Zinc powder producing and manufacturing device and manufacturing process thereof

Abstract

The zinc powder producing and manufacturing device comprises a smashing box, a feeding hopper is fixedly arranged on the upper portion of the smashing box, a discharging hopper is fixedly arranged on the lower portion of the smashing box, a smashing box supporting rod is fixedly arranged at the other end of the lower portion of the smashing box, and a smashing roller is rotationally connected into the smashing box; a motor base is fixedly arranged on the side face of the crushing box, and a driving motor is fixedly arranged above the motor base; raw material particles meeting the size of screen meshes are effectively screened out through the screening plate, it is guaranteed that the sizes of the raw material particles entering the next procedure are uniform, meanwhile, the raw material particles larger than the size of the screen meshes can slide into the collecting box, secondary smashing is facilitated, full utilization of raw materials is guaranteed, and waste is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of zinc powder production equipment, in particular to a zinc powder production device and a manufacturing process thereof. Background Art

[0002] Some existing equipment may lack some screening mechanisms and rapid cleaning and drying devices for raw materials when producing zinc powder. The lack of efficient screening mechanisms makes it impossible to accurately screen out zinc powder raw material particles that meet specific particle size requirements, resulting in uneven raw material particle size in subsequent processing, affecting product quality. The lack of effective cleaning devices makes it impossible to completely remove impurities and contaminants in the raw materials, affecting the purity and quality of the zinc powder. The imperfect drying device may lead to poor drying effect and excessive residual moisture in the raw materials, which in turn affects subsequent grinding and packaging processes. Summary of the invention

[0003] The purpose of the present invention is to provide a zinc powder production device and a manufacturing process thereof in order to solve the problems mentioned in the above background.

[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a zinc powder production and manufacturing device, comprising: a crushing box, a feed hopper is fixedly arranged above the crushing box, a discharge funnel is fixedly arranged below the crushing box, a crushing box support rod is fixedly arranged at the other end below the crushing box, a crushing roller is rotatably connected inside the crushing box, a motor base is fixedly arranged on the side of the crushing box, and a driving motor is fixedly arranged above the motor base.

[0005] As a further solution of the present invention: a water washing box is fixedly connected below the crushing box support rod, a sub-screen plate support rod is fixedly provided at one end above the water washing box, a sub-screen plate is fixedly provided on the side of the sub-screen plate support rod, a collecting box is fixedly provided on the other side of the water washing box, a collecting box support bottom plate is fixedly provided below the collecting box, the collecting box support bottom plate is fixedly connected to the side of the water washing box, and a cleaning door is provided on one side of the collecting box.

[0006] As a further solution of the present invention: a water washing box inlet is provided above the water washing box, a wastewater storage tank is fixedly provided below the water washing box, a cleaning inclined plate is fixedly provided on the inside of the water washing box, a nozzle is fixedly provided on the other end of the inside of the water washing box, one end of the outside of the water washing box is fixedly connected to a water supply pipe, the other end of the water supply pipe is fixedly connected to a water tank, the water tank is fixed to the other end of the outside of the water washing box, a water replenishment hole is provided on one side of the water washing box, a drainage hole is provided on one side of the wastewater storage tank, a water washing box outlet is provided on the other side of the water washing box, a drying box is fixedly connected to the other side of the drying box, and a docking pipe is fixedly connected to the other side of the drying box.

[0007] As a further solution of the present invention: a grinding basin is fixedly connected to the other side of the docking pipe, a bottom plate is fixedly arranged below the grinding basin, a grinding disc support seat is fixedly arranged on the side of the bottom plate, a telescopic rod is fixedly arranged below the grinding disc support seat, a rotating shaft is rotatably connected below the telescopic rod, a grinding disc is rotatably connected below the rotating shaft, a rolling disc is fixedly arranged inside the grinding basin, and a suction pump is fixedly connected to the other side of the grinding basin.

[0008] As a further solution of the present invention: the driving motor drives the crushing roller to rotate through the outer side of the crushing box, and the crushing roller can preliminarily crush the raw materials, and the crushed raw materials are discharged from the discharge funnel.

[0009] As a further solution of the present invention: the sub-screen plate is located directly below the discharge funnel, and the discharge funnel has a certain inclination angle, so that the raw material particles larger than the mesh size of the sub-screen plate can slide into the collecting box, and the cleaning door on the side of the collecting box is convenient for taking out the separated raw materials, and then putting them into the crushing box again for secondary crushing, and the raw material particles that meet the size fall into the water washing box from the sub-screen plate and the water washing box entrance.

