Integrated vertical sand bin thickening experimental system and method based on machine vision

Abstract

The application discloses an integrated vertical sand bin thickening experimental system and method based on machine vision. In view of the problems of single function, inability to simulate compression section pressure, lack of visual observation and unreplaceable module of the existing device, the system is provided with a transparent column body, the upper end of which is detachably connected with a feeding well module or a sealing top cover, the lower end of which is detachably connected with a cone bottom module or a sealing bottom cover, and a pressurizing piston unit with a drain hole is detachably installed in the column body; a machine vision unit comprises a visible light camera and a near-infrared camera, which respectively shoot flocculation body settling images and wall or cone bottom images; a sensor unit comprises a pressure sensor and a displacement sensor embedded in the inner wall; a data acquisition and processing unit is electrically connected with each component and runs an algorithm, and extracts flocculation body contour, settling velocity, diffusion angle and wall gray entropy, texture roughness and other characteristics. The system is used for vertical sand bin feeding well optimization, compression section height determination and cone bottom discharge verification experiment.

Description

Technical Field

[0001] This invention relates to the field of mine backfilling technology. More specifically, this invention relates to an integrated vertical sand silo thickening experimental system and method based on machine vision. Background Technology

[0002] Vertical sand bins are the core concentration equipment in paste filling systems, and their design directly determines the underflow concentration, solids throughput, and operational stability of the filling system. Currently, the design of vertical sand bins mainly relies on experience-based selection and analogy methods, and systematic experimental verification methods are lacking for key design aspects such as feed well structure, compression section height, cone bottom and cone angle, and slurry making system.

[0003] Existing tailings thickening experimental devices have the following shortcomings: First, small settling columns have limited functionality and cannot simulate the pressure effect of the mud layer in the compression section; second, while consolidation apparatuses with pressurized pistons can measure the pressure-concentration relationship, they are usually not equipped with a visual observation system, making it difficult to obtain dynamic evolution information during the compression process; third, existing devices lack the ability to quickly replace the feed well module and the cone bottom module, making it impossible to conduct comparative experiments on different structural forms on the same device; fourth, existing devices lack real-time monitoring methods for pressure values ​​and displacement during pressurization, making it difficult to accurately control experimental conditions. Therefore, developing an integrated experimental system that combines feed well optimization, compression section pressure measurement, cone bottom discharge verification, and machine vision observation has significant engineering application value. Summary of the Invention

[0004] One object of the present invention is to solve at least the above-mentioned problems and to provide at least the advantages that will be described later.

[0005] Another objective of this invention is to provide an integrated vertical sand silo thickening experimental system based on machine vision.

[0006] To achieve these objectives and other advantages according to the present invention, an integrated vertical sand silo thickening experimental system based on machine vision is provided, comprising:

[0007] A transparent column is used to contain tailings slurry. The upper end of the transparent column is detachably connected to a replaceable feed well module or a sealed top cover, and the lower end of the transparent column is detachably connected to a replaceable cone bottom module or a sealed bottom cover. The lower part of the replaceable cone bottom module is equipped with a discharge valve.

[0008] A pressurizing piston unit, which is detachably located inside a transparent cylinder, includes a force-applying device and a piston head connected to the force-applying device, with multiple drainage holes provided on the piston head.

[0009] The machine vision unit includes a visible light camera, a visible light supplementary light, a near-infrared camera, and a near-infrared supplementary light. The visible light camera, in conjunction with the visible light supplementary light, is directed towards the area below the outlet of the replaceable feed well module to capture the floc settling process. The near-infrared camera, in conjunction with the near-infrared supplementary light, is directed towards the wall of the compression zone to capture images of the compression zone wall. Alternatively, the near-infrared camera, in conjunction with the near-infrared supplementary light, is directed towards the cone bottom area to observe the fluidization state and discharge effect of the tailings at the bottom during the pulping process.

[0010] The sensor unit includes a pressure sensor embedded in the inner wall of the transparent cylinder with the probe end face flush with the inner wall and a displacement sensor mounted on the force application mechanism, which are used to monitor the pressure value applied to the mud layer and the piston displacement in real time, respectively.

[0011] The data acquisition and processing unit, which is electrically connected to the pressure sensor, displacement sensor, visible light camera, and near-infrared camera, is used to run image processing algorithms. It extracts floc contours, settling velocity, and radial diffusion angles from floc settling images acquired by the visible light camera; it extracts grayscale entropy, texture roughness, and fluid channel density features from compression zone wall images acquired by the near-infrared camera to characterize the stability of the mud layer structure near the wall; or it extracts grayscale entropy, texture contrast, and motion pixel ratio features from cone bottom area images acquired by the near-infrared camera to comprehensively judge the fluidization state and discharge potential inside the cone bottom.

[0012] Preferably, the replaceable feed well module includes at least two of the following: straight pipe feed well, guide channel feed well, and self-dilution channel feed well.

[0013] Preferably, the cone angle of the replaceable cone bottom module is at least two of 45°, 60°, and 75°, and the cone bottom is provided with a pulping interface for replaceable nozzles.

[0014] Preferably, the force-applying device is a pneumatic pressurization type, and the output end of the force-applying device is connected to a piston rod. The piston rod is vertically arranged and has a hollow structure. The lower end of the piston rod is connected to the piston head, and multiple drainage holes are connected to the hollow part of the piston rod. A support net is provided on the lower surface of the piston head, and filter cloth is covered on the upper and lower surfaces of the support net.

[0015] This invention also provides an integrated vertical sand silo density testing method based on machine vision. The integrated vertical sand silo density testing system based on machine vision of this invention includes the following three independent testing modes:

[0016] Mode 1: Replaceable feed well module selection experiment

[0017] Step A1: Install a sealed bottom cover at the bottom of the transparent column and install a replaceable feed well module at the top;

[0018] Step A2: Prepare tailings slurry and flocculant solution in two mixing tanks respectively, and continuously feed them into the transparent column through a feed pump;

[0019] Step A3: Acquire floc settling images of the area below the outlet of the replaceable feed well module using a visible light camera, extract the floc contours using an image processing algorithm, track the floc movement trajectory using a nearest neighbor matching algorithm, and calculate the axial settling velocity and radial diffusion angle.

[0020] Step A4: Replace with replaceable feed well modules of different structures. Repeat steps A1 to A3 and compare to determine the optimal replaceable feed well module. The module with the highest settling velocity and the smallest standard deviation of diffusion angle and coefficient of variation is the optimal one.

[0021] Mode 2: Compression Section Pressurization Experiment

[0022] Step B1: Using the operations of steps A1 and A2, continuously feed material into the transparent cylinder to form an initial mud layer, stop feeding, remove the top replaceable feeding well module, and install the pressurized piston unit into the transparent cylinder from the top opening, and make the piston head contact the surface of the mud layer.

[0023] Step B2: Apply pressure gradually from low to high using the pressurizing piston unit, while continuously acquiring images of the compression zone wall using a near-infrared camera. Extract grayscale entropy, texture roughness, and fluid channel density characteristic parameters, and monitor the mud layer compression rate using a displacement sensor. Finally, combine the mud layer compression rate and characteristic parameters to determine the paste state. If the paste state has reached the target paste state, record the currently applied pressure value as the target pressure P. target And samples were taken to determine the underflow concentration;

[0024] Step B3: Based on the target pressure P target The required compression section height for reverse-push vertical sand bins;

[0025] Mode 3: Cone-bottom and pulping experiment

[0026] Step C1: Install a replaceable cone-bottom module at the bottom of the transparent column and a replaceable feed well module at the top;

[0027] Step C2: Continuously feed material into the transparent column to form a mud layer using the operation in step A2, stop feeding and complete settling, then make slurry, observe the fluidization state of the tailings at the bottom using a near-infrared camera, and qualitatively judge the slurry making effect.

