Monitoring flotation cells

Abstract

A computer-implemented method (100) for monitoring flotation cell / s (e.g., 12.1, 12.2, 12.3) and associated system (10) are disclosed. A server (14) receives digital images captured by a camera (16.1, 16.2, 16.3 or 62) at the flotation cell and accesses a machine learning model (20) which is trained with training image / s. The received digital images are input into the machine learning model (20) and analysed to identify bubbles and to generate data relating to a characteristic of the bubbles. A reagent dosing parameter (30) is generated based on the analysis by the machine learning model (20), and it is output for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

Description

[0001] MONITORING FLOTATION CELLS

[0002] CROSS-REFERENCE TO RELATED APPLICATIONS

[0003] This application claims priority from South African provisional patent application number 2024 / 09481 filed on 10 December 2024, which is incorporated by reference herein.

[0004] FIELD

[0005] This disclosure relates to systems and methods for monitoring flotation cells. More specifically, but not exclusively, the present disclosure relates to systems and methods for monitoring and controlling flotation cells in a mineral processing process.

[0006] BACKGROUND

[0007] Froth flotation is a process for separating minerals from gangue by taking advantage of differences in their hydrophobicity. Hydrophobicity differences between valuable minerals and waste gangue are increased through the use of surfactants and wetting agents.

[0008] The flotation process typically takes place in an open cell and consists of the pulp phase which can be described as the ‘reactor’ and the froth phase which can be termed the ‘separator’. In the pulp phase, hydrophobic particles preferentially attach to rising air bubbles which form a froth at the top of the pulp phase and are recovered as concentrate. Sub-processes such as bubbleparticle collision, attachment, detachment and entrainment dominate. These sub-processes have an overall effect of transporting particles, mostly hydrophobic particles, to the froth phase. In the froth phase, the process of froth formation and transport determines the kind of sub-processes that take place. Froth phase sub-processes such as thinning of bubble films, bubble coalescence and froth drainage result in an increase in bubble sizes, particles detaching from bubbles and draining back into the pulp phase. Froth phase sub-processes may lead to cleaning / separating action if there is preferential re-attachment of the draining particles to the available bubble surface area. This cleaning action determines the overall grade and recovery of the flotation process. The initial separating action of the froth phase starts at the pulp-froth interface.

[0009] While numerous studies have attempted to correlate froth structure and characteristics with flotation performance, achieving consistent, reliable relationships remains elusive and challenging due to the complex, dynamic nature of froth behaviour and its interaction with mineral recovery. Various methods for measurement of the bubble size distribution (BSD) in the pulp phase have been developed. These methods, however, suffer from a variety of drawbacks. For example, available methods conduct offline or batch measurements of the pulp bubble properties. Most existing offline approaches also involve extracting bubbles from the flotation cell into a separate chamber, where the extracted matter is analysed through the use of camera imaging. The separate chamber is typically filled with clean water (using valve systems) to enable effective imaging of the bubbles. Mineral build-up also occurs in these chambers causing viewing ports to become fouled and necessitating regular cleaning, typically by way of flushing. The clean water inside the chamber gets displaced by the air contained in the bubbles and needs to be refilled after each measurement. Since the bubble collection or draw-off points of these offline devices are inside the cell and the measurement points above or otherwise outside the cell, the pressure differences between the two environments need to be compensated for. This in turn complicates the calculation process to obtain accurate results.

[0010] Moreover, assays or ore testing may be delayed, as testing metal, minerals or ore in a laboratory takes time. An operator of the flotation process may also have a subjective experience, and he or she may not have any visibility of the pulp or froth. This may lead to poor process control and, in turn, a poor mineral recovery grade or an inefficient process in general.

[0011] The applicant considers there to be room for improvement.

[0012] The preceding discussion of the background is intended only to facilitate an understanding of the present disclosure. It should be appreciated that the discussion is not an acknowledgment or admission that any of the material referred to was part of the common general knowledge in the art as at the priority date of the application.

[0013] SUMMARY

[0014] In accordance with an aspect of the disclosure there is provided a computer-implemented method for monitoring a flotation cell, the method comprising: by a server, receiving digital images captured by a camera at the flotation cell, and accessing a machine learning model which is trained with one or more training images; inputting the digital images into the machine learning model, and by the machine learning model, analysing the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; generating a reagent dosing parameter based on the analysis by the machine learning model; and outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

[0015] The characteristic of the bubbles may, for example, be a gas characteristic of the bubbles.

[0016] The method may include generating data relating to any one or more of: a bubble size distribution of bubbles in the received digital images; a gas holdup of bubbles in the received digital images; a gas velocity of bubbles in the received digital images; and / or a colour of the received digital images.

[0017] The method may include monitoring one or more, or a plurality of flotation cells. Outputting the reagent dosing parameter may include outputting it for use in dosage of reagents in any one or more of the plurality of flotation cells.

[0018] Outputting the reagent dosing parameter may include outputting a reagent dosing set-point parameter.

[0019] Additionally, or alternatively, the method may include generating a gas set-point parameter based on the analysis by the machine learning model, and outputting the gas set-point parameter, for use in controlling aeration of the flotation cell / s. The gas set-point parameter may be an air setpoint parameter, e.g., when air is used as an aeration gas to produce bubbles in the process. It will be appreciated that other gases may also be used for aeration or to generate bubbles.

[0020] The reagent dosing parameter and / or the air or gas set-point parameter may be outputted to a system operator that operates a local computing device. The local computing device may e.g., be connected to one or more controller / s or machines, such as one or more of a frother, a collector, a depressant, or a mineral separator machine or a controller associated therewith. The server may optionally be in data communication directly with any one or more of these controller / s or machines, e.g., to transmit any one or more of the parameters / data of the present disclosure to these controller / s.

[0021] The air or gas set-point parameter may for example be outputted by the server to a controller associated with air or gas flow of the process associated with the flotation cell / s. Optionally, the air or gas set-point parameter may be outputted by the server to the local computing device of the operator, wherefrom it may be relayed to the controller associated with air or gas flow of the process associated with the flotation cell / s. Optionally, the air or gas set-point parameter may be outputted by the server to the controller associated with air or gas flow of the process associated with the flotation cell / s.

[0022] The method may include outputting the air / gas set-point parameter so as to enable automated adjustment of aeration or gas input in the process associated with the flotation cell / s.

[0023] The method may include outputting the reagent dosing parameter, or outputting the reagent dosing set-point parameter, so as to enable automated adjustment of reagent addition to the process associated with the flotation cell / s.

[0024] The method may include outputting data by the machine learning model. The data output by the machine learning model may be utilised as an input to a level controller, e.g., so as to enable control of a level of froth in the flotation cell / s.

[0025] The method may include providing one or more controller / s for controlling one or more machines associated with the flotation cell / s. The server and / or the local computing device may be in data communication with the one or more controllers, and arranged to output the reagent dosing parameter and / or the air / gas set-point parameter to the one or more controllers.

[0026] The method may include, by the machine learning model, generating data relating to a bubble surface area flux, e.g., based on the analysis of the received digital images. Any data outputted by the machine learning model may additionally or alternatively be based on an analysis of the bubble surface area flux.

[0027] The method may include, utilising a probe inside each of the one or more flotation cell / s. The probe may be at least partially submerged with a viewing pane and an operative end of the probe inside the flotation cell. The probe may include the camera, and at least one light source. The viewing pane may be positioned at an oblique angle relative to vertical. The method may include, by the camera, capturing intermittent digital images of bubbles inside the flotation cell under illumination from the light source through the viewing pane. The method may further include, transmitting the captured digital images to the server for further processing.

[0028] The method may include utilising a froth sensor that may include the camera. The froth sensor may be positioned to capture top-down imagery of the flotation cell / s froth surface.

[0029] The method may include, by the server, classifying froth states in the received digital images. The froth states may include dry froth, wet froth, and pulp exposure zone / s. The method may include, by the froth sensor, utilising a bubble segmentation model accessible by the froth sensor. The bubble segmentation model may be configured to analyse the captured images to extract parameters including one or more of bubble coverage and bubble count. The extracted parameters may be processed by the server to derive a wet froth indicator that quantifies relative froth wetness and entrainment.

[0030] In a further aspect, the disclosure may provide for the deployment of a froth sensor. The froth sensor may include one or more camera / s. The froth sensor may, for example, be positioned to capture top-down imagery of the flotation cell froth surface. This may enable classification of froth states into dry froth, wet froth, and pulp exposure zone / s. The froth sensor may include an integrated bubble segmentation model which may be configured to analyse the images to extract parameters such as bubble coverage and bubble count. These parameters may be processed to derive a wet froth indicator that quantifies relative froth wetness and entrainment. This wet froth indicator may be utilized by flotation control system / s to maintain the froth within an optimal state through near real-time adjustments of air flow, froth or pulp levels, and reagent dosing. By promoting dry froth stability while minimizing wet froth and pulp carryover, the systems and methods of the present disclosure may enhance recovery of high-grade mineral concentrate and / or reduce entrainment of unwanted materials.

[0031] In accordance with a further aspect of the disclosure there is provided a system for monitoring a flotation cell, the system comprising: a non-transitory computer-readable storage medium; and one or more processors coupled to the non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium comprises program instructions that, when executed on the one or more processors, cause the system to perform operations comprising: by a server, receiving digital images captured by a camera at the flotation cell, and accessing a machine learning model which is trained with one or more training images; inputting the digital images into the machine learning model, and by the machine learning model, analysing the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; generating a reagent dosing parameter based on the analysis by the machine learning model; and outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

[0032] The characteristic of the bubbles may, for example, be a gas characteristic of the bubbles. The machine learning model may be arranged to analyse the received digital images and to generate data relating to any one or more of: a bubble size distribution of bubbles in the received digital images; a gas holdup of bubbles in the received digital images; a gas velocity of bubbles in the received digital images; and / or a colour of the received digital images.

[0033] In accordance with a further aspect of the disclosure there is provided a system for monitoring a flotation cell, the system including a memory for storing computer-readable program code and a processor for executing the computer-readable program code, the system comprising: a server which is arranged to receive digital images captured by a camera at the flotation cell; a machine learning model which is accessible by the server, the machine learning model being trained with one or more training images and operable to receive the digital images as input and to analyse the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; wherein the machine learning model is arranged for generating a reagent dosing parameter based on the analysis by the machine learning model; and wherein the server is arranged for outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

[0034] The characteristic of the bubbles may, for example, be a gas characteristic of the bubbles.

[0035] The machine learning model may be arranged to analyse the received digital images and to generate data relating to any one or more of: a bubble size distribution of bubbles in the received digital images; a gas holdup of bubbles in the received digital images; a gas velocity of bubbles in the received digital images; and / or a colour of the received digital images.

[0036] Further features of the system may include components that are arranged to carry out any one or more of the steps of the method / s defined above.

[0037] In accordance with a further aspect of the disclosure there is provided a computer program product for monitoring a flotation cell, the computer program product comprising a computer- readable medium having stored computer-readable program code for performing the steps of: by a server, receiving digital images captured by a camera at the flotation cell, and accessing a machine learning model which is trained with one or more training images; inputting the digital images into the machine learning model, and by the machine learning model, analysing the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; generating a reagent dosing parameter based on the analysis by the machine learning model; and outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

[0038] The characteristic of the bubbles may, for example, be a gas characteristic of the bubbles.

[0039] Generating data by the machine learning model, e.g., during the analysis of the received digital images, may include generating data relating to any one or more of: a bubble size distribution of bubbles in the received digital images; a gas holdup of bubbles in the received digital images; a gas velocity of bubbles in the received digital images; and / or a colour of the received digital images.

[0040] Further features may provide for the computer-readable medium to be a non-transitory computer- readable medium and for the computer-readable program code to be executable by a processing circuit.

[0041] Embodiments of the technology will now be described, by way of example only, with reference to the accompanying drawings.

