Industrial Flotation Plant Research Articles

Data-driven system for intelligent monitoring and optimization of froth flotation circuits using Artificial Neural Networks and Genetic Algorithms

In minerals processing, the froth flotation is one of the widely used process that separates valuable mineral components from their associated gangue materials. The efficiency of this process relies on several factors, such as feed characteristics, particle size, pulp flow rate, pH, conditioning time, aeration, reagents system and many other affecting parameters. These processing parameters significantly impact the overall performance of the flotation process and influence the quality of the final concentrate. For instance, improper pulp flow and reagent dosing systems can result in metal loss and waste, particularly when dealing with frequently changing ore compositions. In this work, we established an Artificial Intelligence-based system which goal is to intelligently monitor flotation circuits and to recommend set-points for the process’s manipulated variables in order to achieve optimal performance.The system has been developed and evaluated within an industrial flotation plant that processes complex Pb-Cu-Zn sulfide ores. Leveraging an Artificial Neural Network-based Mixture of Experts (MoEs) predictive model, the system accurately estimates the mineral grades in the final concentrate and tailing of the flotation circuit. Moreover, using a Genetic Algorithms-based optimization pipeline, the system recommends set-points for the manipulated variables of the process for a maximum recovery and optimal product quality.The industrial validation of the predictive component demonstrated a 94% accuracy with a rapid 3s response time. Furthermore, the hypothetical simulation of the optimization component indicated a potential 5% increase in circuit recovery and a 4% increase of lead (Pb) grade in the circuit’s final concentrate. This developed system aims to enhance the control of froth flotation process, stabilize the product quality, and improve the overall economic benefits of production efficiency. This research contributes to the field of manufacturing systems by providing practical data-driven application for the advanced monitoring, optimization and control of industrial processes with a specific emphasis on the froth flotation process.

Flotation Froth Image Recognition Using Vision Transformers

The application of computer vision systems on industrial flotation plants has benefited considerably from advances in deep learning over the last decade, mostly based on the use of convolutional neural networks and transfer learning. More recently, vision transformers (ViTs) have attracted strong interest since their first appearance in 2017, compared to the popular convolutional neural networks (CNNs). Although becoming well-established in many areas, they have not yet been considered meaningfully in machine vision or signal processing applications in mineral processing, despite the obvious benefits that their application could realize. In this paper, it is demonstrated that ViTs are neural network architectures highly capable of discriminating between different froth flotation images. A customized ViT model and a pretrained ViT model using transfer learning were studied and compared. The former achieved satisfactory performance and the latter achieved near perfect performance, both at a significantly lower computational cost than CNNs. These results suggest that ViTs can be a competitive alternative to CNNs in the advancement of computer vision systems on industrial flotation plants.

Comparison of Fuzzy and Neural Network Computing Techniques for Performance Prediction of an Industrial Copper Flotation Circuit

This paper presents the development and validation of five different soft computing methods for flotation performance prediction: (1) two models based on fuzzy logic (Mamdani and Takagi-Sugeno fuzzy inference system) and (2) three models based on artificial neural networks. Copper content in the ore feed, collector dosage in the rougher and the scavenger flotation circuits, slurry pH in the rougher flotation circuit and frother consumption were selected as input parameters to estimate the copper grade and recovery of final concentrate, as well as the copper content in the final tailings of the flotation plant. The training and evaluation of the proposed models were performed on the basis of real process data collected by the multiannual monitoring of industrial flotation plant operation in “Veliki Krivelj Mine”. The results showed that the proposed soft computing-based models well describe the behavior of the industrial flotation plant in a wide range of circumstances. Among the proposed algorithms, artificial neural networks gave the most accurate predictions for the final copper concentrate grade and recovery (R2 = 0.98 and R2 = 0.99, respectively) and copper content in final tailings (R2 = 0.87). At some points, fuzzy logic models are almost equally efficient, but artificial neural networks had lower values for all error functions.