[0028] Step C3: Open the discharge valve, continue to observe the discharge process through a near-infrared camera, sample and measure the discharge concentration, and use grayscale entropy, texture contrast and motion pixel ratio fusion to determine the overall fluidization state.

[0029] Step C4: Close the discharge valve, replace with different replaceable cone bottom modules, and repeat steps C1 to C3 to determine the optimal cone bottom angle based on the overall fluidization state.

[0030] Preferably, the method for judging the state of the ointment is as follows:

[0031] Criterion a: Observe the displacement sensor data. If the mud layer compression rate is consistently less than 0.05 mm / min within a continuous 10-minute time window, it indicates that the macroscopic compression of the mud layer has nearly stopped and the pore water discharge is basically complete.

[0032] Criterion b: When criterion a is satisfied, observe the near-infrared wall image obtained by the near-infrared camera, check the liquid flow channels in the wall image and calculate the gray entropy. When the liquid flow channels disappear and the maximum fluctuation of the gray entropy within a continuous 10-minute time window is less than 10%, then criterion b is satisfied, and it is determined that the mud layer has reached the target paste state.

[0033] If criterion a is met but criterion b is not met within the preset time, it indicates that there is still a local drainage channel at the wall surface. The pressure can be increased and the observation can be repeated.

[0034] If criterion a cannot be satisfied, the experiment is terminated and the current state is recorded.

[0035] Preferably, in step B2, the current pressure value P when the paste-like state is reached is recorded. sample Sampling and measuring the underflow concentration C sample If C sample The target underflow concentration C has been reached or exceeded as determined by the design requirements. target Then this pressure value is the standard pressure P. target If C sample Not reached C target Continue increasing the pressure until the target underflow concentration C is reached. target Record the corresponding compliance pressure, which is the compliance pressure P. target .

[0036] Preferably, the target pressure P is recorded based on visual criteria. target and its corresponding undercurrent concentration C target Mark the point on the pressure-concentration coordinate system, and determine whether verification is required according to the following rules:

[0037] The situations requiring verification include:

[0038] When pressure fluctuations, surface cracks in the mud layer, and repeated liquid flow channels on the wall occur during the experiment, the reliability of the target points needs to be verified.

[0039] The verification method is as follows:

[0040] Under the pressure of meeting the standard P targetBased on this, continue increasing the pressure according to the original pressurization gradient to the upper limit. After each pressure level stabilizes, sample and measure the underflow concentration, and plot a complete pressure-concentration curve. The target point (P...) target C target Compare with the pressure-concentration curve:

[0041] If the target point is located in a section where the slope of the curve decreases significantly, then the target point is reliable.

[0042] If the target point is located on a section of the curve with a steep slope, that point is unreliable. Then, move along the curve towards higher pressure and take the pressure value corresponding to the point where the curve slope first drops below 50% of the peak slope as the target pressure P. target ;

[0043] If the curve shows that after reaching the equipment's rated maximum working pressure, the concentration change rate corresponding to the two adjacent pressure levels remains higher than 50% of the peak slope for 30 consecutive minutes, it indicates that the tailings have not yet reached the compression limit under the existing equipment conditions. The design target should be reset or the pressurization equipment should be replaced.

[0044] Preferably, the method for qualitatively judging the slurry-making effect is to observe the fluidization state of the bottom tailings using a near-infrared camera:

[0045] If pixel changes are continuously observed in the wall image within a 30-second observation window by comparing consecutive frame images, it is determined that there is obvious flow.

[0046] If there is no obvious pixel change in the wall image within the 30-second observation window, it is determined that there is no obvious flow.

[0047] If the area at the bottom corner of the cone remains static, without any pixel changes and with no change in wall shape for 2 minutes, it is determined to be a dead zone.

[0048] Record whether there is obvious flow on the wall surface and whether there are dead zones to qualitatively judge the pulping effect.

[0049] Preferably, the method for obtaining the motion pixel ratio is as follows:

[0050] Motion pixel ratio = number of motion pixels / total number of pixels in the image × 100%. The motion pixel ratio is calculated by the data acquisition and processing unit.

[0051] The method for determining fluidization state by fusing grayscale entropy, texture contrast, and moving pixel ratio is as follows:

[0052] If the proportion of moving pixels is higher than the first preset threshold and the grayscale entropy remains high within a continuous 10-second observation window, it is determined that the fluidization is active, the fluidization state is good, and the discharge potential is high.

[0053] If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy drops by more than 50% from a high value, it is determined that the mud layer is static or there is a dead zone. In this case, the overall fluidization state is poor, and there may be risks of poor material discharge or failure to meet the concentration standard. The first preset threshold is greater than the second preset threshold.

[0054] If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy increases, it is determined to be interference caused by changes in lighting or surface texture. In this case, the interference needs to be eliminated and the determination needs to be made again.

[0055] The present invention has at least the following beneficial effects:

[0056] First, this invention integrates four major functions—feed well structure optimization, compression section height determination, cone bottom discharge verification, and machine vision observation—into a single device, enabling one-stop experimental verification of key design aspects of vertical sand silos.

[0057] Secondly, the feed well module, pressurized piston system and conical bottom module of this invention all adopt quick-installation interfaces, which can be replaced within minutes, making it convenient to carry out comparative experiments of different structural forms and improving experimental efficiency.

[0058] Third, this invention employs a dual-camera system of visible light and near-infrared. The visible light camera is used to observe the floc settling process at the feed well outlet, extracting floc trajectory, settling velocity, and radial diffusion angle. The near-infrared camera is used to observe images of the compression zone wall, capturing wall slip marks and fluid channel distribution, and extracting features such as grayscale entropy, texture roughness, and channel density to characterize the stability of the mud layer structure near the wall. These wall feature parameters, together with displacement sensor data, constitute a dual criterion for the paste state, upgrading experimental evaluation from qualitative observation to quantitative analysis, and providing a reliable pressure basis for sand silo height design.

[0059] Fourth, the pressurized piston system of this invention can realistically simulate the pressure conditions of mud layers at different heights, providing reliable experimental pressure data for the design of the compression section height.

[0060] Fifth, this invention uses a pressure sensor installed on the inner wall of the transparent cylinder and a displacement sensor on the pressurized piston system to obtain pressure values ​​and piston displacement in real time, so that the loading process can be quantified and recorded, and the experimental data can be reproduced.

[0061] Sixth, the present invention divides the experimental method into three independent experimental modes, each mode corresponding to a specific technical problem and experimental configuration, avoiding mutual interference between different experimental objectives, and making the experimental process clear and repeatable.

[0062] Seventh, this invention obtains the mud layer pressure value required to achieve the target paste state through experiments, and uses this value to deduce the height of the vertical sand silo compression section, thus transforming the design of the compression section height from empirical analogy to a method of determination based on experimental data.

[0063] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0064] Figure 1 This is a schematic diagram of the system structure under the three independent experimental modes of the present invention;

[0065] Figure 2 This is a schematic diagram of the replaceable feed well module in this invention;

[0066] Figure 3 This is a schematic diagram of the replaceable cone-bottom module in this invention;

[0067] Figure 4 This is a flowchart of the experimental method of the present invention. Detailed Implementation

[0068] The present invention will be further described in detail below with reference to embodiments, so that those skilled in the art can implement it based on the description.

[0069] like Figure 1 As shown, this invention provides an integrated vertical sand silo thickening experimental system based on machine vision, comprising:

[0070] A transparent column 1 is used to contain tailings slurry. The upper end of the transparent column 1 is detachably connected to a replaceable feed well module 2 or a sealed top cover 11. The lower end of the transparent column 1 is detachably connected to a replaceable cone bottom module 3 or a sealed bottom cover 12. The lower part of the replaceable cone bottom module 3 is provided with a discharge valve 9.

