen indicated in Equation (2) and its percentage in the process gas, which is the first control for a target temperature of the operation. Having an online indicator whether of the type of charge (Figure 8) or, ideally, based on information on their species (Figure 9) would allow establishing a mass and energy balance online and adjusting the control variables based on this information. The situation today is that the operator only knows the results of the operation several hours later, having only the measurements of the temperature of the gas line as the most relevant information for the control. 4.1.5. Preliminary Industrial Tests To date, some measurements in an industrial environment made it possible to validate some of the results proposed from the laboratory studies summarized in the previous points. The setup considered the installation of an optical probe in the center of a burner in an industrial flash smelting furnace, which is a location analogous to that of the laboratory (Figure 10). The probe was installed in the space for the fuel lance that flash burners usually have. The fuel lance is concentric to the concentrate burner and aims to provide a radiant heat source from the center of the volume where the flame appears. The probe was placed instead of the fuel lance. Due to the process of air and oxygen injection in this area, the temperature is low enough to fit a standard optical probe. The objective of using this radiometric probe in the control structure of a smelting furnace is to provide on-line and real-time information on the combustion conditions of the charge fed to the burner. As a first result, it can be precise that the probe and its digital system give an on-line temperature measurement, which is within the orders of magnitude of the values obtained in the laboratory. This simple fact is relevant because it allows the validity of the experimental setup (Figure 4) to be established in order to study the flash oxidation reactions of copper concentrates.

Minerals 2021, 11, 1 12 of 17

The main variables of this control are the amount of oxygen indicated in Equation (2) and its percentage in the process gas, which is the ﬁrst control for a target temperature of the operation. Having an online indicator whether of the type of charge (Figure8) or, ideally, based on information on their species (Figure9) would allow establishing a mass and energy balance online and adjusting the control variables based on this information. The situation today is that the operator only knows the results of the operation several hours later, having only the measurements of the temperature of the gas line as the most relevant information for the control.

4.1.5. Preliminary Industrial Tests To date, some measurements in an industrial environment made it possible to validate some of the results proposed from the laboratory studies summarized in the previous points. The setup considered the installation of an optical probe in the center of a burner in an industrial flash smelting furnace, which is a location analogous to that of the laboratory (Figure 10). The probe was installed in the space for the fuel lance that flash burners usually have. The fuel lance is concentric to the concentrate burner and aims to provide a radiant Minerals 2021, 11, x heat source from the center of the volume where the flame appears. The probe was placed14 of 19 instead of the fuel lance. Due to the process of air and oxygen injection in this area, the temperature is low enough to fit a standard optical probe.

FigureFigure 10.10. PositionPosition ofof thethe probeprobe inin aa industrialindustrial ﬂashflash smeltingsmelting burner.burner.

TheA relevant objective objective of using for this the radiometric operation of probe a smelting in the controlfurnace structure is to maintain of a smelting a stable furnaceprocess isfor to the provide sulfide on-line oxidation and (Equation real-time (2)). information In the case on of the a flash combustion smelting conditions furnace, one of thecan chargesay that fed the to objective the burner. is Asto achieve a first result, stable it combustion can be precise (Figure that the 11 probeshows and a drawing its digital of systemsome conditions give an on-line out of temperature standard). It measurement, is widely known which that is any within instability the orders in the of magnitude flame pro- ofduces the valuesa series obtained of disturbances in the laboratory. into the operat Thision, simple perhaps fact is the relevant most relevant because itbeing allows the the in- validitycrease in of dust the generation experimental that setup can generate (Figure4 )accretions to be established inside the in furn orderace to or study in the the gas flash line oxidationthat are difficult reactions to handling of copper impacting concentrates. the reactor operating time. A relevant objective for the operation of a smelting furnace is to maintain a stable process for the sulfide oxidation (Equation (2)). In the case of a flash smelting furnace, one can say that the objective is to achieve stable combustion (Figure 11 shows a drawing of some conditions out of standard). It is widely known that any instability in the flame produces a series of disturbances into the operation, perhaps the most relevant being the

(a) (b) (c)

Figure 11. Some conditions out of standard for the flame. (a): well developed at a distance h; (b): h too far from the burner; (c): h too far and non-symmetric flame.

