Estimating recovery losses due to coarse material using cyclonetrac tm PST technology and machine learning

Abstract

The present invention provides a new and unique methodology for quantifying the impact of different oversize events on copper recovery, e.g., using historical PST and process data. For example, from historical PST data, an average deviation from a high limit of the control mesh was calculated and then used to train different machine learning models to predict the decrease in the recovery due to oversize events. The best machine learning model for this method was a Random Forest Regressor with a mean absolute error of 0.78 percent point and coefficient of determination (r2) equal to 0.901.

Description

[0001] WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO ESTIMATING RECOVERY LOSSES DUE TO COARSE MATERIAL USING CYCLONETRACTMPST TECHNOLOGY AND MACHINE LEARNING CROSS-REFERENCE TO RELATED APPLICATIONS This application claims benefit to provisional patent application serial nos. 63 / 626,648 (WFMB no.712-002.475 (CCS-0224)), filed 30 January 2024, which are both incorporated by reference in its entirety. This application also relates to the following: US 12,121,906, (WFMB no.712-002.464-1-1 / / CCS-0207WO), and corresponding PCT / US2019 / 022943, which claimed benefit to US provisional application no.62 / 664,672, filed 19 March 2018; US 10,830,623 (WFMB no.712-002.419-1-1 / / CCS-0135WO), and corresponding PCT / US2016 / 016721, which claimed benefit to US provisional application no.62 / 112,433, filed 5 February 2015; and US 10,309,887 (WFMB no.712-002.406-1-1 / / CCS-0120WO), and corresponding PCT / US2014 / 012510 , which claimed benefit to US provisional application no.61 / 755,305, filed 22 January 2013, which are all incorporated by reference in their entirety. BACKGROUND OF THE INVENTION 1. Field of Invention This invention relates to a technique for estimating recovery losses due to coarse material using the Assignee's CYCLONEtracTMPST technology. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 2. Description of Related Art Mineral processing systems for monitoring overflow streams in a battery of cyclones are known in the art and controlling the operation of one or more cyclones in the battery of cyclones. Such mineral processing systems may include the use of the Assignee's known CYCLONEtracTMParticle Size Tracking (PST) technology that provides real-time granulometry measurements of overflow streams of individual cyclones, allowing continuous monitoring of the individual classification performance of each one. The introduction of Assignee's known CYCLONEtracTMPST technology on concentrator plants has enabled monitoring the performance of each cyclone of a battery in real time, the detection of oversize events on a particular cyclone and the ability to take actions to correct it. In spite of this introduction and the advantages obtained therefrom, there remains a need for a better way to process PST signaling in such a mineral processing system and control the operation of cyclones in a battery of cyclones based upon the PST signaling processed. SUMMARY OF THE INVENTION The Basic Invention The present invention provides a new and unique methodology for quantifying the impact of different oversize events on copper recovery, e.g., using historical PST and process data. For example, from historical PST data, an average deviation from a high limit of the control mesh was calculated and then used to train different machine learning models to predict the decrease in the recovery due to oversize events. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO The best machine learning model for this method was a Random Forest Regressor with a mean absolute error of 0.78 percent point and coefficient of determination (r2) equal to 0.901. The results set forth below show that a decrease in recovery between 1 and 4.7 percent points could be expected due to oversize events for the data used in this study. Particular Embodiments In particular, and according to some embodiments, the present invention may include, or take the form of, apparatus for monitoring overflow streams in a battery of cyclones of a mineral recovery process, comprising: a signal processor or processing module configure to: receive particle size tracking (PSTi) signaling containing information about particle sizes of coarse particles provided by overflows of cyclones in a battery of cyclones, and historic PSTisignaling containing information about a consolidated database having time-based entries of plant recovery percentages and deviations from a particle size high limit (HL) in a predetermined time interval (to, t1) for each cyclone; determine if the particle size tracking (PSTi) signaling contains a particle size of coarse particles provided by an overflow of a cyclone (Ci) in the battery of cyclones that is an outlier in relation to the historic PSTisignaling, and if so, then also determine if the cyclone (Ci) has an average deviation (Dl) of a particle size tracking (PSTi) that is above the particle size high limit (HL) in the predetermined time interval (to, t1); and WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO provide corresponding signaling containing information to control the operation of the cyclone (Ci) in the battery of cyclones, based upon the signaling received and the average deviation (Dl) determined. The apparatus may include one or more of the following features: The average deviation may be determined by the following equation: The consolidated dataset may include rows of dates and columns for each cyclone Ci, where the columns including a CU_Rec (%), C1_dev, C2_dev, ..., Cn- 1_dev and Cn_dev for each date, where the CU_Rec (%) is a percentage of copper recovery, and where Ci_dev is a deviation of the percentage of copper recovery for a respective cyclone Ci. The CU_Rec (%) may be a percentage of copper recovery determined by the following equation: . The corresponding signaling may include control signaling containing information to change the mineral recovery process, including to change either water addition to, or decreasing tonnage of, crushed ore being processed; or opening another cyclone in the battery of cyclones and then closing the cyclone having oversize coarse particles in an overflow stream. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO The apparatus may include PST sensors, each PST sensor arranged in relation to a respective overflow of a respective cyclone in the battery of cyclones and configured to sense and track the particle sizes of the coarse particles provided by each cyclone's overflow and provide the PSTi signaling. The apparatus may include a consolidated dataset memory module configured to store the consolidated database of the particle sizes of the coarse particles provided by each cyclone's overflows, including daily plant recovery percentage (%) and deviations from the particle size high limit (HL) for each cyclone. The apparatus may include a machine learning model module configured to receive the PSTi signaling, process the PSTi signaling based a machine learning model, and provide the historical PSTi signaling for storing as the consolidated database, including updating the consolidated database over time. By way of example, the machine learning model may include one of the following: ^ Multiple linear regression, ^ Random Forest Regressor, ^ Support Vector Machine Regressor, and ^ Neural Network. The machine learning model may include using a Singular Value Decomposition (SVD) that decomposes a matrix into component matrices.

