Integrated unloading automation system and method

Abstract

Method and integrated system for automation of discharge to a mining crusher that includes the analysis of oversize and crushable objects from the extraction truck, and that tracks by means of images associated with a trained convolutional neural network, the oversize and crushable objects as they fall and move through the crusher bowl, analyzing the material flow conditions in the crusher, ordering the start of the crusher hammer in the bowl when there is an oversized material, such as a rock, stopping the discharge from the truck automatically and when it is released and the flow is resumed in the crusher, it generates the order for a truck to start or continue the discharge to the crusher, achieving an auto- discharge system to the crusher. The analysis of the crushable objects from the extraction truck is carried out through artificial vision comprising a deep learning tool, for detection of non-crushable and oversize during the fall of the material into the crusher bowl (or discharge chute).

Description

[0001] INTEGRATED AUTOMATED UNLOADING SYSTEM FOR A MINING CRUSHER (PREVIOUS ART)

[0002] The unloading of heavy-duty trucks carrying extracted material to crushers in the mining industry is a crucial function in the process of obtaining valuable minerals. Ideally, unloading should be carried out quickly and without interruption, ensuring that the extraction process, truck movement, and crushing are as uniform and continuous as possible. However, the material extracted from the mine faces is not uniform, which sometimes results in oversized rocks being transported to the crusher. Furthermore, during mining operations, the shovels used to load the trucks occasionally lose parts of their buckets, such as teeth, pieces of the bucket itself, or similar components.Both oversized and metallic elements mentioned above can create a bottleneck in the function of the crusher, especially metallic elements since these are uncrushable and cannot even be chopped by the hammer crushers that are arranged in the crusher bowl to reduce the oversized elements.

[0003] In light of the above, it is vitally important to be able to detect in a timely manner, from the unloading of the truck to the crusher, the elements that may cause stoppages in the regular operation of the crusher that affect the flow of the entire process, since the lack of continuity in the unloading and crushing generates an overall delay downstream in the grinding stages and subsequent processes, generating a decrease in production and leading to a significant economic loss.

[0004] Currently, radar is used to detect uncrushable objects. While radar is used, it has limited resolution and can produce inaccurate measurements when there is a high presence of metallic material in suspension or within the material itself. In addition, camera solutions have recently been implemented to visualize objects as they are loaded onto the truck. However, while existing solutions represent a significant advance in the early detection of unwanted elements in the crusher, they do not address the need for an automated truck unloading process—that is, a method with an associated system that enables automatic unloading into the crusher.

[0005] Prior art demonstrates solutions that utilize cameras to view the surface of the haul truck, as well as cameras to visualize what occurs in the crusher bowl. For example, WO2018148832 describes a method for measuring the size distribution and / or hardness of freely falling rock pieces. This method involves projecting at least one laser line onto the falling rock pieces using a laser device; capturing images of the falling rock pieces at an angle from at least one laser line using at least one camera; and obtaining size distribution data for the falling rock pieces based on data derived from a topographic map generated from the captured images.Certain embodiments further comprise: obtaining at least one volume and area measurement of individual rock pieces from the topographic map; performing data analysis on at least one of the rock piece volume and area measurements to reduce at least one of the sampling and measurement errors; determining the size distribution of falling rock pieces based on the data analysis; and optionally, evaluating a rock hardness index. A method is further provided comprising: producing two topographic maps of the pieces from captured images; and obtaining the volume of pieces from the topographic map by summing the average volumes from each of the topographic maps.

[0006] The document US2020049517 (equivalent to Chilean registration CL66.237) describes systems and methods for estimating the magnitude of an ore load, determining the particle size distribution of the ore, recognizing foreign material in ore images, monitoring the ore, and monitoring mineral processing equipment. An image of an ore haul truck is received from an image capture device. Data related to the ore load on the haul truck are detected with a scanner. An estimated magnitude of the ore load is calculated by accessing the ore load data, the image data, and the ore haul truck data. A machine learning module is used to identify a region of ore in an ore image and to recognize ore particles or objects within the image. A particle size value for the ore is calculated.An indicator is generated that estimates the condition of the mineral processing equipment using the calculated value and operating parameters.