[0042] BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In the drawings:

[0044] Figure 1 is a high-level block diagram of an exemplary system for monitoring one or more flotation cell / s;

[0045] Figure 2 is a high-level swim-lane flow diagram which illustrates an exemplary method for monitoring one or more flotation cell / s in accordance with aspects of the present disclosure;

[0046] Figure 3 is a schematic illustration of an exemplary flotation cell utilising a probe and camera, the flotation cell forming part of the exemplary system of Figure 1 ; Figure 4 is an exemplary photograph of an image of bubbles taken by a test probe in accordance with the present disclosure under laboratory conditions;

[0047] Figure 5 is an exemplary photograph of an image of bubbles taken by a test probe in accordance with the present disclosure under production, field conditions;

[0048] Figure 6 is a high-level flow diagram illustrating exemplary parameters that may be input to an advanced process control (APC) system forming part of the system of Figure 1 ;

[0049] Figure 7 is a photograph of an exemplary probe that may be used in the system of Figure 1 ;

[0050] Figure 8 is a photograph of the exemplary probe of Figure 7 being partially submerged into the flotation cell in accordance with aspects of the present disclosure;

[0051] Figure 9 is an example of a digital image of bubbles captured by a camera viewing a froth phase of a flotation cell (e.g., as shown in Figure 11), the digital image forming part of a video stream of digital images analysed by the system of the present disclosure;

[0052] Figure 10 is an example of a digital image of bubbles captured by a camera viewing a pulp phase of a flotation cell (e.g., as shown in Figure 8), the digital image forming part of a video stream of digital images analysed by the system of the present disclosure;

[0053] Figure 11 is a schematic illustration of an exemplary flotation cell utilising a probe and camera for imaging a pulp phase of the flotation cell, as well as a camera for imaging a froth phase of the flotation cell which may form part of the exemplary system of Figure 1 ;

[0054] Figure 12 is a system diagram illustrating exemplary control parameters that may be implemented by the system of Figure 1 ;

[0055] Figure 13 is a high-level block diagram illustrating exemplary aspects of the present disclosure which may enable an increase in mineral recovery and an increase in flotation, while alleviating lower grade recovery and alleviating high entrainment in the flotation cell;

[0056] Figure 14 is an exemplary graph illustrating frother concentration and simulated critical coalescence concentration (CCC) for a normal case where reagent dosing advisor aspects of the system are disabled, where frother concentration within the pulp remains too low and subsequently a sub-optimal pulp bubble size distribution is observed;

[0057] Figure 15 is a graph similar to that of Figure 14, but showing exemplary reagent dosing parameter or advisor aspects of the system enabled, with a corresponding increase in frother concentration, e.g., due to increased reagent feed to address differing operating conditions and as such optimal pulp bubbles size can be maintained;

[0058] Figure 16 is a further development of Figure 14, with CCC drift (e.g., due to ore changes, or operational disturbances) showing that the frother addition needs to be adjusted, otherwise a sub optimal bubble size distribution will be experienced;

[0059] Figure 17 is a further development of Figure 15, with CCC drift (e.g., due to ore changes, or operational disturbances) where the exemplary system’s reagent advisor again inspects all plant and sensor variables and determines the froth dosage needs to change leading to optimal or near-optimal pulp bubble size;

[0060] Figure 18 is a yet further development of Figure 14, with CCC drift (e.g., due to ore changes, or operational disturbances) where the frother addition needs to be adjusted otherwise overdosing of reagents is possible;

[0061] Figure 19 is a yet further development of Figure 15, with CCC drift (e.g., due to ore changes, or operational disturbances) where the reagent advisor of the exemplary system again inspects all plant and sensor variables and determines the froth dosage needs to change leading to optimal (or near- optimal) pulp bubble size while reducing reagent costs;

[0062] Figure 20 is a block diagram illustrating exemplary froth state features that may be implemented by the system of Figure 1 ;

[0063] Figure 21 is a block diagram illustrating exemplary pulp state features that may be implemented by the system of Figure 1 ;

[0064] Figure 22 is a flow diagram illustrating exemplary froth state advisories expert rules that may be implemented by the system of Figure 1 ;

[0065] Figure 23 is a flow diagram illustrating exemplary details of froth state advisories expert rules that may be implemented by the system of Figure 1 ;

[0066] Figure 24 is a high-level block diagram of an exemplary froth state dashboard that may be implemented in accordance with aspects of the system of Figure 1 ;

[0067] Figure 25 is an exemplary chart illustrating a percentage of time in froth states that may be implemented in accordance with aspects of the present disclosure;

[0068] Figure 26 is an exemplary chart illustrating 7-day and 14-day MassPull that may be implemented in accordance with aspects of the present disclosure;

[0069] Figure 27 is an exemplary chart illustrating overall MassPull daily times in various states that may be implemented in accordance with aspects of the present disclosure;

[0070] Figure 28 is an exemplary chart illustrating overall MassPull time in various states per sump that may be implemented in accordance with aspects of the present disclosure;

[0071] Figure 29 is a MassPull control chart showing concentrate MassPull against time that may be implemented in accordance with aspects of the present disclosure;

[0072] Figure 30 is another MassPull control chart showing concentrate MassPull against time that may be implemented in accordance with aspects of the present disclosure;

[0073] Figures 31-32 are diagrams that illustrate overall MassPull and froth state performance metrics in accordance with exemplary aspects of the present disclosure;

[0074] Figures 33-34 are exemplary phase diagrams that illustrate froth states that may be detected by a froth sensor and camera;

[0075] Figure 35 illustrates an example of a computing device in which various aspects of the disclosure may be implemented; and

[0076] Figure 36 illustrates an overview of training and use of machine learning models that may be implemented by aspects of the present disclosure.

[0077] DETAILED DESCRIPTION WITH REFERENCE TO THE DRAWINGS

[0078] Having an accurate model of a flotation process and obtaining near real-time measurements of the froth phase can aid in predicting the grade and recovery of the flotation cell as a whole. This, in turn, may be useful for plant optimisation.

[0079] Important parameters for process control of such froths may be the bubble size distribution (BSD and D32), superficial gas velocity (Jg) gas holdup (Eg) in the pulp phase and pulp colour. By having (near) real-time measurements of these parameters, the process control of the froth floatation process can be greatly optimised. The applicant’s granted South African patents with application numbers ZA2019 / 03223 and ZA2020 / 01710 describe the methods, apparatus and systems for measuring properties of the pulp phase of a flotation cell and computer implemented methods and systems for monitoring froth flotation processes including multiple flotation cells on a plant level, respectively, and are incorporated herein in their entirety by reference.

[0080] Figure 1 is a schematic diagram which illustrates an exemplary system 10 for monitoring a flotation cell. In the present exemplary embodiment, a plurality of flotation cells 12.1 , 12.2, 12.3 are monitored. It will be appreciated that any number of flotation cells may be monitored. The system 10 may include server 14 in data communication with one or more cameras e.g., 16.1,

[0081] 16.2, 16.3. The cameras may, for example be provided at each of the flotation cells 12.1 , 12.2,

[0082] 12.3.

[0083] The server 14 may include a processor 18 for executing the functions of components described herein, which may be provided by hardware or by software units executing on the server 14. The software units may be stored in a memory component 16 and instructions may be provided to the processor 18 to carry out the functionality of the described components. In some cases, for example in a cloud computing implementation, software units arranged to manage and / or process data on behalf of the server may be provided remotely. The memory 16 may be arranged for storing computer-readable program code and the processor 18 may be arranged for executing the computer-readable program code.

[0084] The server may implement, or it may have access to, a machine learning (ML) model 20. Further features of the ML model 20 are described in more detail below.

[0085] The ML model may also be termed an artificial intelligence model. This model may be trained by one or more image processing algorithms. The ML model may be trained, or pre-trained. The machine learning model or module may be a neural network (NN), and it may have a deep learning network architecture. The deep learning network may be a convolutional neural network (CNN), or a fully convolutional deep neural network (FCDNN or FCNN). Other types of artificial neural networks may also be implemented by the present disclosure. The server or backend may implement a machine learning model, or it may have access to a cloud-based machine learning model. An input image (such as image data received by the server or backend from the camera / s) or a sequence of images may be input to the machine learning module or model according to aspects of the present disclosure. The machine learning model or module may be pre-trained or trained to provide an output that is indicative of one or more parameters or characteristic / s relating to bubbles in the images. In other words, predefined images or image data may have been previously input into the machine learning model / module in order to train it to intelligently identify bubbles or other data in received image data. This intelligently identified data may include, but is not limited to a bubble size distribution of bubbles in the received digital images; a gas holdup of bubbles in the received digital images; a gas velocity of bubbles in the received digital images; and a colour of the received digital images (e.g., a colour of froth and / or pulp present in the image / s). The output of the machine learning model / module may be fed back to the server or backend for further processing. The machine learning model may be trained using machine learning techniques (such as, but not limited to NN, CNN, DNN, FCDNN etc.) to improve accuracy of its predictions and / or the accuracy of the identified parameters.

[0086] The system 10 may hence include the server 14 which may be arranged to receive digital images captured by one or more camera / s 16.1 , 16.2, 16.3 at the flotation cell / s 12.1 , 12.2, 12.3. The machine learning model may be accessible by the server 14, and it 14 may be trained with one or more training images. The ML model 20 may further be operable to receive the digital images as input and to analyse the received digital images to identify bubbles and to generate data relating to the digital images. The data relating to the digital images, that may be generated, may include data relating to a characteristic of the bubbles, such as a gas characteristic of the bubbles identified in the images. The data relating to the characteristic / gas characteristic of the identified bubbles (or data relating to a fluid characteristic of the bubbles) may be generated by the ML model 20 based on a number of factors described in the present disclosure. The characteristic of the identified bubbles may be derived or estimated based on the analysis by the (trained or pretrained) ML model.

[0087] The ML model 20 may be arranged to analyse the received digital images to identify bubbles and to generate data relating to any one or more of: a bubble size distribution (BSD) 22 of bubbles in the received digital images. Optionally, the BSD may be indicated / measured in millimetres, but other units of measurements may also be used. The BSD may be indicated by the label D32 (also see Figure 12 described in further detail below); a gas holdup 24 of bubbles in the received digital images. The gas holdup may be labelled Eg (also see Figure 12 described in further detail below); a gas velocity 26 of bubbles in the received digital images. Optionally, the gas velocity 26 may be termed a superficial gas velocity, and it may be indicated / measured in cm / s, but other units of measurements may also be used. The gas velocity may be labelled Jg(also see Figure 12 described in further detail below.); and / or a colour 28 of the received digital images. This may either be a colour of a pulp phase of the flotation cell, or the colour of a froth phase of the flotation cell, as will be described in more detail below.

[0088] It will be appreciated that any one or more of these features of the ML model 20 may be implemented by the system 10 either individually or in conjunction with any one or more of the other features of the ML model.

[0089] The ML model may be arranged for generating a reagent dosing parameter 30 based on the analysis by the machine learning model 20. The server 14 may also have a database, e.g., for storing one or more parameters and / or the received image / s. The server may be arranged for outputting the reagent dosing parameter 30 for use in controlling a dosage of one or more reagents in a process associated with the flotation cell / s 12.1 , 12.2, 12.3. The reagent dosing parameter may in certain cases also be termed a reagent dosing set-point. However, it will be understood that the reagent dosing parameter is not limited to a set-point.

[0090] Optionally, the system may include various machines that perform or facilitate mineral processing.

[0091] For example, a frother 32, a collector 34, and a depressant 36 may be implemented. These machines may each have an associated controller, which may receive the generated reagent dosing set-point parameter from the server 14. Optionally, the reagent dosing parameter may be sent by the server 14 to a local computing device 38 of an operator 40. However, it will be appreciated that in some cases the output of the server may be transmitted directly to control equipment relating to the one or more flotation cells. A feed 42 (e.g., an ore feed or other mineral feed including minerals or metals to be processed) may be provided, in communication with a feed control valve or a feed flow pump 44. The pump 44 may be arranged to be controlled, e.g., by adjusting the pump speed (this may, e.g., be automated - i.e. , the pump may be controlled in accordance with aspects of the system). An air supply 46 may be provided, and it may be in fluid flow communication with one or more valves or pumps 48.1 , 48.2, 48.3. These may be operable to control aeration and / or air flow toward each of the flotation cells 12.1 , 12.2, 12.3. In the present version of the system 10, downstream valves or pumps 50.1 , 50.2, 50.3 may be provided, downstream of their respective flotation cells 12.1 , 12.2, 12.3. Tails 52 may exit, or be outlet from, the system 10 downstream of the third / last flotation cell 12.3. Concentrate 54 may exit, or be outlet from, the system 10 downstream of each of the flotation cells.