[0002] WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO The Method The present invention may include, or take the form of, a method for monitoring overflow streams in a battery of cyclones of a mineral recovery process, featuring: configuring a signal processor or processing module to: receive particle size tracking (PSTi) signaling containing information about particle sizes of coarse particles provided by overflows of cyclones in a battery of cyclones, and historic PSTi signaling containing information about a consolidated database having time-based entries of plant recovery percentages and deviations from a particle size high limit (HL) in a predetermined time interval (to, t1) for each cyclone; determine if the particle size tracking (PSTi) signaling contains a particle size of coarse particles provided by an overflow of a cyclone (Ci) in the battery of cyclones that is an outlier in relation to the historic PSTisignaling, and if so, then also determine if the cyclone (Ci) has an average deviation (Dl) of a particle size tracking (PSTi) that is above the particle size high limit (HL) in the predetermined time interval (to, t1); and provide corresponding signaling containing information to control the operation of the cyclone (Ci) in the battery of cyclones, based upon the signaling received and the average deviation (Dl) determined. The method may include one or more other steps for implementing the other features disclosed herein. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO BRIEF DESCRIPTION OF THE DRAWING The drawing includes Figures 1-7B, which are not necessarily drawn to scale, as follows: Figure 1 shows a graph of PST signals in relation to time in a normal and desired operation in a mineral processing system. Figure 2 shows a graph of PST signals in relation to time having an oversize event. Figure 3 shows a graph of variations in copper recovery in relation to total PST deviation due to oversized events. Figure 4 shows a signal processor or processing module for implementing the signal processing, according to some embodiments of the present invention. Figure 5 shows other components for implementing a mineral recovery process in conjunction with the signal processor or processing module shown in Figure 4, e.g., including Fig.5A showing a battery of cyclones, Fig.5B showing PST sensors, Fig.5C showing a consolidated dataset memory module, and Fig.5D showing a machine learning model module, all according to some embodiments of the present invention. Figure 6 shows steps for implementing the method, according to some embodiments of the present invention. Figure 7A is a photograph of a CYCLONEtracTMPST particle sizing sensor mounted on a hydrocyclone overflow pipe, e.g., that is known in the art. Figure 7B is a diagram of a CYCLONEtracTMPST plant scale installation, e.g., using individual hydrocyclone overflow sensors (AKA PST particle sizing sensor), that is known in the art. The graphs in Figures 1-3 are shown in color. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO DETAILED DESCRIPTION OF BEST MODE OF THE INVENTION Introduction As set forth above, the introduction of the Assignee's CYCLONEtracTMPST technology on concentrator plants has enabled monitoring the performance of each cyclone of a battery in real time, the detection of oversize events on a particular cyclone and the ability to take actions to correct it. These actions could be making changes to the process (i.e. changes in water addiction, decrease tonnage, etc.) or opening another cyclone and then closing the one with oversize in the overflow. However, when no action is taken, the coarse material sent to the flotation stage inevitably reduces the recovery, given the well know inverse relation between this variable and the particle size. This work presents a methodology for quantifying the effect of the oversize generation in cyclones on the recovery using PST signals and machine learning techniques. Methodology The PST technology delivers real-time percentage of retained mass for up to five particle sizes – including the control mesh size – for each cyclone on a battery, enabling monitoring the individual performance of each one. Error! Reference source not found. shows the PST signals for control mesh of a battery, in this case +70# (212 µm), where all signals are below the high limit for the control mesh (red dashed line). Error! Reference source not found. shows a normal and desired operation, where none of the cyclones have oversize on the overflow. However, given the WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO intrinsic variability of the grinding / classification process (ore hardness, mineralogy, changes on pulp density, etc.) sometimes there are some oversize events that can be detected from the PST signals. Error! Reference source not found. shows two and a half days of data, with five different oversize events where the particle size in the overflow of two cyclones was coarse, having the PST signal above the high limit. In addition, it is desirable to catch cyclones whose behaviour is clearly different from the rest. This is achieved by calculating a dynamic outlier detection limit, which in this case was performed using the criteria described by Leys et. al. (2013) and presented in black on Error! Reference source not found.. It can be seen in Error! Reference source not found. that C8 had four oversize events, lasting in total around 23 hours of the analysed period, while C2 had one event, lasting around 20 hours. The cause of the great duration of these events could be due to PST signals not incorporated yet into the Advanced Process Control (APC) of the plant, given the recent installation and commissioning of the PST System. Since oversize events have an impact on the recovery downstream, it is desired to quantify that impact. For achieving this objective, the data was used to train different machine learning models that allow us to have relationships as shown in the equation (1): (1) The approach used in this work is the following: 1. Detect when a cyclone is an outlier. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 2. Check if the PST signal is above the high limit. Calculate the average deviation from the high limit up on an interval. The average deviation above the high limit of a cyclone in an interval is given by (2): (2) In addition to historical PST signals, a dataset with eight-month process datain a daily basis was used. The equation (2) was then applied to each cyclone (batteries, cyclones each) for each day of the process’ dataset to obtain a dataset with a column for the recovery and columns, one for each cyclone with the daily average deviation from the of the control mesh. After consolidating process data with PST signals available for the 8-month period, the dataset contained 180 rows (each one representing a day). Table 1 shows an extract of the consolidated dataset. Table 1 Date Cu_Rec (%) C1_dev C2_dev … Cn-1_dev Cn_dev 2 WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 2023-05-29 83.67 0.3850 0.5320 … 0 0.0228 2 2 1 474 g Four different machine learning models where trained using the consolidated dataset. The models used in this work are: ^ Multiple linear regression ^ Random Forest Regressor ^ Support Vector Machine Regressor ^ Neural Network The consolidated dataset corresponds to a sparse dataset, that is, it contains many values equal zero in the independent variables. Given that, a dimensionality reduction was performed using Truncated Singular Value Decomposition (SVD), which is a method that decomposes a matrix into three other matrices, reducing the number of dimensions while preserving the most valuable information. It allows to keep components that explain most of the variance of the data. The number of components used in this work was selected for having at least 90% of the variance explained, giving for this dataset, seven components. The four models were trained using the full consolidated dataset and the truncated dataset, having in total eight models. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO shows the metrics of the eight models, where it can be seen the Random Forest Regressor had the best performance, thus, it was the model selected for the next stage of the study.