[0007] US2008192987 describes a method for measuring the size distribution and / or hardness of freely falling rock fragments. The method comprises projecting at least one laser line onto the falling rock fragments using a laser device; capturing images of the falling rock fragments at an angle to the at least one laser line using at least one camera; and obtaining size distribution data of the falling rock fragments based on data obtained from a topographic map generated from the captured images.Certain embodiments further comprise: obtaining at least one volume and area measurement of individual rock pieces from the topographic map; performing data analysis on at least one of the volume and area measurements of the rock pieces to reduce at least one of the sampling and measurement errors; determining the size distribution of the falling rock pieces based on the data analysis; and optionally, evaluating a rock hardness index for the rock. Additionally, a method is provided comprising: producing two topographic maps of the pieces from captured images; and obtaining the volume of the pieces from the images.

[0008] Unlike prior art, the present invention provides a method and system capable of characterizing, through machine vision (MV) and artificial intelligence (AI), the behavior of the ore being unloaded from mining trucks. This involves identifying uncrushable objects, tracking larger rocks that could cause a process stoppage, and detecting process equipment such as rock hammers operating within the crusher when required, as well as trucks, among others. In a coordinated manner, and integrated with radar technology, the self-unloading method and system of the present invention optimally and safely controls the unloading commands in the mining process, alerts to ore obstruction or the presence of uncrushable material, and stops the crushing process—for example, the crusher motor—preventing damage to the asset and significantly reducing process downtime.

[0009] Furthermore, the solution of the present invention observes the crusher's discharge area, whereas in the prior art, specifically document CL66237, the sensors observe the truck hoppers and conveyor belts, not the crusher bowl. Therefore, the observation method of the present invention operates in a different spectral band, reducing the impact of dust and improving its visibility performance.

[0010] The self-discharge system of the invention considers active lighting for night operation, whereas in the prior art nothing similar is described for cameras.

[0011] The download method of the invention is based on different stages that generate different information and the detection stage is based on neural networks: this means that it uses a hybrid of real and simulated data for the detection of rocks, rock pick and uncrushable rocks in different lighting conditions.

[0012] A computer vision stage uses the detections to track each object and map its trajectory over time. It also allows for the generation of morphological classifications. All of this is for the automatic control of the crusher's discharge and status.

[0013] In the prior art, for example, document CL66237, a deep learning algorithm is used for material anomaly detection and segmentation. No mention is made of object tracking over time.

[0014] The self-discharge method of the invention uses a combination of optical detection and flow measurement to estimate the proportion of material discharged in real time and the material flow rate in the cup. These calculations are not mentioned in the prior art.

[0015] BRIEF DESCRIPTION OF THE FIGURES

[0016] Figure 1: represents a flowchart that illustrates the stages involved in the self-discharge method of the invention.

[0017] Figure 2: represents a general scheme of the method of the invention

[0018] Figure 3: is an image that illustrates an example of visibility states.

[0019] Figure 4: This image illustrates the detection capabilities of Omnidirectional Vision. In the magnified area, the tracking stage assigns an identification to the rock and traces its trajectory through the video. Figure 5: The image on the left represents an example of optical flow for calculating material flow; the arrows are velocity vectors. The image on the right shows the segmentation of the moving region for calculating occupancy and granularity at a 26% level.