[0010] As a further solution of the present invention: the cleaning inclined plate also has a certain angle, which is convenient for the raw materials to slide out of the water washing box from the water washing box outlet and enter the drying box. The nozzle is aimed at the cleaning inclined plate for spraying, and the raw materials will directly fall on the cleaning inclined plate from the water washing box inlet. Water flow holes are opened on the surface of the cleaning inclined plate, and the cleaned sewage will fall into the wastewater storage tank through the water flow holes. The drying box can dry the cleaned raw materials, and the dried raw materials are transmitted into the grinding basin through the docking pipe.

[0011] As a further solution of the present invention: the grinding disc divides the interior of the grinding basin into two parts, the raw material will be sprinkled onto the grinding disc through the docking pipe, and the grinding disc rotates through the rotating shaft to grind the raw material particles on the grinding disc, and the ground zinc powder falls into the lower part of the grinding basin through the through holes on the surface of the grinding disc, and then the extraction pump extracts the ground zinc powder.

[0012] The invention also discloses a zinc powder production process:

[0013] First, the raw material enters the crushing box through the feed hopper, and the driving motor drives the crushing roller to rotate inside the crushing box through an external connection to perform preliminary crushing on the raw material. The crushed raw material is discharged through a feed hopper with a certain inclination angle and falls onto the screening plate. The screening plate is located directly below the feed hopper and can screen out raw material particles that meet the mesh size of the screen and make them fall into the water washing box through the inlet of the water washing box. Raw material particles larger than the mesh size of the screen slide into the collection box, and these raw material particles can be easily taken out through the cleaning door for secondary crushing.

[0014] Secondly, the raw material particles entering the washing box will first fall on the cleaning inclined plate. The nozzle draws water from the water tank through the water pipe to spray and clean the raw materials on the cleaning inclined plate. The washed sewage flows into the wastewater storage tank through the water flow holes on the surface of the cleaning inclined plate, and the washed raw materials slide out along the inclined angle of the cleaning inclined plate and enter the drying box for drying.

[0015] Finally, the dried raw materials enter the grinding basin through the connection between the drying box and the docking pipe. In the grinding basin, the raw materials fall on the grinding disc, and the grinding disc divides the inside of the grinding basin into two parts, the upper and lower parts. The grinding disc rotates through the rotating shaft connected by the telescopic rod to grind and crush the raw material particles on the grinding disc. The ground and crushed zinc powder falls into the lower part of the grinding basin through the through holes on the surface of the grinding disc, and the ground and crushed zinc powder is extracted by the extraction pump to complete the entire zinc powder production process.

[0016] As a further solution of the present invention, the intelligent control process of the preliminary crushing process of the raw materials is as follows:

[0017] Step 1: Install a high-resolution camera and sensor inside the feed hopper (2) to capture images and related data of the raw materials in real time, and use the sensor to collect information on the rotation speed, humidity, temperature, grinding pressure of the grinding roller (5) and particle size of the raw material particles in real time;

[0018] Step 2: Use the image data obtained by the camera and the raw material feature data collected by the sensor to establish and annotate the data set, classify different types of raw material particles, and use the region-based fast convolutional network algorithm Faster R-CNN to generate candidate boxes through the region recommendation network RPN. Its output is a series of candidate boxes and their corresponding scores. The loss function of RPN consists of two parts: classification loss and bounding box regression loss.

[0019]

[0020] Among them, L cls represents the classification loss, L reg Bounding box regression loss, N cls is the total number of samples, N regThe total number of bounding boxes, C represents the number of categories, and p i c represents the probability that the i-th sample in the true label belongs to the c-th category, Indicates the probability that the i-th sample belongs to the c-th category in the model prediction, t i The actual bounding box parameters, Predicted bounding box parameters L smooth The model’s prediction error for the bounding box, λ is the weight of the bounding box regression loss; x represents the difference between the predicted value and the true value, |x| is the absolute value of x;

[0021] The total loss function of RPN is the weighted sum of classification loss and bounding box regression loss:

[0022] L total =L cls +λL reg

[0023] Among them, L cls represents the classification loss, L reg Bounding box regression loss, λ is the weight of the bounding box regression loss;