[0003] WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO Table 2: Metrics for machine learning models. RMSE Model r2score MAE (%) (%) Results And Discussion From the selected model, it was possible to estimate the base recovery, i.e., when all deviations from the high limit where equal to zero and then, estimate the recovery for every day of the consolidated dataset. The base recovery from the model is equal to 80.94% and the variations for each day of the consolidated dataset are shown on the Error! Reference source not found.. The x-axis shows the “Total PST deviation” meaning the summation of deviations for all cyclones for a particular date, while the y-axis shows the variation of recovery to the base recovery of the model. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO On Error! Reference source not found., grey points were considered as non-relevant since they are inside the error margin for both PST signals (commonly with an error around 3%) and the model predictions (with MAE equal to 0.78 percent points). From the relevant points, four are outliers (marked in blue), hence, they are not considered valid either. From the rest of the points, even though there is no correlation between the Total PST Deviation and the variation in the recovery, a decrease in recovery between 1 and 4.7 percent points could be expected due to oversize events. A possible reason for this low correlation is the fact that for this plant, most but not all cyclone batteries had PST installed, but all batteries were feeding the same flotation area from which the daily plant-level recoveries were obtained. Thus, the expected decrease in plant-level recovery from an oversize event in only one battery would be diluted by the other batteries operating normally. Future work will seek out a plant process design where ideally a single battery feeds a single flotation area, and recovery information is available at a frequency higher than daily, e.g. hourly, which should produce a more measurable correlation between oversize events and recovery. Conclusion A methodology was developed to predict changes on recovery from historical operational and PST data, which enables quantifying the effect of cyclone oversize events on recovery. The expected decrease in the recovery due to oversize events varies from 1 to 4.7 percent points. Even if it is well accepted that oversize affects flotation recovery, the plant design can have a masking effect when some cyclones send coarse material WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO downstream, since different batteries’ overflows are combined before feeding the flotation stage and only one consolidated recovery value is measured. The error of the selected model is acceptable for the purpose of this work, but it must be improved if it is meant to be used for production. This improvement could be achieved by incorporating other process variables in the model and / or using a database with an hourly basis for having more data that can make the model more robust. Figure 4 Figure 4 shows apparatus generally indicated as 10 according to some embodiments of the present invention. The apparatus 10 may include a signal processor or processing module 100 that receives particle size tracking (PSTi) signaling containing information about particle sizes of coarse particles provided by overflows of cyclones in a battery of cyclones, and historic PSTisignaling containing information about a consolidated database having time-based entries of plant recovery percentages and deviations from a particle size high limit (HL) in a predetermined time interval (to, t1) for each cyclone; determine if the particle size tracking (PSTi) signaling contains a particle size of coarse particles provided by an overflow of a cyclone (Ci) in the battery of cyclones that is an outlier in relation to the historic PSTisignaling, and if so, then also determine if the cyclone (Ci) has an average deviation (Dl) of a particle size tracking (PSTi) that is above the particle size high limit (HL) in the predetermined time interval (to, t1); and provide corresponding signaling containing information to control the operation of the cyclone (Ci) in the battery of cyclones, based upon the signaling received and the average deviation (Dl) determined. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO By way of example, and consistent with that described herein, the functionality of the signal processor or processing module 100 may be implemented using hardware, software, firmware, or a combination thereof, although the scope of the invention is not intended to be limited to any particular embodiment thereof. In a typical software implementation, the signal processor would be one or more microprocessor-based architectures having a microprocessor, a random access memory (RAM), a read only memory (ROM), input / output devices and control, data and address buses connecting the same. A person skilled in the art would be able to program such a microprocessor-based implementation to perform the