Automatic control method of rare and precious metal intelligent separation and purification system

Abstract

The present application belongs to the technical field of metallurgical intelligent equipment, and particularly relates to an automatic control method of a rare and precious metal intelligent separation and purification system. The method comprises: starting the rare and precious metal intelligent separation and purification system, completing parameter initialization and whole machine hardware state self-checking; using a two-stage fuzzy PID feeding control system to complete automatic weighing and flat scraping of raw materials; based on a self-walking device and a depth camera, using a PID self-disturbance visual controller to complete automatic alignment of pushing a boat into a furnace; after the boat is put into the furnace, using a Hebb supervised single neuron furnace temperature control system to realize temperature control reduction of the raw materials; based on the self-walking device and the depth camera, completing automatic tracking and alignment of pulling the boat out of the furnace; and breaking a consolidated layer, quality inspection and rare and precious metal recovery of the reduced materials in the boat. The present application effectively improves the accuracy of feeding, effectively resists interference such as furnace tube deformation and stress release, ensures the positioning accuracy of the boat, and solves problems such as high production cost, low yield and high energy consumption of rare and precious metal production.

Description

Technical Field

[0001] This invention belongs to the field of metallurgical intelligent equipment technology, and in particular relates to an automatic control method for an intelligent separation and purification system for rare and precious metals. It is applicable to the high-temperature reduction and purification of rare and precious metals such as ruthenium, rhodium, palladium, iridium, platinum, and rhenium in a boat-type process, and can be applied to the secondary resource recovery of rare and precious metals and the continuous reduction production of high-purity precious metal powders. Background Technology

[0002] With the rapid development of my country's high-tech industries, the demand for rare and precious metals is increasing. These metals have wide applications in aerospace, defense, chemical energy, environmental protection, pharmaceuticals, and consumer electronics. For a long time, my country has failed to achieve large-scale production in the secondary recycling and purification of rare and precious metals such as gold, silver, ruthenium, rhodium, palladium, osmium, iridium, platinum, gallium, vanadium, titanium, indium, germanium, and tellurium. The level of intelligent purification equipment is still in its basic stage, and there is a significant gap compared to foreign countries in terms of quality control and production automation. Currently, there is an urgent need to develop secondary recycling, value-added processing, and regeneration of rare and precious metals. This will not only alleviate the depletion of my country's strategic rare and precious metal resources but also be of great significance for enhancing my country's reserve strength, the development of strategic emerging industries, and the level of high technology.

[0003] Patent CN114540623A discloses an intelligent separation and purification system and method for rare and precious metals. This system includes a reduction device, a material boat propulsion mechanism, and an automatic control device. However, it lacks specific control algorithms and implementation methods for core aspects of the purification process, such as precise control of feeding and weighing accuracy, accurate alignment of the material boat entering and exiting the furnace, and adaptive adjustment of the reduction furnace temperature. The material boat propulsion relies solely on conventional mechanical conveying, which cannot cope with interference from high-temperature conditions such as thermal expansion and contraction of the furnace tubes and material boat deformation. Furthermore, the feeding and temperature control processes struggle to ensure parameter consistency in the rare and precious metal reduction process, failing to address the problems of unstable purification quality and low yield.

[0004] The patent with publication number CN117073361A proposes an automated feeding system and its control method in vacuum smelting. It uses fuzzy PID to calculate the material taking amount, but its application scenario is the feeding control of vacuum smelting furnace, which cannot take into account both feeding efficiency and weighing accuracy in small weighing scenarios. Moreover, this solution only involves a single feeding link and cannot form a complete automatic control system for rare and precious metal purification equipment.

[0005] Patent CN113843417A discloses an automated loading and unloading production line for a pushboat furnace. It automates loading and unloading through conveyor rollers and robotic arms. However, its positioning relies on the rigid positioning of mechanical baffles and claws, making it unable to detect the thermal expansion and contraction deformation and stress release of the furnace tubes caused by high temperatures. Its application scenario is limited to powder metallurgy loading and unloading stations at room temperature, and it cannot adapt to the dynamic deformation conditions of high-temperature reduction furnaces for rare and precious metals, making it difficult to guarantee the positional accuracy and environmental adaptability of the loading boats entering the furnace.

[0006] The patent with publication number CN121386364A proposes an automated control method and system for precious metal purification processes. It achieves automated control by monitoring dendrite growth through sound and light fusion. Its technical route is wet electrolytic purification, which does not involve feeding and conveying of material boats, nor does it involve temperature control of reduction furnaces. Therefore, it cannot be applied to the automatic control of reduction purification equipment.

[0007] The patent with publication number CN101349892A proposes a fuzzy-PID composite control system for adding water during sintering. It switches between fuzzy control and PID control by means of deviation threshold. However, its application scenario is the adjustment of the water valve in the sintering process, which is completely different from the control object and process requirements of weighing and adding rare and precious metal boats.

[0008] Currently, the technology and automation of rare and precious metal purification equipment are generally low. Poor weighing accuracy during purification and separation, easy interference from high-temperature deformation in the material boat conveying and positioning, and insufficient stability of furnace temperature control result in high production costs, low yields, high energy consumption, poor product uniformity, and unstable quality in rare and precious metal purification. At the same time, the production process involves a high proportion of manual operation, resulting in high labor intensity for operators and susceptibility to dust pollution from the industrial environment. These issues have become important problems that urgently need to be addressed for the comprehensive utilization of rare and precious metals. Summary of the Invention

[0009] The purpose of this invention is to provide an automatic control method for an intelligent separation and purification system for rare and precious metals. A two-stage fuzzy proportional-integral-derivative (PID) automatic feeding control system is developed to automatically weigh the raw materials in the feeding boat, making the feeding more precise. Based on a self-propelled device and a depth camera, a PID self-disturbance rejection vision controller is developed to automatically align the boat during the furnace feeding process, effectively resisting thermal expansion and contraction deformation and stress release of the furnace tubes, ensuring the adaptability and accuracy of the boat's position in the reduction furnace. A Hebb-based supervised single-neuron furnace temperature control system achieves precise reduction of the raw materials, ensuring the quality of rare and precious metal purification. This invention can reduce the labor intensity of operators, reduce dust pollution in the industrial environment during rare and precious metal purification, and achieve the goal of digitally empowering the intelligent separation and purification system for rare and precious metals.

