Impurity removing and feeding method, system and equipment for metal gallium production and storage medium

Abstract

The invention relates to the technical field of metal material processing. The invention provides an impurity removing and charging method, system and equipment for metal gallium production and a storage medium. The method comprises the following steps: inputting a production task through an intelligent control system, and presetting a material addition amount and process parameters; the remaining amount of the material storage bin is detected through a material level monitoring sensor, and the materials are conveyed to an adding actuator through a conveying device after it is confirmed that the remaining amount is sufficient; scanning a feeding area by means of a visual identification system and a laser scanner, obtaining coordinate data of a target feeding point, planning a motion path of the robot, and controlling the robot to accurately position; material adding is executed in stages, the concentration or liquid level change is monitored in real time so as to dynamically adjust the adding amount, meanwhile, the feeding state is continuously monitored through visual recognition, and the posture and parameters of the robot are corrected in real time for deviation or abnormity. The problems that in the impurity removal process of metal gallium production, the precision of traditional manual feeding is greatly influenced by operation experience, material proportion deviation is likely to be caused, the labor intensity of workers is high, and safety risks exist are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal material processing, and in particular to a method, system, equipment and storage medium for removing impurities and adding materials for metal gallium production. Background Art

[0002] Gallium, a strategically important metal, plays an irreplaceable and important role in electronics, new energy, aerospace, and medical diagnostics. In the semiconductor sector, it serves as a core raw material for the preparation of compound semiconductors such as gallium arsenide and gallium nitride, supporting the research and development and production of integrated circuits, light-emitting diodes, and power devices. In the new energy sector, gallium-based materials are key components of photovoltaic cells and energy storage devices, directly impacting energy conversion efficiency and storage performance. In the aerospace sector, the high-temperature stability and corrosion resistance of gallium alloys are crucial to the reliability of aerospace equipment, while radioactive isotopes of gallium are also used in medical diagnostics for tumor detection.

[0003] In the production process of gallium metal, impurity removal is a critical step in determining product purity and quality. Currently, most companies still rely on manual addition of materials such as lime and hydrogen peroxide during the impurity removal process. This traditional method has significant drawbacks: the accuracy of manual addition is significantly affected by operator experience, which can easily lead to deviations in material ratios, thereby affecting impurity removal efficiency and gallium purity. The high labor intensity and tedious repetitive operations not only restrict production efficiency but also pose safety risks such as material leakage and exposure to corrosive reagents. Summary of the Invention

[0004] The purpose of the present invention is to provide a method, system, equipment and storage medium for removing impurities and adding materials in gallium metal production, aiming to solve the problems of traditional manual adding materials in the process of removing impurities in gallium metal production, such as the accuracy being greatly affected by operating experience, the easy occurrence of material ratio deviation, high manual labor intensity and safety risks.

[0005] The present invention is achieved through the following technical solutions:

[0006] A method for removing impurities and adding materials for producing metallic gallium comprises the following steps:

[0007] Input production task information through the intelligent control system and preset the target material addition amount and process parameters;

[0008] The material level monitoring sensor detects the remaining information of the material storage bin in real time. After confirming that there is enough material, the conveying device is started to transport the material to the corresponding adding actuator of the robot;

[0009] The feeding area is scanned by a visual recognition system, and the position, shape and size information of the target feeding point are obtained by combining a laser scanner with an image processing algorithm, and coordinate data is generated;

[0010] Based on the coordinate data of the target feeding point, the robot motion path is planned and the robot is controlled to move to the target feeding point;

[0011] According to the preset target material addition amount and process parameters, the intelligent control system sends instructions to the addition actuator to perform material addition operations in stages, and monitors the concentration or liquid level changes during the addition process in real time to dynamically adjust the addition amount;

[0012] The material addition status is continuously monitored through the visual recognition system. If deviation or abnormality is detected, the robot posture is corrected or parameters are added in real time through the intelligent control system.

[0013] Optionally, the specific process of scanning the feeding area by a visual recognition system, combining a laser scanner with an image processing algorithm, obtaining the position, shape and size information of the target feeding point, and generating coordinate data is as follows:

[0014] The industrial camera array collects multi-angle image data of the feeding area, while the laser scanner emits structured grating to perform 3D contour scanning of the target area to obtain laser point cloud data;

[0015] The multi-angle image data and the laser point cloud data are fused, and the mapping relationship between the image pixel coordinate system and the laser scanning coordinate system is established using calibration parameters;

[0016] An improved Canny-Sobel composite operator is used to enhance the detection of image edges and extract the geometric contour features of the feeding port. At the same time, the laser point cloud is registered using the ICP algorithm to generate a 3D topological model of the target area.

[0017] The center coordinates of the feeding port are calculated based on the feature matching algorithm, and the continuous preset frame data are weightedly fused using the multi-frame sliding window optimization algorithm to output the final coordinate data.

[0018] Optionally, the specific process of planning the robot motion path based on the coordinate data of the target feeding point and controlling the robot to move to the target feeding point is:

[0019] Establish the homogeneous transformation matrix between the robot base coordinate system and the visual coordinate system, as shown in the following formula (1):

[0020]

[0021] in, represents the homogeneous transformation matrix; R represents the rotation matrix; t represents the translation vector;

[0022] The target feeding point coordinates are converted to the robot base coordinate system as shown in the following formula (2):

[0023]

[0024] Among them, P base Indicates the position of the target feeding point in the robot base coordinate system; P vision Indicates the position of the target feeding point in the visual coordinate system;

[0025] Based on the robot inverse kinematics model, the angular displacement of each joint is calculated as shown in the following formula (3):

[0026]

[0027] Among them, θ i represents the angular displacement of the i-th joint; φ represents the end posture constraint, as shown in the following formula (4):

[0028]

[0029] in, Represents the rotation matrix from the tool coordinate system to the robot base coordinate system; Indicates the normal vector of the target feeding point; Represents the z-axis unit vector of the tool coordinate system; θ pitch represents the pitch angle of the end effector; θ roll represents the roll angle of the end effector; θ P_max and θ R_max They represent the maximum absolute values ​​of the pitch angle and roll angle respectively; ω represents the angular velocity vector of the end effector; represents the angular acceleration vector of the end effector; ω max and Represent the maximum values ​​of angular velocity and angular acceleration respectively; P tool Represents the position coordinates of the end effector; P obs Indicates the location coordinates of the obstacle; d safe Indicates the safety distance threshold;

[0030] Construct the path optimization function as follows (5):

[0031]

[0032] Where m represents the number of nodes; q k represents the joint angle vector of the kth node; q k-1 represents the joint angle vector of the k-1th node; ‖q k -q k-1 ‖ 2 Represents the square norm of the joint angle difference between adjacent nodes, which is used to constrain the smoothness of joint motion; represents the joint angular acceleration vector of the kth node; Represents the square norm of the joint angular acceleration, which is used to constrain the acceleration amplitude during movement; w1 and w2 represent the weight coefficients of smoothness and acceleration, respectively;

[0033] The joint space trajectory is generated by quintic polynomial interpolation, as shown in the following equation (6):