[0071] The pressurizing piston unit 4 is detachably disposed inside the transparent column 1. It includes a force application device 43 and a piston head 41 connected to the force application device 43. The piston head 41 is provided with multiple drainage holes.

[0072] The machine vision unit 5 includes a visible light camera 51, a visible light supplement light 52, a near-infrared camera 53, and a near-infrared supplement light 54. The visible light camera 51, in conjunction with the visible light supplement light 52, is directed towards the area below the outlet of the replaceable feed well module 2 to capture the floc settling process. The near-infrared camera 53, in conjunction with the near-infrared supplement light 54, is directed towards the wall of the compression zone to capture images of the compression zone wall. Alternatively, the near-infrared camera 53, in conjunction with the near-infrared supplement light 54, is directed towards the cone bottom area to observe the fluidization state and discharge effect of the tailings at the bottom during the pulping process.

[0073] The sensor unit includes a pressure sensor 44 embedded in the inner wall of the transparent column 1 with the probe end face flush with the inner wall and a displacement sensor 45 mounted on the force application mechanism, which are used to monitor the pressure value applied to the mud layer and the piston displacement in real time, respectively.

[0074] The data acquisition and processing unit 8, which is electrically connected to the pressure sensor 44, displacement sensor 45, visible light camera 51, and near-infrared camera 53, is used to run image processing algorithms. It extracts floc contours, settling velocity, and radial diffusion angle from the floc settling images acquired by the visible light camera 51; it extracts grayscale entropy, texture roughness, and fluid channel density features from the compression zone wall images acquired by the near-infrared camera 53 to characterize the stability of the mud layer structure near the wall; or it extracts grayscale entropy, texture contrast, and motion pixel ratio features from the cone bottom area images acquired by the near-infrared camera 53 to comprehensively judge the fluidization state and discharge potential inside the cone bottom.

[0075] Specifically, the transparent column 1 is made of plexiglass, with an inner diameter of up to 200mm and a total height of up to 2000mm. The top of the transparent column 1 has a quick-install flange 13 interface for installing the replaceable feed well module 2; the bottom has a quick-install flange 13 interface for installing the replaceable conical bottom module 3. The transparent column is mounted on a steel bracket 10. A sealing ring 49 is provided around the piston head.

[0076] The material in the mixing tank 7 is continuously fed into the transparent column 1 at a set flow rate by the feed pump 6. There are two mixing tanks 7, one containing tailings slurry and the other containing flocculant.

[0077] The visible light camera 51 continuously acquired images of the area below the feed well outlet at a frame rate of 60fps for 10 minutes.

[0078] The acquired images are transmitted to the data acquisition and processing unit 8, where Gaussian filtering is performed to remove noise, histogram equalization is performed to enhance contrast, Otsu adaptive threshold segmentation is performed to extract floc regions, and morphological opening and closing operations are performed to filter out small noise points.

[0079] The sensor system includes a pressure sensor 44 and a displacement sensor 45. The pressure sensor 44 is arranged along the height of the side wall of the transparent cylinder 1, with the sensor probe end face flush with the inner wall of the cylinder and not protruding from the inner wall surface. It is used to measure the pore water pressure at different heights within the mud layer. The side wall of the transparent cylinder 1 has threaded holes for installing the sensor. The sensor is screwed in from the outside of the cylinder through these threaded holes, and the mud layer directly contacts the flush probe end face inside the cylinder, preventing interference with the sensor when the piston moves downwards. The displacement sensor 45 is mounted on a force-applying device 43 to monitor the piston displacement in real time and calculate changes in mud layer height. The force-applying device is a pneumatic or hydraulic cylinder.

[0080] Machine vision unit 5, such as Figure 1As shown, a visible light camera 51, equipped with white LED supplementary lighting, is installed at a height of 200-500 mm below the feed well outlet to capture the floc settling process and obtain floc contours, settling velocity, and radial diffusion angle. A near-infrared camera 53, equipped with an 850nm NIR LED supplementary lighting, is installed in the mud layer region in the middle of the compression section to capture images of the compression zone wall, capturing wall slippage traces and liquid flow channel distribution. The acquired wall images undergo feature extraction by the data acquisition and processing unit 8, calculating parameters such as grayscale entropy, texture roughness, and channel density to characterize the stability of the mud layer structure near the wall. When the near-infrared camera 53 and its matching NIR LED supplementary lighting are installed in the cone bottom region, they are used to observe the fluidization state of the bottom tailings and the discharge effect during the pulping process.

[0081] Gray-level entropy is obtained by calculating the entropy value of the image's gray-level histogram. Gray-level entropy reflects the degree of disorder in the pixel gray-level distribution in a wall image, and its mathematical formula is as follows:

[0082]

[0083] Where: L is the total number of gray levels in the image, and p(i) is the probability of gray level i appearing in the image.

[0084] Contrast ratio is commonly used as a quantitative indicator of texture roughness, and its mathematical formula is as follows:

[0085]

[0086] Where: P(i,j) are elements of the gray-level co-occurrence matrix.

[0087] The channel density is directly obtained from the statistical results of the edge detection algorithm, and its mathematical formula is as follows:

[0088]

[0089] Where: N is the number of pixels at the edge of the fluid channel detected in the near-infrared wall image, and A is the total pixel area of ​​the region of interest in the image.

[0090] Figure 1 There are three independent working modes. If you need to conduct the test of the feed well module 2 in mode one, connect the components of the transparent column 1 according to the connection method of mode one and connect the outlet end of the feed pump 6 to the transparent column 1. If you need to conduct the pressurization test of the compression section in mode two, connect the transparent column 1 according to the connection method of mode two and connect the outlet end of the feed pump 6 to the transparent column 1. If you need to conduct the cone bottom and slurry making test in mode three, connect the components of the transparent column 1 according to the connection method of mode three and connect the outlet end of the feed pump 6 to the transparent column 1.

[0091] This invention integrates four functions: feed well optimization, compression section pressurization, cone-bottom discharge verification, and machine vision observation. A transparent column with detachable upper and lower connectors facilitates switching between different experimental modules. A visible light camera is used to observe floc settling, while a near-infrared camera is used to observe the wall surface or cone bottom condition. Pressure and displacement sensors record pressure and displacement data in real time. After image processing, algorithms can extract characteristic parameters such as floc settling velocity, diffusion angle, grayscale entropy, and texture roughness, providing quantitative data for subsequent analysis.

[0092] According to another embodiment of the present invention, the replaceable feed well module 2 includes at least two of the following: a straight pipe feed well, a guide channel feed well, and a self-dilution channel feed well.

[0093] Specifically, the replaceable feed well module 2 has three structural forms, such as... Figure 2 As shown: (a) Straight pipe type feed well, with a simple structure, serving as a benchmark control group; feed well 21 is a straight cylinder; (b) Guide channel type feed well, with a guide channel 22 at the outlet of feed well 21. The guide channel 22 has a small diameter at the upper end and a large diameter at the lower end, used to buffer the impact of flocs; (c) Self-dilution channel type feed well, with a self-dilution port 23 on the side wall of feed well 21. The self-dilution port 23 is a dilution water inlet, used to add dilution water to adjust the feed concentration. All three modules are connected to the transparent column 1 using a quick-connect flange 13, allowing for replacement within 10 minutes.

[0094] The feed well module of this invention includes at least two of three structures: straight pipe, guide channel, and self-dilution channel. All three structures are connected to the column via quick-connect interfaces, allowing for rapid replacement. The effects of different structures on floc settling velocity and diffusion angle can be directly compared experimentally, and the selected optimal structure is supported by measured data.