[0010] To achieve the above-mentioned objectives, the present invention adopts the following technical solution:

[0011] An automatic control method for a rare and precious metal intelligent separation and purification system includes the following steps:

[0012] S1: The intelligent separation and purification system for rare and precious metals starts up, completing parameter initialization and self-check of the overall hardware status;

[0013] S2: The two-stage fuzzy PID feeding control system is used to complete the automatic weighing and leveling of raw materials;

[0014] S3: Based on a self-propelled device and a depth camera, an automatic alignment for pushing the boat into the furnace is achieved using a PID self-disturbance vision controller;

[0015] S4: After the material boat is put into the furnace, the raw material temperature control reduction is achieved by using the Hebb supervised single-neuron furnace temperature control system;

[0016] S5: Based on a self-propelled device and a depth camera, it completes automatic tracking and alignment during the pulling of the boat out of the furnace;

[0017] S6: Perform cracking of the caking layer, quality inspection, and recovery of rare and precious metals on the reduced material in the material boat.

[0018] The automatic control method for the intelligent separation and purification system of rare and precious metals includes step S1: preparing rare and precious metal raw materials and material boats, turning on the power, starting the system, performing self-checks on the status of the self-propelled device and cylinders, and initializing various parameters.

[0019] The automatic control method for the intelligent separation and purification system of rare and precious metals, step S2 includes:

[0020] S21: In the first stage, coarse feeding is carried out and the raw material is weighed in real time. When the real-time weighed feeding amount m is less than or equal to 0.85 times the given amount M, coarse feeding continues. When the real-time weighed feeding amount m exceeds 0.85M, the second stage of fuzzy PID feeding control is carried out to complete fine feeding.

[0021] S22: Initialize error value at time 0 Virtualized "-1" time value The control law of a conventional PID controller in the fine feeding stage is as follows:

[0022] (1)

[0023] In the above formula, This is the output value of the feed weight, i.e., the motor speed control. Assume that the relationship between the motor speed and the weight being weighed is linear, with a proportionality constant of 1 / 2. Error value Input value and output value The difference, It is the proportional coefficient of a standard PID controller. It is the integral time constant of a conventional PID controller. It is the differential time constant of a conventional PID controller; with respect to time t, the integral term... and differential terms Discretize the data as shown in equation (2):

[0024] (2)

[0025] In the above formula, k is the discrete sampling number, and the sampling period T = 1~3s. Therefore, the expression for positional discrete PID control is:

[0026] (3)

[0027] In the above formula, m(k) is the feed control output at the k-th sampling time, e(k) is the weighing error value at the k-th sampling time, and K P0 The baseline proportional coefficient is used, and the standard PID integral coefficient is... The derivative coefficient of a conventional PID controller is ; Complete the three parameters of a conventional PID according to equation (3) , , Initially, an incremental PID controller was constructed based on the positional PID controller, defining the increment of the feeding output:

[0028] (4)

[0029] Further simplification yields:

[0030] (5)

[0031] in, , , ;

[0032] S23: Utilizing fuzzy PID to dynamically and in real-time correct traditional PID coefficients, based on weighing error values. and error change rate To be respectively the proportional coefficient correction amount Integral coefficient correction amount Differential coefficient correction amount Construct fuzzy rules based on practical work experience and weighing error values. The language values ​​{NB, NM, NS, O, PS, PM, PB} are selected from the range [-1.5M, 0]. Similarly, the error rate of change is used to select the language values. Select the corresponding language values ​​{NB, NM, NS, O, PS, PM, PB} within the range [-1.5M, 1.5M], and adjust them according to the actual PID coefficient range. , , Select the corresponding language value {NB, NM, NS, O, PS, PM, PB};

[0033] S24: Obtain the parameters for regular PID correction based on the fuzzy rule table. , , ,

[0034] (6)

[0035] The three coefficients K of the corrected PID P K I K D By replacing the coefficients in equation (5), we obtain the corrected fuzzy PID output equation:

[0036] (7)

[0037] In the above formula, Δm(k) is the increment of the feed weight control at the k-th sampling time; T is the unified sampling period of the system, k is the discrete sampling number, and e(k) is the weighing error value at the k-th sampling time;

[0038] The incremental speed control value is obtained by using the relationship between the output speed of the feeding motor and the weight of the material being fed:

[0039] (8)

[0040] Thus, the speed output of the feeding motor is obtained based on the given material weight M. This is to complete the control of the automatic weighing and feeding control system.

[0041] The automatic control method for an intelligent separation and purification system for rare and precious metals uses fuzzy rule language definition: input NB represents severe overfeeding, NM represents moderate overfeeding, NS represents slight overfeeding, O represents weight matching, PS represents slight underfeeding, PM represents moderate underfeeding, and PB represents severe underfeeding; output NB represents a significant decrease in PID coefficient, NM represents a moderate decrease, NS represents a slight decrease, O represents no adjustment, PS represents a slight increase, PM represents a moderate increase, and PB represents a significant increase.

[0042] The automatic control method for the intelligent separation and purification system of rare and precious metals, wherein step S3 includes:

[0043] S31: Constructing a World Coordinate System Depth camera coordinate system and furnace door coordinate system In space, the three sets of coordinate systems are parallel to each other, and the positions of the front and rear end markers of the material boat are set as follows: and The depth camera is mounted at a position that remains constant relative to the furnace door and at a suitable angle. The depth camera determines the real-time position of the material boat by identifying the marks before and after the material boat.