[0092] Additionally, or alternatively, the system and method of the present disclosure may include generating a gas set-point parameter based on the analysis by the machine learning model 20. The gas set-point parameter may e.g., be output for use in controlling aeration of the flotation cell / s 12.1 , 12.2, 12.3. The gas set-point parameter may be an air set-point parameter 56, e.g., when air is used as an aeration gas to produce bubbles in the process. It will be appreciated that other gases may also be used for aeration or to generate bubbles. One or more controllers may be implemented by the system. For example, an air / gas controller may be used, and / or a level controller may be used. In the present example, a Proportional - Integral - Derivative (PID) 58 controller may be used. The controller / s 58 may, e.g., be in data communication with the pumps and / or valves 48.1 , 48.2, 48.3, 50.1 , 50.2, 50.3. This may, in turn facilitate aeration and / or level control in the flotation cells. The relevant set-point or gas control parameter may also be generated by the ML model and / or by the server, e.g., in response to the analysis of the received image data by the ML model 20.

[0093] The reagent dosing parameter 30 may be utilised by the system to control the addition of one or more reagent / s to the process. This may be performed automatically, e.g., by way of a controller causing the frother 32, collector 34 or depressant 36 to release reagent / s into (and / or mixed with) the stream of processed material / minerals (which may be termed a feed stream 42). In other words, control signals may be transmitted to these machines based on the analysis of the ML model, and the outputted reagent dosing parameter 30. The reagent may in certain cases be termed a soap or another additive or substance that may be introduced so as to increase or decrease, or have an effect on, the sizes of bubbles formed in the flotation cell / s. In other words, feedback data or received digital images of bubbles may be analysed by the ML model 20 to determine whether the bubbles forming (whether in the pulp phase, or in the froth phase, or both) are either too big, or too small, or whether the BSD is either too large or too small, and the reagent parameter may be adjusted or generated dynamically in near real-time to address this.

[0094] Additionally, or alternatively, the air / gas parameter 56 generated by the ML model 20 (e.g., in conjunction with the server 14) may be used to adjust the aeration or gas supply to each of the flotation cells, e.g., via electrically controlled pumps or valves 48.1 , 48.2, 48.3. In other words, control signals may be transmitted to these pumps or valves based on the analysis of the ML model, and the outputted air / gas parameter 56 (e.g., via controller 58).

[0095] In other words, feedback data 60 may be provided by the system to the server, for example in the form of image data captured by the one or more camera / s. The server may also be termed an Advanced Process Control (APC) system in accordance with exemplary aspects of the present disclosure.

[0096] The system 10 described above may implement a method for monitoring a flotation cell. An exemplary method 100 for monitoring a flotation cell is illustrated in the swim-lane flow diagram of Figure 2, in which respective swim-lanes delineate steps, operations or procedures performed by respective entities or devices. The method may include, providing a camera at the flotation cell. In the example system of Figure 1 , a camera 16.1 , 16.2, 16.3 is provided for each flotation cell, and the camera may be submerged into the pulp phase. Optionally a camera 62 or sensor (see Figure 11) may also be provided for each of the flotation cells, and it may be capable of capturing images of a flotation phase. In other words, one or more cameras may capture images of the pulp phase, and one or more cameras may capture images of the froth phase of each flotation cell, and image data of one or more of these cameras may be fed / transmitted to the server 14 for further processing.

[0097] The method 100 may include, capturing 110 digital images of bubbles by one or more of the camera / s 16.1 , 16.2, 16.3 or 62. The digital image / s may be transmitted 112 to the server 14 where they may be received 114. The images may optionally be received as a video feed and / or over a communications network such as the Internet. These image / s may be received in near real-time and or as a live or near-live video feed at the server. The server 14 may be arranged to access 116 the machine learning model 20 (which may be trained 118, or pre-trained, with one or more training images, e.g., of bubbles). In other words, the method may include inputting the digital images into the machine learning model 20, and by the machine learning model, analysing 120 the received digital images to identify bubbles. The method 100 may further include generating 122 (by the ML model 20 and / or by the server 14) data relating to a characteristic I gas characteristic of the bubble / s. The method 100 may include generating data relating to any one or more of: a bubble size distribution 22 of bubbles in the received digital images; a gas holdup 24 of bubbles in the received digital images; a gas velocity 26 of bubbles in the received digital images; and / or colour 28 of the received digital images.

[0098] Optionally, the method 100 may include generating 124 a reagent dosing parameter 30 based on the analysis by the machine learning model 20. The method 100 may further include outputting 126 the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell / s 12.1 , 12.2, 12.3.

[0099] The method 100 may include monitoring a plurality of flotation cells 12.1 , 12.2, 12.3. Outputting 126 the reagent dosing parameter may include outputting it for use in dosage of reagents in any one or more of the plurality of flotation cells. Outputting 126 the reagent dosing parameter may include outputting a reagent dosing set-point parameter. Additionally, or alternatively, the method 100 may include generating a gas / air set-point parameter based on the analysis by the machine learning model 20. The gas / air set-point parameter may e.g., be output for use in controlling aeration or bubble forming (by introduced gas / air) in the flotation cell / s 12.1 , 12.2, 12.3.

[0100] The reagent dosing parameter 30 and / or the air or gas set-point parameter 56 may optionally be outputted to a system operator 40 that operates a local computing device 38. The local computing device 38 may e.g., be connected to one or more controller / s or machines, such as one or more of the frother 32, collector 34, depressant 36, or a mineral separator machine or a controller associated with any one or more of the machines, pumps or valves of the system 10.

[0101] The air or gas set-point parameter 56 may for example be outputted by the server to the controller 58 associated with air or gas flow of the process associated with the flotation cell / s. Optionally, the air or gas set-point parameter may be outputted by the server to the local computing device 38 of the operator 40, wherefrom it may be relayed to the controller associated with air or gas flow of the process associated with the flotation cell / s. Optionally, the local computing device 38 may act as a controller for either reagent dosing, or aeration / gas control. The method 100 may optionally include outputting the air set-point parameter 56 so as to enable automated adjustment of aeration in the process associated with the flotation cell / s. The method 100 may further include outputting the reagent dosing parameter 30, or outputting the reagent dosing set-point parameter, so as to enable automated adjustment of reagent addition to the process associated with the flotation cell / s. The method 100 may include outputting data by the machine learning model may be utilised as an input to a frother 32 level controller (which may also be termed a level controller), so as to enable control of a level of froth in the flotation cell / s 12.1 , 12.2, 12.3. The method may include providing one or more controller / s for controlling one or more machines associated with the flotation cell / s. The server 14 and / or the local computing device 38 may be in data communication with the one or more controllers, and arranged to output the reagent dosing parameter and / or the air / gas set-point parameter to the one or more controllers.

[0102] The method 100 may include, by the machine learning model 20, generating data relating to a bubble surface area flux, e.g., based on the analysis of the received digital images. Any data outputted by the machine learning model 20 may additionally or alternatively be based on an analysis of the bubble surface area flux.

[0103] The method 100 may include, utilising a probe inside each of the one or more flotation cell / s. An enlarged schematic diagram of one of the flotation cells 12.1 is shown in Figure 3. The probe 13 may be at least partially submerged with a viewing pane 64 and an operative end of the probe inside the flotation cell 12.1. The probe 13 may include the camera 16.1 , and at least one light source 67. The viewing pane 64 may be positioned at an oblique angle relative to vertical. This may inhibit fouling, and it may enable high resolution images to be captured by the camera 67 which are then transmitted in near real-time to the server 14. The method may include, by the camera 16.1 , capturing intermittent digital images of bubbles inside the flotation cell 12.1 under illumination from the light source 67 through the viewing pane 64. The method may further include, transmitting the captured digital images to the server for further processing. It will be appreciated that the further flotation cells and probes may be similar to that of Figure 3.

[0104] In Figure 4 is shown an exemplary photograph of an image of bubbles taken by a test probe in accordance with the present disclosure under laboratory conditions. In Figure 5 is shown an exemplary photograph of an image of bubbles taken by a test probe in accordance with the present disclosure under production and / or under field conditions. These are merely examples of photographs that may be transmitted digitally to the server for further processing.

[0105] Various components may be provided for implementing the method and system described above with reference to Figures 1 and 2, or any of the other systems or methods of the present disclosure. The local computing device, and / or controller / s (e.g., 58) may each include a processor for executing the functions of components described herein, which may be provided by hardware or by software units executing on the local computing device or controller. The software units may be stored in a memory component and instructions may be provided to the processor to carry out the functionality of the described components. In some cases, for example in a cloud computing implementation, software units arranged to manage and / or process data on behalf of the local computing device 38 or controller / s 58 may be provided remotely.

[0106] The system / s and method / s of the present disclosure may be referred to as a reagent advisor, or an online flotation pulp sensor control system / method. Figure 6 shows Advanced Process Control (APC) automation features that may be implemented by aspects of the present disclosure, for example by the server 14 and / or ML model 20. The system 10 may be arranged to take pulp measurements and / or pulp images that are analysed by the backend (e.g., by the server 14 and / or the ML model 20). Air Proportional - Integral - Derivative (PID) setpoints may be determined and utilised for process control, and air valve adjustments may be performed based on data received from the APC backend 14, 20, and based on ML 20 analysis of gas hold-up and / or superficial gas velocity.

[0107] Frother Dosing and Dosing Pump Speed adjustments may be performed based on APC data received from the backend 14, 20 based on BSD analysis by the ML model. Examples may include Methyl Isobutyl Carbinol (MIBC) or Polypropylene Glycol Methyl Ether (PPGME).

[0108] Pulp colour may also be analysed by the APC backend 14, 20, and Valuable Mineral or Mineral of Interest Inferential may be used by the system to generate and output data relating to Collector Dosing and Pump Speed adjustments. Examples may include Xanthates and / or Dithiophosphates.

[0109] Moreover, pulp colour may be analysed and processed by the APC backend, 14, 20 in relation to a Gangue Mineral or an Unwanted Mineral Inferential, so as to generate and output data relating to Depressant Dosing and Pump Speed adjustments. Examples may include Lime (Calcium Hydroxide, Ca(OH)2), Sodium Cyanide (NaCN), Zinc Sulphate (ZnSO4), and / or Sodium Sulphite (Na2SO3).

[0110] The system 10 may implement a probe and camera (also termed a pulp sensor in some instances) submerged inside the flotation cell (e.g., inside the pulp phase). This is illustrated in Figure 8. An example image captured by the camera is shown in Figure 10 from which live (real-time or near real-time) bubble information can be derived by the system. Alternatively, or in addition, a probe and / or camera 62 may be arranged above the froth phase as is diagrammatically illustrated in Figure 11. Images captured by this froth phase camera (also, e.g., termed a froth sensor 62) may also be transmitted to the backend or server 14. An example of such an image (of the froth phase) is shown in Figure 9. Figures 9-10 also illustrate machine learning (ML) techniques being applied to these images to generate data relating to a characteristic of the bubbles, or data relating to gas or other fluid inside each of the bubbles that are in the captured image. In other words, the ML model 20 may be arranged to identify and / or detect bubbles in the images, as well as their BSD, velocity, height, colour, size, their outline / s or shape / s, and other sensed or measured parameters or metrics such as mass pull, air recovery, froth stability, grade etc.

[0111] Figure 12 is an exemplary system diagram illustrating exemplary control parameters that may be implemented by the system 10 of Figure 1 , or by any of the other systems or methods of the present disclosure.

[0112] The system 10 may implement one or more of the following formulae in order to determine the Pulp Bubble surface area flux [Sb]:

[0113] > ^gas

[0114] £g ~ vgas - ~I- Tv7 slurry

[0115] Where Egis the Pulp Gas hold up;

[0116] Where Vgasis the volume of the bubbles;

[0117] Where Vsiurryis volume of the slurry (pulp and bubbles);

[0118] Jg= ug£g

[0119] Where ugis the Pulp Bubble velocity;

[0120] Where Sb is the Pulp Bubble surface area flux

[0121] Where d32 is the Pulp Bubble size; and

[0122] Where jgis the Pulp Superficial gas velocity.