[0034] θ(t)=a0+a1t+a2t 2 +a3t 3 +a4t 4 +a5t 5 (6)

[0035] Where θ(t) represents the angular displacement of the robot joint at time t; a0, a1, a2, a3, a4, and a5 represent the coefficients of the quintic polynomial, which are determined by the boundary conditions of trajectory planning and are used to fit the smooth joint motion trajectory;

[0036] The Lyapunov function is used to verify the trajectory stability, as shown in the following equation (7):

[0037]

[0038] in, represents the Lyapunov function; e represents the error vector of joint angular displacement; represents the joint angular velocity error vector; K p Proportional gain matrix, used to adjust the weight of the error vector of joint angular displacement; K d represents the differential gain matrix, which is used to adjust the weight of the joint angular velocity error vector;

[0039] The joint angle θ is fed back in real time by the encoder during the movement actual , the proportional-integral-differential control law is used to adjust the driving torque, as shown in the following formula (8):

[0040]

[0041] Where τ represents the driving torque vector of the robot joint, which is used to drive the movement of each joint so that the end effector can track the target trajectory; J T represents the transpose of the robot Jacobian matrix; K i represents the integral gain matrix;

[0042] θ desired represents the desired angular displacement vector of the joint; θ actual Represents the actual angular displacement vector of the joint;

[0043] Until the positioning accuracy of the end effector meets ‖P actual -P base ‖≤ε, where P actualrepresents the actual position coordinate of the robot's end effector; ε represents the positioning accuracy threshold.

[0044] Optionally, the specific process of sending instructions to the adding actuator through the intelligent control system according to the preset target material addition amount and process parameters to perform the material adding operation in stages is as follows:

[0045] According to the preset process parameters, the material addition process is divided into the initialization pre-addition stage, the main flow stable addition stage and the precise compensation addition stage;

[0046] During the initialization pre-addition phase, the intelligent control system calculates the pre-addition threshold based on the volume of the target addition point and the initial concentration of the material, and controls the addition actuator to inject the material into the reaction vessel at a first preset rate until it detects that the liquid level or concentration in the vessel reaches the pre-addition trigger condition;

[0047] During the main flow stable addition stage, the real-time theoretical addition rate is calculated based on the material balance equation. The mapping relationship between the addition rate and the impurity removal efficiency is established in combination with the reaction kinetics model. The flow output of the addition actuator is adjusted through the proportional integral differential control algorithm to stabilize the material addition rate within the theoretical value ±5%;

[0048] During the precise compensation addition stage, the characteristic parameters in the container are collected in real time, the compensation addition amount is calculated through the fuzzy control algorithm, and the addition actuator is controlled to perform micro-addition at a second preset rate until the characteristic parameter change rate is detected to be less than 0.3% / min and the cumulative addition amount reaches 95% to 98% of the preset target value.

[0049] Optionally, the specific process of real-time monitoring of concentration or liquid level changes during the addition process and dynamically adjusting the addition amount is as follows:

[0050] The concentration sensor and liquid level sensor arranged in the reaction vessel collect the solution conductivity, turbidity and liquid level data in real time at a frequency of not less than 20Hz, and the data is input into the intelligent control system after the noise is removed by the Kalman filter algorithm;

[0051] In the initialization pre-addition stage, when the liquid level monitoring value reaches 15% to 20% of the target addition point volume, the main flow stable addition stage start instruction is triggered;

[0052] During the main flow stable addition stage, the deviation between the actual material addition rate and the theoretical addition rate is calculated based on real-time concentration data. The incremental PID control algorithm outputs the adjustment parameters to control the flow fluctuation of the addition actuator within ±3% of the rated flow rate.

[0053] In the precise compensation addition stage, a fuzzy rule base for concentration change rate and addition amount is established. When the characteristic parameter change rate exceeds 0.5% / min for 30 consecutive seconds, the pulse micro-addition mode is activated, with a single pulse addition amount of 0.1mL to 0.5mL, until the characteristic parameter change rate stabilizes below 0.3% / min and the cumulative addition amount error converges to within ±2% of the target value.

[0054] During the entire process, if the concentration or liquid level monitoring value exceeds the preset safety threshold, the intelligent control system automatically sends an emergency stop command and triggers an audible and visual alarm.

[0055] Optionally, the specific process of continuously monitoring the material addition status through the visual recognition system and correcting the robot posture or adding parameters in real time through the intelligent control system if deviation or abnormality is detected is as follows:

[0056] The visual recognition system continuously collects image data of the feeding area at a preset frequency, and uses a deep learning-based target detection algorithm to perform real-time analysis of the material feeding status, identifying the material feeding trajectory, flow pattern, and material distribution characteristics around the feeding port;

[0057] When it is detected that the material addition trajectory deviates from the preset path, the flow pattern fluctuates abnormally, or material deviation or abnormal conditions occur around the feeding port, the abnormal handling mechanism is immediately triggered;

[0058] For robot posture correction, the intelligent control system uses the deviation data fed back by the visual recognition system and the current posture information of the robot to recalculate the angular displacement of each joint of the robot through the inverse kinematics model, generate posture correction instructions and send them to the robot controller to control the robot to adjust the spatial position and posture of the end effector;

[0059] For additive parameter correction, the preset parameter adjustment strategy library is called according to the abnormality type; if a flow abnormality is detected, the flow output of the additive actuator is dynamically adjusted through the proportional integral differential control algorithm; if a material distribution abnormality is detected, the material addition rate or total amount is adjusted according to the real-time concentration or liquid level data;

[0060] During the correction process, the system continuously monitors the correction effect. If the abnormal state is not eliminated, the correction parameters are iteratively optimized until the material addition state returns to normal and meets the process requirements.

[0061] Optionally, the specific process of using a deep learning-based target detection algorithm to perform real-time analysis on the material addition status and identify the material addition trajectory, flow pattern, and material distribution characteristics around the feeding port is as follows:

[0062] A material state detection model based on the improved YOLOv5 is constructed. The loss function of the material state detection model is shown in the following formula (9):

[0063]

[0064] in, represents the total loss function of the material state detection model; cls Represents the classification loss weight coefficient; A represents the total number of grids in the feature map; B represents the number of bounding boxes predicted for each grid; represents the characteristic function; C ij Indicates the true category confidence; Represents the predicted category confidence; λ loc represents the positioning damage weight coefficient; p ijk Represents the real bounding box coordinate parameters; Represents the predicted bounding box coordinate parameters; SmoothL1 loss function for bounding box localization;

[0065] The continuous image frames are processed by the spatiotemporal feature fusion module, and the optical flow constraint equation is defined as shown in the following equation (10):

[0066] I(x,y,t)=I(x+Δx,y+Δy,t+Δt)(10)

[0067] Where I(x,y,t) represents the brightness value of the pixel with coordinates (x,y) in the image at time t; Δx and Δy represent the displacement of the pixel in the x and y directions within the time interval Δt, respectively.