[0044] S32: Complete the calibration of the intrinsic and extrinsic parameters of the D435i depth camera;

[0045] S33: Given the initial position of the trolley when the furnace door is activated. Let the location of the marker on the material boat, captured by the depth camera, be... Combining coordinate system relationships and constructing a Denavit-Hartenberg (DH) model Where T is the transformation matrix and P is the position. Given a third-order identity matrix, position the boat marker as... Substituting this into P yields the transformation matrix of the boat marker relative to the depth camera. Since the positions of the depth camera and the furnace door are relatively fixed, the coordinate system of the depth camera can be determined. Relative furnace door coordinate system Transformation matrix The coordinate transformation is performed according to the following formula (9) to obtain the position of the marker at the front end of the material boat. Or the location of the marker at the rear of the boat Relative to the furnace door coordinate system Real-time conversion matrix:

[0046] (9)

[0047] Extracting the marker position at the front end of the material boat using the DH model Or the location of the marker at the rear of the boat Relative to the furnace door coordinate system Real-time location ; The initial position of the trolley when the furnace door is activated. Real-time position of the marker on the material boat, captured in real time by a depth camera. The difference is calculated, and the derivative is taken with respect to time to obtain the position error of the material boat. and error rate :

[0048] (10)

[0049] S34: Let e(t) = By analogy with equation (2), the PID output speed can also be obtained. The expression for PID control is:

[0050] (11)

[0051] In the above formula, It is the sampling period. It is the discrete sampling sequence number. It is the PID proportional coefficient. It is the integral time constant of a conventional PID controller. It is the derivative time constant of a conventional PID controller;

[0052] S35: Due to the variable factors of material boat deformation, furnace tube thermal expansion and contraction, and stress release, in order to enhance the system's anti-interference capability, a classic active disturbance rejection controller control function Z is added, assuming... To accelerate speed control, a post-predicted speed value is obtained by using a traditional EKF extended Kalman filter based on a depth camera for speed prediction. The speed control value of the self-propelled device is obtained. :

[0053] (12)

[0054] In the above formula, v ee v is the walking speed after disturbance compensation by the active disturbance rejection controller. p The velocity prediction value is the output of the EKF extended Kalman filter; v B The final control speed is sent to the servo drive of the self-propelled device; based on the above control quantity, the self-propelled device is driven to move, and the automatic alignment of the boat-pushing-in-furnace process is completed.

[0055] The automatic control method for the intelligent separation and purification system of rare and precious metals, step S4 includes:

[0056] S41: Input is The output is , Let be the discrete sampling sequence number, and define the temperature error as . ,but

[0057] (13)

[0058] The increment of the control quantity is obtained by discretizing the PID controller. :

[0059] (14)

[0060] In the above formula, It is a proportionality coefficient. It is the integral coefficient. These are differential coefficients;

[0061] S42: Add monitoring items The post-single neuron algorithm is as follows:

[0062] (15)

[0063] In the above formula, , , These are the weights corresponding to the three inputs of the neuron at time k. , , These are the learning rates for the integral channel, proportional channel, and derivative channel. , , These are the three input quantities of the neuron, which are constructed from the temperature error and its difference:

[0064] (16)

[0065] S43: Let K be the proportional coefficient of the neuron output, then the temperature control quantity of a single neuron output... Increment for:

[0066] (17)

[0067] Adjusting weights using a single-neuron supervised Hebb learning algorithm , , Initialize weights =0, =0, =0, learning rate , , Self-regulation, setting , ,but Update according to equation (13) ,set up =0, ,Depend on renew Update according to equation (16) , , Substitution (15) update , , Set the value of K (K>0) and substitute it into equation (17) to obtain the increment of the control quantity of the neuron system. .

[0068] The automatic control method of the intelligent separation and purification system for rare and precious metals includes step S5: when the reduction time is up, prepare the boat to be unloaded from the furnace, and use the PID self-disruption vision controller consistent with step S3 to complete the automatic tracking and alignment in the boat unloading operation based on the self-propelled device and the depth camera D435i.

[0069] The aforementioned automatic control method for an intelligent separation and purification system for rare and precious metals is applicable to the production of rare and precious metal powders such as ruthenium, rhodium, palladium, platinum, or rhenium using a boat-type hydrogen reduction purification process.

[0070] The design concept of this invention is:

[0071] In existing technologies, the weighing accuracy of raw material feeding is insufficient, making it difficult to balance feeding efficiency and weighing accuracy, resulting in deviations in raw material ratios and fluctuations in purification quality. In the feeding stage, this invention adopts a two-stage fuzzy PID control architecture, using 85% of the target setpoint (the set weight of material in a single boat) as a threshold to switch between coarse feeding and fine feeding stages. Coarse feeding ensures feeding efficiency, while the fine feeding stage dynamically corrects incremental PID parameters through fuzzy rules to achieve precise control of weighing accuracy.

[0072] In existing technologies, the alignment accuracy of pushing the boat into the furnace and pulling it out is easily affected by factors such as thermal expansion and contraction of the furnace tubes and deformation of the boat during use, resulting in poor positioning adaptability under high-temperature conditions. In the process of conveying the boat, this invention uses a multi-coordinate system DH transformation model constructed based on a depth camera to obtain the real-time position of the boat. It integrates PID control, active disturbance rejection controller and EKF speed prediction to form a PID active disturbance rejection vision controller, which actively compensates for interferences such as thermal expansion and contraction of the furnace tubes and deformation of the boat, thereby achieving automatic tracking and alignment of pushing the boat into the furnace and pulling it out.

[0073] In existing technologies, the temperature control of reduction furnaces is not robust enough and cannot adapt to the nonlinear and large hysteresis characteristics of rare and precious metal reduction, resulting in poor reduction quality stability. In the reduction process, this invention adopts a Hebb supervised single-neuron furnace temperature control system, using temperature error, error rate of change and second-order error change as neuron inputs. Through supervised learning, the weights are adjusted to adaptively match the temperature control requirements of the reduction process.

[0074] Thus, through the coordinated operation of two-stage fuzzy PID feeding control, PID self-disturbance rejection visual boat alignment, and single-neuron adaptive furnace temperature control, automatic control is achieved from system self-inspection, feeding, boat pushing, reduction, furnace discharge to post-processing, realizing intelligent automatic control of purification, improving feeding accuracy, positioning adaptability and reduction quality stability, and reducing manual labor intensity and dust pollution.