[0020] DESCRIPTION OF THE INVENTION

[0021] The integrated method for automating unloading to a mining crusher, which includes analyzing oversized and uncrushable objects from the haul truck, involves a controller capturing high-speed images in the near-infrared (NIR) spectrum as the ore falls from the haul truck's hopper at a given site. The images are captured in real time and processed by a trained deep learning system. This system detects the camera's visibility of the unloading, identifies the images, and records events such as oversized rocks and uncrushable elements, along with the corresponding unloading truck. This information is then added to the deep learning system's database. Near-infrared (NIR) spectroscopy is a technology that analyzes the interaction between light and matter to generate a spectrum.In the present invention, the chosen spectral operating range was NIR. Image sensors more sensitive to this spectrum are used, and the observation system incorporates 850nm illumination and 850nm filters to avoid light contamination from other spectral bands that could dazzle the cameras. Furthermore, this band is non-invasive to the human eye, allowing for high-power illumination without affecting operators in the surrounding area. To this end, the image sensors are strategically placed both in the discharge area and in certain areas of the crusher to optimize observation of the discharge process.

[0022] Once the discharge surface has been identified and the events recorded, a morphological classification of the detected events is performed. A monitoring stage is then carried out on the ore as it passes through the crusher bowl, focusing on oversized or uncrushable material recorded as events, thus generating an early warning system. If an oversized or uncrushable element is detected that could obstruct the flow to the crusher, a change in the discharge conditions is implemented, potentially halting the process or indicating the step that must be taken when the material is discharged to the crusher.

[0023] With the information provided during the oversize or uncrushable material identification stage, the controller authorizes the discharge into the crusher bowl. A high-resolution near-infrared (NIR) image capture stage then begins, recording the ore's fall from the truck hopper to the crusher bowl. During the fall, any uncrushable or oversize material already identified on the surface of the haul truck hopper is monitored. If the oversize exceeds the crushing setting (crush post opening or height), the system automatically initiates a rock-breaking stage to attempt to reduce the oversize detected during the fall from the hopper to the crusher bowl. The images captured in real time are processed by a deep learning system that detects visibility, which can present different states, as illustrated in Figure 3.With adequate visibility, discharge events are identified, including both oversized and uncrushable material, as well as whether the crusher is operational. If the rock-breaking hammer has been activated, it must be clear of the crusher bowl for discharge to continue. To perform the aforementioned steps, the method involves capturing images of the crusher bowl and spider.

[0024] To determine the discharge flow to the crusher, the method involves morphologically characterizing the material falling into the crusher bowl, as well as analyzing the material dynamics to identify its occupancy within the bowl. The monitoring process includes tracking any oversized or uncrushable material detected through real-time image capture, as illustrated by an amorphous line in Figure 4. If the morphological characterization corresponds to an oversized, uncrushable, or bowl blockage event, the method generates alerts and alarm triggers (Trip), prompting a change in the discharge decision or the creation of a virtual barrier using the rock pick.These alert stages configure the crusher's safety, in which the self-unloading controller determines whether to continue unloading, change the unloading speed, or stop the operation. This integrates self-unloading from the moment the extraction truck parks in the crusher's unloading area, as well as within the crusher itself, triggered by tracking an event from the truck's hopper or detecting it during the fall into the bowl. This results in a much more dynamic process, making unloading faster, reducing the time between unloading operations, and increasing the crusher's availability.

[0025] The crusher's detection and alert system acts in response to obstructions, allowing identification of when the crusher's cup or spider is obstructed by rocks, characterizing their location and size.

[0026] The method also includes determining the percentage of occupancy in the crusher cup, making it possible to deliver real-time information on the percentage of the area occupied with ore in the crusher cup.

[0027] The morphological characterization performed by the method of the invention comprises calculating the percentage of particle size distribution by material type within the crusher bowl in real time as the ore moves and breaks down (changing its shape) during its advance to the center of the bowl to be crushed. An example of this is shown in Figure 5. This characterization allows for the generation of statistics on material dynamics (Identification, Labeling, Tracking) through a real-time video window, identifying and labeling the rocks of interest, showing their trajectory, speed, and a graphical alarm in case they come to a standstill.