[0068] Solve the dense optical flow field using the Horn-Schunck algorithm And extract the material trajectory vector Where N represents the number of optical flow sampling points; Represents the optical flow vector of the i-th sampling point; when When , it is determined that the trajectory deviates; represents the reference trajectory vector, δ teaj Indicates the trajectory deviation threshold;

[0069] A dual-branch convolutional network is used to extract traffic morphology features and material distribution features respectively. The traffic morphology branch extracts multi-scale spatiotemporal features through a dilated convolutional layer, as shown in the following formula (11):

[0070]

[0071] in, Represents the characteristic value of the traffic shape characteristic map of the lth layer at the spatial coordinate (x, y); Represents the weight coefficient of the hole convolution kernel of the lth layer at the relative coordinate (i, j); I(x+d·i,y+d·j) is the pixel value of the input image at the absolute coordinate (x+d·i,y+d·j); d represents the expansion rate of the hole convolution; k represents the half-side length of the convolution kernel; (i,j) represents the relative coordinates within the convolution kernel, i,j∈[-k,k];

[0072] The material distribution branch generates a pixel-level segmentation mask through the U-Net structure and calculates the material coverage around the feeding port, as shown in the following formula (12):

[0073]

[0074] Where ρ represents the material coverage rate around the feeding port; |R| represents the total number of pixels in the preset target area R around the feeding port; M(x,y) represents the pixel-level segmentation mask function. When the pixel (x,y) belongs to the material coverage area, M(x,y) = 1, otherwise M(x,y) = 0;

[0075] When |ρ-ρ th |≥Δρ crit When , the distribution abnormality warning is triggered; where ρ th Indicates the preset material coverage threshold; Δρ crit represents the critical deviation threshold;

[0076] A dynamic model of traffic morphology is established through a temporal convolutional network to predict the traffic fluctuation threshold, as shown in the following formula (13):

[0077]

[0078] in, represents the traffic value predicted by the temporal convolutional network at time t; L represents the historical time step considered in the temporal convolutional network; W s Represents the weight matrix of the sth historical time step; y t-s represents the measured flow value at time ts; b represents the bias term;

[0079] When the measured flow rate meets the following conditions within three consecutive sampling periods: When , it is determined to be abnormal fluctuation; among them, y t Represents the material flow value measured at time t; σ flow Indicates the standard deviation of material flow.

[0080] Based on the same inventive concept, the present invention also provides an impurity removal and feeding system for gallium metal production, which is used to implement the impurity removal and feeding method for gallium metal production, comprising:

[0081] Intelligent control system, used to input production task information, preset target material addition amount and process parameters, receive data from each module and send control instructions;

[0082] A material level monitoring module, including a material level monitoring sensor, is used to detect the remaining information of the material storage bin in real time and transmit the remaining information to the intelligent control system;

[0083] A material conveying module, comprising a conveying device, electrically connected to the intelligent control system, and configured to convey the material to the adding actuator after the intelligent control system confirms that the material is sufficient;

[0084] A visual recognition module includes an industrial camera array, a laser scanner, and an image processing unit. The industrial camera array is used to collect multi-angle image data of the feeding area. The laser scanner is used to emit structured gratings to obtain laser point cloud data of the target area. The image processing unit is used to perform feature fusion, edge detection, point cloud registration, and coordinate calculation on the image data and laser point cloud data, and transmit the coordinate data of the target feeding point to the intelligent control system.

[0085] The robot system includes a robot body, a robot controller, and an adding actuator. The robot controller is electrically connected to the intelligent control system and is used to receive motion path planning instructions generated by the intelligent control system based on the coordinate data of the target feeding point, calculate the angular displacement of each joint through an inverse kinematics model, and control the robot body to move to the target feeding point. The adding actuator is used to perform the material adding operation;

[0086] A process monitoring module, comprising a concentration sensor and a liquid level sensor arranged in the reaction vessel, for collecting characteristic parameters of the solution in real time and transmitting the data to the intelligent control system after Kalman filtering;

[0087] The intelligent control system is also used to send phased material addition instructions to the addition actuator based on the preset target material addition amount and process parameters, monitor the concentration or liquid level changes during the addition process in real time to dynamically adjust the addition amount, and continuously monitor the material addition status through the visual recognition module, and correct the robot posture or addition parameters in real time when deviation or abnormality is detected.

[0088] Based on the same inventive concept, the present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the above-mentioned impurity removal and feeding method for metal gallium production.

[0089] Based on the same inventive concept, the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the above-mentioned impurity removal and feeding method for metal gallium production is implemented.

[0090] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0091] Significantly improve feeding accuracy and impurity removal efficiency: The intelligent control system presets the addition amount and process parameters, combines the visual recognition system and laser scanner to accurately locate the position and shape of the feeding point, and dynamically adjusts the addition amount in stages. This can avoid the experience-dependent deviation of manual operation, ensure accurate material ratios, optimize the impurity removal process from the root, and improve the purity and impurity removal efficiency of metal gallium.

[0092] Reduce labor intensity and operational risks: Full-process automated control replaces repetitive manual feeding operations, reducing direct contact with corrosive reagents such as lime and hydrogen peroxide, avoiding the risk of material leakage, and significantly improving production safety. At the same time, it eliminates tedious manual operations, freeing up manpower for more efficient work, in line with the trend of industrial intelligent production.

[0093] Achieve dynamic monitoring and real-time correction of the production process: Use material level monitoring sensors to confirm material remaining in real time, visual recognition systems to continuously monitor feeding status, dynamically adjust feeding parameters based on concentration or liquid level changes, and make real-time corrections to the robot's posture to build a closed-loop feedback control system that effectively responds to sudden deviations in the production process and ensures process stability and product quality consistency.

[0094] Improve production efficiency and intelligence: Automated conveying, precise path planning, and adaptive control shorten feeding time, reduce process interruptions caused by human intervention, and improve the controllability of production rhythm; the system integrates multiple sensors and intelligent algorithms to provide key technical support for the digital and intelligent upgrade of gallium metal production, adapting to the demand for refined processes in high-end manufacturing.

[0095] Enhanced production safety and reliability: Avoid safety hazards caused by material misjudgment, dosage deviation, or operational errors during manual operation. Through full-process automation and real-time monitoring, the risk of accidents is reduced, stable equipment operation and production continuity are guaranteed, and a reliable technical solution is provided for the large-scale production of high-purity gallium metal. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 Schematic diagram of the process of the impurity removal and feeding method for gallium metal production according to an embodiment of the present invention;

[0097] Figure 2 Schematic diagram of the structure of the impurity removal and feeding system for gallium metal production according to an embodiment of the present invention. DETAILED DESCRIPTION

[0098] The following is a specific implementation method with reference to the accompanying drawings.

[0099] Reference Figure 1 A method for removing impurities and adding materials for producing metallic gallium comprises the following steps:

[0100] Step 1: Input production task information through the intelligent control system and preset the target material addition amount and process parameters.