[0075] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0076] 1. This invention provides an automatic control method for an intelligent separation and purification system for rare and precious metals. First, a two-stage fuzzy PID automatic feeding control system is used to automatically weigh the raw materials in the feeding boat, making the feeding more precise. Second, a PID self-disturbance rejection vision controller is developed based on a self-propelled device and a depth camera to automatically align the boat during the pushing into and pulling out of the furnace, effectively resisting interference such as thermal expansion and contraction deformation and stress release of the furnace tube, ensuring the adaptability and accuracy of the boat's position in the reduction furnace. Finally, a Hebb supervised single-neuron furnace temperature control system is used to achieve precise reduction of the raw materials, ensuring the quality of rare and precious metal purification.

[0077] 2. This invention adopts a two-stage fuzzy PID automatic feeding control system. By segmenting the feeding into coarse and fine feeding, it significantly improves the accuracy of raw material weighing while ensuring feeding efficiency, ensuring the flatness and proportion accuracy of raw materials in the material boat, and providing stable material conditions for subsequent reduction and purification.

[0078] 3. This invention constructs a PID active disturbance rejection vision controller based on a self-propelled device and a depth camera. It obtains the precise position of the material boat through multi-coordinate system transformation and integrates active disturbance rejection control and extended Kalman filter velocity prediction. It can effectively resist interference such as furnace tube thermal expansion and contraction deformation, stress release, and material boat deformation during use, ensuring the adaptability and accuracy of the material boat's position in the reduction furnace and improving the reliability of equipment operation under high temperature conditions.

[0079] 4. This invention adopts Hebb's supervised single-neuron furnace temperature control system, which learns and adaptively adjusts the temperature control parameters to adapt to the nonlinear and large hysteresis characteristics of the rare and precious metal reduction process, thereby achieving precise reduction of raw materials and effectively ensuring the quality stability of rare and precious metal purification.

[0080] 5. This invention realizes automated control from system self-inspection, automatic feeding, pushing the boat into the furnace, temperature-controlled reduction, pulling the boat out of the furnace to post-processing, reducing the labor intensity of operators, reducing the impact of industrial dust pollution on personnel and products, realizing digital empowerment of the intelligent separation and purification system for rare and precious metals, and improving overall production efficiency and product quality consistency. Attached Figure Description

[0081] Figure 1 This is a flowchart of the automatic control method for an intelligent separation and purification system for rare and precious metals in this invention.

[0082] Figure 2 This is a flowchart illustrating the principle of the two-stage fuzzy PID automatic feeding control system of the present invention.

[0083] Figure 3 This is the control block diagram of the second stage of the automatic feeding control system of the present invention, which is a fuzzy PID feeding control.

[0084] Figure 4 This is a schematic diagram illustrating the establishment of the coordinate system during the automatic boat pushing process of the present invention.

[0085] Figure 5 This is a block diagram of the PID self-disturbance rejection visual control for the automatic boat pushing process of the present invention.

[0086] Figure 6 This is a block diagram of the Hebb supervised single-neuron furnace temperature control system of the present invention. Detailed Implementation

[0087] In its specific implementation, this invention is based on the automatic control method proposed in "An Intelligent Separation and Purification System and Method for Rare and Precious Metals (Publication No. CN114540623A)". The hardware utilizes existing waste electrolysis devices, detection devices, separation devices, purification devices, reduction devices, and automatic control devices. The only improvement is the addition of a high-precision weighing sensor and a D435i depth camera as supporting data acquisition units for the control method. Furthermore, a supporting intelligent control algorithm and closed-loop regulation logic are added to the existing hardware. The improvements are in the intelligent control logic for feeding, visual boat pushing, and furnace temperature. The control method includes the following steps:

[0088] S1: The intelligent separation and purification system for rare and precious metals is started, completing the initialization of the overall parameters and the self-check of the hardware status of the sensors and actuators;

[0089] S2: A two-stage fuzzy PID control system is used to complete the automatic weighing and filling of raw materials and the leveling and scraping of the material surface.

[0090] S3: Based on the D435i depth camera and self-propelled device, a PID self-disturbance-resistant vision controller is used to complete the automatic alignment of the material boat into the furnace.

[0091] S4: After the material boat is put into the furnace, the Hebb supervised single neuron control system is used to realize the adaptive temperature control reduction of the reduction furnace.

[0092] S5: Reuses the same PID self-disturbance rejection vision control logic as S3 to complete the automatic tracking and alignment of the material boat pulling out of the furnace;

[0093] S6: The furnace material is crushed and the caking layer is inspected online. Rare and precious metal powders are recovered. The empty material boat is returned to the feeding station for cyclic operation.

[0094] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. The specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention. It should also be noted that, for ease of description, only the parts relevant to the present invention are shown in the accompanying drawings.

[0095] like Figure 1 As shown, this invention proposes an automatic control method for an intelligent separation and purification system for rare and precious metals, comprising the following steps:

[0096] Step 1: The intelligent separation and purification system for rare and precious metals is started, and parameter initialization and self-check of the overall hardware status are completed.