[0123] These parameters and variables may be calculated and / or estimated by the backend, server 14 and / or ML model 20. Essentially, Sb, i.e., the Pulp Bubble surface area flux may be a direct measure of pulp zone flotation efficiency. Figure 12 may demonstrate how the present disclosure may facilitate an increase in true flotation while also implementing a reduction in entrainment of the flotation cell / s.

[0124] The present disclosure may enable an increase in mineral recovery / efficiency with APC by controlling the below in real-time or near real-time (e.g., by way of the system 10): Gas hold-up [Eg]: Volume of gas in the pulp zone. This may be used by the system 10 and backend 14, 20 to control residence time available for flotation;

[0125] V Bubble size [D32]: May be critical to generate bubbles of the correct diameter. Smaller bubbles are generally required for fine particle flotation;

[0126] V Superficial gas velocity [Jg]: Upward velocity relative to the cell cross-sectional area. This may provide information to the system 10 (and backend 14, 20) on the gas dispersion efficiency and early warning of agitator mechanism wear. Excessive air addition may increase bubble size and lower air dispersion efficiency. In other words, controlling the air rate within an optimal range by the system may be implemented.

[0127] V Bubble surface area flux [Sb]: Bubble surface area rising up a flotation cell per cross- sectional area may be implemented by aspects of the system 10 (and backend 14, 20).

[0128] Figure 13 is a high-level block diagram illustrating exemplary aspects of the present disclosure which may enable an increase in mineral recovery and an increase in flotation, while alleviating lower grade recovery and alleviating high entrainment in the flotation cell. In other words, the figure explains why it may be advantageous to implement pulp sensor / s 16.1 , 16.2, 16.3 into the system 10. The APC backend (server and ML model) may be implemented to facilitate reagent addition by the system 10. This may optimise or increase the efficiency of mineral recovery by more effective attachment of minerals to bubbles. This may reduce entrainment and increase true flotation and improve recovery.

[0129] The system 10 illustrated in the exemplary block diagram of Figure 1 may implement (preferably automated, or partially automated) reagent dosing features, which may also be termed a reagent advisor. An exemplary reagent advisor process flow may be implemented by the system 10 as follows:

[0130] 1 . Ingest, or receive at the server 14, plant online data (e.g., densities, flow rates, pulp levels, aeration rates, reagent feed rates, ore assays, online assays, mill power, mill load, mill grind, pH, agitator current draw, froth velocities, froth bubble size, froth colour, froth states and pulp and froth image / s);

[0131] 2. Combine the received data with pulp sensor data (e.g., superficial gas velocity JG, bubble size distribution, gas holdup, rise velocity);

[0132] 3. Determine an optimal or near-optimal adjustment to reagents such as frother by comparing the above values to expert rules, fundamental and machine learning model / s, e.g., by way of the ML model 20; and / or

[0133] 4. Write out, or output, one or more suggestion / s to APC and / or to client’s (e.g., operator’s 40) supervisory control and data acquisition (SCADA), notification system, e.g., for use in client action or further processing.

[0134] Turning to Figures 14 to 19, a number of exemplary graphs are shown, that illustrate aspects of the present disclosure, for an advisor being active (Figs. 15, 17, and 19), or the advisor being disabled (Figs. 14, 16, and 18). The “advisor” may be referred to as the backend, e.g., including the server 14 and ML model 20. Stated differently, the reagent dosing parameter may be output by the server 14, and this reagent dosing parameter may be implemented by aspects of the system 10, e.g., to control or adjust dosage of reagent / s to components of the system 10, and this may be called “advisor active”. On the other hand, if the advisor is not active, the reagent dosing parameter may e.g., not be outputted, and in such cases the system would operate less efficiently. The advisor active may thus facilitate more efficient operation of the system 10, as can be seen from Figures 15, 17, and 19.

[0135] Referring to Figure 14, there is shown an exemplary graph 1400 illustrating frother concentration and simulated critical coalescence concentration (CCC) for a normal case where reagent dosing advisor aspects of the system 10 are disabled. In other words, the graph 1400 illustrates an exemplary case where frother concentration within the pulp remains too low and subsequently a sub-optimal pulp bubble size distribution is observed.

[0136] Figure 15 is a graph 1500 similar to that of Figure 14, but showing exemplary reagent dosing parameter or advisor aspects of the system 10 enabled, with a corresponding increase in frother concentration, e.g., due to increased reagent feed to address differing operating conditions and as such optimal pulp bubbles size can be maintained. In other words, frother concentration within the pulp may increase due to an increased reagent feed to address differing operating conditions, and as such, optimal or near-optimal, or an improved pulp bubble size may be maintained by aspects of the system 10.

[0137] Figure 16 is a graph 1600 similar to Figure 14, and it illustrates a further development of Figure 14. In this case, the reagent advisor aspects or reagent dosing parameter output of the system 10 may again be disabled. With CCC drift (e.g., due to ore changes, and / or operational disturbances), the frother addition needs to be adjusted otherwise a sub optimal bubble size distribution may be experienced. Figure 17 is a graph 1700 similar to Figure 15, and it illustrates a yet further development of Figure 15. In this case the, the reagent advisor aspects or reagent dosing parameter output of the system 10 may be enabled. In other words, the reagent dosing parameter may be output by the server 14 and ML module 20. With the Advisor active, and with CCC drift (e.g., due to ore changes, and / or operational disturbances), the reagent advisor (e.g., the system 10 and ML model 20) again inspects all plant and sensor variables and it may be determined that the froth dosage needs to change leading to optimal (or improved, or near-optimal) pulp bubble size / s.

[0138] Figure 18 is a graph 1800 showing a yet further development of Figure 14, with the advisor aspects of the system 10 disabled. With CCC drift (e.g., due to ore changes, and / or operational disturbances) the frother addition may be adjusted, otherwise overdosing of reagents may be possible.

[0139] Figure 19 is a graph 1900 showing a yet further development of Figure 15, with the advisor aspects of the system 10 active. With CCC drift (e.g., due to ore changes, and / or operational disturbances) the reagent advisor features of the system may again inspect all plant and sensor variables, and the system 10 may determine that the froth dosage needs to change, which may lead to optimal (or near-optimal, or improved) pulp bubble size while reducing reagent costs.

[0140] The present disclosure may implement system / s and method / s involving using a camera and lights inside flotation cells to capture images of bubbles in the pulp, which are then analysed to determine bubble size, gas velocity, and gas holdup. This information may be used by the system 10, and this may facilitate optimisation of a flotation process. The probe / s 16.1 , 16.2, 16.3 may be termed Online Pulp Sensor / s (OPS). Feedback from these OPS’s may be used to enhance the Reagent Advisor, which may optimize reagent dosing and monitor reagent addition throughout the entire processing plant. The integration of the OPS into Advanced Process Control (APC) systems may enable real-time adjustments to key process parameters, such as reagent dosing and aeration control, and this may lead to improved mineral recovery.

[0141] A number of technical problems may be addressed by the present disclosure:

[0142] 1. Measurement of Pulp Properties by the system 10: Measuring pulp properties, such as bubble size distribution 22, superficial gas velocity (Jg) 26, and gas holdup 24. The present disclosure may make use of machine vision to measure these parameters in situ, i.e., directly at or in the flotation cell.

[0143] 2. Optimization of Reagent Dosing: The correct addition of reagents may be crucial for the efficient recovery of minerals. Over-dosing or under-dosing can lead to suboptimal recovery rates and increased operating costs. By combining bubble size parameters with frother knowledge, the reagent dosages can be optimised.

[0144] 3. Reagent Management Over Multiple Banks: Managing reagents across the entire processing plant (see e.g., Figure 1) may involve coordinating multiple control points and stages. The present disclosure may provide a holistic approach to reagent management, and this may facilitate that deviations and disturbances are quickly identified and corrected.

[0145] 4. Process Upset Reporting and Alerts: By monitoring critical KPIs, the system 10 can proactively identify anomalies and alert operators 40 to issues, such as dosing irregularities or system disturbances, that need attention. This may enhance the automation and control capabilities of the flotation process / es.

[0146] 5. Integration with Advanced Process Control Systems: The system 10 may be designed to be integrated into APC systems, allowing for seamless real-time adjustments in response to changes in feed characteristics and operational demands. This may enhance the system's ability to maintain efficient and effective flotation processes, which may be crucial for optimizing mineral recovery.

[0147] As mentioned in the background of the present disclosure, traditional offline methodologies for bubble measurement in flotation cells typically involve transferring bubbles from the cell to a separate chamber, often filled with clean water, to facilitate image capture. This process requires valve systems for water management and regular maintenance to address mineral buildup that obscures viewing ports. Furthermore, the displacement of clean water by the extracted air necessitates frequent chamber refilling, and pressure discrepancies between the collection and measurement points must be accounted for due to their different locations.

[0148] In contrast, the present disclosure conducts measurement directly within the pulp of the flotation cell (e.g., 12.1 , 12.2, 12.3), thereby eliminating the need for a separate imaging chamber, reducing maintenance for cleaning, and avoiding the complexities of refilling and pressure adjustment. This may facilitate streamlining the process, allowing for continuous in-situ measurements that are more efficient and representative of actual flotation conditions in the field.

[0149] The present disclosure may offer an advantageous approach to monitoring flotation cells, directly imaging bubbles within the pulp to eschew the need for a separate clean water chamber, eliminating associated valve systems and maintenance. Its capability for continuous and pressure-independent operation enables real-time, precise analysis of bubble size, gas velocity, and holdup without downtime, optimising reagent dosing and flotation efficiency. Integrated with Advanced Process Control (APC) systems, it may deliver seamless, dynamic process adjustments, and this may significantly improve mineral recovery, reducing waste, and enhancing the overall effectiveness of mineral processing operations.

[0150] An optional alternative approach would be to automate refilling and flushing of the chamber by automating valve operation. However, this may result in difficulties to have the device operate reliably all the time, because of the harsh environment. Measurement or sensing inside the pulp may simplify the measurement / sensing / imaging process / es significantly.

[0151] Exemplary APC Input, e.g., input to the server 14 and / or ML model 20:

[0152] Online pulp sensor (OPS) technology may be implemented so as to holistically encompass its pivotal role within Advanced Process Control (APC) systems. The OPS's advent may represent a significant advancement in the field of mineral processing, offering unprecedented real-time insights into pulp properties that may be crucial for the optimization of flotation operations.

[0153] By capturing pulp metrics — such as pulp gas hold-up, superficial gas velocity (Jg), bubble size distribution, bubble surface area flux, and pulp colour — the OPS may provide robust and actionable data that empower APC systems to intelligently and dynamically adjust key process parameters. Such agility in response is pivotal for maintaining efficient and effective flotation processes in the face of fluctuating feed characteristics and operational demands.

[0154] The integration of the online pulp sensor OPS into APC systems may be implemented for fine- tuning reagent dosing strategies, which may include the precise management of frothers 32, collectors 34, and depressants 36. Notably, the sensor 16.1 , 16.2, 16.3 (and / or 62) may enable granular control of frother levels to adjust the bubble size distribution. This may, in turn, facilitate optimally sized bubbles which may be crucial for effective mineral recovery. This dynamic adaptability may further be facilitated by real-time colorimetric data from the OPS (or camera 16.1 , 16.2, 16.3, 62), which reflects changes in mineral composition, paving the way for automated reagent dosing tailored to the nuanced requirements of the ore in process. The sensor's essential role in elevating APC capabilities may be comprehensively safeguarded. The scope of the present disclosure may enable the OPS's (or other sensor / camera’s) contributions to orchestrating the intricate balance of flotation parameters, particularly the synergy of frother dosage with bubble size measurements and the broader spectrum of reagent management. This may encapsulate the OPS's influence on the APC strategies aimed at achieving improved mineral recovery.