Long short-term memory-based grade monitoring in froth flotation using a froth video sequence

With the rapid development of machine vision technology, machine vision systems are being widely used in flotation plants for online grade monitoring. In an industrial flotation plant, the sample rate of a froth video is considerably higher than that of the labelled grade data obtained using special detection devices; this results in several unlabelled froth videos. However, these unlabelled videos are valuable and should not be discarded, as they can provide temporal information regarding the video sequence, which can reflect the variation trend of the grade. Therefore, a long short-term memory (LSTM)-based network is proposed herein to estimate the tailing grade of the first rougher from a zinc flotation circuit. First, the froth video representation is obtained by calculating the average of the visual features and integrating the rougher feed grades. Then, the representations of the froth video sequence are introduced to the LSTM network to match the labelled grade. Through network training with the historical data, a grade monitoring model can be constructed, in which the froth video information is utilised adequately, and the problem of different sample rates is also solved. Experimental results show the effectiveness of the proposed grade monitoring model. Compared with those of the traditional neural network without unlabelled froth videos, the root mean squared error of the proposed model decreases by 8.48% and the R-squared score of the proposed model increases by 9.32%.

Carrier Flotation: State of the Art and its Potential for the Separation of Fine and Ultrafine Mineral Particles

Critical raw materials (CRMs) are of primary importance for energy storage systems as needed for electromobility. Many mineral deposits which contain CRMs are low-grade ores. To liberate the CRMs, a grinding of the mineral ores to very fine sizes below 20 µm particle size is necessary. However, the present class of industrial flotation plants fail to extract such fine and ultrafine particles. To improve the recovery in fine particle flotation, techniques have been developed which attempt to agglomerate the fine valuable particles into larger aggregates which subsequently can be separated by established technologies such as froth flotation. Carrier flotation is one of these techniques. The present work reviews the state of the art of this technique for the recovery of fines and ultrafines.

Adapting a Phenomenological Model of a Rougher Flotation Circuit to Industrial Historical Operating Data Base

The use of simulators is a powerful tool to train plant operators and to be also incorporated in the development and test of supervisory control strategies. However, the phenomenological models describing the process are relatively complex, characterized by nonlinear relationships and whose parameters are depending on many local factors, such as, the plant configuration, the individual characteristics of the equipment, the availability of on-line measurements and the characteristics of the feed, among others.In this work, the previously developed phenomenological model is adapted to the particular characteristics of the rougher circuit of an industrial flotation plant, considering its particular layout and the available information to feed the simulator. The rougher circuit consists of three lines of 8 mechanical cells processing a feed of 4000 tons/h.The new model predictions were tested for a family of feed characteristics, including variation of mineralogical species under different operating conditions. Since some variables are commonly unmeasured during the operation, additional data were incorporated to improve the model predictability. The use of the simulator is illustrated in several examples, as well a discussion of the model prediction limitations due to some particularities found in the historical operating data.

Froth image analysis by use of transfer learning and convolutional neural networks

Deep learning constitutes a significant recent advance in machine learning and has been particularly successful in applications related to image processing, where it can already surpass human accuracy in some cases. In this paper, the use of a convolutional neural network, AlexNet, pretrained on a database of images of common objects was used as is to extract features from flotation froth images. These features could subsequently be used to predict the conditions or performance of the flotation systems. Two case studies are considered. In the first, froth regimes in an industrial flotation plant could be identified significantly more reliably with the features generated by AlexNet than with previous state-of-the-art approaches, such as wavelets, grey level co-occurrence matrices or local binary patterns. In the second case study, the arsenic concentration in the batch flotation of realgar-orpiment-quartz mixtures could be predicted more accurately than was possible with features extracted by wavelets, grey level co-occurrence matrices, local binary patterns or by use of colour. These results suggest that feature extraction with convolutional neural networks trained on complex data sets from other domains can serve as more reliable methods than previous state-of-the-art approaches to froth image analysis.