[0097] Step 2: Utilize a two-stage fuzzy PID feeding control system to automatically weigh and level the raw materials. For example... Figure 2As shown, the automatic loading command begins. The first stage involves coarse feeding, with real-time weighing of the raw materials. When the real-time weighed feeding amount m is less than or equal to 0.85 times the given amount M, coarse feeding continues. When the real-time weighed feeding amount m exceeds 0.85M, the second stage, fuzzy PID feeding control, is implemented to complete fine feeding. Figure 3 As shown, the initial error value at time 0. Virtualized "-1" time value The control law of a conventional PID controller in the fine feeding stage is as follows:

[0098] (1)

[0099] In the above formula, the error value Input value and output value The difference represents the deviation between the target weight and the actual output weight; r(t) is the input value of the target weight of the feed. This is the output value of the added weight, v M (t) represents the speed of the feeding motor, satisfying the motor speed control requirement. Let's assume that the relationship between the speed of the feeding motor and the weight is linear, with a proportionality constant of 1 / 2. , It is the proportional coefficient of a standard PID controller. It is the integral time constant of a conventional PID controller. This is the derivative time constant of a conventional PID controller. With respect to time t, the integral term... and differential terms Discretize the data as shown in equation (2):

[0100] (2)

[0101] In the above formula, k is the discrete sampling index (k=0,1,2……), i is the index of the integral summation, and the sampling period T ranges from 1 to 2. The feasible range for the sampling period T is 1~3s. Within this range, the fuzzy PID control system can operate stably during the fine feeding stage (the control effect is optimal when the sampling period T=1s). Therefore, the expression for the positional discrete PID control can be obtained as follows:

[0102] (3)

[0103] In the above formula, m(k) is the feed control output at the k-th sampling time, e(k) is the weighing error value at the k-th sampling time, and K P0 The baseline proportional coefficient is used, and the standard PID integral coefficient is... The derivative coefficient of a conventional PID controller is Based on equation (3), complete the three parameters of a conventional PID controller. , , Initially, an incremental PID was constructed based on the positional PID, defining the increment of the feeding output. Let's assume...

[0104] (4)

[0105] Further simplification yields:

[0106] (5)

[0107] in, , , Traditional PID controllers with fixed parameters struggle to adapt to time-varying disturbances during powder feeding, failing to simultaneously balance dynamic response speed and steady-state weighing accuracy. This invention utilizes fuzzy PID to dynamically and in real-time correct the coefficients of traditional PID controllers. Based on the weighing error value... and error change rate To solve for the proportional coefficient correction amount separately Integral coefficient correction amount Differential coefficient correction amount Construct fuzzy rules based on practical work experience and weighing error values. The language values ​​{NB, NM, NS, O, PS, PM, PB} are selected from the range [-1.5M, 0]. Similarly, the error rate of change is used to select the language values. Select the corresponding language values ​​{NB, NM, NS, O, PS, PM, PB} within the range [-1.5M, 1.5M], and adjust them according to the actual PID coefficient range. , , Select the corresponding language values ​​{NB, NM, NS, O, PS, PM, PB}. The following are three sets of fuzzy rules, as shown in Tables 1, 2, and 3:

[0108] Table 1 Fuzzy rule table

[0109]

[0110] Table 2 Fuzzy rule table

[0111]

[0112] Table 3 Fuzzy rule table

[0113]

[0114] Based on the error value e(k) = set material weight - current material weight, the physical meanings of each language are explained in two categories: input and output. Input includes the weighing error value e(k) and the error change rate e. c (k) indicates that the negative value is extremely large, the current material weight is much greater than the set material weight, and the material is being overloaded significantly; NM indicates that the negative value is moderate, the current material weight is slightly overloaded, and there is a moderate degree of overloading; NS indicates that the negative value is very small, the actual weight is slightly overloaded, and there is only a trace amount of overloading; O indicates that the value deviation is approximately equal to 0, the current material weight is basically consistent with the set material weight, the weighing deviation is minimal, and the expected weight range is reached; PS indicates that the positive value is very small, the actual weight is slightly lower than the target, and there is a slight shortage of material; PM indicates that the positive value is moderate, the current material is significantly insufficient, and there is a moderate shortage of material; PB indicates that the positive value is extremely large, the material in the material boat is far less than the set material weight, and there is a severe shortage of material. The output correction includes the proportional coefficient correction. Integral coefficient correction amount Differential coefficient correction amount NB indicates a significant reduction in the corresponding PID baseline coefficient, NM indicates a moderate reduction in the corresponding PID baseline coefficient, NS indicates a slight reduction in the corresponding PID baseline coefficient, O indicates no adjustment to the PID baseline coefficient, PS indicates a slight increase in the corresponding PID baseline coefficient, PM indicates a moderate increase in the corresponding PID baseline coefficient, and PB indicates a significant increase in the corresponding PID baseline coefficient.

[0115] The parameters for correcting a conventional PID can be obtained from the fuzzy rule table. , , ,

[0116] (6)

[0117] In the above formula, K P0 K I0 K D0 ΔK represents the initial reference coefficients for the PID control during the fine feeding stage. P ΔK I ΔK D These represent the correction increments for the proportional coefficient, integral coefficient, and differential coefficient of the fuzzy inference output; K P K I K D These are the adaptive PID real-time coefficients at the current sampling time after online correction.

[0118] The three coefficients K of the corrected PID P K I K DBy replacing the coefficients in equation (5), we can obtain the corrected fuzzy PID output equation:

[0119] (7)

[0120] In the above formula, Δm(k) is the increment of the feed weight control at the k-th sampling time; T is the unified sampling period of the system, k is the discrete sampling number, and e(k) is the weighing error value at the k-th sampling time.

[0121] Based on the linear relationship between the output speed of the feeding motor and the weight of the material being fed, the incremental speed control value of the feeding motor can be obtained as follows:

[0122] (8)

[0123] In the above formula, Δv m (k) represents the speed output of the feeding motor, which can be obtained based on the given material weight M. This is to complete the control of the automatic weighing and feeding control system.