[0155] Moreover, the technology of the present disclosure may be instrumental in facilitating automated adjustments to aeration. By continuously adjusting aeration (e.g., based on the outputted air set- point 56 from the server 14 / ML model 20) in an effort to maintain superficial gas velocity within an optimal range, the APC system 10 may optimise air recovery, enhancing flotation hydrodynamics. Aeration control may form a fundamental aspect of advanced process controls, as it can significantly impact the flotation efficiency and recovery rates. The OPS (sensor or camera 16.1-3 or 62) may play a key role in regulating this variable. By providing instantaneous Superficial gas velocity (Jg) measurements, the APC system can optimize aeration rates on a cell-by-cell basis, altering the airflow to reduce entrainment of gangue and increase the selective recovery of valuable minerals.

[0156] In the technology of the present disclosure the OPS's (or sensor / s or camera / s of the system 10) may function as part of an integrated system that closes the loop with APC, capacitating minute- by-minute, closed-loop optimizations via a combination of advanced algorithms and real-time data feed (or near real-time data feedback received by the server 14 and ML model 20).

[0157] The sensor's (e.g., 16.1 , 62) design, may feature a specialized camera 16.1 and lighting arrangement 67. This may facilitate or ensure that the high-definition images required for this level of control are captured consistently, even in the challenging environment of a flotation cell.

[0158] The present disclosure may enable a methodology by which the OPS aids in dynamic process optimization via APC and the detailed mechanism through which real-time data impact on reagent dosing, aeration control, and overall flotation efficiency. Both the hardware elements and the advanced software components of the OPS within APC applications may be implemented, and the present disclosure may enable operators 40 (or service providers, e.g., operating the server 14 and / or ML model 20) to obtain a competitive edge in the realm of mineral processing technology.

[0159] Exemplary Reagent Advisor (RA) Inputs:

[0160] Features of the system 10 may be referred to as Reagent Advisor (RA) features. It will be appreciated that the present disclosure extends to other outputs generated by the server, so it will be understood that RA may refer more broadly to the system 10.

[0161] Online Pulp Sensor (OPS) technology may be augmented to include the Reagent Advisor (RA), which may not only optimize reagent dosing within flotation operations, but it may also serve as an advanced monitoring system 10 for comprehensive reagent management across the whole plant. These features may capitalize on real-time or near real-time sensor / camera / OPS data to enhance the precision and efficacy of mineral processing activities. Through the integration of the Reagent Advisor, the system 10 may gain the ability to monitor reagent addition throughout the entire processing plant. By analysing data streams from various control points, the Reagent Advisor can continuously evaluate the status of reagent dosing across different stages of the flotation process. Its sophisticated software algorithms may enable it to detect any deviation from expected dosing patterns, thus identifying under-dosing or over-dosing events, along with other disturbances or faults that could impact process outcomes.

[0162] Beyond optimizing dosing in real-time, the Reagent Advisor may be tasked with exception reporting. This functionality may allow the system 10 to maintain routine operations without intervention but provides alerts or notifications when parameters fall outside predefined thresholds. These exceptions might include significant changes in bubble size distribution, gas hold-up, superficial gas velocity (Jg), or other critical metrics that could indicate a dosing irregularity or system disturbance.

[0163] By consolidating the data across all sensors and process points, the Reagent Advisor 10 can also generate an overview report of plant-wide reagent use, highlighting areas where adjustments have been made and where further attention may be required. This may streamline decisionmaking processes for plant operators, overseeing reagent application with precision and delivering actionable insights only when necessary.

[0164] Incorporating the Reagent Advisor into the OPS may extend the potential of the technology to not only measure and optimize but also monitor and report on aspects that may be critical to the flotation process. The present disclosure may reflect the inclusion of an intelligent notification system within the OPS framework, aiming to proactively identify, communicate, and rectify exceptions in the dosing process, thereby aiding continuous process optimization and system integrity over the long term. These features may signify an evolutionary step in the application of mineral processing technologies, offering a more nuanced and responsive approach to reagent management and overall plant performance.

[0165] The technology of the present disclosure may aid monitoring and sustaining operations. Counter measures or actions may be implemented to correct anomalies or inefficiencies or errors at the plant / flotation cells. These corrections may be implemented automatically and in near real-time. Hence, the system may identify plant inefficiencies and automatically cause them to be improved by outputting one or more of the reagent dosing parameter / s 30 and the air / gas parameter or set-point 56. Monitoring and diagnostics may be implemented in an automated fashion by the system 10. Features of the system may include Data Science Services & Offline Studies, Training Courses for Process Data Analytics Powered by Domain Knowledge, Experience, Flotation Simulator / s and an Artificial Intelligence or Machine Learning technology called StoneGPT which may be a technology implementing Generative Pre-Trained Transformers which may form part of the ML model 20 in accordance with exemplary aspects.

[0166] Exemplary components of the system 10 may include one or more of the following:

[0167] 1 . Sensor data health monitoring and tracking dashboard;

[0168] 2. Live video streams of smart sensors with analytics overlays highlighting suboptimal cells;

[0169] 3. PID + APC Performance Monitoring: Detect & diagnose poorly performing controllers;

[0170] 4. Froth States: Cell not producing states (below lip, static, etc.), dry froth and cell not optimal states (pulping, wet froth, fizzy, etc.);

[0171] 5. Deep Dives: Data Science Services; Custom data-driven investigation for process troubleshooting;

[0172] 6. Online Mass-Balance: Critical stream information displayed near real-time; What-if- analysis;

[0173] 7. Mass-Pull performance tracking and insight;

[0174] 8. Flotation Performance Overview: Overall Flotation Performance Dashboard with Insights; KPI tracking, Process shifts and event detection alerts with advisories;

[0175] 9. Reagent Advisor: Optimal reagent dosage for ore type & operational targets; data driven insight(s); and / or

[0176] 10. Flotation Simulator: Model for fast prototyping of solutions; stepping-stone towards a future real-time model.

[0177] Exemplary actions that may be implemented by the system 10 may include any one or more of the following:

[0178] Flotation Reagent Decision Tree and modelling output / s;

[0179] - All inputs may be aggregated;

[0180] Trends and anomalies may be generated;

[0181] These trends o anomalies may be compared to expected and desired behaviour;

[0182] Notification / s may be sent (e.g., by the server 14) based on abnormal events;

[0183] - An action case may be opened for information tracking; and / or Suggestion / s on reagent dosage levels may be outputted.

[0184] It is also envisaged that a Reagent Simulator may be implemented by the system 10. The Reagent Simulator may provide a tool which can enable operators to investigate how different pulp level control responds, how will froth height respond, how will tailings gate valve / s (e.g., 50.3) respond, etc. It may also, through addition of frother reagent / s impact all the factors mentioned previously. The Reagent Advisor system 10 may make use of the sensing instrument / s 16.1 which may be in the form of a near real-time pulp sensor, enabling it to calculate unique derivatives for optimal froth dosage (i.e. monitoring and / or controlling one or more reagents in flotation). Because of the (near) real-time nature of the instruments and calculations these setpoints can be provided in (near) real-time and integrated with an Advanced Process Control system. Similarly, the Reagent Advisor system 10 may make use of pulp sensor / s 16.1 , 16.2, 16.3 measurements, to, when combined with the froth sensor 62 (e.g., see Fig. 11), enables it to calculate optimal or near- optimal collector 34 dosing. The present disclosure may enable a Reagent Dosage Recipe Model that takes ore changes, process states and the froth and pulp sensor metrics into account to advise on reagent dosage to ensure near optimal recovery using live or near-live data. A User Interface dashboard may be provided, e.g., to operator 40 with prescriptive analytics for reagent dosage recipe decision support. Heuristics and codified knowledge base may be implemented by the system to give / output data driven insight / s on flotation chemistry. Pro-active froth and pulp metrics may be implemented by the system so as to indicate reagent opportunity or issue / s. This may optionally provide a single source of truth regarding reagent performance. Data driven insights and adjustment recommendations may be provided.

[0185] Referring to Figure 20, there is shown a high-level block diagram 2000 illustrating exemplary froth state features that may be implemented by the system 10 of Figure 1. A broad overview 2010 is shown. The ML model 20 may receive 2012 data, e.g., from froth sensor / s, e.g., such as image data 2014. This may be termed real-time froth state detection 2016, but this data may also be received by the server 14 in near real-time. An exemplary process flow diagram is shown for a case where a cell (e.g., one of the flotation cell / s 12.1 , 12.2, 12.3) is not producing 2018. This may be detected by the system 10, and it may be e.g., due to a “below lip froth” 2020 scenario, e.g., when the froth is detected or estimated by the system 10, by way of the received image data, to be below a threshold vertical level. Alternatively, or in addition, the cell may not be producing due to static froth 2022 being detected, or due to the detection of the system of froth flowing away 2024 from a lip of the cell. This state 2018 of the cell not producing may be automatically detected by the ML model 20 and the server 14, and any one or more of the cell / s 16.1 , 16.2, 16.3, etc., may be assigned this state if these froth characteristics are detected.

[0186] A further state 2026 may be assigned or detected by the system 10, e.g., if the cell is not optimal, or not optimally producing. This may, e.g., be due to pulping froth 2028; fizzy froth 2030; wet froth; or unstable froth velocity 2032 being detected or estimated by the ML model 20 or Froth Sensor outputs by way of the received image data. One or more deliverables 2034 (also termed output / s) may be provided by the system 10, e.g., when one or more cell / s are assigned either to the cell not producing state 2018, or to the cell not optimal state 2026. The froth state / s may be provided to a SCADA system via tag / s where appropriate. Froth state / s may be displayed, e.g., via one or more overlays, or the state / s may be superimposed over live video streams or feeds as shown at 2038. One or more of these features may for example be made available to the client or operator 40, e.g., via their computing device 38. A dashboard of froth state / s may be provided 2040 by the system 10. Froth states advisories 2042 may also be output by the system 10, e.g., by outputting advisory actions (either as suggestions or as automatically implemented actions), or by outputting the reagent dosing parameter / s in accordance with aspects of the present disclosure. The reagent dosing parameter / s may be outputted, e.g., by the Advanced Process Control (APC) system.

[0187] Site integration 2044 may also be implemented. Site actions 2046 that may be outputted by the system 10 and the ML model 20 may include any one or more of:

[0188] Pulp level I Airflow Set-Pont Adjustment;

[0189] Instrumentation I Equipment Maintenance;

[0190] Regulatory Controller T uning;

[0191] APC Implementation;

[0192] Clean cell buildup;

[0193] Investigate upstream disturbances; and / or Reagent or grinding recipe adjustment.

[0194] These are exemplary site actions that can, e.g., follow after the froth states advisories are outputted to site. These may be the actions that site personnel can action (e.g., operator 40 using computing device 38, or another operator using control system / s at the location of the flotation cell / s), or the APC may automatically control or implement one or more of these actions.

[0195] One or more outcomes 2048 may be achieved, or aimed to be achieved, by the implementation of the system 10 and the ML model 20 of the present disclosure. These outcomes may include any one or more of:

[0196] Close to 100% (or near-optimal) froth production;

[0197] Reduced Time spent in undesired froth state;

[0198] Representative insight on cell availability;

[0199] Real-time (or near real-time) advisory / ies;

[0200] Decreased time to action; and / or

[0201] Giving back time to process personnel.

[0202] Referring to Figure 21 , there is shown a high-level block diagram 2100 illustrating exemplary pulp state features that may be implemented by the system 10 of Figure 1. A broad overview 2110 is shown where the ML model 20 receives 2112 data and / or captured images (e.g., image 2114 that may e.g., form part of a video feed) from pulp sensor / s. One or more pulp state / s may be detected 2116 by the system 10 in near real-time or in real-time. Examples 2118 of pulp state / s that may be detected include: 2120 Size class & Bubble size distribution; 2122 Bubble Loading; 2124 High fraction of wasted air; 2126 Unstable pulp (e.g., erratic pulp); and / or 2128 Above pulp / froth interface (e.g., when this interface is above a certain vertical threshold level).

[0203] One or more pulp sensor “health” states may be generated 2130 by the system 10, e.g., 2132 No bubbles are present or detected or identified by the system 10; and / or buildup on the sensor 2134 (e.g., of fouling or another obstruction or buildup of sediment) is detected by the system. Buildup of material may sometimes occur on the pulp sensor camera view, causing incorrect readings. This may form part of a state to indicate the sensor “health” or camera lens “heath”. No buildup, or minimal buildup is preferred.