[0024] Step 4: Use deep reinforcement learning to achieve intelligent control of the grinding roller speed and grinding force; take the speed, humidity, temperature, grinding pressure and particle size information data of the raw material particles as input and express them as state vectors s, s t represents the state vector at time step t; the action of adjusting the speed, humidity, temperature, grinding pressure and particle size information parameters of the raw material particles is represented by the action vector a, a t represents the action vector at time step t;

[0025] Choose multilayer perceptron as the basic structure of deep Q network;

[0026] Input layer: 5 neurons, corresponding to the dimension of the state space;

[0027] Hidden layer: 128 hidden layers are set, each hidden layer contains 8 neurons;

[0028] Output layer: 5 neurons, corresponding to the dimension of the action space;

[0029] The Q-value function of the deep reinforcement learning network is defined as:

[0030] Q(s t ,a t ;θ)

[0031] Among them, θ is the network parameter, a t represents the action vector at time step t, s trepresents the state vector at time step t;

[0032] Formulate the reward function R(s t ,a t ,s t+1 ), used to measure the agent in state s t Execute action a t Then transfer to state s t+1 The degree of pros and cons; using the experience replay mechanism to store the agent’s experience in the environment (s t ,a t ,r t ,s t+1 ), and randomly sample from them for training, r t Perform action a at time step t t Immediate rewards after

[0033] Step 5: Calculate the target Q value Q according to the Bellman equation target (s t ,a t ), used to update the network parameters θ;

[0034] Q target (s t ,a t )=r t +γmaxa t+1 Q(s t+1 ,a t+1 θ - )

[0035] Among them, r t Perform action a at time step t t The immediate reward obtained after, γ represents the discount factor, which is used to balance the importance of immediate rewards and future rewards, and θ - Represents the target network parameter, maxa t+1 Q(s t+1 ,a t+1 θ - ) means in state s t+1 Take the best action a t+1 The maximum Q value that can be obtained;

[0036] Define the mean square error as the loss function L(θ) to measure the difference between the Q value output by the network and the target Q value;

[0037]

[0038] in, represents the expected operation, D represents the experience set stored in the experience playback, Q target (s t ,a t) represents the target Q value, Q(s t ,a t ; θ) represents the current Q value, which is estimated by the deep reinforcement learning network, and θ is the network parameter;

[0039] Use stochastic gradient descent SGD to minimize the loss function and update the network parameters θ:

[0040] θ - =θ-α▽ θ L(θ)

[0041] Among them, θ - represents the updated parameter vector; θ represents the current parameter vector; α is the learning rate, which is used to control the step size of parameter update; It represents the gradient of the loss function L with respect to the parameter θ, that is, the rate of change of the loss function with respect to the parameter.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] 1. In the present invention, the raw material particles that meet the mesh size of the sieve are effectively screened out through the sub-screen plate, ensuring that the raw material particles entering the next process are uniform in size. At the same time, the raw material particles larger than the mesh size of the sieve can slide into the collection box, which is convenient for secondary crushing, ensuring full utilization of the raw materials and reducing waste.

[0044] 2. The design of the cleaning inclined plate enables the raw material particles to be evenly distributed on the inclined plate. The nozzle draws water from the water tank through the water pipe to fully spray and clean the raw materials, effectively removing impurities in the raw materials and improving the purity of the zinc powder. The cleaned sewage flows into the wastewater storage tank through the water flow holes on the surface of the inclined plate, which is convenient for subsequent treatment or reuse.

[0045] 3. The drying box uses efficient drying technology to quickly dry the cleaned raw material particles, reducing the moisture content and providing favorable conditions for subsequent grinding. The grinding disc rotates through the shaft connected by the telescopic rod to finely grind the raw material particles, ensuring the uniform particle size and stable quality of the final product.