On the mineral recovery estimation in Cu/Mo flotation plants

A numerical conditioning analysis for mineral recovery estimation was performed for industrial flotation plants, considering the copper (Cu), molybdenum (Mo) and iron (Fe) separability. A modified relative condition number, χ, was presented that allowed sensitivity analysis to be evaluated for the component recovery by means of an analytical formula. This closed form made it possible for the error propagation to be determined from feed, concentrate and tail grades with different orders of magnitude. The χ parameter can be evaluated using available grade data from the design criteria or historical mass balances, in which the variability is typically unknown. Reconciled data from different Cu/Mo concentrators were employed to evaluate the effect of small numerical disturbances in the grade data on the mineral recovery estimation. Higher error propagation was typically observed for Fe. The Mo minerals presented numerical problems, mainly in second cleaners and in the first cell of rougher banks. Lower condition numbers were observed for Cu due to the higher flotation rates. Mass-balance data reconciliation without redundancy was evaluated for a typical Cu/Mo flotation circuit using relative error minimization. Significant relative errors in the mineral recovery estimation were obtained with nonreconciled data in ill-conditioned problems. Negligible improvements in the mineral recovery estimation because of the data reconciliation with regard to the nonreconciled approach were obtained in well-conditioned problems. In addition, the improvements in mineral recovery estimation by using Cu, Mo and Fe in the data reconciliation were nonsignificant with respect to using only the best-conditioned component in well- and ill-conditioned problems. Despite the effort in data reconciliation and data repetition, poor performance may be obtained in ill-conditioned problems, which can deteriorate the flotation rate characterization. The error propagation has a negative impact on the mineral recovery of the first cell, which may significantly bias the flotation rate characterization of both valuable and nonvaluable elements.

Estimation of Copper and Molybdenum Grades and Recoveries in the Industrial Flotation Plant Using the Artificial Neural Network

In this paper, prediction of copper and molybdenum grades and their recoveries of an industrial flotation plant are investigated using the Artificial Neural Networks (ANN) model. Process modeling has done based on 92 datasets collected at different operational conditions and feed characteristics. The prominent parameters investigated in this network were pH, collector, frother and F-Oil concentration, size percentage of feed passing 75 microns, moisture content in feed, solid percentage, and grade of copper, molybdenum, and iron in feed. A multilayer perceptron neural network, with 10:10:10:4 structure (two hidden layers), was used to estimate metallurgical performance. To obtain the optimal hidden layers and nodes in a layer, a trial and error procedure was done. In training and testing phases, it achieved quite correlations of 0.98 and 0.93 for Copper grade, of 0.99 and 0.92 for Copper recovery, of 0.99 and 0.92 for Molybdenum grade and of 0.99 and 0.94 for Molybdenum recovery prediction, respectively. The proposed neural network model can be applied to determine the most beneficial operational conditions for the expected Copper and Molybdenum grades and their recovery in final concentration of the industrial copper flotation process.

Modelling Of Flotation Processes By Classical Mathematical Methods – A Review

Abstract Flotation process modelling is not a simple task, mostly because of the process complexity, i.e. the presence of a large number of variables that (to a lesser or a greater extent) affect the final outcome of the mineral particles separation based on the differences in their surface properties. The attempts toward the development of the quantitative predictive model that would fully describe the operation of an industrial flotation plant started in the middle of past century and it lasts to this day. This paper gives a review of published research activities directed toward the development of flotation models based on the classical mathematical rules. The description and systematization of classical flotation models were performed according to the available references, with emphasize exclusively given to the flotation process modelling, regardless of the model application in a certain control system. In accordance with the contemporary considerations, models were classified as the empirical, probabilistic, kinetic and population balance types. Each model type is presented through the aspects of flotation modelling at the macro and micro process levels.

Modeling of residence time distribution in regrinding Vertimill

The regrinding stage is necessary to achieve the particle size (liberation) for the final upgrade in the cleaning flotation process. At present, vertical stirred mills are increasingly employed at industrial scale. The regrinding process characterization requires mixing regime knowledge along with the mean residence time of the liquid and solid phases. In this work, residence time distribution (RTD) measurements were performed in two different industrial flotation plants. Both plants operate with Vertimills of 1000–1500HP in the regrinding process with different circuit arrangements according to the plant treatment capacity. The RTD was measured using the radioactive tracer technique, which allows for non-invasive tracer detection in real time. It was found that mixing was not perfect in agitated Vertimills and consequently the RTD was better modeled by the more flexible Large and Small Tanks in Series (LSTS) model which allowed the characterization of the different mixing regimes for both liquid and solids in the regrinding mills. Effective mean residence times in the range of 1.7–12.5min were obtained with RTD shapes similar to those obtained in tumbling ball mills. It was found that residence times of liquid and solid particles were similar. RTD estimations are useful for better understanding the Vertimill behavior and to identify the actual breakage rates, as well as for improving Vertimill modeling and simulation.