[0204] One or more deliverables 2136 (also termed output / s) may be provided by the system 10. The pulp state / s may be provided 2138 to a SCADA system via tag / s where appropriate. Pulp state / s may be displayed, e.g., via one or more overlays, or the state / s may be superimposed over live video streams or feeds as shown at 2140. One or more of these features may for example be made available to the client or operator 40, e.g., via their computing device 38. A dashboard of pulp state / s may be provided 2142 by the system 10. Pulp state / s advisory / ies 2144 may also be output by the system 10, e.g., by outputting advisory actions (either as suggestions or as automatically implemented actions), or by outputting the reagent dosing parameter / s in accordance with aspects of the present disclosure.

[0205] Site integration 2146 may also be implemented. Site actions 2148 that may be outputted by the system 10 and the ML model 20 may include any one or more of:

[0206] Airflow Set-Point Adjustment;

[0207] Instrumentation / Equipment Maintenance;

[0208] Regulatory Controller T uning;

[0209] APC Implementation;

[0210] Investigate upstream disturbances; and / or

[0211] Reagent or Grinding Recipe Adjustment.

[0212] One or more outcomes 2148 may be achieved, or aimed to be achieved, by the implementation of the system 10 and the ML model 20 of the present disclosure. These outcomes may include any one or more of: Increased time in optimal (or near-optimal) production state;

[0213] Reduced Time spent in undesired pulp state;

[0214] Significant reduction in wasted air;

[0215] Optimal (or near-optimal) bubble size distribution in pulp phase;

[0216] Real time (or near-real time) advisory / ies;

[0217] Decreased time to action; and / or

[0218] Giving back time to process personnel.

[0219] In Figure 22 is shown an exemplary flow diagram illustrating exemplary froth state advisories expert rules 2200 that may be implemented by the system of Figure 1 , and / or by the ML model 20. For a not producing froth state (e.g., 2018 in Fig. 20 described above), one or more observations may be performed by implementing an expert rules decision logic. This may include receiving a notification at the server 14 of the relevant froth state after which a detected cause may be estimated or determined 2210. One or more recommended action / s may then be output 2212 by the system 10. For example, if an underperforming mill is detected 2214, an exemplary output of the ML model 20 may be an advisory indicating investigation of mill control or performance 2216. If a manual operation of the cell is detected 2218, the system 10 may output an actionable advisory 2220 such as: switch cell to auto control mode if mechanical issue is fixed. If bank feed inconsistencies are detected 2222, the system may output an action such as 2224 Investigate upstream feed supply sump / tank control. If 2226 Faulty level instrumentation, 2228 or valve obstruction / valve leaking is detected by the system 10, a recommended action 2230 such as: Inspect level sensor and associated level equipment, or Inspect valve I Dart valve for obstructions I leakages may be outputted.

[0220] Air supply inconsistencies 2232 may be detected, in which case the action output may be 2234 Inspect blower / air supply for damages. If 2236 Possible air line leaks I Air valve faulty is detected, an output action 2238 may be: Inspect air supply lines and air flow valve. If 2240 it is detected by the system that a Cell mechanism has a mechanical issue, an exemplary output action 2242 may be: Investigate cell mechanism and repair mechanical issues.

[0221] If 2244 Control limits not optimal; and / or if 2246 Controller too aggressive / slow acting; and / or if 2248 valve stiction is detected; the system 10 may output a recommended action 2250 such as: Review control variable limits (LL and HH) and APC controller parameters I Also check for valve stiction.

[0222] If 2252 an unidentified cause is detected, the system may output 2254 an advisory recommended action such as: The exact cause was undetermined, possible causes include: Ore characteristic changes, level and air set points too low or reagent dosage changes.

[0223] In Figure 23 is shown a high-level flow diagram of Froth State Advisories Expert rules, showing exemplary further details of the exemplary Expert rules of Figure 22. As before, a notification of the froth state may be received at the server 14. One or more observations 2310 may be performed by the ML model, one or more causes may be detected 2312, and one or more recommended actions may be outputted 2314 by the system 10.

[0224] One or more examples of observations of the system may include: is 2316 Mill power within control range more than 90% of time or not?; is 2318 a Cell in Manual operation mode?; is 2320 a Control variable (level PV) outside a +-5% range?; is 2322 Control variable (Air PV) outside +-5% range?; is 2324 Cell (motor) current deviating from historic values?; is 2326 Air supply (Blower) pressure on target?; is 2328 Manipulated variable (Air valve) constantly saturated?; is 2330 Manipulated variable (Flow valve) constantly saturated?; is 2332 Is the bank feed within control limits?.

[0225] One or more examples of detected causes, i.e., detected by the system 10 may include: an underperforming mill 2334; manual operation 2336 of one or more cell / s; an unidentified 2338 cause; a cell mechanism has a mechanical issue 2340;

[0226] Air supply inconsistencies 2342;

[0227] Possible air line leaks I Air valve faulty 2344;

[0228] Controller too aggressive / slow acting 2346;

[0229] Control limits not optimal 2348;

[0230] Valve Stiction 2350;

[0231] Bank feed inconsistencies 2352;

[0232] Faulty level instrumentation 2354;

[0233] Valve obstruction / Valve leaking 2356.

[0234] One or more examples of recommended actions that may be outputted by the system 10 may include:

[0235] Investigate Mill control / performance 2358;

[0236] Switch cell to auto control mode if mechanical issue is fixed 2360; The exact cause was undetermined, possible causes include: Ore characteristic changes, level and air set points too low or reagent dosage changes 2362;

[0237] Investigate cell mechanism and repair mechanical issues 2364;

[0238] Inspect blower / air supply for damages 2366;

[0239] Inspect air supply lines and air flow valve 2368;

[0240] Review control variable limits (LL and HH) and APC controller parameters

[0241] Also check for valve stiction 2370;

[0242] Investigate upstream feed supply sump / tank control 2372;

[0243] Inspect level sensor and associated level equipment Inspect valve / Dart valve for obstructions / leakages 2374.

[0244] An exemplary use case of the froth state dashboard showing cell availability is shown in the diagram 2400 of Figure 24. If 2410 the cells generating a concentrate stream, then the system may detect that the relevant cell is okay 2412. If 2410 the cell is not generating a concentrate stream, then the system may detect that the cell is not producing 2414.

[0245] One or more management metric / s may be provided by the dashboard of the system 10:

[0246] Better understanding of production; engineering performance assisting; and / or planning and maintenance.

[0247] This may be relevant to operator / s 40 or client / s, e.g., because of one or more of the following: Cell availability can be an important metric to understand, and it can quantify flotation process stability;

[0248] Increased stability may lead to an improvement in overall plant recovery;

[0249] Cell availability is historically typically a metric that is obtained manually and tediously: a metallurgist or similar process personnel will have to physically walk to a flotation section to observe the cell availability. In other words, without implementing the system of the present disclosure, Cells would need to be observed one by one, at a point in time. The present disclosure may improve or alleviate these disadvantages of known systems and methods.

[0250] One or more benefits of the present disclosure may include:

[0251] Reduction in number of on site, in person checks required: in an exemplary case, the present disclosure may provide an opportunity to save a large number (e.g., 66 hours per month per resource or more);

[0252] Incorporation of near real-time data: decreased time to action, information available daily across all shifts. Dashboard view may enable a continuous, overall view.

[0253] Remote monitoring of flotation cells may facilitate decreased travel time, and / or standby access.

[0254] An unbiased metric calculation may be provided by the system 10 (e.g., less prone to human error).

[0255] The system may facilitate a better use of Metallurgist time: e.g., providing more time to focus on high priority tasks.

[0256] An exemplary chart of percentages of time in various froth states per bank is shown in Figure 25. The system 10 and the ML model 20 may determine if a cell is okay, if a cell is not producing, if a cell is offline, or if a sensor reading is unavailable. This data may be presented in a chart similar to that of Figure 25, data relating to which may be outputted to the operator computing device 38 (as with the other graphs, charts or diagrams in the drawings).

[0257] The present disclosure may facilitate more productive flotation cells, ideally closer to 100% productive. The following problems may be alleviated by the present disclosure:

[0258] Non-producing (not pulling) cells impacting recovery;

[0259] Not having real-time visibility on non-producing cells;

[0260] Manual inspection and calculation of cell availability which is time intensive and only a snapshot in time. No insight over entire period would thus have been provided;

[0261] Unclear insights on root cause / s and required action / intervention for nonproducing cells; Prioritization and resolution of issues can be difficult and very time consuming and may only be addressed ad-hoc by known systems.

[0262] The present disclosure may enable system / s and method / s that can facilitate close to 100% froth production. This can give back time to process personnel, and it can facilitate a shortened mean time to do repair / s. The systems and methods can also help with fault finding and giving a more representative view of the reality on site at the flotation cells. This may allow operators to keep plant running closer to designed residence time. A reduced time spent in undesired froth states may be facilitated by the system. Real-time or near real-time visibility and automated alerts on non-producing cells with associated actionable advisories may be provided. Decreased time to action, with clear directive may be provided.

[0263] The systems and methods of the present disclosure may provide the following advantages:

[0264] Using froth sensor and process data;

[0265] Near real-time froth states detection;

[0266] Not producing froth state - either below lip, static froth or froth flowing away from the lip; Data output with OPC integration; Automated notifications (SCADA I Live feed, etc);

[0267] Automated online advisories on worst performing cells; and / or Ongoing support services and performance tracking reports.

[0268] In Figures 26-30 are shown an exemplary charts and graphs showing overall masspull performance metrics that may be implemented by the system 10.

[0269] One or more insights may be determined by the system from the data in these Figures may include:

[0270] A 13% increase in Mass Pull (MP) time in state (within 5% of target) is noted, see Figure 26;

[0271] An Increase in overall MP performance due to SP increases in Rougher Bank A and B can be noted (see Figure 27) from control charts (see Figure 29);

[0272] Sump A shows poor SP tracking and unable to reach target (see Figure 30).

[0273] One or more actionable advisories that may be implemented by the system 10 include:

[0274] Compare process conditions between last 2 weeks and this week to identify cause of time within 5% of target to increase by 13%.

[0275] Due to positive improvement in M P time in state noted when Rougher Bank setpoints were changed, site is advised to remain at a setpoint of 1.20 for Rougher Mass pull.

[0276] Exemplary Correlated effects:

[0277] 1 . Sump A low feed % solids (Insight from site discussion)

[0278] 2. Low feed rate into sump A

[0279] Proposed action: Site is advised to inspect ISA Mill discharge flow and density.

[0280] Figures 31 and 32 are exemplary diagrams that illustrate overall MassPull and froth state performance metrics in accordance with exemplary aspects of the present disclosure.

[0281] One or more insights may be determined by the system from the data in Figures 31-32 may include:

[0282] 1. From the froth states dashboard an exemplary cell (e.g. cell no. FT037) was identified as the worst performing cell as it spent 40% time in the not producing state for this week. Historical data captured by the system 10 indicates FT037 spent 10% in a not producing state.

[0283] 2. Investigating level timeseries trends for both periods shows that a setpoint between 65 to 70 for level results in a lower time in not producing state.

[0284] 3. From Analytics images captured by the system of exemplary cell no. FT037 shows the froth is loaded, however, frequently below lip and therefore potential losses are realised. In the images of Figures 31 and 32 there may be a lot of build-up detected by the system, and / or by the ML model 20, e.g., by way of the ML model analysing the received image data from the relevant sensor or camera.

[0285] An actionable advisory that may be outputted by the system 10 based on data relating to Figures 31-32 (and an analysis by the ML model 20) may be as follows: If there are no concerns of cell slimming, the site is advised to increase the level setpoint of FT037 from 60 to 70. Additionally, the site (i.e., the operator 40, or a computing device at the flotation cell / s) may be advised to implement or action a cleaning process of the buildup detected by the system 10.

[0286] In addition to the froth sub-processes described, the froth phase may be further distinguished into discrete zones. These discrete zones may be characterized as dry froth, wet froth, and pulp exposure. These froth zones may demonstrate distinct metallurgical characteristics that impact flotation performance. The dry froth zone may generally be characterized by lower entrainment and reduced water content, which may correlate with higher mineral grade in the froth. In contrast, the wet froth zone may exhibit a higher water content and entrainment, typically resulting in lower mineral grades in the froth. Notably, within the dry froth, frequent bubble burst events may indicate a fragile bubble structure susceptible to rapid breakdown. Transitional phases between dry and wet froth states are identifiable and present opportunities for enhanced process control to optimize mineral recovery.