functionality set forth in the signal processing block 100, as well as other functionality described herein without undue experimentation. The scope of the invention is not intended to be limited to any particular implementation using technology now known or later developed in the future. Moreover, the scope of the invention is intended to include the signal processor being a stand alone module, as shown, or in the combination with other circuitry for implementing another module. It is also understood that the apparatus 10 may include one or more other modules, components, circuits, or circuitry for implementing other functionality associated with the apparatus that does not form part of the underlying invention, and thus is not described in detail herein. By way of example, the one or more other modules, components, circuits, or circuitry may include random access memory, read only memory, input / output circuitry and data and address buses for use in relation to implementing the signal processing functionality of the signal processor 100, or devices or components related to mixing or pouring concrete in a ready-mix concrete truck or adding chemical additives, etc. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO Figure 5 The apparatus 10 may include other components and modules, e.g., like elements 20, 30, 40, 50 that are shown in further detail in Figure 5. For example, Figure 5A shows a battery of cyclones 20 configured to process crushed ore, each cyclone having an overflow O (Fig.7A) for providing coarse particles for further processing. Batteries of cyclones like element 20 are known in the art; and the scope of the invention is not intended to be limited to any particle type or kind thereof either now known or later developed in the future. Figure 5B shows PST sensors 30, each PST sensor 30 being arranged in relation to a respective overflow O (Fig.7A) of a respective cyclone in the battery of cyclones and configured to sense and track the particle sizes of the coarse particles provided by each cyclone's overflow O and provide the PSTi signaling. PST sensors like element 30, including CYCLONEtracTMPST technology, known in the art; and the scope of the invention is not intended to be limited to any particle type or kind thereof either now known or later developed in the future. Figure 5C shows a consolidated dataset memory module 40 configured to store the consolidated database of the particle sizes of the coarse particles provided by each cyclone's overflows O (Fig.7A), including daily plant recovery percentage (%) and deviations from the particle size high limit (HL) for each cyclone. One skilled in the art would appreciate and understand how to configure a consolidated dataset memory module like element 40 without undue experimentation to store and update over time a consolidated dataset like that shown in Table 1. Figure 5D shows a machine learning model module 50 configured to receive the PSTisignaling, process the PSTisignaling based a machine learning model, and provide the historical PSTisignaling for storing and / or updating over time as the WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO consolidated database. One skilled in the art would appreciate and understand how to configure a consolidated dataset, e.g., by processing the PSTi signaling received based a machine learning model, and providing the historical PSTi signaling for storing and / or updating as the consolidated database. Moreover, machine learning models like a Random Forest Regressor, Neural Network, Multiple Linear Regression, and Support Vector Machine Regressor with or with implementing the same using SVC as shown in Table 1 are known in the art; and the scope of the invention is not intended to be limited to any particle type or kind thereof either now known or later developed in the future. For example, one skill in the art would appreciate that the present invention may be implemented using other types or kinds of machine learning models either now known or later developed in the future, e.g., other than a Random Forest Regressor, Neural Network, Multiple linear Regression, and Support Vector Machine Regressor, again without undue experimentation. Detect When a Cyclone is An Outlier. By way example, Figure 3 shows variations in recovery due to oversized events, e.g., when one or more outliers are detected. The scope of the invention is not intended to any particular type, kind or way of detecting of one or more outliers, which may includes types, kinds or ways of detecting of one or more outliers either now known or later developed in the future. Figure 6: The Method Figure 6 shows a method generally indicated as 150 for monitoring overflow streams in a battery of cyclones of a mineral recovery process, according to the present invention, having steps 152, 154 and 156. WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO In particular, the method 150 may include configuring a signal processor or processing module like element 100 (Fig. to: receive in step 162 particle size tracking (PSTi) signaling containing information about particle sizes of coarse particles provided by overflows of cyclones in a battery of cyclones, and historic PSTi signaling containing information about a consolidated database having time-based entries of plant recovery percentages and deviations from a particle size high limit (HL) in a predetermined time interval (to, t1) for each cyclone; determine in step 154 if the particle size tracking (PSTi) signaling contains a particle size of coarse particles provided by an overflow of a cyclone (Ci) in the battery of cyclones that is an outlier in relation to the historic PSTi signaling, and if so, then also determine if the cyclone (Ci) has an average deviation (Dl) of a particle size tracking (PSTi) that is above the particle size high limit (HL) in the predetermined time interval (to, t1); and provide in step 156 corresponding signaling containing information to control the operation of the cyclone (Ci) in the battery of cyclones, based upon the signaling received and the average deviation (Dl) determined. The method may include one or more other steps for implementing the other features disclosed herein. Figures 7A and 7B: The PST Particle Sizing Sensor By way of example, Figure 7A shows a PST particle sizing sensor 30 mounted on a hydrocyclone overflow pipe O; and Figure 7B shows a PST plant scale installation diagram, e.g., using individual hydrocyclone overflow sensors (aka PST particle sizing sensor). In Figure 7A and 7B, the PST particle sizing sensor includes WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO a mounting bracket MB configured to arrange the PST particle sizing sensor 30 on the hydrocyclone classifier overflow pipe O of at least one hydrocyclone in a hydrocyclone battery. In Figure 7A and 7B, the PST particle sizing sensor 30 is the CYCLONEtracTMPST particle sizing sensor, which was developed, manufactured and distributed by the Assignee of the present invention. The PST particle sizing sensor 30 is disclosed in the aforementioned US 12,121,906 (WFMB no.712-002.464-1-1 / / CCS- 0207WO), US 10,830,623 (WFMB no.712-002.419-1-1 / / CCS-0135WO), and US 10,309,887 (WFMB no.712-002.406-1-1 / / CCS-0120WO), which are all incorporated by reference in their entirety. Although the present invention is disclosed using the Assignee's CYCLONEtracTMPST particle sizing sensor 30, embodiments are envisioned, and the scope of the invention is intended to include, e.g., using other types or kind of particle sizing sensor configured to be arranged on arranged on a hydrocyclone classifier overflow pipe O of at least one hydrocyclone in a hydrocyclone battery that are both now known or later developed in the future within the spirit of the present invention. Applications By way of example, the present invention may be used in, or form part of, or used in conjunction with, industrial processes like a mineral extraction processing system for extracting or separating minerals in a fluidic medium that are either now known or later developed in the future, including any mineral process, such as those related to processing substances or compounds that result from inorganic processes of nature and / or that are mined from the ground, as well as including either other WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO extraction processing systems or other industrial processes, where the extraction, or separating, or sorting, or classification, of product by size, or density, or some electrical characteristic, is critical to overall industrial process performance. Nomenclature Particle Size Tracking average deviation of PST signal from the high limit for cyclone . beginning of the analyzed period. end of the analyzed period. PST signal of control mesh for cyclone high limit of control mesh. coefficient of determination. root of the mean squared error. mean absolute error. number of batteries. number of cyclones per battery. References A. S. K. Christoffersen et al. (2021). Benchmarking Machine Learning Algorithms for Greenhouse Gas Flux Estimation from Sparse Data. Environmental Modelling & Software, 143, 105011. https: / / doi.org / 10.1016 / j.envsoft.2021.105011 Leys, C., Ley, C., Klein, . Detecting outliers: Do not use standard deviation around the mean, use absolute deviation WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO around the median. Journal of Experimental Social Psychology, 49(4), 764-766. https: / / doi.org / 10.1016 / j.jesp.2013.03.013 Tomlin. (1985). A Comparative Study of Sparse Matrix Transactions on Mathematical Software, 11(1), 1-26. https: / / doi.org / 10.1145 / 4080.4091 Li. (2018). A Comparative Study of Sparse Learning Algorithms for the Classification of Healthcare Data. International Journal of Environmental Research and Public Health, 15(6), 1123. https: / / doi.org / 10.3390 / ijerph15061123 Sepúlveda, J., Maron, R., Estrada, M., Bruna, R., (2019) ‘On-Line Detection of Abnormal Cyclone Performance using Particle Size Tracking (PST) Technology’, Proceedings of PROCEMIN 2019, GECAMIN, Chile.