[0028] Furthermore, the generation of an alarm for the presence of uncrushable material during the truck unloading process and in the crusher bowl is carried out in two stages: early on by observing the process of pouring ore from the truck and during a second stage when it moves in the crusher bowl.

[0029] This alarm is transmitted to the controller to alert and potentially stop the crusher motor, and is captured and displayed graphically on the controller screen.

[0030] The method involves categorizing and tracking uncrushable materials during the truck unloading process and in the crusher cup, categorized according to type and impact on the process, and through a graphical interface projecting the tracking of the detected elements.

[0031] The virtual barrier for the rock crusher hammer is designed to identify and track the hammer during the operation of the primary crusher. The safety and operational benefits are:

[0032] • Picaroca hammer in parking area, avoid unloading with the hammer in the cup.

[0033] • Compensate for the lack of a parking sensor on the rock drill. • Supplement with a rock drill parking sensor (physical sensor).

[0034] • Asset care.

[0035] Avoid downtime at the plant for replacement of damaged rock spikes due to this risk.

[0036] The method involves using a virtual truck presence sensor to identify mining trucks and support equipment on the unloading platform(s) and monitor their ore unloading process into the crusher bowl. It also includes using a virtual haul truck parking sensor to supervise the truck's parking process on the platform, providing information on distance, location, and deviation of the truck, identifying potential unloading outside the operating area, and triggering a safety interlock.

[0037] Additionally, the method includes activating and deactivating water sprinklers to the crusher and fulfills the function of activating and deactivating the need for operation of the dust mitigation (pollution) system based on the conditions of the environment such as identifying equipment operating in the area, incorporating a criterion for activating sprinklers and avoiding mud.

[0038] The integrated automated unloading system for a mining crusher includes means for detecting oversized or uncrushable objects from a haul truck. It consists of a unloading controller, which includes connections to high-speed cameras located in the haul truck parking area or slab at the crusher. It also includes connections to high-resolution cameras around the crusher bowl. Both types of cameras are integrated into a machine vision system assisted by near-infrared (NIR) spectrum capture and deep learning for the detection, tracking, and measurement of various materials, as well as the identification of oversized, uncrushable, and rock-breaking objects. The system includes a high-power infrared (IR) lighting system, allowing 24 / 7 monitoring without affecting operator visibility.In addition, the system includes a Vision Free-Fall module, designed to detect and track material as it falls from the truck into the crusher bowl. It also includes a Vision All-aware module, which enables detailed analysis of the material in the crusher bowl.

[0039] The Vision Free-Fall module includes a logic function to automatically monitor the visibility of each camera, based on histograms and decision thresholds. This logic function enables or disables detection processes to prevent false positives in low-visibility conditions. The visibility detector can differentiate between poor visibility and an ongoing lightning strike (see Figure 3). The latter is used as an event indicator for the temporal processing of the images.

[0040] Next, a convolutional neural network processes the images captured by the high-speed camera to detect the following classes of objects:

[0041] - Parked trucks.

[0042] - Rocks in free fall.

[0043] - Unbreakable in free fall.

[0044] When an object is detected, properties such as its relative position and surface area are extracted. Based on these characteristics, the method described above allows for the classification of rocks based on their size into two categories: Small rock

[0045] - Oversized rock

[0046] Then, the system, through the controller, uses this information to catalog and track each object in real time.

[0047] The Vision All-aware module comprises a deep convolutional network for the detection of different classes:

[0048] • Rocks

[0049] • Rock pickaxe

[0050] • Unbreakable

[0051] • Crusher spider

[0052] An example of the conscious vision module is illustrated in Figure 4.

[0053] When the convolutional neural networks detect one of the classes, they generate properties such as the relative position and surface area of ​​each detected object. Based on these characteristics, a classification stage allows rocks to be differentiated based on their size into two categories:

[0054] Small rock

[0055] Oversized rock

[0056] The tracking stage then uses this information to catalog each object and track it over time (example in Figure 4). This allows the system to monitor the hammer's status and generate a rock-pepper hammer status signal.