[0287] Correlating froth structure and characteristics with flotation performance, and achieving consistent, reliable relationships may be facilitated by aspects of the present disclosure, in spite of the complex, dynamic nature of froth behaviour and its interaction with mineral recovery.

[0288] Figures 33-34 are phase diagrams illustrating exemplary advanced froth states that may be detected by the froth sensor (e.g., 62 in Figure 11) which may form part of the system 10 of the present disclosure. This phase diagram delineates the operational regions or zones corresponding to dry froth 3310, wet froth 3312, dry to wet transitions 3314, and high bubble burst rate zone / s 3316 within the dry froth zone for the flotation cell froth phase. Pulping events may form part of wet froth. It graphically represents transitions between these states based on characteristic metrics such as bubble coverage, bubble count, and wetness indicators derived from bubble segmentation analysis. The diagram in Figure 33 may serve as a process visualization tool by indicating the current froth state and its proximity to transition boundaries, thereby providing actionable insights for process control. Operators 38, 40 and automated control systems can utilize data relating to phase diagram / s such as that of Figure 33 to rapidly assess froth condition stability, and to implement control strategies aimed at maintaining the flotation process within the desired optimal froth region to maximize concentrate grade and recovery while minimizing entrainment.

[0289] Figure 34 illustrates exemplary images that may be captured by the camera associated with the froth sensor 62. The captured exemplary images are shown superimposed over the relevant zones, namely the dry froth 3310, wet froth 3312, dry to wet transitions 3314, and high bubble burst rate zone / s 3316.

[0290] In a further exemplary aspect, the disclosure may provide for the deployment of one or more froth sensor / s (e.g., 62 in Figure 11) in conjunction with the system 10. The froth sensor / s may include one or more camera / s. The froth sensor may, for example, be positioned to capture top-down imagery of the flotation cell froth surface. This may enable classification of froth states into the dry froth, wet froth, and pulp exposure zone / s. The froth sensor / s 62 may include an integrated bubble segmentation model which may be configured to analyse the images (e.g., see Figure 34) to extract parameters such as bubble coverage and bubble count. These parameters may be processed, e.g., by the server 14 to derive a wet froth indicator that quantifies relative froth wetness and entrainment. This wet froth indicator may be utilized by flotation control system / s to maintain the froth within an optimal state through near real-time adjustments of air flow, froth or pulp levels, and reagent dosing. By promoting dry froth stability while minimizing wet froth and pulp carryover, the systems and methods of the present disclosure may enhance recovery of high-grade mineral concentrate and / or reduce entrainment of unwanted materials.

[0291] The technology may include, by the server 14, classifying froth states (e.g., 3310, 3312, 3314, and / or 3316) in the received digital images. The froth states may include dry froth, wet froth, and pulp exposure zone / s. The technology may include, by the froth sensor / s 62, utilising a bubble segmentation model accessible by the froth sensor 62. Alternatively, or additionally, the bubble segmentation model may be accessible by the server 14, and / or it may form part of the ML model 20 accessible by the server (or accessible by the froth sensor 62, e.g., via the server 14). The bubble segmentation model may be configured to analyse the captured images to extract parameters including one or more of bubble coverage and bubble count. The extracted parameters may be processed by the server 14 to derive a wet froth indicator that quantifies relative froth wetness and entrainment.

[0292] The froth sensor 62 may incorporate an integrated bubble segmentation model, which may analyse captured images to quantify froth characteristics. Metrics derived from the bubble segmentation, which may include bubble coverage, bubble count, and related parameters, may be processed to compute a wet froth indicator, representing the relative wetness of the froth. This wetness indicator may allow the flotation control system to maintain the froth within an optimal froth state by implementing targeted control actions based on near real-time froth condition monitoring and automated control actions. Specifically, upon detection of wet froth or pulp at the froth interface, control actions may be triggered with the APC controller (e.g., of system 10), or recommendations are generated by the reagent advisor, which may include adjustment of air flow, froth or pulp levels, or reagent amounts.

[0293] By promoting the formation and stability of dry froth while minimizing wet froth and pulp carryover, this enhanced froth state measurement and control capability may improve flotation performance. The integration of froth sensor-based froth state classification with bubble segmentation may enable near real-time control, resulting in maximized recovery of high-grade concentrate and reduced entrainment of unwanted material.

[0294] In addition to the parameters and control features described herein, additional froth states, which may be referred to as advanced froth states (e.g., 3310, 3312, 3314, 3316), may be utilized in advanced process control or the reagent advisor, as described by the phase diagrams in Figures 33-34.

[0295] These advanced froth states (e.g., 3310, 3312, 3314, 3316) may exhibit distinct metallurgical characteristics. The dry froth zone 3310, 3316 may be characterized by low entrainment and reduced water content, and typically corresponds to a higher mineral grade relative to wet froth and exposed pulp. Conversely, the wet froth zone 3312 may be associated with elevated entrainment and higher water content, and correspondingly lower mineral grades. Within the dry froth zone, a high incidence of bubble burst events may occur (e.g., at 3316), indicative of a fragile bubble structure prone to rapid breakdown. Furthermore, transitional phases / zones 3314 between dry froth and wet froth states may be identifiable or detectable by aspects of the technology.

[0296] Advanced froth states can, e.g. be detected using only some of the froth sensor outputs:

[0297] Dry froth - low entrainment, low water content;

[0298] Wet froth - high entrainment, high water content; and / or Possibly relating froth structure to flotation performance.

[0299] Figure 35 illustrates an example of a computing device (3500) in which various aspects of the disclosure may be implemented. The computing device (3500) may be embodied as any form of data processing device including a personal computing device (e.g. laptop or desktop computer), a server computer (which may be self-contained, physically distributed over a number of locations), a client computer, or a communication device, such as a mobile phone (e.g. cellular telephone), satellite phone, tablet computer, personal digital assistant or the like. Different embodiments of the computing device may dictate the inclusion or exclusion of various components or subsystems described below.

[0300] The computing device (3500) may be suitable for storing and executing computer program code. The various participants and elements in the previously described system diagrams may use any suitable number of subsystems or components of the computing device (3500) to facilitate the functions described herein. The computing device (3500) may include subsystems or components interconnected via a communication infrastructure (3505) (for example, a communications bus, a network, etc.). The computing device (3500) may include one or more processors (3510) and at least one memory component in the form of computer-readable media. The one or more processors (3510) may include one or more of: CPUs, graphical processing units (GPUs), microprocessors, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs) and the like. In some configurations, a number of processors may be provided and may be arranged to carry out calculations simultaneously. In some implementations various subsystems or components of the computing device (3500) may be distributed over a number of physical locations (e.g. in a distributed, cluster or cloud-based computing configuration) and appropriate software units may be arranged to manage and / or process data on behalf of remote devices.

[0301] The memory components may include system memory (3515), which may include read only memory (ROM) and random access memory (RAM). A basic input / output system (BIOS) may be stored in ROM. System software may be stored in the system memory (3515) including operating system software. The memory components may also include secondary memory (3520). The secondary memory (3520) may include a fixed disk (3521), such as a hard disk drive, and, optionally, one or more storage interfaces (3522) for interfacing with storage components (3523), such as removable storage components (e.g. magnetic tape, optical disk, flash memory drive, external hard drive, removable memory chip, etc.), network attached storage components (e.g. NAS drives), remote storage components (e.g. cloud-based storage) or the like.

[0302] The computing device (3500) may include an external communications interface (3530) for operation of the computing device (3500) in a networked environment enabling transfer of data between multiple computing devices (3500) and / or the Internet. Data transferred via the external communications interface (3530) may be in the form of signals, which may be electronic, electromagnetic, optical, radio, or other types of signal. The external communications interface (3530) may enable communication of data between the computing device (3500) and other computing devices including servers and external storage facilities. Web services may be accessible by and / or from the computing device (3500) via the communications interface (3530).

[0303] The external communications interface (3530) may be configured for connection to wireless communication channels (e.g., a cellular telephone network, wireless local area network (e.g. using Wi-Fi™), satellite-phone network, Satellite Internet Network, etc.) and may include an associated wireless transfer element, such as an antenna and associated circuitry.

[0304] The computer-readable media in the form of the various memory components may provide storage of computer-executable instructions, data structures, program modules, software units and other data. A computer program product may be provided by a computer-readable medium having stored computer-readable program code executable by the central processor (3510). A computer program product may be provided by a non-transient or non-transitory computer- readable medium, or may be provided via a signal or other transient or transitory means via the communications interface (3530).

[0305] Interconnection via the communication infrastructure (3505) allows the one or more processors (3510) to communicate with each subsystem or component and to control the execution of instructions from the memory components, as well as the exchange of information between subsystems or components. Peripherals (such as printers, scanners, cameras, or the like) and input / output (I / O) devices (such as a mouse, touchpad, keyboard, microphone, touch-sensitive display, input buttons, speakers and the like) may couple to or be integrally formed with the computing device (3500) either directly or via an I / O controller (3535). One or more displays (3545) (which may be touch-sensitive displays) may be coupled to or integrally formed with the computing device (3500) via a display or video adapter (3540).

[0306] The terms artificial intelligence (Al), machine learning, and deep learning may be used interchangeably throughout this disclosure when referring to trained Al models. Machine learning may be considered a sub-branch of Al and deep learning may be considered a sub-branch of machine learning. Deep learning is a form of machine learning that uses a layered network, referred to as an artificial neural network (ANN). Any Al, machine learning and deep learning system may rely on an underlying model. The model may be tailored to a specific use case. Although Al is considered the broadest of term, it is common that any Al system includes some form of machine learning, with some systems further including deep learning.

[0307] Some examples of machine learning models may include, but are not limited to: decision trees, random forest regression, support-vector machines, K-means clustering, regression analysis, Gaussian processes, and the like.

[0308] Machine learning models (as well as deep learning models) may be categorised into classification or regression. Classification models may classify an input into one or more of a set of classifications, with the output being one of discrete classifications. Regression models may determine an output that may be a value or output across a continuous output range. A regression model may estimate a relationship between an input to an output.

[0309] ANNs may include a variety of structures, which are referred to as architectures. Different architectures are suitable for different use cases. Examples of ANN architectures may include convolutional neural networks or recurrent neural networks. Convolutional neural networks may be suitable for image-based data or multi-dimensional input data. Recurrent neural networks, such as long-short term memory networks, may be more suitable for time series applications.

[0310] An ANN may consist of interconnected units, commonly referred to as neurons, as they are inspired by and resemble neurons of the brain. The units may be made up of nodes and edges forming a connected network. The edges may connect nodes together. ANNs may be configured in the form of a layered structure with an input at the first layer and an output provided by the final layer. The layers between the first layer and final layer are hidden layers.

[0311] The input layer may include one or more nodes. An edge may extend from each node. Each edge may be connected to a node in a subsequent hidden or output layer. Each node may include more than one edge that connects the node to a plurality of other nodes in other layers. In some examples, an edge may feed back into a previous node in a preceding layer (a node not in subsequent layers but in a further layer), or to a different node in the same layer.

[0312] The output of a node may be computed by an activation function, which may be a linear or a nonlinear function of the sum of the inputs into each node in each layer. The output value of each node in the preceding layer is multiplied by a weighting value, which determines the strength of each nodes’ output value. Finally, the value that is determined at the node(s) of the final layer is the output of the ANN. For regression type ANNs, the output may contain only a single node with a value, or many nodes. For classification type ANNs, the output may include multiple nodes, where each node is an output of the probability of a classification type.

[0313] More complex ANNs are better suited to specific tasks. In addition to the weights and activation functions of a regular ANN, a convolutional neural network applies a filter (or a kernel) onto a two- dimensional data structure, which may reduce the number of edges between the hidden layers in the neural network. This may in turn reduce the number of weights within the neural network. A convolutional neural network may find application in image-based tasks, where image data may be structured as a two-dimensional data structure. A convolutional neural network may be extended into further dimensions by increasing the dimensions of the filter / kernel to match the number of dimensions of the input data.