[0101] In some embodiments, the operator uses the intelligent control system's user interface, which can be a computer software interface installed in the monitoring room or a dedicated application interface on a handheld mobile device. The operator selects the "Production Task Input" function module on the interface. Within this module, the operator manually enters relevant information for the current production task, such as the production task number, product name, and production batch. The operator then presets the target material addition amount within the corresponding settings interface of the intelligent control system. The operator enters the precise target material addition amount based on production process requirements, product specifications, and past production experience. For example, if the production of a certain type of gallium metal product requires the addition of a specific type and weight of impurity remover according to the recipe, the operator accurately enters the desired weight of the impurity remover into the system. The system can provide unit selection, such as kilograms or grams, to facilitate operator selection based on actual needs. The system also checks the entered amount for plausibility. If the entered value exceeds the preset reasonable range, a prompt prompts the operator to re-enter the value. The operator presets the addition rate at different stages of the material addition process based on the target material's characteristics and production process. For example, during the initial pre-addition phase, a higher addition rate is preset to quickly achieve a specific material concentration. During the main flow stabilization phase, a relatively stable addition rate is preset to ensure stable and accurate material addition. During the precise compensation phase, a lower addition rate is preset. Depending on the specific production process, other parameters may also be preset, such as the temperature control range and stirring speed during material addition. The operator enters these values ​​in the corresponding parameter setting area of ​​the system, and the system records and stores the input parameters. By presetting the target material addition amount and process parameters, the material addition amount and addition process are ensured to strictly comply with the predetermined requirements during each production run. This helps ensure stable and consistent product quality and reduces product quality issues caused by inaccurate material addition amounts or fluctuating process parameters. For example, in gallium metal production, precise impurity removal agent addition amounts and appropriate addition process parameters can effectively remove impurities, improve gallium metal purity, and ensure that it meets specific quality standards. After the intelligent control system presets the parameters, the operator only needs to perform simple parameter input before production, without the need for complex professional knowledge or extensive operating experience. This reduces the requirements for operators and reduces operational errors caused by human factors. For example, a novice operator simply needs to enter parameters according to the system prompts, and the system will automatically complete subsequent material addition control, avoiding errors caused by unfamiliar operation. The system records preset production task information, target material addition amounts, and process parameters. This data serves as a vital record of the production process. If product quality issues arise or production analysis is required, these records can be queried to trace the various parameter settings during the production process, facilitating identification of the cause and implementation of improvements.For example, if it is found that the purity of a batch of gallium metal does not meet the requirements, the parameter settings during the production of the batch can be checked to analyze which link has the problem and take corresponding measures to improve it.

[0102] Step 2: Use the material level monitoring sensor to detect the remaining information of the material storage bin in real time. After confirming that there is enough material, start the conveying device to transport the material to the corresponding adding actuator of the robot.

[0103] In some embodiments, appropriate level monitoring sensors, such as capacitive, radar, or ultrasonic sensors, are selected based on the structure and material characteristics of the material storage bin. For example, a capacitive level sensor may be more suitable for powdered gallium metal production raw materials, while a radar level sensor may be more advantageous for liquid raw materials. Level monitoring sensors are installed at appropriate locations within the material storage bin to ensure accurate detection of the remaining material level within the bin. Sensors can be installed at key locations where material may accumulate or flow, avoiding interference from the bin structure or other factors. Level monitoring sensors collect real-time information about the remaining material level within the bin and transmit this data to the intelligent control system. The sensors collect data at a specific frequency (e.g., once per second or as required) to ensure timely and accurate data. The collected data can be analog or digital, and the intelligent control system has the appropriate signal processing capabilities to receive and interpret different types of signals. A standard threshold for material sufficiency is preset within the intelligent control system. This threshold is determined based on factors such as production process requirements, the capacity of the actuator to be added, and the needs of subsequent production processes. When the remaining material level information collected by the material level monitoring sensor reaches or exceeds a preset sufficient threshold, the intelligent control system determines that the material is sufficient. For example, if the preset threshold is 80% of the storage bin capacity, the sensor detects that the current material level reaches or exceeds this ratio, indicating that the material is sufficient. After confirming that the material is sufficient, the intelligent control system sends a start command to the conveying device. The conveying device can be a belt conveyor, screw conveyor, pneumatic conveyor, etc., depending on the nature of the material and production site conditions. The conveying device starts operating according to preset parameters (such as conveying speed and conveying volume), transporting the material from the material storage bin to the corresponding dosing actuator of the robot. During the conveying process, the conveying device may be equipped with appropriate monitoring equipment such as speed sensors and flow sensors to ensure stable and accurate conveying. By monitoring the remaining material level in the material storage bin in real time, the material adequacy can be promptly detected, avoiding production interruptions caused by material shortages. In the gallium metal production process, each link is closely connected, and the timely supply of materials is critical to ensuring continuous production. If material shortages are not detected in time, it may lead to imbalanced material ratios in the reaction vessel, affecting product quality and production efficiency. Therefore, this step ensures the continuity of the production process and improves production efficiency.

[0104] Step 3: Scan the feeding area through the visual recognition system, combine the laser scanner with the image processing algorithm to obtain the position, shape and size information of the target feeding point, and generate coordinate data.

[0105] In some embodiments, the feeding area is scanned by a visual recognition system, and the position, shape, and size information of the target feeding point are obtained by combining a laser scanner with an image processing algorithm, and the specific process of generating coordinate data is as follows:

[0106] The industrial camera array collects multi-angle image data of the feeding area, while the laser scanner emits structured grating to perform 3D contour scanning of the target area to obtain laser point cloud data;

[0107] The multi-angle image data and the laser point cloud data are fused, and the mapping relationship between the image pixel coordinate system and the laser scanning coordinate system is established using calibration parameters;

[0108] An improved Canny-Sobel composite operator is used to enhance the detection of image edges and extract the geometric contour features of the feeding port. At the same time, the laser point cloud is registered using the ICP algorithm to generate a 3D topological model of the target area.

[0109] The center coordinates of the feeding port are calculated based on the feature matching algorithm, and the continuous preset frame data are weightedly fused using the multi-frame sliding window optimization algorithm to output the final coordinate data.

[0110] Step 4: Based on the coordinate data of the target feeding point, plan the robot's motion path and control the robot to move to the target feeding point.

[0111] In some embodiments, based on the coordinate data of the target feeding point, the robot motion path is planned, and the specific process of controlling the robot to move to the target feeding point is as follows:

[0112] Establish the homogeneous transformation matrix between the robot base coordinate system and the visual coordinate system, as shown in the following formula (1):

[0113]

[0114] in, represents the homogeneous transformation matrix; R represents the rotation matrix; t represents the translation vector;

[0115] The target feeding point coordinates are converted to the robot base coordinate system as shown in the following formula (2):

[0116]

[0117] Among them, P baseIndicates the position of the target feeding point in the robot base coordinate system; P vision Indicates the position of the target feeding point in the visual coordinate system;

[0118] Based on the robot inverse kinematics model, the angular displacement of each joint is calculated as shown in the following formula (3):

[0119]

[0120] Among them, θ i represents the angular displacement of the i-th joint; φ represents the end posture constraint, as shown in the following formula (4):

[0121]