[0057] The Vision All-aware module comprises a submodule that focuses on acquiring morphological and dynamic characteristics of each processed image. This submodule generates and controls regions of interest within the crusher bowl to measure the concave and blockage status based on decision thresholds. It also calculates the optical flow of the images to obtain a spatiotemporal description of the material, differentiating whether it is stationary or moving. Furthermore, the optical flow density is used to estimate the percentage of small material being discharged into the bowl in real time.

[0058] However, the information obtained by both modules is sent to a download decision block that considers aspects such as:

[0059] Picaroca state

[0060] Concave block

[0061] Presence of uncrushable materials

[0062] In addition to generating the following alerts:

[0063] Virtual barrier of picaroca

[0064] Alert for blockage in cup

[0065] Alert for oversize detection during download

[0066] Alert for oversized item stuck in cup.

[0067] Additionally, trip alerts are generated in the following cases:

[0068] Early trip due to detection of uncrushable material in free fall.

[0069] Trip due to detection of uncrushable material in cup.

[0070] Trip due to oversize detection in concave.

[0071] Trip through fortification.

Claims

RECLIN DICATION IS 1. Integrated method for automating the unloading of a haul truck to a mining crusher that automates the decision to unload haul trucks to the crusher, avoiding errors due to poor observation of events and optimizing unloading times and crusher availability, CHARACTERIZED in that it comprises the following stages: a. Analyze oversized or uncrushable objects coming from the extraction truck by means of a controller to capture images in the NIR spectrum at high speed in the fall of ore from the hopper of said ore extraction truck; where the images are captured in real time and processed through a trained deep learning system, through which the visibility of the discharge of the camera is detected, the images are identified and the events that may be given by oversized rocks, uncrushable elements are recorded as to which discharge truck they correspond to, which are part of the database of the deep learning system; b. Once the truck unloading has been identified and the events recorded, perform a morphological classification of the detected events and carry out a monitoring stage of the ore in transit through the crusher cup of the oversized or crushable pieces recorded as events, generating an early warning of the same; c. Once the events have been identified, the controller authorizes the discharge into the crusher bowl, and high-resolution NIR images of the mineral falling from the truck hopper to the crusher bowl are captured; d. monitor these events, in which, if an oversize exceeds the crushing adjustment, automatically initiate a stage in which a rock-breaking hammer intervenes to try to reduce the oversize detected both on the surface of the hopper and during the fall from it to the crusher cup; e. determine if the crusher is in a condition to operate without the rock-breaking hammer being in the vicinity of the crusher bowl so that the discharge can continue; f. To morphologically characterize the material that falls into the crusher bowl and analyze the dynamics of the material to identify its occupation in the bowl; g. track the morphologically characterized material through real-time image capture; and generate an alert if the morphological characterization corresponds to any oversize, uncrushable or blockage event in the cup; h. Change the unloading decision or generate a virtual barrier with a rock-breaking hammer; i. Determine through said unloading controller whether to continue unloading, change the unloading speed, put on hold or stop the operation, achieving the integration of self-unloading from the very moment the extraction truck parks in the unloading area to the crusher.

2. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it comprises that if upon detecting an oversized or uncrushable element that may cause an obstruction in the flow to the crusher, it generates a change in the unloading conditions, being able to stop the process, put on hold or slow the speed of the crusher when the unloading to the crusher is generated.

3. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that said images are captured in real time and processed by a logic function within a convolutional neural network that detects visibility.

4. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it comprises the capture of images in the cup and in the spider of the crusher.

5. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that said detection and alert identifies when the crusher cup or spider is obstructed by rocks or accumulation of material, characterizing its location and size.

6. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it further comprises determining the percentage of occupancy in the crusher cup by delivering real-time information on the percentage of the area occupied with ore in the crusher cup.

7. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that morphological characterization comprises calculating the percentage of particle size by material type within the crusher cup in real time as the ore moves and breaks (changing its shape) during its advance to the center of it to be crushed.

8. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that said morphological characterization comprises capturing in a real-time video window and identifying and labeling the rocks of interest, showing their trajectory, speed and graphic alarm in case they stop.

9. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it further comprises generating an alarm for the presence of uncrushable material during the unloading process of the truck and in the crusher bowl by means of two stages: early on by observing the process of pouring ore from the truck and during a second stage when it moves in the crusher bowl; where this alarm is transmitted to the controller to alert and potentially stop the crusher motor.

10. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 9, CHARACTERIZED in that said alarm is captured and presented graphically on a controller screen.

11. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it comprises the categorization and tracking of uncrushable material during the unloading process of the truck and in the crusher cup, categorized according to type and impact for the process and through a graphical interface projecting the tracking of the detected elements.

12. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it further comprises a virtual barrier for the rock-breaking hammer to identify and follow the hammer during the operation of the primary crusher.

13. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it further comprises the use of a virtual truck presence sensor, support equipment on the unloading slab(s) and its ore unloading process into the crusher cup.

14. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it further comprises using a virtual sensor for the loading of the haul truck and monitoring the process of parking the truck on the slab.

15. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that it further comprises activating and deactivating water sprinklers to the crusher and a dust (pollution) mitigation system based on the environmental conditions.

16. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 15, CHARACTERIZED in that it comprises identifying equipment operating in the environment, incorporating a criterion for activating sprinklers and avoiding mud.

17. Integrated method for automating the unloading of a haul truck to a mining crusher according to claim 1, CHARACTERIZED in that said trained deep learning system is a trained neural convolutional network.

18. Integrated system for automating the unloading of a haul truck to a mining crusher that automates the decision to unload haul trucks to the crusher, avoiding errors due to poor observation of events and optimizing unloading times and crusher availability, CHARACTERIZED in that it comprises a unloading controller, in which said unloading controller includes high-speed cameras arranged in the parking area or slab of the haul trucks at a crusher; furthermore, said controller comprises the connection of high-resolution cameras around the crusher bowl, where both types of cameras are integrated into a machine vision system assisted by capture in the NIR spectrum and a learning system for the detection, tracking and measurement of various materials, as well as the identification of events such as oversize, uncrushable material and rock hammer;Connected to said controller is a high-power infrared (IR) lighting system, as well as a free-fall vision module, which is intended to detect and track the material during its fall from the truck to the crusher bowl; and an all-aware vision module, which allows for detailed analysis of the material in the crusher bowl; a virtual sensor for the presence of trucks, support equipment on the unloading slab(s) and their ore unloading process into the crusher bowl; and a virtual sensor for the approach of the haul truck and to monitor the truck parking process on the slab.

19. Integrated system for automating the unloading of a haul truck to a mining crusher according to claim 18, CHARACTERIZED in that said Vision Free-Fall module comprises a logic function to automatically monitor the visibility of each camera, based on histograms and decision thresholds.

20. Integrated system for automating the unloading of a haul truck to a mining crusher according to claim 19, CHARACTERIZED in that it further comprises a visibility detector to differentiate between poor visibility and / or an unloading operation.

21. Integrated system for automating the unloading of a haul truck to a mining crusher according to claim 19, CHARACTERIZED in that it comprises a convolutional neural network for processing the images captured by the high-speed camera to detect different kinds of objects, such as parked trucks, free-falling rocks, or uncrushable objects in free fall.

22. Integrated system for automating the unloading of a haul truck to a mining crusher according to claim 18, CHARACTERIZED in that said vision-aware module comprises a deep convolutional network for the detection of different classes, such as rocks, rock hammer, uncrushable, and crusher spider.

23. Integrated system for automating the unloading of a haul truck to a mining crusher according to claim 18, CHARACTERIZED in that said Vision All-aware module comprises a submodule that focuses on acquiring morphological and dynamic characteristics of each processed image.

Images