Modelling and simulation of rougher flotation circuits

At present, large mechanical flotation cells of 100 to 300m3 are used in rougher operation in different industrial flotation plants around the world. However, in spite of the advances in fundamental research and the notable growth in equipment size, there is still a lack of reliable data for industrial flotation modelling and simulation. In this work, a procedure for modelling and simulating rougher flotation banks that is based on operating variables and parameters fitted by empirical data from plant measurements is presented.Recently, a new methodology for describing industrial flotation by separating the collection and froth zones has been developed. This approach consists of using a new apparatus for direct bubble load measurement below the pulp–froth interface in industrial cells. The procedure allows independently estimating the froth recovery and the collection zone recovery. Metallurgical characterisation is developed by plant sampling, mass balance adjustment and applying a short-cut method for kinetic characterisation of flotation banks. In addition, the effect of grinding is characterised using flotation models with distributed flotation rates (for each particle size class) and a low number of parameters.From the experimental data, a non-linear distributed model to simulate rougher flotation banks was developed, which allows describing the industrial operation as a function of the main operating variables (e.g., mass flow rate, solid percentage, grinding degree, pulp level and superficial air rate). The simulator was calibrated and validated using experimental data from the rougher operation at El Teniente Division, Codelco-Chile.The new approach is a practical tool that is useful for flotation operation diagnosis, for robust design and analysis of new control and optimisation strategies and for evaluation of the effect that design and operating variables have upon the collection and froth zone performance in large flotation cells. This methodology can be applied to other flotation operations, such as cleaner and scavenger circuits that use mechanical flotation cells.

A continuous pilot-scale flotation rig for the systematic study of flotation variables

A pilot-scale flotation apparatus has been developed for the purpose of determining floatability parameters for an ore. It has the capability to vary many of the important flotation variables allowing them to be investigated independently. Adjustable factors include: cell size, froth depth, feed, tailings and air volumetric flow rates and reagent addition rates. The flotation rig was commissioned successfully at BHP Billiton’s Cannington mine in North-West Queensland, Australia. It was found to supply a stable feed stream with air well distributed across the cell. At the test conditions, the relative standard deviation of the calculated galena overall recovery was found to be only 1%. Such an apparatus could prove invaluable in flotation research, allowing live streams from an operating industrial flotation plant to be processed continuously. The versatility of the design and accuracy of performance allow the investigation of most major flotation variables with confidence.

Predicting air recovery in flotation cells

In this paper, a simple theoretical model for the fractional air recovery to the concentrate in flotation froths is presented. The model assumes that two forces are exerted on the films, one from the curvature of Plateau borders and the other from the stretching of the films. It is further assumed that there is a critical pressure above which the films will fail. This model predicts that there is expected to be a maximum in the air recovery as the air rate is increased, which is indeed observed experimentally at a number of industrial flotation plants. This maximum in air recovery does not correspond to a maximum in the inherent stability of the films and occurs even if this stability is constant, though having a film stability that changes with air rate causes some interesting changes in the behaviour of the system.

Characterisation of flotation froth colour and structure by machine vision

It is well known and well recognised that flotation is a process that is complex to monitor and study if a classical approach based on the evaluation of the signals resulting from sensors is adopted. Sensors are usually strategically positioned in the bank cells and detect global process variables such as pH, reagent addition, froth level, on-stream chemical analysis, particle size distribution, etc. In the last ten years several studies have been carried out with the main goal to utilise imaging techniques to detect froth bubbles characteristics and to evaluate the flotation process performance. In this paper an approach of this type is described. More specifically, image processing techniques to automatically measure the colour and the structure of the froth bubbles are presented and the results are discussed. All the investigations are carried out on digital sample images collected in an industrial flotation plant operating in steady-state conditions. The colour analysis is performed on the whole surface of the froth images considering different colour reference systems (RGB, HSV, HSI); the morphological measurements are obtained after the application of selected enhancement and segmentation techniques, necessary to consider the bubbles as separate domains. The multiple correlation analysis performed between froth mineral concentrations (Cu, MgO, Zn and Pb content) and the extracted colour and structure parameters are good in most situations.