The first objective was, therefore, to propose a parameter that would inform the operator about the stability of the flame, following the intensity measured by the probe over time. If this measurement is below a certain threshold of reference for a certain period in a previous interval (for example, the last 60 or 90 min), the combustion could be out of standard, and the operator must take action to monitor the function of the burner as a priority. Figure 12 gives an example of this indicator, being very relevant to identify mo- ments in which the measurement dropped abruptly, even without energy measurements, which may mean that the flame did not form or it is outside the field of view of the probe. There are conditions of this type where, however, the feed of the concentrate and O2(g) was maintained.

Minerals 2021, 11, x 14 of 19

Figure 10. Position of the probe in a industrial flash smelting burner.

A relevant objective for the operation of a smelting furnace is to maintain a stable Minerals 2021, 11, 1 process for the sulfide oxidation (Equation (2)). In the case of a flash smelting furnace,13 of one 17 can say that the objective is to achieve stable combustion (Figure 11 shows a drawing of some conditions out of standard). It is widely known that any instability in the flame produces a series of disturbances into the operation, perhaps the most relevant being the in- increasecrease in in dust dust generation generation that that can can generate generate accretions accretions inside the furnacefurnace oror inin thethe gasgas lineline thatthat areare difﬁcultdifficult toto handlinghandling impactingimpacting the the reactor reactor operating operating time. time.

(a) (b) (c)

Figure 11. Some conditions out of standard for the flame. (a): well developed at a distance h; (b): h Figure 11. Some conditions out of standard for the flame. (a): well developed at a distance h; (b): h tootoo farfar fromfrom thethe burner;burner; ((cc):): hh tootoo farfar andand non-symmetricnon-symmetric flame.flame. The first objective was, therefore, to propose a parameter that would inform the The first objective was, therefore, to propose a parameter that would inform the op- operator about the stability of the flame, following the intensity measured by the probe erator about the stability of the flame, following the intensity measured by the probe over over time. If this measurement is below a certain threshold of reference for a certain period time. If this measurement is below a certain threshold of reference for a certain period in in a previous interval (for example, the last 60 or 90 min), the combustion could be out a previous interval (for example, the last 60 or 90 min), the combustion could be out of of standard, and the operator must take action to monitor the function of the burner as astandard, priority. and Figure the 12 operator gives an must example take action of this to indicator, monitor beingthe function very relevant of the burner to identify as a momentspriority. Figure in which 12 the gives measurement an example dropped of this indicator, abruptly, evenbeing without very relevant energy measurements,to identify mo- Minerals 2021, 11, x ments in which the measurement dropped abruptly, even without energy measurements,15 of 19 which may mean that the flame did not form or it is outside the field of view of the probe. which may mean that the flame did not form or it is outside the field of view of the probe. There are conditions of this type where, however, the feed of the concentrate and O2(g) wasThere maintained. are conditions of this type where, however, the feed of the concentrate and O2(g) was maintained.

(a) (b)

FigureFigure 12. Value 12. Value over overtime timeof the of % the of % “good of “good combusti combustion”on” during during the last the last60 min. 60 min. In case In case (a), the probe(a ),shows the probe the variability shows the of variability this parameter of this parameterand as well and a no as negligible well a no negligible proportion proportion of time with a valueof under time withthe reference. a value under For the the (b reference.) there is more For the stable (b) thereoperation is more and stable with almost operation the andentire day with the parameter close to 100%. with almost the entire day with the parameter close to 100%.

AA secondsecond stagestage forfor the the analysis analysis of of the the collected collected information information considers considers the the design design of of onlineonline algorithms.algorithms. OfflineOffline PCAPCA waswas alreadyalready appliedapplied inin thethe previouslypreviously discusseddiscussed results,results, wherewhere it it was was possible possible to to establish establish that that the the signals signals of of FeO, FeO, CuO CuOx,x and, and Fe Fe3O3O4 4appear appear in in the the measuredmeasured spectra. spectra. AsAs was was done done for for laboratory laboratory tests tests using using PCA, PCA, changes changes in in the the charge charge fed fed to to the the furnace furnace werewere identified.identified. Figure 13 shows the the evolution evolution in in time time of of the the passage passage from from one one charge charge to toanother. another. This This identification identification is isrelevant relevant since since the the arrival arrival of of a a new type ofof chargecharge toto thethe burnerburner is is established established by by a a methodology methodology considering considering the the mass mass balance balance of of the the hoppers, hoppers, and and thethe change change of of the the control control parameters parameters of of the the furnace furnace is is programmed programmed accordingly, accordingly, without without having any field measurements of this process. The effective moment when the charge changes is not certain. It was found that, in effect, this change differs at the start from the program, having a difference of up to 25 min. The change itself takes at least 10 min, so the furnace was in inappropriate control conditions for at least 35 min, and if the charge changes several times per day, there is a risk of the furnace having inappropriate control parameters several hours a day.