[0004] WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO The Scope of the Invention While the invention has been described with reference to an exemplary embodiment, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof without departing from the scope of the invention. In addition, may modifications may be made to adapt a particular situation or material to the teachings of the invention without departing from the essential scope thereof. Therefore, it is intended that the invention not be limited to the particular embodiment(s) disclosed herein as the best mode contemplated for carrying out this invention.

Claims

WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO WHAT IS CLAIMED IS:

1. Apparatus for monitoring overflow streams in a battery of cyclones of a mineral recovery process, comprising: a signal processor or processing module configure to: receive particle size tracking (PSTi) signaling containing information about particle sizes of coarse particles provided by overflows of cyclones in a battery of cyclones, and historic PSTi signaling containing information about a consolidated database having time-based entries of plant recovery percentages and deviations from a particle size high limit (HL) in a predetermined time interval (to, t1) for each cyclone; determine if the particle size tracking (PSTi) signaling contains a particle size of coarse particles provided by an overflow of a cyclone (Ci) in the battery of cyclones that is an outlier in relation to the historic PSTi signaling, and if so, then also determine if the cyclone (Ci) has an average deviation (Dl) of a particle size tracking (PSTi) that is above the particle size high limit (HL) in the predetermined time interval (to, t1); and provide corresponding signaling containing information to control the operation of the cyclone (Ci) in the battery of cyclones, based upon the signaling received and the average deviation (Dl) determined.WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 2. Apparatus according to claim 1, wherein the average deviation is determined by the following equation:

3. Apparatus according to claim 1, wherein the consolidated dataset includes rows of dates and columns for each cyclone Ci, where the columns including a CU_Rec (%), C1_dev, C2_dev, ..., Cn-1_dev and Cn_dev for each date, where the CU_Rec (%) is a percentage of copper recovery, and where Ci_dev is a deviation of the percentage of copper recovery for a respective cyclone Ci.

4. Apparatus according to claim 3, wherein the CU_Rec (%) is a percentage of copper recovery determined by the following equation: .WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 5. Apparatus according to claim 1, wherein the corresponding signaling includes control signaling containing information to change the mineral recovery process, including to change either water addition to, or decreasing tonnage of, crushed ore being processed; or opening another cyclone in the battery of cyclones and then closing the cyclone having oversize coarse particles in an overflow stream.