[0046] 4. Through real-time monitoring, the system can quickly and accurately obtain various parameters of raw materials, providing a basis for subsequent data processing and intelligent control. This real-time data collection ensures accurate grasp of the raw material status during the production process, which helps to improve the consistency and stability of product quality. Through efficient and accurate classification of raw materials by intelligent algorithms, the system can adjust production parameters according to different types of raw materials, improve the efficiency and accuracy of screening, reduce human intervention and errors, and improve the level of automation; through the application of deep reinforcement learning algorithms, the system can dynamically adjust parameters such as the speed, humidity, temperature, and grinding pressure of the crushing roller to achieve optimal production control. This not only improves production efficiency and reduces energy consumption, but also ensures the consistency and stability of product quality. At the same time, the system can adaptively adjust production strategies to cope with different raw material characteristics and changes in the production environment, and improve the flexibility and intelligence level of production. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the overall structure of a zinc powder production and manufacturing device and a manufacturing process thereof according to the present invention;

[0048] Figure 2 It is a schematic structural diagram of a zinc powder production and manufacturing device and a manufacturing process thereof according to the present invention;

[0049] Figure 3 It is a structural schematic diagram of a crushing box in a zinc powder production and manufacturing device and a manufacturing process thereof according to the present invention;

[0050] Figure 4 It is a schematic diagram of the structure inside the water washing box in a zinc powder production and manufacturing device and a manufacturing process thereof according to the present invention;

[0051] Figure 5 It is a structural schematic diagram of the side surface of a water washing box in a zinc powder production and manufacturing device and a manufacturing process thereof according to the present invention;

[0052] Figure 6 It is a schematic structural diagram of the outer portion of the grinding disc in a zinc powder production and manufacturing device and a manufacturing process thereof according to the present invention;

[0053] Figure 7 The invention discloses a zinc powder production and manufacturing device and a schematic diagram of the structure of the interior of a grinding disc in its manufacturing process.

[0054] Figure 8 The invention discloses a flow chart of a control algorithm in a zinc powder production and manufacturing device and a manufacturing process thereof.

[0055] In the figure: 1. crushing box; 2. feeding hopper; 3. unloading funnel; 4. crushing box support rod; 5. crushing roller; 6. motor base; 7. driving motor; 8. washing box; 9. support rod of sub-screen plate; 10. sub-screen plate; 11. collecting box; 12. supporting bottom plate of collecting box; 13. cleaning door; 14. inlet of washing box; 15. wastewater storage tank; 16. cleaning inclined plate; 17. nozzle; 18. water pipeline; 19. water tank; 20. water replenishment hole; 21. drainage hole; 22. outlet of washing box; 23. drying box; 24. docking pipe; 25. grinding basin; 26. bottom plate; 27. grinding disc support seat; 28. telescopic rod; 29. ​​rotating shaft; 30. grinding disc; 31. grinding disc; 32. suction pump. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, which are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", and "set" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or it can be the internal connection of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The following is an explanation of an embodiment of the present invention based on the overall structure of the present invention.

[0058] Reference Figures 1 to 7In an embodiment of the present invention, a zinc powder production and manufacturing device includes: a feed hopper 2 is fixedly arranged above a crushing box 1, a discharge funnel 3 is fixedly arranged below the crushing box 1, a crushing box support rod 4 is fixedly arranged at the other end below the crushing box 1, a crushing roller 5 is rotatably connected inside the crushing box 1, a motor base 6 is fixedly arranged on the side of the crushing box 1, and a driving motor 7 is fixedly arranged above the motor base 6.

[0059] The above scheme is adopted: the crushing roller 5 rotatably connected inside the crushing box 1 crushes the raw material into smaller particles through high-speed rotation and friction between the raw material and the raw material. This design ensures that the raw material can be fully crushed, providing high-quality raw materials for subsequent processing steps.

[0060] A water washing box 8 is fixedly connected below the crushing box support rod 4, a sub-screen plate support rod 9 is fixedly arranged at one end above the water washing box 8, a sub-screen plate 10 is fixedly arranged on the side of the sub-screen plate support rod 9, a collecting box 11 is fixedly arranged on the other side of the water washing box 8, a collecting box support bottom plate 12 is fixedly arranged below the collecting box 11, the collecting box support bottom plate 12 is fixedly connected to the side of the water washing box 8, and a cleaning door 13 is arranged on one side of the collecting box 11.

[0061] The above scheme is adopted: the sub-screen plate 10 is fixed to the upper end of the washing box 8 through the sub-screen plate support rod 9, and is used to screen the raw materials that have been preliminarily crushed and cleaned. This design can ensure that the particle size of the raw materials meets the preset requirements, further improving the quality of the product.