Minerals 2021, 11, x 15 of 19

(a) (b) Figure 12. Value over time of the % of “good combustion” during the last 60 min. In case (a), the probe shows the variability of this parameter and as well a no negligible proportion of time with a value under the reference. For the (b) there is more stable operation and with almost the entire day with the parameter close to 100%.

A second stage for the analysis of the collected information considers the design of online algorithms. Offline PCA was already applied in the previously discussed results, where it was possible to establish that the signals of FeO, CuOx, and Fe3O4 appear in the measured spectra. As was done for laboratory tests using PCA, changes in the charge fed to the furnace Minerals 2021, 11, 1 were identified. Figure 13 shows the evolution in time of the passage from one charge14 of 17to another. This identification is relevant since the arrival of a new type of charge to the burner is established by a methodology considering the mass balance of the hoppers, and the change of the control parameters of the furnace is programmed accordingly, without havinghaving anyany ﬁeldfield measurementsmeasurements ofof thisthis process.process. TheThe effectiveeffective momentmoment whenwhen thethe chargecharge changeschanges isis notnot certain.certain. ItIt waswas foundfound that,that, inin effect,effect, thisthis changechange differsdiffers atat thethe startstart fromfrom the the program,program, havinghaving aa differencedifference ofof upup toto 2525 min.min. TheThe changechange itselfitself takestakes atat leastleast 1010 min,min, soso thethe furnacefurnace waswas inin inappropriateinappropriate controlcontrol conditionsconditions forfor atat leastleast 3535 min,min, andand ifif thethe charge charge changeschanges severalseveral timestimes perper day,day, therethere isis aa riskrisk of of the the furnace furnace having havinginappropriate inappropriatecontrol control parametersparameters severalseveral hourshours aa day.day.

Figure 13. Two principal components (PC) showing the change of the charge from Blend 1 to Blend 2. With circles are the evolution of those PC. The change was schedule at 08:00 and started at 08:10 and took more than 9 min.

4.2. Non-Reactive Systems The previous sections have shown that radiometric measurements can follow dynamic aspects of the oxidation of sulfides and, in particular, mineral sulfides. The analysis of the characteristics of these measurements, which are non-invasive, at a distance, and with increasingly affordable components, can be extended for the application of spatial recognition of the molten phases. Based on technological surveillance and prospecting of the scientific and technical literature, there is no commercial instrument available for copper smelters, that quantifies online and without contact the percentage of copper (% Cu) in the molten metallic phases during tapping operations, nor the possibility of discriminating the phases during tapping, which is of a superior interest for these operations. Indeed, the discrimination of matte on the slag during slag tapping in smelting reactors or metallic copper during slag tapping during the converting stage can impact Cu recovery for the whole smelter. This action is, again, based solely and exclusively on the experience of ground operators. Just as a starting point for reactive system measurements was established in developments in the steel industry, development in spatial recognition of high-temperature molten phases has also been established. In the last decade the spectral behavior of pig iron, and slag in pig iron production processes in blast furnaces have been studied to estimate tapping temperature [28,29]. The models presented describe the parameters contained in the iron-slag mixture in the blast furnace tapping, essentially describing parameters such as iron emissivity, tapping depth, slag layer thickness, slag absorption coefficients and radiometric parameters such as reflectance at the iron–slag interface. The method consists in determining a spectral range in which the spectral radiation of the molten phases tapped is high, facilitating the detection by a silicon CCD camera, sensitive in the visible spectral Minerals 2021, 11, 1 15 of 17