[0095] According to another embodiment of the present invention, the cone angle of the replaceable cone bottom module 3 is at least two of 45°, 60°, and 75°, and the cone bottom 31 is provided with a pulping interface for a replaceable nozzle 32, one of the detachable connection methods being a bolt connection. Specifically, the replaceable cone bottom module 3 has three cone angle specifications, such as... Figure 3 As shown: (a) is 45°, (b) is 60°, and (c) is 75°. The cone-bottom module is equipped with a pulping interface with a replaceable nozzle 32.

[0096] The cone-bottom module of this invention provides at least two of three cone angles: 45°, 60°, and 75°, and is equipped with a pulping interface with interchangeable nozzles. By changing different cone angles and nozzles, the discharge smoothness and underflow concentration under each combination can be compared, providing a reference for on-site cone-bottom selection and setting.

[0097] According to another embodiment of the present invention, the force-applying device 43 is a pneumatic pressurization type. The output end of the force-applying device 43 is connected to a piston rod 42. The piston rod 42 is vertically arranged and has a hollow structure forming a hollow channel. The lower end of the piston rod 42 is connected to the piston head 41, and multiple drainage holes are connected to the hollow part of the piston rod 42. The hollow part is a hollow channel. The lower surface of the piston head 41 is provided with a support net 46, and the upper and lower surfaces of the support net 46 are covered with filter cloth 47.

[0098] Specifically, the pressurized piston unit 4, as shown in the figure... Figure 1 As shown, the device includes a piston head 41, a piston rod 42, a force-applying device 43, a stainless steel support mesh 46, and a filter cloth 47. The piston head 41 is made of stainless steel and has multiple drainage holes. Below the piston head 41, the stainless steel support mesh 46 and the filter cloth 47 are laid sequentially. The stainless steel support mesh 46 provides structural support to prevent the filter cloth 47 from being crushed under pressure; the filter cloth 47 traps tailings particles, allowing only water to pass through. The top of the piston rod 42 is connected to the output end of the force-applying device 43. The upper side wall of the piston rod 42 has a water outlet 48, which communicates with an external water collection device. The pressed-out water enters the hollow channel of the piston rod 42 through the drainage holes and flows into the external water collection device through the water outlet 48. The force-applying device 43 uses a pneumatic pressurization method, controlling the pressure through a precision pressure regulating valve. The pressurization range is 0~200kPa, with an accuracy of ±1kPa.

[0099] The force-applying device of this invention employs a pneumatic pressurization method, with pressure controlled by a pressure regulating valve. The piston rod has a hollow structure; the pressurized water enters the hollow channel through the drain hole and is discharged from the side wall outlet 48. The support mesh and filter cloth below the piston head prevent tailings particles from passing through, allowing only water to drain, while also preventing the filter cloth from breaking under pressure.

[0100] This invention provides an integrated vertical sand silo density testing method based on machine vision. The integrated vertical sand silo density testing system based on machine vision of this invention is characterized by including the following three independent testing modes:

[0101] Mode 1: Replaceable feed well module 2 - Select Experiment

[0102] Step A1: Install a sealed bottom cover 12 at the bottom of the transparent column 1 and a replaceable feed well module 2 at the top;

[0103] Step A2: Prepare tailings slurry and flocculant solution in two mixing tanks 7 respectively, and continuously feed them into the transparent column 1 through the feed pump 6;

[0104] Step A3: Acquire floc settling images of the area below the outlet of the replaceable feed well module 2 using visible light camera 51, extract floc contours using image processing algorithms, track floc movement trajectories using nearest neighbor matching algorithms, and calculate axial settling velocity and radial diffusion angle.

[0105] Step A4: Replace the replaceable feed well module 2 with different structures, repeat steps A1 to A3, and compare to determine the optimal replaceable feed well module 2. The one with the largest settling velocity and the smallest standard deviation of diffusion angle and coefficient of variation is the optimal one.

[0106] Mode 2: Compression Section Pressurization Experiment

[0107] Step B1: Using the operations of steps A1 and A2, continuously feed material into the transparent column 1 to form an initial mud layer, stop feeding, remove the top replaceable feeding well module 2, and install the pressurized piston unit 4 into the transparent column 1 from the top opening, and make the piston head 41 contact the surface of the mud layer.

[0108] Step B2: Apply pressure gradually from low to high using the pressurizing piston unit 4. Simultaneously, continuously acquire images of the compression zone wall using the near-infrared camera 53, extracting grayscale entropy, texture roughness, and fluid channel density characteristic parameters. Monitor the mud layer compression rate using the displacement sensor 45. Finally, combine the mud layer compression rate and characteristic parameters to determine the paste state. If the paste state has reached the target paste state, record the currently applied pressure value as the target pressure P. target And samples were taken to determine the underflow concentration;

[0109] Step B3: Based on the target pressure P target The required compression section height for reverse-push vertical sand bins;

[0110] Mode 3: Cone-bottom and pulping experiment

[0111] Step C1: Install a replaceable cone bottom module 3 at the bottom of the transparent column 1 and a replaceable feed well module 2 at the top;

[0112] Step C2: Using the operation of step A2, continuously feed material into the transparent column 1 to form a mud layer, stop feeding and complete the settling, and then make slurry. Observe the fluidization state of the tailings at the bottom through the near-infrared camera 53 to qualitatively judge the slurry making effect.

[0113] Step C3: Open the discharge valve 9, continue to observe the discharge process through the near-infrared camera 53, sample and measure the discharge concentration, and use grayscale entropy, texture contrast and motion pixel ratio fusion to determine the overall fluidization state.

[0114] Step C4: Close the discharge valve 9, replace with a different replaceable cone bottom module 3, and repeat steps C1 to C3 to determine the optimal cone bottom angle based on the overall fluidization state.

[0115] Specifically, the nearest neighbor matching algorithm is used to track the movement trajectory of the flocs: the centroid coordinates of the flocs in each frame of the image are calculated, and the flocs with the smallest Euclidean distance between adjacent frames are matched to form a complete trajectory. During the matching process, if the displacement direction of a matching point deviates from the displacement direction of the previous and next frames by more than 90° or the displacement amplitude exceeds twice the displacement corresponding to the maximum settling velocity of the tailings, it is determined to be an abnormal matching point and is removed.

[0116] Calculate the settling velocity of the floc: Extract the axial displacement component in the trajectory, divide it by the time interval Δt between adjacent frames, Δt = 1 / fps, where fps is the camera frame rate, to obtain the instantaneous settling velocity of the floc. Statistically analyze the instantaneous settling velocities of more than 100 flocs at various times, take the average value as the average settling velocity under this working condition, and simultaneously analyze the velocity distribution.

[0117] Calculate the radial diffusion angle: Taking the center of the feed well outlet as the origin, calculate the ratio of the radial displacement to the axial displacement of the floc trajectory, take the arctangent to obtain the diffusion angle, and statistically calculate the average diffusion angle. Repeat the experiment three times for each structure, and use the statistical average for comparison. The uniformity of the diffusion angle is evaluated by calculating the standard deviation and coefficient of variation of the floc radial diffusion angle in multiple experiments. Based on comprehensive comparison, the optimal feed well structure is the one with the highest settling velocity and the smallest standard deviation and coefficient of variation of the diffusion angle.

[0118] According to the target pressure P target The required compression section height for a vertical sand silo is calculated as follows: H = P target / γ′, where γ′ is the effective unit weight of the tailings, γ′=γ s -γ w γ s γ represents the saturated unit weight of tailings. w The specific weight of water is taken as 9.81 kN / m³. 3 γ s The saturated specific gravity is obtained through the saturated mud method: After fully saturating the tailings sample, it is placed in a container of known volume, weighed, and the total weight is subtracted from the container weight, then divided by the volume. The height H is the core basis for the overall height design of the sand silo.