[0314] Recurrent neural networks include a recurrent unit. This recurrent unit may maintain a hidden state over time, thereby providing a pseudo-memory capability. Such models may find application in time series or sequential operations, such as speech or text. Multiple recurrent units may be connected to each other, where the output of one unit at a first timestep may be used as an input into another recurrent unit at a second timestep. Examples of recurrent neural networks include, but are not limited to, long short-term memory networks, and gated recurrent units. Transformers are another form of deep learning architectures well suited for sequential based data. Transformers may utilise a self-attention mechanism instead of recurrence (such as in a recurrent neural network).

[0315] A trained Al model may be configured to run on a computing device, such as a Raspberry Pi ™, a NVIDIA Jetson Nano ™ developer kit, or a standard personal computer (PC) including a graphical processing unit (GPU). Example computing devices may be designed to perform specific computational tasks, which may include running multiple neural networks in parallel for applications including image classification, object detection, segmentation, and speech processing. In some cases, a trained Al model may be configured to run on a computing device in the form of a large computing system, such as a computing cluster (such as that found in a data centre).

[0316] T raining Al models may be a computationally intensive and time consuming. Models may thus be trained on a computing device provided by a large computing infrastructure or a cloud computing infrastructure that can be accessed over a network. These resources allow for dynamic computing resources to be dedicated to training a deep neural network, after which the trained model can be downloaded to run on a separate application.

[0317] Figure 36 illustrates a general overview (3600) of training and use of machine learning models (3614) in accordance with aspects of the present disclosure. The training may include a data preparation process (3611) that formats a raw incoming data (3610). The data preparation process (3611) may prepare training data (3612). The data preparation process (3611) may involve labelling the raw incoming data (3610). Labelling the raw incoming data (3610) may include labelling each input data of the raw incoming data (3610). Labelling the data may include providing a known value or solution that must be output by the model when a specific data is input into the model. The data preparation process (3611) may include formatting the raw incoming data (3610) into a format suitable for the type of model or in the case of an ANN, suitable for the model architecture. The data preparation process (3611) may generate the training data (3612). The training data (3612) may include a subset of data called validation data. The model may be trained using the training data (3612), but excluding the validation data. The validation data may be used within a training process (3613) to determine an accuracy level of the model on “unseen” input data. The validation data may be applied to the model during and after training.

[0318] The training process (3613) may receive the training data (3612) and iteratively update the model until a predefined quality criteria and / or accuracy criteria are achieved. The model (3614) may be output at the end of the training process (3613) to be used in a runtime process (3622).

[0319] The model (3614) may be trained using a training method. The training method may include any one of: supervised learning, unsupervised learning, semi-supervised, and reinforcement learning. The training process (3613) may include using a plurality of training methods.

[0320] Supervised learning may require labelled training data (3612), such that the correct output is known for each training data input. The task of the training process (3613) is to minimize the difference (or error) between the output of the model (3614) and the known output (for example, due to the labelling process) of the training data (3612). In some examples, the output may be verified as the output must satisfy a provided formula, such as with physics-informed models. The training procedure modifies the machine learning model (3613) such that the difference (or error) is minimized.

[0321] Unsupervised learning may be configured to extract features or patterns from unlabelled data. Unsupervised learning may be used when the raw incoming data (3610) is too large to be labelled. For example, unsupervised learning may be used for auto-encoders, where the aim is for the model output to match the model input by encoding the input data, and decoding the encoded input data.

[0322] When the model (3614) is in use, an input (3621) may be received into the runtime process (3622) that uses the model (3614) to obtain an output (3623) that may be used in a downstream process (3624). The computation of the runtime process (3622) is often referred to as ‘inference’.

[0323] The training process (3613) may be computationally demanding and time consuming. To successfully train a machine learning model, very large datasets may be used which are stored on a database. The training process may be performed on a computing device in the form of a large computing cluster which may access the database to obtain the training data when required. Additionally, the trained machine learning model (3614) may be stored on the database. The runtime process (3622) may run on an end user computing device by downloading the machine learning model (3614) over a network from a database, or the runtime process (3622) may run on a large computing infrastructure such as a computing cluster. An example embodiment of interacting with the machine learning model (3614) may include an end user computing device, such as a mobile device or a computer which may obtain or be the source of the input data (3621), transmit the input data (3621) over a network to a computing cluster to perform the runtime process (3622). Alternatively, an end user computing device may obtain the machine learning model from a database over a network and store the machine learning model locally on the device. The end user device may obtain an input data (3621) and perform the runtime process (3622) locally on the device to obtain an output (3632). By performing the runtime process (3622) locally on the device, the input data (3621) does not need to be transmitted over a network, reducing bandwidth usage. This may be referred to as ‘on-the-edge’ computing.

[0324] The foregoing description has been presented for the purpose of illustration; it is not intended to be exhaustive or to limit the technology to the precise forms disclosed. Persons skilled in the relevant art can appreciate that many modifications and variations are possible in light of the above disclosure.

[0325] Any of the steps, operations, components or processes described herein may be performed or implemented with one or more hardware or software units, alone or in combination with other devices. Components or devices configured or arranged to perform described functions or operations may be so arranged or configured through computer-implemented instructions which implement or carry out the described functions, algorithms, or methods. The computer- implemented instructions may be provided by hardware or software units. In one embodiment, a software unit is implemented with a computer program product comprising a non-transient or non- transitory computer-readable medium containing computer program code, which can be executed by a processor for performing any or all of the steps, operations, or processes described. Software units or functions described in this application may be implemented as computer program code using any suitable computer language such as, for example, Java™, C++, or Perl™ using, for example, conventional or object-oriented techniques. The computer program code may be stored as a series of instructions, or commands on a non-transitory computer-readable medium, such as a random-access memory (RAM), a read-only memory (ROM), a magnetic medium such as a hard-drive, or an optical medium such as a CD-ROM. Any such computer-readable medium may also reside on or within a single computational apparatus, and may be present on or within different computational apparatuses within a system or network.

[0326] Flowchart illustrations and block diagrams of methods, systems, and computer program products according to embodiments are used herein. Each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, may provide functions which may be implemented by computer readable program instructions. In some alternative implementations, the functions identified by the blocks may take place in a different order to that shown in the flowchart illustrations.

[0327] Some portions of this description describe the examples in terms of algorithms and symbolic representations of operations on information. These algorithmic descriptions and representations, such as accompanying flow diagrams, are commonly used by those skilled in the data processing arts to convey the substance of their work effectively to others skilled in the art. These operations, while described functionally, computationally, or logically, are understood to be implemented by computer programs or equivalent electrical circuits, microcode, or the like. The described operations may be embodied in software, firmware, hardware, or any combinations thereof.

[0328] The language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the present disclosure be limited not by this detailed description, but rather by any claims that issue on an application based hereon. Accordingly, the present disclosure is intended to be illustrative, but not limiting, of the scope of any accompanying claims.

[0329] Finally, throughout the specification and any accompanying claims, unless the context requires otherwise, the word ‘comprise’ or variations such as ‘comprises’ or ‘comprising’ will be understood to imply the inclusion of a stated integer or group of integers but not the exclusion of any other integer or group of integers.

Claims

CLAIMS:1 . A computer-implemented method for monitoring a flotation cell, the method comprising: by a server, receiving digital images captured by a camera at the flotation cell, and accessing a machine learning model which is trained with one or more training images; inputting the digital images into the machine learning model, and by the machine learning model, analysing the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; generating a reagent dosing parameter based on the analysis by the machine learning model; and outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

2. The method of claim 1 , wherein the method includes generating data relating to any one or more of: a bubble size distribution of bubbles in the received digital images; a gas holdup of bubbles in the received digital images; a gas velocity of bubbles in the received digital images; and / or a colour of the received digital images.

3. The method of claim 1 or claim 2, wherein outputting the reagent dosing parameter includes outputting a reagent dosing set-point parameter.

4. The method of any one of the preceding claims, wherein the method includes monitoring one or more flotation cell / s, and wherein outputting the reagent dosing parameter includes outputting it for use in dosage of reagents in any one or more of the flotation cell / s.

5. The method of claim 4, wherein the method includes generating the gas set-point parameter based on the analysis by the machine learning model, and outputting the gas set-point parameter for use in controlling aeration of the flotation cell / s.

6. The method of claim 5, wherein the gas set-point parameter is outputted by the server to a controller associated with air or gas flow of a process associated with the flotation cell / s, so as to enable automated adjustment of aeration or gas input in the process associated with the flotation cell / s.

7. The method of claim 5, wherein the method includes outputting the reagent dosingparameter so as to enable automated adjustment of reagent addition to a process associated with the flotation cell / s.

8. The method of any one of claims 4 to 7, wherein the reagent dosing parameter and / or the gas set-point parameter is / are outputted to one or more controller / s associated with one or more of a frother, a collector, a depressant, and / or a mineral separator machine.

9. The method of any one of claims 4 to 8, wherein the method includes outputting data by the machine learning model, and wherein the data output by the machine learning model is utilised as an input to a level controller so as to enable control of a level of froth in the flotation cell / s.

10. The method of any one of claims 4 to 9, wherein the method includes, by the machine learning model, generating data relating to a bubble surface area flux based at least partially on the analysis of the received digital images.

11. The method of any one of claims 4 to 10, wherein the method includes utilising a probe inside each of the one or more flotation cell / s, wherein the probe is at least partially submerged with a viewing pane and an operative end of the probe inside the flotation cell, and wherein the probe includes the camera and at least one light source.

12. The method of claim 11, wherein the viewing pane is positioned at an oblique angle relative to vertical, and wherein the method includes, by the camera, capturing intermittent digital images of bubbles inside the flotation cell under illumination from the at least one light source through the viewing pane.

13. The method of claim 12, wherein the method includes, transmitting the captured digital images to the server for further processing.

14. The method of any one of claims 4 to 10, wherein the method includes utilising a froth sensor that includes the camera, and wherein the froth sensor is positioned to capture top-down imagery of the flotation cell / s froth surface.

15. The method of claim 14, wherein the method includes, by the server, classifying froth states in the received digital images, and wherein the froth states include dry froth, wet froth, and pulp exposure zone / s.

16. The method of claim 14 or claim 15, wherein the method includes, by the froth sensor,utilising a bubble segmentation model accessible by the froth sensor, the bubble segmentation model being configured to analyse the captured images to extract parameters including one or more of bubble coverage and bubble count.

17. The method of claim 16, wherein the extracted parameters are processed by the server to derive a wet froth indicator that quantifies relative froth wetness and entrainment.

18. A system for monitoring a flotation cell, the system comprising: a non-transitory computer- readable storage medium; and one or more processors coupled to the non-transitory computer- readable storage medium, wherein the non-transitory computer-readable storage medium comprises program instructions that, when executed on the one or more processors, cause the system to perform operations comprising: by a server, receiving digital images captured by a camera at the flotation cell, and accessing a machine learning model which is trained with one or more training images; inputting the digital images into the machine learning model, and by the machine learning model, analysing the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; generating a reagent dosing parameter based on the analysis by the machine learning model; and outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

19. A system for monitoring a flotation cell, the system including a memory for storing computer-readable program code and a processor for executing the computer-readable program code, the system comprising: a server which is arranged to receive digital images captured by a camera at the flotation cell; a machine learning model which is accessible by the server, the machine learning model being trained with one or more training images and operable to receive the digital images as input and to analyse the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; wherein the machine learning model is arranged for generating a reagent dosing parameter based on the analysis by the machine learning model; and wherein the server is arranged for outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

20. A computer program product for monitoring a flotation cell, the computer program productcomprising a computer-readable medium having stored computer-readable program code for performing the steps of: by a server, receiving digital images captured by a camera at the flotation cell, and accessing a machine learning model which is trained with one or more training images; inputting the digital images into the machine learning model, and by the machine learning model, analysing the received digital images to identify bubbles and to generate data relating to a characteristic of the bubbles; generating a reagent dosing parameter based on the analysis by the machine learning model; and outputting the reagent dosing parameter for use in controlling a dosage of one or more reagents in a process associated with the flotation cell.

Images