[0062] A water washing box inlet 14 is provided above the water washing box 8, a wastewater storage tank 15 is fixedly arranged below the water washing box 8, a cleaning inclined plate 16 is fixedly arranged on the inside of the water washing box 8, a nozzle 17 is fixedly arranged on the other end of the inside of the water washing box 8, a water supply pipe 18 is fixedly connected to one end of the outside of the water washing box 8, the other end of the water supply pipe 18 is fixedly connected to a water tank 19, the water tank 19 is fixed to the other end of the outside of the water washing box 8, a water replenishment hole 20 is provided on one side of the water washing box 8, a drainage hole 21 is provided on one side of the wastewater storage tank 15, a water washing box outlet 22 is provided on the other side of the water washing box 8, a drying box 23 is fixedly connected to the other side of the water washing box 8, and a docking pipe 24 is fixedly connected to the other side of the drying box 23.

[0063] The above scheme is adopted: the setting of the cleaning inclined plate 16 allows the material to slide along the inclined plate during the cleaning process, increasing the contact area between the material and water and the cleaning time, thereby improving the cleaning effect. The nozzle 17 is connected to the water tank 19 through the water pipe 18 to provide a stable and sufficient water source for the cleaning process. This design ensures the efficient cleaning process and can adjust the water flow and cleaning time according to actual needs.

[0064] A grinding basin 25 is fixedly connected to the other side of the docking pipe 24, a bottom plate 26 is fixedly arranged below the grinding basin 25, a grinding disc support seat 27 is fixedly arranged on the side of the bottom plate 26, a telescopic rod 28 is fixedly arranged below the grinding disc support seat 27, a rotating shaft 29 is rotatably connected below the telescopic rod 28, a grinding disc 30 is rotatably connected below the rotating shaft 29, a grinding disc 31 is fixedly arranged inside the grinding basin 25, and a suction pump 32 is fixedly connected to the other side of the grinding basin 25.

[0065] The above scheme is adopted: the grinding basin 25 provides a closed and stable grinding environment, and the grinding disc 31 inside it works together with the grinding disc 30 below to finely grind the material. This double-disc design can ensure that the material is subjected to uniform and sufficient grinding force during the grinding process, thereby obtaining a more delicate and uniform grinding effect. The grinding disc support seat 27 is connected to the grinding disc 30 through a telescopic rod 28 and a rotating shaft 29. This design allows the grinding disc 25 to be fine-tuned during the grinding process to adapt to different materials or different grinding requirements.

[0066] The driving motor 7 drives the crushing roller 5 to rotate through the outer side of the crushing box 1. The crushing roller 5 can preliminarily crush the raw materials, and the crushed raw materials are discharged from the discharge funnel 3.

[0067] With the above solution, the design of the discharge hopper 3 enables the raw materials that have been preliminarily crushed to be smoothly discharged from the crushing box 1, and due to the inclination angle of the hopper, the raw materials can slide naturally, reducing the possibility of blockage.

[0068] The sub-screen plate 10 is located directly below the discharge funnel 3. The discharge funnel 3 has a certain inclination angle, so that the raw material particles larger than the mesh size of the sub-screen plate 10 can slide into the collecting box 11. The cleaning door 13 on the side of the collecting box 11 is convenient for taking out the separated raw materials, and then putting them into the crushing box 1 again for secondary crushing. The raw material particles that meet the size fall into the water washing box 8 from the sub-screen plate 10 and the water washing box inlet 14.

[0069] The above scheme is adopted: the collecting box 11 is located on the other side of the water washing box 8, and is used to collect raw material particles that do not meet the particle size requirements after screening. The collecting box support bottom plate 12 is fixedly connected to the side of the water washing box 8, providing stable support for the collecting box 11. Through the design of the cleaning door 13, the staff can easily open the cleaning door and take out the raw material particles in the collecting box for secondary processing or recycling, thereby realizing the effective utilization of resources.

[0070] The cleaning inclined plate 16 also has a certain angle, which is convenient for the raw materials to slide out of the water washing box 8 from the water washing box outlet 22 and enter the drying box 23. The nozzle 17 is aimed at the cleaning inclined plate 16 for spraying, and the raw materials will fall directly onto the cleaning inclined plate 16 from the water washing box inlet 14. Water flow holes are opened on the surface of the cleaning inclined plate 16, and the cleaned sewage will fall into the wastewater storage tank 15 through the water flow holes. The drying box 23 can dry the cleaned raw materials, and the dried raw materials are transmitted into the grinding basin 25 through the docking pipe 24.