[0119] The fluidization state is determined by fusing multiple indicators, including gray-level entropy, texture contrast, and moving pixel ratio. Gray-level entropy reflects the disorder of the image's gray-level distribution; texture contrast reflects the roughness of the wall texture; and moving pixel ratio reflects the activity level of particles near the wall.

[0120] This invention divides the experiment into three independent modes, targeting feed well selection, compression section height determination, and cone bottom discharge optimization, respectively. Each mode has corresponding system configurations and operating procedures, and they do not interfere with each other. Mode 1 uses visible light images to calculate settling velocity and diffusion angle to select the optimal feed well; Mode 2 obtains the target pressure through stepwise pressurization and then reverse-calculates the compression section height; Mode 3 optimizes the cone bottom parameters by observing the pulping and discharge processes using near-infrared images.

[0121] According to another embodiment of the present invention, the method for determining the state of the paste is as follows:

[0122] Criterion a: Observe the data from displacement sensor 45. If the mud layer compression rate is consistently less than 0.05 mm / min within a continuous 10-minute time window, it indicates that the macroscopic compression of the mud layer has nearly stopped and the pore water discharge is basically complete.

[0123] Criterion b: When criterion a is satisfied, observe the near-infrared wall image acquired by the near-infrared camera 53, check the liquid flow channels in the wall image and calculate the gray entropy. When the liquid flow channels disappear and the maximum fluctuation of the gray entropy within a continuous 10-minute time window is less than 10%, then criterion b is satisfied, and it is determined that the mud layer has reached the target paste state.

[0124] If criterion a is met but criterion b is not met within the preset time, it indicates that there is still a local drainage channel at the wall surface. The pressure can be increased and the observation can be repeated.

[0125] If criterion a cannot be satisfied, the experiment is terminated and the current state is recorded.

[0126] According to another embodiment of the present invention, in step B2, the current pressure value P when the paste state is reached is recorded. sample Sampling and measuring the underflow concentration C sample If C sample The target underflow concentration C has been reached or exceeded as determined by the design requirements. target Then this pressure value is the standard pressure P. target If C sample Not reached C target Continue increasing the pressure until the target underflow concentration C is reached. target Record the corresponding compliance pressure, which is the compliance pressure P. target .

[0127] According to another embodiment of the present invention, the target pressure P is recorded based on visual criteria. target and its corresponding undercurrent concentration C target Mark the point on the pressure-concentration coordinate system, and determine whether verification is required according to the following rules:

[0128] The situations requiring verification include:

[0129] When pressure fluctuations, surface cracks in the mud layer, and repeated liquid flow channels on the wall occur during the experiment, the reliability of the target points needs to be verified.

[0130] The verification method is as follows:

[0131] Under the pressure of meeting the standard P target Based on this, continue increasing the pressure according to the original pressurization gradient to the upper limit. After each pressure level stabilizes, sample and measure the underflow concentration, and plot a complete pressure-concentration curve. The target point (P...) target C target Compare with the pressure-concentration curve:

[0132] If the target point is located in a section where the slope of the curve decreases significantly, then the target point is reliable.

[0133] If the target point is located on a section of the curve with a steep slope, that point is unreliable. Then, move along the curve towards higher pressure and take the pressure value corresponding to the point where the curve slope first drops below 50% of the peak slope as the target pressure P. target ;

[0134] If the curve shows that after reaching the equipment's rated maximum working pressure, the concentration change rate corresponding to the two adjacent pressure levels remains higher than 50% of the peak slope for 30 consecutive minutes, it indicates that the tailings have not yet reached the compression limit under the existing equipment conditions. The design target should be reset or the pressurization equipment should be replaced.

[0135] Specifically, pressure fluctuations are determined by the data acquisition and processing unit 8 through real-time monitoring of the pressure sensor 44. A pressure fluctuation is defined as a deviation from the preceding and following trend exceeding ±5 kPa for more than 30 seconds. Cracks on the mud layer surface are identified through abrupt changes in the wall image texture. Repeated appearances of fluid flow channels on the wall are identified by the phenomenon of channel density increasing again during a decrease. All of the above determinations are automatically completed by the data acquisition and processing unit 8.

[0136] This invention can verify the reliability of the target pressure. When pressure fluctuations, cracks, or repeated flow paths occur during experiments, the target point may be unreliable. By plotting a complete pressure-concentration curve and comparing the target point with the curve, if it falls within a steeply rising slope, the curve is corrected towards higher pressure. This ensures that the final target pressure is not too low or too high due to local anomalies, and avoids overly conservative or overly aggressive design of the compression section height.

[0137] According to another embodiment of the present invention, the method for qualitatively judging the slurry-making effect by observing the fluidization state of the bottom tailings using a near-infrared camera 53 is as follows:

[0138] If pixel changes are continuously observed in the wall image within a 30-second observation window by comparing consecutive frame images, it is determined that there is obvious flow.

[0139] If there is no obvious pixel change in the wall image within the 30-second observation window, it is determined that there is no obvious flow.

[0140] If the area at the bottom corner of the cone remains static, without any pixel changes and with no change in wall shape for 2 minutes, it is determined to be a dead zone.

[0141] Record whether there is obvious flow on the wall surface and whether there are dead zones to qualitatively judge the pulping effect.

[0142] This invention uses near-infrared image pixel changes as the basis for judging the pulping effect, replacing the traditional manual visual inspection method. By setting observation windows of 30 seconds and 2 minutes, the wall flowability and cone bottom dead zone are judged separately. The results are objective, repeatable, and consistent among different operators.

[0143] According to another embodiment of the present invention, the method for obtaining the motion pixel ratio is as follows:

[0144] The motion pixel ratio = number of motion pixels / total number of pixels in the image × 100%, and the motion pixel ratio is calculated by the data acquisition and processing unit 8.

[0145] The method for determining fluidization state by fusing grayscale entropy, texture contrast, and moving pixel ratio is as follows:

[0146] If the proportion of moving pixels is higher than the first preset threshold and the grayscale entropy remains high within a continuous 10-second observation window, it is determined that the fluidization is active, the fluidization state is good, and the discharge potential is high.

[0147] If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy drops by more than 50% from a high value, it is determined that the mud layer is static or there is a dead zone. In this case, the overall fluidization state is poor, and there may be risks of poor material discharge or failure to meet the concentration standard. The first preset threshold is greater than the second preset threshold.

[0148] If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy increases, it is determined to be interference caused by changes in lighting or surface texture. In this case, the interference needs to be eliminated and the determination needs to be made again.

[0149] The fusion determination method of this invention comprehensively utilizes two indicators, the proportion of moving pixels and grayscale entropy, to distinguish between three situations: active fluidization, static mud layer, and external interference. When the proportion of moving pixels is low but the grayscale entropy increases, it can be identified that the interference is caused by changes in lighting or texture rather than a true change in fluidization state, thus avoiding misjudgment and improving the accuracy of fluidization state assessment.