[0122] in, Represents the rotation matrix from the tool coordinate system to the robot base coordinate system; Indicates the normal vector of the target feeding point; Represents the z-axis unit vector of the tool coordinate system; θ pitch represents the pitch angle of the end effector; θ roll represents the roll angle of the end effector; θ P_max and θ R_max They represent the maximum absolute values ​​of the pitch angle and roll angle respectively; ω represents the angular velocity vector of the end effector; represents the angular acceleration vector of the end effector; ω max and Represent the maximum values ​​of angular velocity and angular acceleration respectively; P tool Represents the position coordinates of the end effector; P obs Indicates the location coordinates of the obstacle; d safe Indicates the safety distance threshold;

[0123] Construct the path optimization function as follows (5):

[0124]

[0125] Where m represents the number of nodes; q k represents the joint angle vector of the kth node; q k-1 represents the joint angle vector of the k-1th node; ‖q k -q k-1 ‖ 2 Represents the square norm of the joint angle difference between adjacent nodes, which is used to constrain the smoothness of joint motion; represents the joint angular acceleration vector of the kth node; Represents the square norm of the joint angular acceleration, which is used to constrain the acceleration amplitude during movement; w1 and w2 represent the weight coefficients of smoothness and acceleration, respectively;

[0126] The joint space trajectory is generated by quintic polynomial interpolation, as shown in the following equation (6):

[0127] θ(t)=a0+a1t+a2t 2 +a3t 3 +a4t 4 +a5t 5 (6)

[0128] Where θ(t) represents the angular displacement of the robot joint at time t; a0, a1, a2, a3, a4, and a5 represent the coefficients of the quintic polynomial, which are determined by the boundary conditions of trajectory planning and are used to fit the smooth joint motion trajectory;

[0129] The Lyapunov function is used to verify the trajectory stability, as shown in the following equation (7):

[0130]

[0131] in, represents the Lyapunov function; e represents the error vector of joint angular displacement; represents the joint angular velocity error vector; K p Proportional gain matrix, used to adjust the weight of the error vector of joint angular displacement; K d represents the differential gain matrix, which is used to adjust the weight of the joint angular velocity error vector;

[0132] The joint angle θ is fed back in real time by the encoder during the movement actual , the proportional-integral-differential control law is used to adjust the driving torque, as shown in the following formula (8):

[0133]

[0134] Where τ represents the driving torque vector of the robot joint, which is used to drive the movement of each joint so that the end effector can track the target trajectory; J T represents the transpose of the robot Jacobian matrix; K i represents the integral gain matrix;

[0135] θ desired represents the desired angular displacement vector of the joint; θ actual Represents the actual angular displacement vector of the joint;

[0136] Until the positioning accuracy of the end effector meets ‖P actual -P base ‖≤ε, where P actual represents the actual position coordinate of the robot's end effector; ε represents the positioning accuracy threshold.

[0137] Step 5. According to the preset target material addition amount and process parameters, the intelligent control system sends instructions to the addition actuator to perform the material addition operation in stages, and monitors the concentration or liquid level changes during the addition process in real time to dynamically adjust the addition amount.

[0138] In some embodiments, according to the preset target material addition amount and process parameters, the intelligent control system sends instructions to the addition actuator, and the specific process of performing the material addition operation in stages is as follows:

[0139] According to the preset process parameters, the material addition process is divided into the initialization pre-addition stage, the main flow stable addition stage and the precise compensation addition stage;

[0140] During the initialization pre-addition phase, the intelligent control system calculates the pre-addition threshold based on the volume of the target addition point and the initial concentration of the material, and controls the addition actuator to inject the material into the reaction vessel at a first preset rate until it detects that the liquid level or concentration in the vessel reaches the pre-addition trigger condition;

[0141] During the main flow stable addition stage, the real-time theoretical addition rate is calculated based on the material balance equation. The mapping relationship between the addition rate and the impurity removal efficiency is established in combination with the reaction kinetics model. The flow output of the addition actuator is adjusted through the proportional integral differential control algorithm to stabilize the material addition rate within the theoretical value ±5%;

[0142] During the precise compensation addition stage, the characteristic parameters in the container are collected in real time, the compensation addition amount is calculated through the fuzzy control algorithm, and the addition actuator is controlled to perform micro-addition at a second preset rate until the characteristic parameter change rate is detected to be less than 0.3% / min and the cumulative addition amount reaches 95% to 98% of the preset target value.

[0143] In some embodiments, the concentration or liquid level changes during the addition process are monitored in real time, and the specific process for dynamically adjusting the addition amount is as follows:

[0144] The concentration sensor and liquid level sensor arranged in the reaction vessel collect the solution conductivity, turbidity and liquid level data in real time at a frequency of not less than 20Hz, and the data is input into the intelligent control system after the noise is removed by the Kalman filter algorithm;

[0145] In the initialization pre-addition stage, when the liquid level monitoring value reaches 15% to 20% of the target addition point volume, the main flow stable addition stage start instruction is triggered;

[0146] During the main flow stable addition stage, the deviation between the actual material addition rate and the theoretical addition rate is calculated based on real-time concentration data. The incremental PID control algorithm outputs the adjustment parameters to control the flow fluctuation of the addition actuator within ±3% of the rated flow rate.

[0147] In the precise compensation addition stage, a fuzzy rule base for concentration change rate and addition amount is established. When the characteristic parameter change rate exceeds 0.5% / min for 30 consecutive seconds, the pulse micro-addition mode is activated, with a single pulse addition amount of 0.1mL to 0.5mL, until the characteristic parameter change rate stabilizes below 0.3% / min and the cumulative addition amount error converges to within ±2% of the target value.

[0148] During the entire process, if the concentration or liquid level monitoring value exceeds the preset safety threshold, the intelligent control system automatically sends an emergency stop command and triggers an audible and visual alarm.

[0149] Step 6: Continuously monitor the material addition status through the visual recognition system. If deviation or abnormality is detected, the robot posture is corrected or parameters are added in real time through the intelligent control system.

[0150] In some embodiments, the visual recognition system continuously monitors the material addition status. If a deviation or abnormality is detected, the intelligent control system corrects the robot posture or adds parameters in real time. The specific process is as follows:

[0151] The visual recognition system continuously collects image data of the feeding area at a preset frequency, and uses a deep learning-based target detection algorithm to perform real-time analysis of the material feeding status, identifying the material feeding trajectory, flow pattern, and material distribution characteristics around the feeding port;

[0152] When it is detected that the material addition trajectory deviates from the preset path, the flow pattern fluctuates abnormally, or material deviation or abnormal conditions occur around the feeding port, the abnormal handling mechanism is immediately triggered;

[0153] For robot posture correction, the intelligent control system uses the deviation data fed back by the visual recognition system and the current posture information of the robot to recalculate the angular displacement of each joint of the robot through the inverse kinematics model, generate posture correction instructions and send them to the robot controller to control the robot to adjust the spatial position and posture of the end effector;

[0154] For additive parameter correction, the preset parameter adjustment strategy library is called according to the abnormality type; if a flow abnormality is detected, the flow output of the additive actuator is dynamically adjusted through the proportional integral differential control algorithm; if a material distribution abnormality is detected, the material addition rate or total amount is adjusted according to the real-time concentration or liquid level data;

[0155] During the correction process, the system continuously monitors the correction effect. If the abnormal state is not eliminated, the correction parameters are iteratively optimized until the material addition state returns to normal and meets the process requirements.