Vision- and model-based control of flotation

A new vision system for monitoring and control of flotation processes has been developed The system produces information of the flotation froth structure and colour. This can be used for estimation and prediction of the state of the process. Due to the complexity and nonlinearity of the flotation process the classification methods have been applied to classify the operation conditions Process models containing image variables have been constructed and applied to simulation. The equipment has been installed on an industrial flotation plant. New control strategies based on image information have been developed and tested.

Development of an image analysis based architecture for the automated control of flotation processes

In the framework of the EU Esprit LTR Project N◦ 24931 ChaCo “The characterisation of flotation froth structure and colour by machine vision”, the step by step procedure adopted in order to develop a software and hardware architecture based on machine vision for the flotation process control is described. The froth images were collected from the industrial flotation plants of Garpenberg (Sweden) and Pyhasalmi (Finland). After that, some data sets of images were extracted from the source images in order to have a handier sample and to create some subsets for the validation of the results. The next step was the definition of processing algorithms able to characterise the froth, and their subsequent implementation (colour, morphological, textural and other parameters). Then a regression analysis to find correlation with the mineral concentration in the froth was performed. The estimated algorithms were applied to the remaining data sets of images for the validation of the model. Once tuned this procedure and verified its validity, the image processing algorithms were integrated in a software and hardware architecture in order to realise the final prototype of the on-line froth analyser.

An empirical model for entrainment in industrial flotation plants

The entrainment and drainage of particles in a flotation froth have long been recognised as important factors affecting both concentrate grade and recovery. In this paper, an empirical partition curve is proposed that describes the degree of entrainment within a conventional flotation cell. Two empirical parameters are defined: • • entrainment parameter (ζ)—the particle size for which the degree of entrainment is 20%. • • drainage parameter (δ)—related to the preferential drainage of coarse particles. The model is flexible enough to fit a broad data set published in the flotation literature. Results from a case study indicate that the model is sufficiently detailed to describe the entrainment and drainage mechanisms. This being so, the effect of process variables which affect the degree of entrainment can be properly quantified and then included in a simulation package, assisting in the optimisation of existing flotation circuits and the design of future ones.

Rapid estimation of floatability components in industrial flotation plants

In the analysis of industrial flotation plant data, it is advantageous to build a reliable computer based flotation model to be used in optimisation and design. A number of models have been proposed by many authors with each model having value depending on the complexity of the situation for which it was developed. Several of these models assume that a floatability distribution of a mineral entering a flotation circuit can be described by a number of floatability classes or components with each of these components (and its flotation rate constant) being conserved around the circuit. A non-linear optimisation procedure for regressing floatability parameters (number of floatability components, and the flotation rate constant of each) from laboratory batch flotation tests conducted on streams around the circuit has recently been devised [1]. This paper outlines a technique to linearise the above regression procedure which improves the ease of solution and the level of confidence in the floatability parameter determination. The technique also offers a strategy for determining the “optimum” number of floatability components. Dummy and real plant data were used to test the validity of the new linear modelling technique. The dummy data were generated using a set flowsheet and a feed stream having two components (with set first order rate constants) to simulate the plant. The simulated data obtained were then fitted with the general linear model technique. The results of this operation show that the linear modelling procedure can obtain the original model parameters very well. The model was then tested on real plant data from three typical flotation plants. Again, there is a very good correlation between stream experimental and model calculated recoveries based on the feed to the circuit. The confidence limits of the model parameters generated using this technique are reported. The ability to estimate parameter confidence intervals indicates where additional data (or an independent method) are required for more accurate parameter determination.

Multivariable Control of Pulp Levels in Flotation Circuits

A system for exceptionally good control of pulp levels in a flotation circuit is described It produces fast and stable control that is much better than is achieved with conventional control loops. The control strategy eliminates the problem of cyclic or interactive disturbances that are commonly present in plants with connected stages and recycled streams. It has been implemented in Mintek’s FloatStar™ controller on industrial flotation plants in South Africa and Australia The strategy described is also applicable to the control of many processes in which material is transferred in one or both directions along a series of stages.