range and part of the near infrared. An optical filter centered at 650 nm was used together with the optics, so the radiation emitted by the slag was partially filtered in the indicated spectral range. The results confirm that the difference in emissivity of iron and slag at 650 nm allows a spatial distinction of the location of these phases, further identifying that the radiation intensity of the molten iron remains practically constant during the process, while the intensity of radiation from the slag is fluctuating. The fluctuation of this radiation intensity would be due to the difference in the thickness of the slag layer, absorbing and transmitting with variable percentages the radiation coming from the steel at high temperature. Moreover, the optical system has been calibrated with a high-temperature black body radiator, allowing the temperature value to be estimated at 1500–1600 ◦C. These results confirm that by using appropriate spectral models, sensitive optoelectronic systems in the spectral molten iron emission band and processing software, it is possible to obtain both laboratory and industrial developments of reliable and robust systems. In addition, it is now possible to find commercial instrumentation for the on-line detection of slag in the steel industry (Slag Detection Systems, LAND instruments) [29]. This instrument measures infrared radiation in the middle band (3–5 micrometer wavelength) with a narrow band optical filter in 3.7 micrometers, generating a video in which the steel of the slag is clearly discriminated in the image. Associated with the equipment, there is an algorithm specially designed to calculate the percentage of slag. This review shows that the possibility to apply spatial recognition of molten phases is already a feasible development based on the same principles we have used until now for reactive systems. The concept is to adopt the fundamentals the define the principle and adapt them to the physical, chemical and optic properties of the molten phases of Cu pyrometallurgy. Iron and copper have similar physical characteristics: bond microstructure and comparable transport properties, especially of the molten phase. At the same time, the steel making slag and the copper slag have similar oxide structure because they have the same constituents: SiO2, CaO, Al2O3, MgO and FeOx.

5. Final Remarks In this first stage, the results obtained at the laboratory scale have been validated in the industrial environment. It is important to point out that the experimental set-up designed to replicate the fundamental phenomena of industrial reactors on a small scale is fully valid for taking radiometric measurements during the evolution of the reactions. The radiometric measurements are the subject of this proposal for the development of optoelectronic sensors for the Cu pyrometallurgy, or in a wider scope, for non-ferrous metallurgy. These radiometric measurements are specially adapted for the chemical processes that take place in the reactors of a copper smelter. As previously specified (Figures2 and3), the scale of the fundamental phenomena is in the order of milliseconds, even though the control of the smelter is based on the monitoring of variables with a sampling rate in the order of a minute. Optoelectronic sensors have the characteristics required to follow at this level the oxidation phenomena that take place during all operational stages of a copper smelter. Undoubtedly, the current reality of operational control in copper smelters presents a considerable gap with the situation of other process industries, and very particularly with the iron and steel industry. However, the new paradigm of Industry 4.0 could be a catalyst to address these necessary developments and thus narrow the gap. The developments presented require work to validate the instruments, not in their functionality (which has already been demonstrated) but in their application, first, to create tools of real value for the operators. The possibility of carrying out an online treatment of the information is something very promising and so require different process calculations to support the operators and tend to develop automatic control strategies. Those predictive calculations require the generation of databases to evaluate different tools within the data analytic techniques that, and in an accelerated manner, have been consolidated as an applied alternative to solve complex problems. In particular, it is interesting to propose the prediction of the output variables of the different reactors: chemical compositions Minerals 2021, 11, 1 16 of 17

and temperatures of the molten phases that are produced in each one of them. From a fundamental perspective, this objective is reasonable considering that the combustion flame of sulfides carries information on the species participating in the oxidation, as was demonstrated at the laboratory and industrial level. Consequently, it is plausible to expect that the information from the flame radiation, perhaps combined with some other variables, could predict the chemical characteristics of the products. The reality of copper smelters is that the information systems must be able to incorporate information from different discrete variables, such as temperature and composition of molten phases tapped out of the reactor to make batch and semi-continuous operations compatible. Although this operation seems trivial it is a source of errors and inconsisten- cies. The development of sensors that continuously measure the chemical composition of the molten phases from the reactors will be of great help to have a uniform and unique database to approach the analysis of modern techniques looking to optimize the processes and the business of the smelter. The results shown are part of a broad research program and progress has also been made in the application of these concepts for bath smelting reactors. Although this part of the study lags behind the application to flash melting furnaces, that represent over 50% of smelting reactors in the world, the first results indicate that the signal from a molten bath will be more stable and with a higher intensity than in the case of the sulfide oxidation flame. Indeed, the flame is not a continuous volume, instead, the molten bath represents a continuous domain for the field of view of the probe.