[0156] In some embodiments, the specific process of using a deep learning-based target detection algorithm to perform real-time analysis of the material addition status and identify the material addition trajectory, flow pattern, and material distribution characteristics around the feed port is as follows:

[0157] A material state detection model based on the improved YOLOv5 is constructed. The loss function of the material state detection model is shown in the following formula (9):

[0158]

[0159] in, represents the total loss function of the material state detection model; cls Represents the classification loss weight coefficient; A represents the total number of grids in the feature map; B represents the number of bounding boxes predicted for each grid; represents the characteristic function; C ij Represents the true category confidence; Represents the predicted category confidence; λ loc represents the positioning damage weight coefficient; p ijk Represents the real bounding box coordinate parameters; Represents the predicted bounding box coordinate parameters; SmoothL1 loss function for bounding box localization;

[0160] The continuous image frames are processed by the spatiotemporal feature fusion module, and the optical flow constraint equation is defined as shown in the following equation (10):

[0161] I(x,y,t)=I(x+Δx,y+Δy,t+Δt)(10)

[0162] Where I(x,y,t) represents the brightness value of the pixel with coordinates (x,y) in the image at time t; Δx and Δy represent the displacement of the pixel in the x and y directions within the time interval Δt, respectively.

[0163] Solve the dense optical flow field using the Horn-Schunck algorithm And extract the material trajectory vector Where N represents the number of optical flow sampling points; Represents the optical flow vector of the i-th sampling point; when When , it is determined that the trajectory deviates; represents the reference trajectory vector, δ traj Indicates the trajectory deviation threshold;

[0164] A dual-branch convolutional network is used to extract traffic morphology features and material distribution features respectively. The traffic morphology branch extracts multi-scale spatiotemporal features through a dilated convolutional layer, as shown in the following formula (11):

[0165]

[0166] in, Represents the characteristic value of the traffic shape characteristic map of the lth layer at the spatial coordinate (x, y); Represents the weight coefficient of the hole convolution kernel of the lth layer at the relative coordinate (i, j); I(x+d·i,y+d·j) is the pixel value of the input image at the absolute coordinate (x+d·i,y+d·j); d represents the expansion rate of the hole convolution; k represents the half-side length of the convolution kernel; (i,j) represents the relative coordinates within the convolution kernel, i,j∈[-k,k];

[0167] The material distribution branch generates a pixel-level segmentation mask through the U-Net structure and calculates the material coverage around the feeding port, as shown in the following formula (12):

[0168]

[0169] Where ρ represents the material coverage rate around the feeding port; |R| represents the total number of pixels in the preset target area R around the feeding port; M(x,y) represents the pixel-level segmentation mask function. When the pixel (x,y) belongs to the material coverage area, M(x,y) = 1, otherwise M(x,y) = 0;

[0170] When |ρ-ρ th |≥Δρ crit When , the distribution abnormality warning is triggered; where ρ th Indicates the preset material coverage threshold; Δρ crit represents the critical deviation threshold;

[0171] A dynamic model of traffic morphology is established through a temporal convolutional network to predict the traffic fluctuation threshold, as shown in the following formula (13):

[0172]

[0173] in, represents the traffic value predicted by the temporal convolutional network at time t; L represents the historical time step considered in the temporal convolutional network; W s Represents the weight matrix of the sth historical time step; y t-s represents the measured flow value at time ts; b represents the bias term;

[0174] When the measured flow rate meets the When , it is determined to be abnormal fluctuation; among them, y t Represents the material flow value measured at time t; σ flow Indicates the standard deviation of material flow.

[0175] Based on the same inventive concept, corresponding to any of the above embodiments, refer to Figure 2The present invention provides a system for removing impurities and feeding for the production of metallic gallium, which is used to implement the aforementioned method for removing impurities and feeding for the production of metallic gallium, comprising:

[0176] Intelligent control system, used to input production task information, preset target material addition amount and process parameters, receive data from each module and send control instructions;

[0177] A material level monitoring module, including a material level monitoring sensor, is used to detect the remaining information of the material storage bin in real time and transmit the remaining information to the intelligent control system;

[0178] A material conveying module, comprising a conveying device, electrically connected to the intelligent control system, and configured to convey the material to the adding actuator after the intelligent control system confirms that the material is sufficient;

[0179] A visual recognition module includes an industrial camera array, a laser scanner, and an image processing unit. The industrial camera array is used to collect multi-angle image data of the feeding area. The laser scanner is used to emit structured gratings to obtain laser point cloud data of the target area. The image processing unit is used to perform feature fusion, edge detection, point cloud registration, and coordinate calculation on the image data and laser point cloud data, and transmit the coordinate data of the target feeding point to the intelligent control system.

[0180] The robot system includes a robot body, a robot controller, and an adding actuator. The robot controller is electrically connected to the intelligent control system and is used to receive motion path planning instructions generated by the intelligent control system based on the coordinate data of the target feeding point, calculate the angular displacement of each joint through an inverse kinematics model, and control the robot body to move to the target feeding point. The adding actuator is used to perform the material adding operation;

[0181] The process monitoring module includes a concentration sensor and a liquid level sensor arranged in the reaction vessel, which are used to collect characteristic parameters such as solution conductivity, turbidity, and liquid level in real time, and transmit the data to the intelligent control system after Kalman filtering;

[0182] The intelligent control system is also used to send phased material addition instructions to the addition actuator based on the preset target material addition amount and process parameters, monitor the concentration or liquid level changes during the addition process in real time to dynamically adjust the addition amount, and continuously monitor the material addition status through the visual recognition module, and correct the robot posture or addition parameters in real time when deviation or abnormality is detected.

[0183] Based on the same inventive concept, corresponding to any of the above embodiments, the present invention provides an electronic device, including a memory and a processor, the memory being used to store a computer program, and the processor running the computer program to enable the electronic device to execute the impurity removal and feeding method for metal gallium production of the embodiment.

[0184] Optionally, the above-mentioned electronic device may be a server.

[0185] In addition, this embodiment further provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the impurity removal and feeding method for metal gallium production of the embodiment is implemented.

[0186] It is understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0187] The method steps in the embodiments of the present invention can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC.

[0188] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a storage medium or transmitted via a storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

Claims

1. A method for removing impurities and adding materials for the production of metallic gallium, characterized in that: The following steps are involved: Input production task information through the intelligent control system and preset the target material addition amount and process parameters; The material level monitoring sensor detects the remaining information of the material storage bin in real time. After confirming that there is enough material, the conveying device is started to transport the material to the corresponding adding actuator of the robot; The feeding area is scanned by a visual recognition system, and the position, shape and size information of the target feeding point are obtained by combining a laser scanner with an image processing algorithm, and coordinate data is generated; Based on the coordinate data of the target feeding point, the robot motion path is planned and the robot is controlled to move to the target feeding point; According to the preset target material addition amount and process parameters, the intelligent control system sends instructions to the addition actuator to perform material addition operations in stages, and monitors the concentration or liquid level changes during the addition process in real time to dynamically adjust the addition amount; The material addition status is continuously monitored through the visual recognition system. If deviation or abnormality is detected, the robot posture is corrected or parameters are added in real time through the intelligent control system.