[0150] The specific implementation method is as follows:

[0151] <Example 1>

[0152] Mode 1: Replaceable feed well module selection experiment

[0153] Step A1: Install a sealed bottom cover 12 at the bottom of the transparent column 1 and a replaceable feed well module 2 at the top;

[0154] Step A2: Prepare tailings slurry and flocculant solution separately in two mixing tanks 7, and feed them using the feed pump 6 at a set flow rate, which can be 1 m³ / min. 3 / h, continuously feed material into transparent column 1, and prepare 20% tailings slurry and 0.01% flocculant in mixing tank 7;

[0155] Step A3: Start the visible light camera 51 and continuously acquire floc settling images of the area below the feed well outlet at a frame rate of 60fps for 10 minutes. Use the nearest neighbor matching algorithm to track the floc movement trajectory: calculate the centroid coordinates of the floc in each frame, find the floc with the smallest Euclidean distance between adjacent frames for matching, and form a complete trajectory. During the matching process, if the displacement direction of a matching point deviates from the displacement direction of the previous and next frames by more than 90° or the displacement amplitude exceeds twice the displacement corresponding to the maximum settling velocity of the tailings, it is judged as an abnormal matching point and is removed. Calculate the floc settling velocity: extract the axial displacement component in the trajectory, divide it by the time interval Δt between adjacent frames, Δt = 1 / fps, where fps is the camera frame rate, to obtain the instantaneous settling velocity of the floc. The instantaneous settling velocities of more than 100 flocs at various times are statistically analyzed, and the average value is taken as the average settling velocity under this working condition. At the same time, the velocity distribution is statistically analyzed. The radial diffusion angle is calculated: with the center of the feed well outlet as the origin, the ratio of the radial displacement to the axial displacement of the floc trajectory is calculated, and the arctangent is taken to obtain the diffusion angle. The average diffusion angle is statistically analyzed.

[0156] Step A4: Replace the replaceable feed well module 2 with different structures, repeat steps A1 to A3, and compare to determine the optimal replaceable feed well module 2. The module with the highest settling velocity and the smallest standard deviation and coefficient of variation of the diffusion angle is the optimal one. The uniformity of the diffusion angle is evaluated by calculating the standard deviation and coefficient of variation of the radial diffusion angle of the flocs in multiple experiments. In summary, the optimal feed well structure is the one with the highest settling velocity and the smallest standard deviation and coefficient of variation of the diffusion angle.

[0157] <Example 2>

[0158] Mode 2: Compression Section Pressurization Experiment

[0159] Before conducting the compression section pressurization experiment using an integrated vertical sand silo thickening experimental system based on machine vision, the target underflow concentration C was first determined according to the design requirements of the filling system. target The target concentration is determined based on the mine backfill design, and its specific sources include: the basic properties of tailings, the strength requirements of the backfill body, the requirements for backfill slurry transportation, and the requirements for stope backfilling technology, etc. (Ct) argetThis is typically an optimized value under the aforementioned multiple constraints, proposed by the mine design unit or owner before this experiment. The target underflow concentration is usually 65%~75%, with the specific value varying depending on the properties of the mine tailings and the backfilling process conditions. For example, for a certain iron ore backfilling station with all tailings, a ash-to-sand ratio of 1:4, and a design strength of 1.5 MPa, the target underflow concentration is determined to be 68% after comprehensively considering the above constraints.

[0160] Step B1: Following the operations of steps A1 and A2, continuously feed material into the transparent column 1 to form an initial mud layer with a height of approximately 500 mm. Stop feeding, remove the top replaceable feed well module 2, and install the pressurized piston unit 4 into the transparent column 1 through the top opening, ensuring that the piston head 41 contacts the surface of the mud layer. At this time, the piston rod 42 protrudes from the top opening of the transparent column 1 and connects to the force application device 43. The bottom of the transparent column 1 is a sealed bottom cover 12. The operations of steps A1 and A2 are as follows: install the sealed bottom cover 12 at the bottom of the transparent column 1, install the replaceable feed well module 2 at the top, prepare tailings slurry and flocculant solution in the two mixing tanks 7 respectively, and continuously feed them into the transparent column 1 through the feed pump 6. For details, please refer to Example 1.

[0161] Step B2: Supply air to the force application device 43 through a precision pressure regulating valve, applying pressure gradually from a low pressure. The recommended pressure sequence is 10, 20, 40, 60, 80, 100, 120, 160, 200 kPa, which can be adjusted according to the actual tailings compression characteristics. This pushes the piston downward to compress the mud layer. During the compression process, water enters the hollow part of the piston rod 42 from the transparent cylinder 1, passes through the hollow part of the piston rod 42, and enters the external water collection device from the outlet 48. At the same time, the near-infrared camera 53 captures the water at 30fps. Images of the compression zone wall are continuously acquired at a frame rate of s. Near-infrared camera 53, in conjunction with near-infrared supplementary light 54, is directed towards the compression zone wall to extract grayscale entropy, texture roughness, and fluid channel density characteristic parameters. The mud layer compression rate is monitored by displacement sensor 45, and piston position changes are recorded by displacement sensor 45. Pressure values ​​are confirmed in real-time by pressure sensor 44. Finally, combining the mud layer compression rate and characteristic parameters, the paste state is determined. If the paste state has reached the target paste state, the currently applied pressure value is recorded as the target pressure P. target And samples were taken to determine the underflow concentration;

[0162] The method for judging the state of a paste is as follows:

[0163] Criterion a: Observe the data from displacement sensor 45. If the mud layer compression rate is consistently less than 0.05 mm / min within a continuous 10-minute time window, it indicates that the macroscopic compression of the mud layer has nearly stopped and the pore water discharge is basically complete.

[0164] Criterion b: When criterion a is satisfied, observe the near-infrared wall image acquired by the near-infrared camera 53, check the liquid flow channels in the wall image and calculate the gray entropy. When the liquid flow channels disappear and the maximum fluctuation of the gray entropy within a continuous 10-minute time window is less than 10%, then criterion b is satisfied, and it is determined that the mud layer has reached the target paste state, indicating that the mud layer structure near the wall has become stable and the internal floc structure has been reorganized.

[0165] Abnormal Handling: If criterion a is met but criterion b is not met within the preset time period (30 minutes), it indicates that a local drainage channel still exists at the wall surface. In this case, increase the pressure by 10 kPa. If criterion b is still not met after 30 minutes, continue increasing the pressure by 10 kPa until criterion b is met or the pressure reaches the upper limit. If the pressure reaches the upper limit and criterion b is still not met, terminate the test. Restart the timing and observation after each increase.

[0166] If criterion a cannot be met, such as if the compression time exceeds 2 hours, the experiment is terminated and the current state is recorded. The current state is recorded as the pressure value, mud layer compression amount, wall image features, bottom flow concentration measured by sampling, and total compression time. Wall image features include grayscale entropy and liquid flow channel density.

[0167] In step B2, record the current pressure value P when the paste-like state is reached. sample Sampling and measuring the underflow concentration C sample If C sample The target underflow concentration C has been reached or exceeded as determined by the design requirements. target Then this pressure value is the standard pressure P. target If C sample Not reached C target Continue increasing the pressure until the target underflow concentration C is reached. target Record the corresponding compliance pressure, which is the compliance pressure P. target .

[0168] Recorded target pressure P target and its corresponding undercurrent concentration C target Mark the point on the pressure-concentration coordinate system, and determine whether verification is required according to the following rules:

[0169] The situations requiring verification include:

[0170] When pressure fluctuations, surface cracks in the mud layer, and repeated liquid flow channels on the wall occur during the experiment, the reliability of the target points needs to be verified.

[0171] The verification method is as follows:

[0172] Under the pressure of meeting the standard P targetBased on the existing pressure gradient, continue increasing the pressure to the upper limit, increasing by 10 kPa per stage. After each pressure stage stabilizes, sample and measure the underflow concentration, and plot a complete pressure-concentration curve. The target point (P...) target C target Compare with the pressure-concentration curve:

[0173] If the target point is located in a section where the slope of the curve decreases significantly, then the target point is reliable; a significant decrease means that the average slope of three consecutive adjacent pressure intervals within that section is less than 50% of the peak slope.