Author Contributions: Conceptualization: R.P., S.T. and D.S.; methodology: R.P. and S.T.; validation: L.A., E.B. and S.T., formal analysis, L.A., E.B., R.P., D.S. and S.T.; writing—original draft preparation, S.T. and R.P.; writing—review and editing: R.P. and funding acquisition, D.S., S.T., and R.P. All authors have read and agreed to the published version of the manuscript. Funding: This research was funded by ANID Agency grant ACM 170008 (Anillo Minería) and FONDEF program grant IT16M 10029. Both FONDEF and ANID are Chilean government agencies. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: Data is contained within the article. Conﬂicts of Interest: The authors declare no conﬂict of interest.

References 1. The 4th Industrial Revolution. Available online: http://www.industrialagilesolutions.com/industry-4-0/ (accessed on 2 Octo- ber 2020). 2. Laurila, T.; Oikari, R.; Joutsenoja, T.; Mikkola, P.; Ranki-Kilpinen, T.; Taskinen, P.; Hernberg, R. Flash Oxidation in a Laminar Flow Reactor. Metall. Mater. Trans. B 2005, 36.[CrossRef] 3. Gillard, P.; Roux, M. Study of the Radiation Emitted During the Combustion of Pyrotechnic Charges. Part I: Non Stationary Measurement of the Temperature by Means of a Two Color Pyrometer. Propellants Explos. Pyrotech. 2002, 27, 72–79. [CrossRef] 4. Gillard, P.; de Izarra, C.; Roux, M. Study of the Radiation Emitted During the Combustion of Pyrotechnic Charges. Part II: Characterization by Fast Visualization and Spectroscopic Measurements. Propellants Explos. Pyrotech. 2002, 27, 80–87. [CrossRef] 5. Weiser, V.; Eisenreich, N. Fast Emission Spectroscopy for a Better Understanding of Pyrotechnic Combustion Behavior. Propellants Explos. Pyrotech. 2005, 30, 67–78. [CrossRef] 6. Queneau, P.E.; Marcuson, S.W. Oxygen Pyrometallurgy at Copper Cliff A Half Century of Progress. JOM 1996, 48, 14–21. [CrossRef] 7. Ball, D.F.; Bullock, J.D.; Filer, D.P.; Samain, M.D.; Shewring, D.; Watts, C. Emission Spectra from Linz-Donawitz Converters. In Proceedings of the Symposium on Process Instrumentation in the Metals Industry, Swansea, UK, 19–22 April 1971. 8. Ball, D.F.; Bullock, J.D.; Filer, D.P.; Orville-Thomas, W.J.; Samain, M.D.; Shewring, D.; Watts, C. Radiation Measurements for the Control of LD Steelmaking. In Proceedings of the International Conference and Exhibition on Industrial Measurement and Control by Radiation Techniques, Surrey, UK, 11–13 April 1972. 9. Chiba, K.; Ono, A.; Ohno, T. Direct Measurement of Manganese in Molten Iron with Observing Spectrum from Hot Spot. Trans. ISIJ 1987, 27, B185. 10. Chiba, K.; Ono, A.; Saeki, M.; Ohno, T.; Yamauchi, M.; Kanamoto, Y. On-line analysis of Molten Steel in converter. Anal. Sci. 1991, 7, 655–658. [CrossRef] Minerals 2021, 11, 1 17 of 17