2. The impurity removal and feeding method for gallium metal production according to claim 1, characterized in that: The specific process of scanning the feeding area by the visual recognition system, combining the laser scanner with the image processing algorithm, obtaining the position, shape and size information of the target feeding point, and generating the coordinate data is as follows: The industrial camera array collects multi-angle image data of the feeding area, while the laser scanner emits structured grating to perform 3D contour scanning of the target area to obtain laser point cloud data; The multi-angle image data and the laser point cloud data are fused, and the mapping relationship between the image pixel coordinate system and the laser scanning coordinate system is established using calibration parameters; An improved Canny-Sobel composite operator is used to enhance the detection of image edges and extract the geometric contour features of the feeding port. At the same time, the laser point cloud is registered using the ICP algorithm to generate a 3D topological model of the target area. The center coordinates of the feeding port are calculated based on the feature matching algorithm, and the continuous preset frame data are weightedly fused using the multi-frame sliding window optimization algorithm to output the final coordinate data.

3. The impurity removal and feeding method for gallium metal production according to claim 1, characterized in that: The specific process of planning the robot motion path based on the coordinate data of the target feeding point and controlling the robot to move to the target feeding point is as follows: Establish the homogeneous transformation matrix between the robot base coordinate system and the visual coordinate system, as shown in the following formula (1): in, represents the homogeneous transformation matrix; R represents the rotation matrix; t represents the translation vector; The target feeding point coordinates are converted to the robot base coordinate system as shown in the following formula (2): Among them, P base Indicates the position of the target feeding point in the robot base coordinate system; P vision Indicates the position of the target feeding point in the visual coordinate system; Based on the robot inverse kinematics model, the angular displacement of each joint is calculated as shown in the following formula (3): Among them, θ i represents the angular displacement of the i-th joint; φ represents the end posture constraint, as shown in the following formula (4): in, Represents the rotation matrix from the tool coordinate system to the robot base coordinate system; Indicates the normal vector of the target feeding point; Represents the z-axis unit vector of the tool coordinate system; θ pitch represents the pitch angle of the end effector; θ roll represents the roll angle of the end effector; θ P_max and θ R_max They represent the maximum absolute values ​​of the pitch angle and roll angle respectively; ω represents the angular velocity vector of the end effector; represents the angular acceleration vector of the end effector; ω max and Represent the maximum values ​​of angular velocity and angular acceleration respectively; P tool Represents the position coordinates of the end effector; P obs Indicates the location coordinates of the obstacle; d safe Indicates the safety distance threshold; Construct the path optimization function as follows (5): Where m represents the number of nodes; q k represents the joint angle vector of the kth node; q k-1 represents the joint angle vector of the k-1th node; ‖q k -q k-1 ‖ 2 Represents the square norm of the joint angle difference between adjacent nodes, which is used to constrain the smoothness of joint motion; represents the joint angular acceleration vector of the kth node; Represents the square norm of the joint angular acceleration, which is used to constrain the acceleration amplitude during movement; w1 and w2 represent the weight coefficients of smoothness and acceleration, respectively; The joint space trajectory is generated by quintic polynomial interpolation, as shown in the following equation (6): θ(t)=a0+a1t+a2t 2 +a3t 3 +a4t 4 +a5t 5 (6) Where θ(t) represents the angular displacement of the robot joint at time t; a0, a1, a2, a3, a4, and a5 represent the coefficients of the quintic polynomial, which are determined by the boundary conditions of trajectory planning and are used to fit the smooth joint motion trajectory; The Lyapunov function is used to verify the trajectory stability, as shown in the following equation (7): in, represents the Lyapunov function; e represents the error vector of joint angular displacement; represents the joint angular velocity error vector; K p Proportional gain matrix, used to adjust the weight of the error vector of joint angular displacement; K d represents the differential gain matrix, which is used to adjust the weight of the joint angular velocity error vector; The joint angle θ is fed back in real time by the encoder during the movement actual , the proportional-integral-differential control law is used to adjust the driving torque, as shown in the following formula (8): Where τ represents the driving torque vector of the robot joint, which is used to drive the movement of each joint so that the end effector can track the target trajectory; J T represents the transpose of the robot Jacobian matrix; K i represents the integral gain matrix; θ desired represents the desired angular displacement vector of the joint; θ actual Represents the actual angular displacement vector of the joint; Until the positioning accuracy of the end effector meets ‖P actual -P basw ‖≤ε, where P actual represents the actual position coordinate of the robot's end effector; ε represents the positioning accuracy threshold.

4. The impurity removal and feeding method for gallium metal production according to claim 1, characterized in that: The specific process of executing the material addition operation in stages by sending instructions to the addition actuator through the intelligent control system according to the preset target material addition amount and process parameters is as follows: According to the preset process parameters, the material addition process is divided into the initialization pre-addition stage, the main flow stable addition stage and the precise compensation addition stage; During the initialization pre-addition phase, the intelligent control system calculates the pre-addition threshold based on the volume of the target addition point and the initial concentration of the material, and controls the addition actuator to inject the material into the reaction vessel at a first preset rate until it detects that the liquid level or concentration in the vessel reaches the pre-addition trigger condition; During the main flow stable addition stage, the real-time theoretical addition rate is calculated based on the material balance equation. The mapping relationship between the addition rate and the impurity removal efficiency is established in combination with the reaction kinetics model. The flow output of the addition actuator is adjusted through the proportional integral differential control algorithm to stabilize the material addition rate within the theoretical value ±5%; During the precise compensation addition stage, the characteristic parameters in the container are collected in real time, the compensation addition amount is calculated through the fuzzy control algorithm, and the addition actuator is controlled to perform micro-addition at a second preset rate until the characteristic parameter change rate is detected to be less than 0.3% / min and the cumulative addition amount reaches 95% to 98% of the preset target value.

5. The impurity removal and feeding method for gallium metal production according to claim 4, characterized in that: The specific process of real-time monitoring of the concentration or liquid level changes during the addition process and dynamically adjusting the addition amount is as follows: The concentration sensor and liquid level sensor arranged in the reaction vessel collect the solution conductivity, turbidity and liquid level data in real time at a frequency of not less than 20Hz, and the data is input into the intelligent control system after the noise is removed by the Kalman filter algorithm; In the initialization pre-addition stage, when the liquid level monitoring value reaches 15% to 20% of the target addition point volume, the main flow stable addition stage start instruction is triggered; During the main flow stable addition stage, the deviation between the actual material addition rate and the theoretical addition rate is calculated based on real-time concentration data. The incremental PID control algorithm outputs the adjustment parameters to control the flow fluctuation of the addition actuator within ±3% of the rated flow rate. In the precise compensation addition stage, a fuzzy rule base for concentration change rate and addition amount is established. When the characteristic parameter change rate exceeds 0.5% / min for 30 consecutive seconds, the pulse micro-addition mode is activated, with a single pulse addition amount of 0.1mL to 0.5mL, until the characteristic parameter change rate stabilizes below 0.3% / min and the cumulative addition amount error converges to within ±2% of the target value. During the entire process, if the concentration or liquid level monitoring value exceeds the preset safety threshold, the intelligent control system automatically sends an emergency stop command and triggers an audible and visual alarm.