[0174] If the target point is located on a section of the curve with a steep slope, that point is unreliable. The pressure is then moved along the curve towards higher pressure, and the pressure value corresponding to the point where the curve slope first drops below 50% of the peak slope is taken as the new design target pressure value P. design and the pressure value P design As the pressure to meet the standard P target The slope of the curve rises sharply at the target point, where the slope is more than 70% higher than the peak slope.

[0175] If the curve shows that after reaching the equipment's rated maximum working pressure, the concentration change rate corresponding to two adjacent pressure levels remains higher than 50% of the peak slope for 30 consecutive minutes, it indicates that the tailings have not yet reached their compression limit under the existing equipment conditions. The design target should be reset, or the pressurizing equipment should be replaced with one whose rated maximum working pressure is greater than the currently used equipment. The design target should then be reset to reduce the concentration target C. target The rate of change in concentration is the slope of the curve.

[0176] Step B3: Based on the target pressure P target The required compression section height for a reverse-engineered vertical sand bin.

[0177] <Example 3>

[0178] Mode 3: Cone-bottom and pulping experiment

[0179] Step C1: Install the replaceable conical bottom module 3 at the bottom of the transparent column 1 and the replaceable feed well module 2 at the top. Select the straight pipe feed well. If it is necessary to screen the replaceable feed well module 2, you can first determine the appropriate replaceable feed well module 2 according to the experimental method of Mode 1. If it is not necessary to screen the replaceable feed well module 2, the replaceable feed well module 2 will be selected as the straight pipe feed well by default.

[0180] Step C2: Following the operation in Step A2, continuously feed material into the transparent column 1 to form a mud layer. Stop feeding and complete the settling process. Allow the material to settle, then start the slurry making system to make slurry. Observe the fluidization state of the tailings at the bottom using a near-infrared camera 53. The near-infrared camera 53, in conjunction with a near-infrared supplementary light 54, is directed towards the cone bottom area to qualitatively judge the slurry making effect. The slurry making effect is determined by whether there is obvious flow and whether there are dead zones. The method for qualitatively judging the slurry making effect by observing the fluidization state of the tailings at the bottom using a near-infrared camera 53 is as follows: Continuously feed material into the transparent column 1 following the operation in Step A2, which involves preparing tailings slurry and flocculant solution in two mixing tanks 7 respectively, using a feed pump 6 at a set flow rate, which can be 1 m³ / s. 3 The material is continuously fed into the transparent column 1 at a rate of / h. A pulping system is used for pulping, with the nozzle angle 32, pressure, and pulse frequency parameters set. At this time, the discharge valve 9 is closed. Specifically, the nozzle angle 32 can be 30°, the pulping pressure can be 0.4MPa, and the pulse frequency can be 0.5Hz. These parameters can be adjusted according to actual working conditions and are not limited to these.

[0181] If pixel changes are continuously observed in the wall image within a 30-second observation window by comparing consecutive frame images, it is determined that there is obvious flow.

[0182] If there is no obvious pixel change in the wall image within the 30-second observation window, it is determined that there is no obvious flow. Perform inter-frame difference on adjacent frames and calculate the total number of pixels whose gray values ​​change in the difference image. If the total number of pixels accounts for less than 5% of the total number of pixels in the image and the duration exceeds 30 seconds, it is determined that there is no obvious pixel change.

[0183] If the area at the bottom corner of the cone remains static, without any pixel changes and with no change in wall shape for 2 minutes, it is determined to be a dead zone.

[0184] Record whether there is obvious flow on the wall surface and whether there are dead zones to qualitatively judge the pulping effect;

[0185] Step C3: Open the discharge valve 9, continue to observe the discharge process using the near-infrared camera 53, sample and measure the discharge concentration, and use grayscale entropy, texture contrast, and motion pixel ratio fusion to determine the overall fluidization state; the method for obtaining the motion pixel ratio is as follows:

[0186] The motion pixel ratio = number of motion pixels / total number of pixels in the image × 100%, and the motion pixel ratio is calculated by the data acquisition and processing unit 8.

[0187] The method for determining fluidization state by fusing grayscale entropy, texture contrast, and moving pixel ratio is as follows:

[0188] If the proportion of moving pixels is higher than the first preset threshold for a continuous 10-second observation window, and the first preset threshold can be 5% and the grayscale entropy remains high, it is determined that the fluidization is active, then the fluidization state is good and the discharge potential is high; the high value is greater than 4.0.

[0189] If the proportion of moving pixels is lower than the second preset threshold, which is 1%, close to 0 and the grayscale entropy drops by more than 50% from a high value, it is considered a significant decrease. This indicates that the mud layer is stagnant or has dead zones, and the overall fluidization state is poor. There may be risks of poor material discharge or substandard concentration. The first preset threshold is greater than the second preset threshold.

[0190] If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy increases, it is determined to be interference caused by changes in lighting or surface texture. The interference must be eliminated before re-evaluation. The first preset threshold, second preset threshold, high grayscale entropy value, and observation window are all recommended values. In practical applications, they can be adjusted according to the tailings characteristics and image quality. A value below the second preset threshold indicates no significant change.

[0191] Step C4: Close the discharge valve 9, stop pulping, replace with different replaceable cone bottom modules 3, repeat steps C1 to C3, determine the optimal cone bottom angle based on the overall fluidization state, with the goal of smooth discharge, no dead zones and the highest underflow concentration, determine the optimal cone bottom angle, and output recommended parameters.

[0192] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and embodiments shown and described herein.

Claims

1. An integrated vertical sand silo thickening experimental system based on machine vision, characterized in that, include: A transparent column is used to contain tailings slurry. The upper end of the transparent column is detachably connected to a replaceable feed well module or a sealed top cover, and the lower end of the transparent column is detachably connected to a replaceable cone bottom module or a sealed bottom cover. The lower part of the replaceable cone bottom module is equipped with a discharge valve. A pressurizing piston unit, which is detachably located inside a transparent cylinder, includes a force-applying device and a piston head connected to the force-applying device, with multiple drainage holes provided on the piston head. The machine vision unit includes a visible light camera, a visible light supplementary light, a near-infrared camera, and a near-infrared supplementary light. The visible light camera, in conjunction with the visible light supplementary light, is directed towards the area below the outlet of the replaceable feed well module to capture the floc settling process. The near-infrared camera, in conjunction with the near-infrared supplementary light, is directed towards the wall of the compression zone to capture images of the compression zone wall. Alternatively, the near-infrared camera, in conjunction with the near-infrared supplementary light, is directed towards the cone bottom area to observe the fluidization state and discharge effect of the tailings at the bottom during the pulping process. The sensor unit includes a pressure sensor embedded in the inner wall of the transparent cylinder with the probe end face flush with the inner wall and a displacement sensor mounted on the force application mechanism, which are used to monitor the pressure value applied to the mud layer and the piston displacement in real time, respectively. The data acquisition and processing unit, which is electrically connected to the pressure sensor, displacement sensor, visible light camera, and near-infrared camera, is used to run image processing algorithms. It extracts floc contours, settling velocity, and radial diffusion angles from floc settling images acquired by the visible light camera; it extracts grayscale entropy, texture roughness, and fluid channel density features from compression zone wall images acquired by the near-infrared camera to characterize the stability of the mud layer structure near the wall; or it extracts grayscale entropy, texture contrast, and motion pixel ratio features from cone bottom area images acquired by the near-infrared camera to comprehensively judge the fluidization state and discharge potential inside the cone bottom.

2. The integrated vertical sand silo thickening experimental system based on machine vision as described in claim 1, characterized in that, The replaceable feed well module includes at least two of the following: straight pipe feed well, guide channel feed well, and self-dilution channel feed well.

3. The integrated vertical sand silo thickening experimental system based on machine vision as described in claim 2, characterized in that, The replaceable cone bottom module has a cone angle of at least two of 45°, 60°, and 75°, and the cone bottom is provided with a pulping interface for replaceable nozzles.