11. SEMTECH. Available online: http://www.semtech.se/ (accessed on 13 October 2020). 12. Wendt, W.; Alden, M.; Bjorkman, B.; Lehner, T.; Persson, W. Controlling Copper Conversion via Optical Spectroscopy. J. Met. 1987, 39, 14–17. [CrossRef] 13. Persson, W.; Wendt, W.; Demetrio, S. Use of optical on-line production control in copper smelters, Copper 99/Cobre 99, Phoenix, 1999. In The Canadian Metallurgical & Materials Landscape 1960 to 2011; Kapusta, J., Mackey, P., Nathan, S., Eds.; The Canadian Institute of Mining, Metallurgy and Petroleum: Quebec, QC, Canada, 2011. 14. Díaz, C.M.; Levac, C.; Mackey, P.J.; Marcuson, S.W.; Schonewill, R.; Themelis, N.J. Innovation in nonferrous Pyrometallurgy- 1961–2011. In The Canadian Metallurgical & Materials Landscape 1960 to 2011; Kapusta, J., Mckey, P., Stubina, N., Eds.; MetSoc of CIM: Montreal, QC, Canada, 2011; pp. 330–360. 15. Sohn, H.Y.; Chaubal, P.C. The ignition and combustion of chalcopyrite concentrate particles under suspension-smelting conditions. Metall. Trans. B 1993, 24, 975–985. [CrossRef] 16. Wilkomirsky, I.; Otero, A.; Balladares, E. Kinetics and reaction mechanisms of high-temperature flash oxidation of molybdenite. Metall. Mater. Trans. B 2010, 41, 63–73. [CrossRef] 17. Bejarano, P.A.; Levendis, Y.A. Single-coal-particle combustion in O2/N2 and O2/CO2 environments. Combust. Flame 2008, 153, 270–287. [CrossRef] 18. Maffei, T.; Khatami, R.; Pierucci, S.; Faravelli, T.; Ranzi, E.; Levendis, Y.A. Experimental and modeling study of single coal particle combustion in O2/N2 and Oxy-fuel (O2/CO2) atmospheres. Combust. Flame 2013, 160, 2559–2572. [CrossRef] 19. Lee, H.; Choi, S. An observation of combustion behavior of a single coal particle entrained into hot gas flow. Combust. Flame 2015, 162, 2610–2620. [CrossRef] 20. Arias, L.; Torres, S.; Toro, C.; Balladares, E.; Parra, R.; Loeza, C.; Villagrán, C.; Coelho, P. Flash smelting copper concentrates spectral emission measurements. Sensors 2018, 18, 2009. [CrossRef][PubMed] 21. Marín, M.; Toro, C.; Arias, L.; Balladares, E. Estimation of Spectral Emissivity and S/Cu Ratio From Emissions of Copper Concentrates at the Flash Smelting Process. IEEE Access 2019, 7, 103346–103353. [CrossRef] 22. Díaz, W.; Toro, C.; Balladares, E.; Parra, V.; Coelho, P.; Reyes, G.; Parra, R. Spectral characterization of copper and iron sulfide combustion: A multivariate data analysis approach for mineral identification on the blend. Metals 2019, 9, 1017. [CrossRef] 23. Villagrán, C. Aplicación de las Mediciones Espectrales al Monitoreo de Variables Críticas en la Pirometalurgia del Cu. Master’s Thesis, Magister en Ingeniería Metalúrgica, Universidad de Concepción, Concepción, Chile, May 2017. 24. González, V. Mediciones Multi-esprectrales Aplicada a la Caracterización de la Combustión Flash de Concentrados a Nivel de Laboratorio. Engineering Undergraduate Thesis, Ingeniería Civil Metalúrgica, Universidad de Concepción, Concepción, Chile, June 2016. 25. Toro, C.; Torres, S.; Parra, V.; Fuentes, R.; Castillo, R.; Díaz, W.; Reyes, G.; Balladeres, E.; Parra, R. On the Detection of Spectral Emissions of Iron Oxides in Combustion Experiments of Pyrite Concentrates. Sensors 2020, 20, 1284. [CrossRef][PubMed] 26. Marín, S.M. Modelación Radiométrica de la combustión de especies minerales y concentrados de cobre. Master’s Thesis, Magíster en Ciencias de la Ingeniería c/m Ingeniería Eléctrica, Universidad de Concepción, Concepción, Chile, April 2019. 27. Sugiura, M.; Otani, Y.; Nakashima, M. Radiation Thermometry for High-temperature Liquid Stream at Blast Furnace. In Proceed- ings of the SICE Annual Conference, Tokyo, Japan, 13–18 September 2011. 28. Ugiura, M.; Otani, Y.; Nakashima, M.; Omoto, N. Continuous Temperature Measurement of Liquid Iron and Slag Tapped from a Blast Furnace. SICE J. Control. Syst. Integr. 2014, 7, 147–151. 29. LAND AMETEK. Available online: https://www.ametek-land.com/products/ir-thermal-imagers-and-line-scanners/sds-slag- detection-system (accessed on 15 October 2020).