[0124] Step 3: Automatic tracking and alignment are achieved during the boat-pulling and unloading operation using a self-propelled device and a depth camera. For example... Figure 4 As shown, construct the world coordinate system Depth camera coordinate system and furnace door coordinate system The three coordinate systems are parallel to each other in space. The positions of the front and rear markers of the material boat are set as follows: and The depth camera is positioned at a suitable angle relative to the furnace door, and it determines the real-time position of the material boat by identifying the markers at the front and rear of the boat. The entire process of pushing the boat into the furnace includes control after identifying the positions of the two markers. During the entry into the furnace, it is necessary to pay attention to the position of the marker at the front of the material boat. To activate the furnace door, after the charging boat has partially entered, pay attention to the position of the marker at the rear of the charging boat. This triggers the closing of the furnace door. The furnace door control process is the same, such as... Figure 5 As shown, the first step is to calibrate the intrinsic and extrinsic parameters of the D435i depth camera, given the initial position of the trolley when the furnace door is activated. Let's assume the location of the marker on the material boat captured by the depth camera is... Combining the coordinate system relationships of this invention with the ability to construct a Denavit-Hartenberg (DH) model Where T is the transformation matrix and P is the position. It is a third-order identity matrix. The position of the material boat marker is... Substituting this into P yields the transformation matrix of the boat marker relative to the depth camera. Since the positions of the depth camera and the furnace door are relatively fixed, the coordinate system of the depth camera can be determined. Relative furnace door coordinate system Transformation matrix The position of the marker at the front end of the material boat can be obtained by performing coordinate transformation according to the following formula (9). (or the location of the marker at the rear of the boat) Relative to the furnace door coordinate system Real-time conversion matrix:

[0125] (9)

[0126] The position of the marker at the front end of the material boat can be extracted using the DH model. (or the location of the marker at the rear of the boat) Relative to the furnace door coordinate system Real-time location The initial position of the trolley when the furnace door is activated. Real-time position of the marker on the material boat, captured in real time by a depth camera. By taking the difference and differentiating it with respect to time, the positional error of the material boat can be obtained. and error rate (i.e., the rate of change of position error over time):

[0127] (10)

[0128] Let e(t) = By analogy with equation (2), the PID output speed can also be obtained. The discrete PID control expression is:

[0129] (11)

[0130] In the above formula, It is the sampling period. It is the discrete sampling sequence number. It is the PID proportional coefficient. It is the integral time constant of a conventional PID controller. This is the derivative time constant of a conventional PID controller. Since external disturbances such as deformation of the material boat over long-term use, thermal expansion and contraction of the furnace tubes, and stress release can affect positioning accuracy, a classic active disturbance rejection controller function Z is added to enhance the system's anti-interference capabilities. Let's assume... To accelerate dynamic response speed control, based on the time-series information of the boat's position continuously acquired by a depth camera, a traditional EKF extended Kalman filter method is used for speed prediction, resulting in a post-predicted speed value. By superimposing the speed after disturbance compensation with the predicted speed, the speed control value of the self-propelled device can be obtained. :

[0131] (12)

[0132] In the above formula, v ee v is the walking speed after disturbance compensation by the active disturbance rejection controller. p The velocity prediction value is the output of the EKF extended Kalman filter; v B This is the final control speed sent to the servo drive of the self-propelled device. Based on the above control quantity, the self-propelled device is driven to move, completing the automatic alignment of the boat-pushing process into the furnace; the boat-pulling process can reuse this set of control formulas, only adjusting the direction of movement of the self-propelled device, and the disturbance observation weight of the active disturbance rejection module can be adjusted according to the high-temperature material loading condition.

[0133] Step 4: After the feed boat is placed into the furnace, the raw material temperature-controlled reduction is achieved using Hebb's supervised single-neuron furnace temperature control system. For example... Figure 6 As shown, the input is The output is , Let be the discrete sampling sequence number, and define the temperature error as . ,but

[0134] (13)

[0135] In the above formula, r T (k) represents the process set temperature at the k-th discrete sampling time, y T (k) represents the measured temperature of the reduction furnace at the k-th sampling time; k is the discrete sampling number, and the system uses the same constant sampling period T as the aforementioned feeding and visual positioning control.

[0136] Based on the discretization of the PID controller, the increment of the control quantity can be obtained as follows: :

[0137] (14)

[0138] In the above formula, It is a proportionality coefficient. It is the integral coefficient. These are the differential coefficients.

[0139] This invention employs supervised Hebb learning rules to achieve adaptive iteration of single neuron weights, adding a supervision term. The single-neuron algorithm is as follows:

[0140] (15)

[0141] In the above formula , , These are the weights corresponding to the three inputs of the neuron at time k. , , These are the learning rates corresponding to the integral channel, proportional channel, and derivative channel. The learning rate can be adaptively adjusted according to the heating, isothermal, and cooling process stages. , , These are the three input quantities of the neuron, which are constructed from the temperature error and its difference:

[0142] (16)

[0143] In the above formula, x1(k) represents the temperature error, x2(k) represents the first-order difference of the temperature error, and x3(k) represents the second-order difference Δ of the temperature error. 2 e T (k).

[0144] according to Figure 6 As shown, let K be the proportional coefficient of the neuron output, then the temperature control quantity output by a single neuron... Increment for:

[0145] (17)

[0146] Adjusting weights using a single-neuron supervised Hebb learning algorithm , , Initialize weights =0, =0, =0, learning rate , , It can be self-adjusting, so let's set it up. , ,but According to equation (13), it can be updated Let's assume =0, , which can be determined by renew According to equation (16), it can be updated , , Substituting equation (15) can update , , Let's set a value for K (K>0) and substitute it into equation (17) to obtain the increment of the control quantity of the neuron system. The supervised Hebb learning algorithm described above is used to update the weights in real time, enabling adaptive temperature control of the reduction furnace.

[0147] Step 5: When the reduction time is up, prepare to unload the charging boat. Based on the self-propelled device and the D435i depth camera, the automatic tracking and alignment during the unloading operation is completed using a PID active disturbance rejection vision controller. The control algorithm for this step 5 is completely consistent with that of step 3.

[0148] Step 6: Break the caking layer of the reduced material in the material boat, perform quality inspection, and recover rare and precious metals.

[0149] Example 1: Purification and Control of Platinum-Palladium Mixed Rare and Precious Metal Powder

[0150] 1. Preparatory work: Platinum-palladium mixed ammonium chloride powder is used as raw material. The material weight of the single material boat is set to M=1000g, the sampling period is fixed at T=1s, and the whole machine is powered on to complete the self-test of the sensor, servo and camera hardware.

[0151] 2. Feeding control: During the coarse feeding stage, the motor runs at high speed and weighs the powder in real time to 850g (0.85M) before switching to fine feeding. During fine feeding, the weighing error and error change rate are calculated in real time. The PID proportional, integral, and derivative coefficients are corrected in real time through fuzzy rules to dynamically reduce the motor speed. Finally, the powder is accurately filled to 1000g, and the leveling mechanism smooths the powder surface.