6. The impurity removal and feeding method for gallium metal production according to claim 1, characterized in that: The specific process of continuously monitoring the material addition status through the visual recognition system and correcting the robot posture or adding parameters in real time through the intelligent control system if deviation or abnormality is detected is as follows: The visual recognition system continuously collects image data of the feeding area at a preset frequency, and uses a deep learning-based target detection algorithm to perform real-time analysis of the material feeding status, identifying the material feeding trajectory, flow pattern, and material distribution characteristics around the feeding port; When it is detected that the material addition trajectory deviates from the preset path, the flow pattern fluctuates abnormally, or material deviation or abnormal conditions occur around the feeding port, the abnormal handling mechanism is immediately triggered; For robot posture correction, the intelligent control system uses the deviation data fed back by the visual recognition system and the current posture information of the robot to recalculate the angular displacement of each joint of the robot through the inverse kinematics model, generate posture correction instructions and send them to the robot controller to control the robot to adjust the spatial position and posture of the end effector; For additive parameter correction, the preset parameter adjustment strategy library is called according to the abnormality type; if a flow abnormality is detected, the flow output of the additive actuator is dynamically adjusted through the proportional integral differential control algorithm; if a material distribution abnormality is detected, the material addition rate or total amount is adjusted according to the real-time concentration or liquid level data; During the correction process, the system continuously monitors the correction effect. If the abnormal state is not eliminated, the correction parameters are iteratively optimized until the material addition state returns to normal and meets the process requirements.

7. The impurity removal and feeding method for gallium metal production according to claim 6, characterized in that: The specific process of using the deep learning-based target detection algorithm to perform real-time analysis of the material addition status and identify the material addition trajectory, flow pattern, and material distribution characteristics around the feeding port is as follows: A material state detection model based on the improved YOLOv5 is constructed. The loss function of the material state detection model is shown in the following formula (9): in, represents the total loss function of the material state detection model; cls Represents the classification loss weight coefficient; A represents the total number of grids in the feature map; B represents the number of bounding boxes predicted for each grid; represents the characteristic function; C ij Represents the true category confidence; Represents the predicted category confidence; λ loc represents the positioning damage weight coefficient; p ijk Represents the real bounding box coordinate parameters; Represents the predicted bounding box coordinate parameters; SmoothL1 loss function for bounding box localization; The continuous image frames are processed by the spatiotemporal feature fusion module, and the optical flow constraint equation is defined as shown in the following equation (10): I(x,y,t)=I(x+Δx,y+Δy,t+Δt) (10) Where I(x,y,t) represents the brightness value of the pixel with coordinates (x,y) in the image at time t; Δx and Δy represent the displacement of the pixel in the x and y directions within the time interval Δt, respectively. Solve the dense optical flow field using the Horn-Schunck algorithm And extract the material trajectory vector Where N represents the number of optical flow sampling points; Represents the optical flow vector of the i-th sampling point; when When , it is determined that the trajectory deviates; represents the reference trajectory vector, δ traj Indicates the trajectory deviation threshold; A dual-branch convolutional network is used to extract traffic morphology features and material distribution features respectively. The traffic morphology branch extracts multi-scale spatiotemporal features through a dilated convolutional layer, as shown in the following formula (11): in, Represents the characteristic value of the traffic shape characteristic map of the lth layer at the spatial coordinate (x, y); Represents the weight coefficient of the hole convolution kernel of the lth layer at the relative coordinate (i, j); I(x+d·i,y+d·j) is the pixel value of the input image at the absolute coordinate (x+d·i,y+d·j); d represents the expansion rate of the hole convolution; k represents the half-side length of the convolution kernel; (i,j) represents the relative coordinates within the convolution kernel, i,j∈[-k,k]; The material distribution branch generates a pixel-level segmentation mask through the U-Net structure and calculates the material coverage around the feeding port, as shown in the following formula (12): Where ρ represents the material coverage rate around the feeding port; |R| represents the total number of pixels in the preset target area R around the feeding port; M(x,y) represents the pixel-level segmentation mask function. When the pixel (x,y) belongs to the material coverage area, M(x,y) = 1, otherwise M(x,y) = 0; When |ρ-ρ th |≥Δρ crit When , the distribution abnormality warning is triggered; where ρ th Indicates the preset material coverage threshold; Δρ crit represents the critical deviation threshold; A dynamic model of traffic morphology is established through a temporal convolutional network to predict the traffic fluctuation threshold, as shown in the following formula (13): in, represents the traffic value predicted by the temporal convolutional network at time t; L represents the historical time step considered in the temporal convolutional network; W s Represents the weight matrix of the sth historical time step; y t-s represents the measured flow value at time ts; b represents the bias term; When the measured flow rate meets the following conditions within three consecutive sampling periods: When , it is determined to be abnormal fluctuation; among them, y t Represents the material flow value measured at time t; σ flow Indicates the standard deviation of material flow.

8. A system for removing impurities and feeding for gallium metal production, used to implement the method for removing impurities and feeding for gallium metal production according to any one of claims 1 to 7, characterized in that: include: Intelligent control system, used to input production task information, preset target material addition amount and process parameters, receive data from each module and send control instructions; A material level monitoring module, including a material level monitoring sensor, is used to detect the remaining information of the material storage bin in real time and transmit the remaining information to the intelligent control system; A material conveying module, comprising a conveying device, electrically connected to the intelligent control system, and configured to convey the material to the adding actuator after the intelligent control system confirms that the material is sufficient; A visual recognition module includes an industrial camera array, a laser scanner, and an image processing unit. The industrial camera array is used to collect multi-angle image data of the feeding area. The laser scanner is used to emit structured gratings to obtain laser point cloud data of the target area. The image processing unit is used to perform feature fusion, edge detection, point cloud registration, and coordinate calculation on the image data and laser point cloud data, and transmit the coordinate data of the target feeding point to the intelligent control system. The robot system includes a robot body, a robot controller, and an adding actuator. The robot controller is electrically connected to the intelligent control system and is used to receive motion path planning instructions generated by the intelligent control system based on the coordinate data of the target feeding point, calculate the angular displacement of each joint through an inverse kinematics model, and control the robot body to move to the target feeding point. The adding actuator is used to perform the material adding operation; A process monitoring module, comprising a concentration sensor and a liquid level sensor arranged in the reaction vessel, for collecting characteristic parameters of the solution in real time and transmitting the data to the intelligent control system after Kalman filtering; The intelligent control system is also used to send phased material addition instructions to the addition actuator based on the preset target material addition amount and process parameters, monitor the concentration or liquid level changes during the addition process in real time to dynamically adjust the addition amount, and continuously monitor the material addition status through the visual recognition module, and correct the robot posture or addition parameters in real time when deviation or abnormality is detected.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the impurity removal and feeding method for metal gallium production according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the impurity removal and feeding method for metal gallium production according to any one of claims 1 to 7 is implemented.

Images