6. Apparatus according to claim 1, wherein the apparatus comprises PST sensors, each PST sensor arranged in relation to a respective overflow of a respective cyclone in the battery of cyclones and configured to sense and track the particle sizes of the coarse particles provided by each cyclone's overflow and provide the PSTi signaling.

7. Apparatus according to claim 1, wherein the apparatus comprises a consolidated dataset memory module configured to store the consolidated database of the particle sizes of the coarse particles provided by each cyclone's overflows, including daily plant recovery percentage (%) and deviations from the particle size high limit (HL) for each cyclone.

8. Apparatus according to claim 1, wherein the apparatus comprises a machine learning model module configured to receive the PSTisignaling, process the PSTisignaling based a machine learning model, and provide the historical PSTisignaling for storing as the consolidated database.WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 9. Apparatus according to claim 8, wherein the machine learning model includes one of the following: ^ Multiple linear regression, ^ Random Forest Regressor, ^ Support Vector Machine Regressor, and ^ Neural Network.

10. Apparatus according to claim 9, wherein the machine learning model includes using a Singular Value Decomposition (SVD) that decomposes a matrix into component matrices.WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 11. A method for monitoring overflow streams in a battery of cyclones of a mineral recovery process, comprising: configuring a signal processor or processing module to: receive particle size tracking (PSTi) signaling containing information about particle sizes of coarse particles provided by overflows of cyclones in a battery of cyclones, and historic PSTi signaling containing information about a consolidated database having time-based entries of plant recovery percentages and deviations from a particle size high limit (HL) in a predetermined time interval (to, t1) for each cyclone; determine if the particle size tracking (PSTi) signaling contains a particle size of coarse particles provided by an overflow of a cyclone (Ci) in the battery of cyclones that is an outlier in relation to the historic PSTi signaling, and if so, then also determine if the cyclone (Ci) has an average deviation (Dl) of a particle size tracking (PSTi) that is above the particle size high limit (HL) in the predetermined time interval (to, t1); and provide corresponding signaling containing information to control the operation of the cyclone (Ci) in the battery of cyclones, based upon the signaling received and the average deviation (Dl) determined.WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 12. A method according to claim 11, wherein the method comprises determining the average deviation by the following equation:

13. A method according to claim 12, wherein the method comprises configuring the consolidated dataset with rows of dates and columns for each cyclone Ci, where the columns including a CU_Rec (%), C1_dev, C2_dev, ..., Cn- 1_dev and Cn_dev for each date, where the CU_Rec (%) is a percentage of copper recovery, and where Ci_dev is a deviation of the percentage of copper recovery for each cyclone Ci.

14. A method according to claim 13, wherein the CU_Rec (%) is a percentage of copper recovery determined by the following equation: .WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 15. A method according to claim 11, wherein the corresponding signaling includes control signaling containing information to change the mineral recovery process, including to change either water addition to, or decreasing tonnage of, crushed ore being processed; or opening another cyclone in the battery of cyclones and then closing the cyclone having oversize coarse particles in an overflow stream.

16. A method according to claim 11, wherein the method comprises configuring PST sensors in relation to overflows of cyclones in the battery of cyclones, including arranging each PST sensor in relation to a respective overflow of a respective cyclone in the battery of cyclones and configured to sense and track the particle sizes of the coarse particles provided by each cyclone's overflow and provide the PSTi signaling.

17. A method according to claim 11, wherein the method comprises storing in a consolidated dataset memory module the consolidated database of the particle sizes of the coarse particles provided by each cyclone's overflows, including daily plant recovery percentage (%) and deviations from the particle size high limit (HL) for each cyclone.

18. A method according to claim 11, wherein the method comprises receiving in a machine learning model module the PSTisignaling, processing the PSTisignaling based a machine learning model, and providing the historical PSTisignaling for storing as the consolidated database.WFMB / CiDRA Docket Nos.712-002.475-1 / CCS-0224WO 19. A method according to claim 18, wherein the machine learning model includes one of the following: ^ Multiple linear regression, ^ Random Forest Regressor, ^ Support Vector Machine Regressor, and ^ Neural Network.

20. A method according to claim 19, wherein the method comprises configuring the machine learning model to use a Singular Value Decomposition (SVD) that decomposes a matrix into component matrices.

Images