4. The integrated vertical sand silo thickening experimental system based on machine vision as described in claim 3, characterized in that, The force-applying device is a pneumatic pressurization type. The output end of the force-applying device is connected to a piston rod, which is vertically set and has a hollow structure. The lower end of the piston rod is connected to the piston head, and multiple drainage holes are connected to the hollow part of the piston rod. A support net is provided on the lower surface of the piston head, and filter cloth is covered on the upper and lower surfaces of the support net.

5. A machine vision-based integrated vertical sand silo density testing method, employing the machine vision-based integrated vertical sand silo density testing system as described in any one of claims 1 to 4, characterized in that... Includes the following three independent experimental modes: Mode 1: Replaceable feed well module selection experiment Step A1: Install a sealed bottom cover at the bottom of the transparent column and install a replaceable feed well module at the top; Step A2: Prepare tailings slurry and flocculant solution in two mixing tanks respectively, and continuously feed them into the transparent column through a feed pump; Step A3: Acquire floc settling images of the area below the outlet of the replaceable feed well module using a visible light camera, extract the floc contours using an image processing algorithm, track the floc movement trajectory using a nearest neighbor matching algorithm, and calculate the axial settling velocity and radial diffusion angle. Step A4: Replace with replaceable feed well modules of different structures. Repeat steps A1 to A3 and compare to determine the optimal replaceable feed well module. The module with the highest settling velocity and the smallest standard deviation of diffusion angle and coefficient of variation is the optimal one. Mode 2: Compression Section Pressurization Experiment Step B1: Using the operations of steps A1 and A2, continuously feed material into the transparent cylinder to form an initial mud layer, stop feeding, remove the top replaceable feeding well module, and install the pressurized piston unit into the transparent cylinder from the top opening, and make the piston head contact the surface of the mud layer. Step B2: Apply pressure gradually from low to high using the pressurizing piston unit, while continuously acquiring images of the compression zone wall using a near-infrared camera. Extract grayscale entropy, texture roughness, and fluid channel density characteristic parameters, and monitor the mud layer compression rate using a displacement sensor. Finally, combine the mud layer compression rate and characteristic parameters to determine the paste state. If the paste state has reached the target paste state, record the currently applied pressure value as the target pressure P. target And samples were taken to determine the underflow concentration; Step B3: Based on the target pressure P target The required compression section height for reverse-push vertical sand bins; Mode 3: Cone-bottom and pulping experiment Step C1: Install a replaceable cone-bottom module at the bottom of the transparent column and a replaceable feed well module at the top; Step C2: Continuously feed material into the transparent column to form a mud layer using the operation in step A2, stop feeding and complete settling, then make slurry, observe the fluidization state of the tailings at the bottom using a near-infrared camera, and qualitatively judge the slurry making effect. Step C3: Open the discharge valve, continue to observe the discharge process through a near-infrared camera, sample and measure the discharge concentration, and use grayscale entropy, texture contrast and motion pixel ratio fusion to determine the overall fluidization state. Step C4: Close the discharge valve, replace with different replaceable cone bottom modules, and repeat steps C1 to C3 to determine the optimal cone bottom angle based on the overall fluidization state.

6. The integrated vertical sand silo thickening experiment method based on machine vision as described in claim 5, characterized in that, The method for judging the state of a paste is as follows: Criterion a: Observe the displacement sensor data. If the mud layer compression rate is consistently less than 0.05 mm / min within a continuous 10-minute time window, it indicates that the macroscopic compression of the mud layer has nearly stopped and the pore water discharge is basically complete. Criterion b: When criterion a is satisfied, observe the near-infrared wall image obtained by the near-infrared camera, check the liquid flow channels in the wall image and calculate the gray entropy. When the liquid flow channels disappear and the maximum fluctuation of the gray entropy within a continuous 10-minute time window is less than 10%, then criterion b is satisfied, and it is determined that the mud layer has reached the target paste state. If criterion a is met but criterion b is not met within the preset time, it indicates that there is still a local drainage channel at the wall surface. The pressure can be increased and the observation can be repeated. If criterion a cannot be satisfied, the experiment is terminated and the current state is recorded.

7. The integrated vertical sand silo thickening experiment method based on machine vision as described in claim 5, characterized in that, In step B2, record the current pressure value P when the paste-like state is reached. sample Sampling and measuring the underflow concentration C sample If C sample The target underflow concentration C has been reached or exceeded as determined by the design requirements. target Then this pressure value is the standard pressure P. target If C sample Not reached C target Continue increasing the pressure until the target underflow concentration C is reached. target Record the corresponding compliance pressure, which is the compliance pressure P. target .

8. The integrated vertical sand silo density test method based on machine vision as described in claim 5, characterized in that, Based on visual criteria, the target stress P is recorded. target and its corresponding undercurrent concentration C target Mark the point on the pressure-concentration coordinate system, and determine whether verification is required according to the following rules: The situations requiring verification include: When pressure fluctuations, surface cracks in the mud layer, and repeated liquid flow channels on the wall occur during the experiment, the reliability of the target points needs to be verified. The verification method is as follows: Under the pressure of meeting the standard P target Based on this, continue increasing the pressure according to the original pressurization gradient to the upper limit. After each pressure level stabilizes, sample and measure the underflow concentration, and plot a complete pressure-concentration curve. The target point (P...) target C target Compare with the pressure-concentration curve: If the target point is located in a section where the slope of the curve decreases significantly, then the target point is reliable. If the target point is located on a section of the curve with a steep slope, that point is unreliable. Then, move along the curve towards higher pressure and take the pressure value corresponding to the point where the curve slope first drops below 50% of the peak slope as the target pressure P. target ; If the curve shows that after reaching the equipment's rated maximum working pressure, the concentration change rate corresponding to the two adjacent pressure levels remains higher than 50% of the peak slope for 30 consecutive minutes, it indicates that the tailings have not yet reached the compression limit under the existing equipment conditions. The design target should be reset or the pressurization equipment should be replaced.

9. The integrated vertical sand silo thickening experiment method based on machine vision as described in claim 5, characterized in that, The method for qualitatively judging the mud-making effect is to observe the fluidization state of the bottom tailings using a near-infrared camera: If pixel changes are continuously observed in the wall image within a 30-second observation window by comparing consecutive frame images, it is determined that there is obvious flow. If there is no obvious pixel change in the wall image within the 30-second observation window, it is determined that there is no obvious flow. If the area at the bottom corner of the cone remains static, without any pixel changes and with no change in wall shape for 2 minutes, it is determined to be a dead zone. Record whether there is obvious flow on the wall surface and whether there are dead zones to qualitatively judge the pulping effect.

10. The integrated vertical sand silo thickening experiment method based on machine vision as described in claim 9, characterized in that, The method for obtaining the motion pixel ratio is as follows: The percentage of moving pixels = number of moving pixels / total number of pixels in the image × 100%. The percentage of moving pixels is calculated by the data acquisition and processing unit.

11. The method for determining fluidization state by fusing grayscale entropy, texture contrast, and moving pixel ratio is as follows: If the proportion of moving pixels is higher than the first preset threshold and the grayscale entropy remains high within a continuous 10-second observation window, it is determined that the fluidization is active, the fluidization state is good, and the discharge potential is high. If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy drops by more than 50% from a high value, it is determined that the mud layer is stagnant or has dead zones. This indicates poor overall fluidization, which may pose risks of poor material discharge or substandard concentration. The first preset threshold is greater than the second preset threshold; If the proportion of moving pixels is lower than the second preset threshold and the grayscale entropy increases, it is determined to be interference caused by changes in lighting or surface texture. In this case, the interference needs to be eliminated and the determination needs to be made again.

Images