[0152] 3. Boat alignment: The D435 camera identifies the position M1 of the front end of the material boat in real time, solves the positioning error through DH coordinate transformation, and outputs the basic walking speed through PID; the self-disturbance module compensates for high temperature deformation disturbances, and the EKF predicts the walking speed and drives the trolley to enter the furnace smoothly. The furnace door is automatically closed when the position M2 of the rear end of the material boat is identified.

[0153] 4. Temperature-controlled reduction: A three-stage heating process is adopted, and a single neuron controller iterates the weights in real time to adaptively adjust the heating power and stably maintain the constant temperature range of each stage of reduction to complete the hydrogen reduction reaction of platinum palladium powder.

[0154] 5. Unloading: After the reduction and heat preservation cycle ends, the material boat is pulled out in reverse using visual self-disturbance control logic.

[0155] 6. Post-processing: Crush agglomerated powder, detect the purity of precious metals online, collect finished products by grade, and return empty material boats to the feeding station.

[0156] The entire process requires no manual intervention, the weighing deviation is minimal, there is no impact or jamming of the material boat, and the furnace temperature fluctuation is significantly reduced.

[0157] Example 2: Purification and Control of Ammonium Rheniumate Rare and Precious Metal Powder

[0158] 1. Preparatory steps: Use ammonium perrylate precursor powder as raw material. Set the material weight M=500g for a single material boat and select T=2s for the sampling period. Power on the whole machine to complete the self-test of the sensor, servo, and camera hardware.

[0159] 2. Feeding control: Coarse feeding continues until 425g (0.85M) and then fine feeding is switched; Fuzzy PID dynamically corrects PID parameters based on the real-time material shortage / over-material level, and finely adjusts the motor speed at low speed to accurately fill to 500g and level the powder surface.

[0160] 3. Pushing the boat into the furnace: The depth camera continuously collects the three-dimensional coordinates of the front and rear end marker positions of the boat, and converts them into the furnace door reference position through the DH matrix; the position error is fed into the PID, and the self-disturbance rejection compensation and EKF predicted speed drive the trolley to smoothly enter the furnace and automatically open and close the furnace door.

[0161] 4. Temperature-controlled reduction: The rhenium reduction low-temperature process is adopted. The single neuron adaptively adjusts the learning rate of the three channels according to the heating, constant temperature and cooling stages, and updates the neuron weights online to stabilize the reaction temperature in the furnace.

[0162] 5. Unloading: After the holding time is reached, the material boat is pulled out by moving in the opposite direction. The vision tracking control logic is exactly the same as that for pushing the boat.

[0163] 6. Post-processing: The crushed material caking occurs, and the high-purity rhenium metal powder is collected by quality inspection. The empty material boat is recycled.

[0164] In this embodiment, a sampling period of 2 seconds is selected, which can still stably complete the automatic control of the entire process. The weighing, positioning, and temperature control effects meet the requirements of industrial production.

[0165] The results show that the present invention effectively improves the accuracy of feeding, effectively resists interference such as furnace tube deformation and stress release, ensures the accuracy of material boat positioning, and solves the problems of high production cost, low yield and high energy consumption of rare and precious metals.

Claims

1. An automatic control method for an intelligent separation and purification system for rare and precious metals, characterized in that, Includes the following steps: S1: The intelligent separation and purification system for rare and precious metals starts up, completing parameter initialization and self-check of the overall hardware status; S2: The two-stage fuzzy PID feeding control system is used to complete the automatic weighing and leveling of raw materials; S3: Based on a self-propelled device and a depth camera, an automatic alignment for pushing the boat into the furnace is achieved using a PID self-disturbance vision controller; S4: After the material boat is put into the furnace, the raw material temperature control reduction is achieved by using the Hebb supervised single-neuron furnace temperature control system; S5: Based on a self-propelled device and a depth camera, it completes automatic tracking and alignment during the pulling of the boat out of the furnace; S6: Perform cracking of the caking layer, quality inspection, and recovery of rare and precious metals on the reduced material in the material boat.

2. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 1, further characterized in that, Step S1 includes: preparing rare and precious metal raw materials and boats, turning on the power, starting the system, performing self-checks on the self-propelled device and cylinder status, and initializing various parameters.

3. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 1, characterized in that, Step S2 includes: S21: In the first stage, coarse feeding is carried out and the raw material is weighed in real time. When the real-time weighed feeding amount m is less than or equal to 0.85 times the given amount M, coarse feeding continues. When the real-time weighed feeding amount m exceeds 0.85M, the second stage of fuzzy PID feeding control is carried out to complete fine feeding. S22: Initialize error value at time 0 Virtualized "-1" time value The control law of a conventional PID controller in the fine feeding stage is as follows: （1） In the above formula, This is the output value of the feed weight, i.e., the motor speed control. Assume that the relationship between the motor speed and the weight being weighed is linear, with a proportionality constant of 1 / 2. Error value Input value and output value The difference, It is the proportional coefficient of a standard PID controller. It is the integral time constant of a conventional PID controller. It is the differential time constant of a conventional PID controller; with respect to time t, the integral term... and differential terms Discretize the data as shown in equation (2): （2） In the above formula, k is the discrete sampling number, and the sampling period T = 1~3s. Therefore, the expression for positional discrete PID control is: （3） In the above formula, m(k) is the feed control output at the k-th sampling time, e(k) is the weighing error value at the k-th sampling time, and K P0 The baseline proportional coefficient is used, and the standard PID integral coefficient is... The derivative coefficient of a conventional PID controller is ; Complete the three parameters of a conventional PID according to equation (3) , , Initially, an incremental PID controller was constructed based on the positional PID controller, defining the increment of the feeding output: （4） Further simplification yields: （5） in, , , ; S23: Utilizing fuzzy PID to dynamically and in real-time correct traditional PID coefficients, based on weighing error values. and error change rate To be respectively the proportional coefficient correction amount Integral coefficient correction amount Differential coefficient correction amount Construct fuzzy rules based on practical work experience and weighing error values. The language values ​​{NB, NM, NS, O, PS, PM, PB} are selected from the range [-1.5M, 0]. Similarly, the error rate of change is used to select the language values. Select the corresponding language values ​​{NB, NM, NS, O, PS, PM, PB} within the range [-1.5M, 1.5M], and adjust them according to the actual PID coefficient range. , , Select the corresponding language value {NB, NM, NS, O, PS, PM, PB}; S24: Obtain the parameters for regular PID correction based on the fuzzy rule table. , , , （6） The three coefficients K of the corrected PID P K I K D By replacing the coefficients in equation (5), we obtain the corrected fuzzy PID output equation: （7） In the above formula, Δm(k) is the increment of the feed weight control at the k-th sampling time; T is the unified sampling period of the system, k is the discrete sampling number, and e(k) is the weighing error value at the k-th sampling time; The incremental speed control value is obtained by using the relationship between the output speed of the feeding motor and the weight of the material being fed: （8） Thus, the speed output of the feeding motor is obtained based on the given material weight M. This is to complete the control of the automatic weighing and feeding control system.

4. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 3, characterized in that, Fuzzy rule definition: Input NB represents severe overfeed, NM represents moderate overfeed, NS represents slight overfeed, O represents weight matching, PS represents slight underfeed, PM represents moderate underfeed, and PB represents severe underfeed; Output NB represents a significant decrease in PID coefficient, NM represents a moderate decrease, NS represents a slight decrease, O represents no adjustment, PS represents a slight increase, PM represents a moderate increase, and PB represents a significant increase.

5. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 3, characterized in that, Step S3 includes: S31: Constructing a World Coordinate System Depth camera coordinate system and furnace door coordinate system In space, the three sets of coordinate systems are parallel to each other, and the positions of the front and rear end markers of the material boat are set as follows: and The depth camera is mounted at a position that remains constant relative to the furnace door and at a suitable angle. The depth camera determines the real-time position of the material boat by identifying the marks before and after the material boat. S32: Complete the calibration of the intrinsic and extrinsic parameters of the D435i depth camera; S33: Given the initial position of the trolley when the furnace door is activated. Let the location of the marker on the material boat, captured by the depth camera, be... Combining coordinate system relationships and constructing a Denavit-Hartenberg (DH) model Where T is the transformation matrix and P is the position. Given a third-order identity matrix, position the boat marker as... Substituting this into P yields the transformation matrix of the boat marker relative to the depth camera. Since the positions of the depth camera and the furnace door are relatively fixed, the coordinate system of the depth camera can be determined. Relative furnace door coordinate system Transformation matrix The coordinate transformation is performed according to the following formula (9) to obtain the position of the marker at the front end of the material boat. Or the location of the marker at the rear of the boat Relative to the furnace door coordinate system Real-time conversion matrix: （9） Extracting the marker position at the front end of the material boat using the DH model Or the location of the marker at the rear of the boat Relative to the furnace door coordinate system Real-time location ; The initial position of the trolley when the furnace door is activated. Real-time position of the marker on the material boat, captured in real time by a depth camera. The difference is calculated, and the derivative is taken with respect to time to obtain the position error of the material boat. and error rate : （10） S34: Let e(t) = By analogy with equation (2), the PID output speed can also be obtained. The expression for PID control is: （11） In the above formula, It is the sampling period. It is the discrete sampling sequence number. It is the PID proportional coefficient. It is the integral time constant of a conventional PID controller. It is the derivative time constant of a conventional PID controller; S35: Due to the variable factors of material boat deformation, furnace tube thermal expansion and contraction, and stress release, in order to enhance the system's anti-interference capability, a classic active disturbance rejection controller control function Z is added, assuming... To accelerate speed control, a post-predicted speed value is obtained by using a traditional EKF extended Kalman filter based on a depth camera for speed prediction. The speed control value of the self-propelled device is obtained. : （12） In the above formula, v ee v is the walking speed after disturbance compensation by the active disturbance rejection controller. p The velocity prediction value is the output of the EKF extended Kalman filter; v B The final control speed is sent to the servo drive of the self-propelled device; based on the above control quantity, the self-propelled device is driven to move, and the automatic alignment of the boat-pushing-in-furnace process is completed.

6. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 5, characterized in that, Step S4 includes: S41: Input is The output is , Let be the discrete sampling sequence number, and define the temperature error as . ,but （13） The increment of the control quantity is obtained by discretizing the PID controller. : （14） In the above formula, It is a proportionality coefficient. It is the integral coefficient. These are differential coefficients; S42: Add monitoring items The post-single neuron algorithm is as follows: （15） In the above formula, , , These are the weights corresponding to the three inputs of the neuron at time k. , , These are the learning rates for the integral channel, proportional channel, and derivative channel. , , These are the three input quantities of the neuron, which are constructed from the temperature error and its difference: （16） S43: Let K be the proportional coefficient of the neuron output, then the temperature control quantity of a single neuron output... Increment for: （17） Adjusting weights using a single-neuron supervised Hebb learning algorithm , , Initialize weights =0, =0, =0, learning rate , , Self-regulation, setting , ,but Update according to equation (13) ,set up =0, ,Depend on renew Update according to equation (16) , , Substitution (15) update , , Set the value of K (K>0) and substitute it into equation (17) to obtain the increment of the control quantity of the neuron system. .

7. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 1, characterized in that, Step S5 includes: when the reduction time is up, prepare the boat to be unloaded from the furnace, and use the PID active disturbance rejection vision controller consistent with step S3 to complete the automatic tracking and alignment in the boat unloading operation based on the self-propelled device and depth camera D435i.

8. The automatic control method for a rare and precious metal intelligent separation and purification system according to claim 1, characterized in that, The method is applicable to the production of rare and precious metal powders such as ruthenium, rhodium, palladium, platinum or rhenium using boat-type hydrogen reduction purification.

Images