[0071] The above scheme is adopted: the wastewater storage tank 15 below the washing box 8 is used to collect wastewater generated during the washing process. The setting of the wastewater storage tank 15 not only avoids the direct discharge of wastewater, but also facilitates the subsequent treatment or recovery of wastewater. The washed material enters the drying box 23 through the washing box outlet 22 for drying. The setting of the drying box 23 enables the material to quickly remove excess moisture, which provides convenience for subsequent processing or storage.

[0072] The grinding disc 31 divides the interior of the grinding basin 25 into two parts, the upper and lower parts. The raw materials will fall onto the grinding disc 31 through the docking pipe 24. The grinding disc 30 rotates through the rotating shaft 29 to grind the raw material particles on the grinding disc 31. The ground zinc powder falls into the lower part of the grinding basin 25 through the through holes on the surface of the grinding disc 31, and then the extraction pump 32 extracts the ground zinc powder.

[0073] The above scheme is adopted: the grinding disc 31 divides the interior of the grinding basin 25 into two parts, the upper and lower parts. This design effectively separates the preparation area before raw material grinding and the collection area after grinding, making the whole grinding process more orderly and efficient. The ground and crushed zinc powder falls into the lower part of the grinding basin 25 through the through holes on the surface of the grinding disc 31. This design enables the ground zinc powder to be quickly and concentratedly collected together, which is convenient for subsequent extraction and processing.

[0074] Adopt the above scheme: the intelligent control process of the preliminary crushing process of the raw materials, the specific information is as follows:

[0075] Step 1: Install a high-resolution camera and sensor inside the feed hopper (2) to capture images and related data of the raw materials in real time, and use the sensor to collect information on the rotation speed, humidity, temperature, grinding pressure of the grinding roller (5) and particle size of the raw material particles in real time;

[0076] Step 2: Use the image data obtained by the camera and the raw material feature data collected by the sensor to establish and annotate the data set, classify different types of raw material particles, and use the region-based fast convolutional network algorithm Faster R-CNN to generate candidate boxes through the region recommendation network RPN. Its output is a series of candidate boxes and their corresponding scores. The loss function of RPN consists of two parts: classification loss and bounding box regression loss.

[0077]

[0078] Among them, L cls represents the classification loss, L reg Bounding box regression loss, N cls is the total number of samples, N reg The total number of bounding boxes, C represents the number of categories, and p i c represents the probability that the i-th sample in the true label belongs to the c-th category, Indicates the probability that the i-th sample belongs to the c-th category in the model prediction, t i The actual bounding box parameters, Predicted bounding box parameters L smooth The model’s prediction error for the bounding box, λ is the weight of the bounding box regression loss; x represents the difference between the predicted value and the true value, |x| is the absolute value of x;

[0079] The total loss function of RPN is the weighted sum of classification loss and bounding box regression loss:

[0080] L total =L cls +λL reg

[0081] Among them, L cls represents the classification loss, L reg Bounding box regression loss, λ is the weight of the bounding box regression loss;

[0082] Step 4: Use deep reinforcement learning to achieve intelligent control of the grinding roller speed and grinding force; take the speed, humidity, temperature, grinding pressure and particle size information data of the raw material particles as input and express them as state vectors s, s t represents the state vector at time step t; the action of adjusting the speed, humidity, temperature, grinding pressure and particle size information parameters of the raw material particles is represented by the action vector a, a t represents the action vector at time step t;

[0083] Choose multilayer perceptron as the basic structure of deep Q network;

[0084] Input layer: 5 neurons, corresponding to the dimension of the state space;

[0085] Hidden layer: 128 hidden layers are set, each hidden layer contains 8 neurons;

[0086] Output layer: 5 neurons, corresponding to the dimension of the action space;

[0087] The Q-value function of the deep reinforcement learning network is defined as:

[0088] Q(s t ,a t ;θ)

[0089] Among them, θ is the network parameter, a t represents the action vector at time step t, s t represents the state vector at time step t;

[0090] Formulate the reward function R(s t ,a t ,s t+1 ), used to measure the agent in state s t Execute action a t Then transfer to state s t+1 The degree of pros and cons; using the experience replay mechanism to store the agent’s experience in the environment (s t ,a t ,r t ,s t+1 ), and randomly sample from them for training, r t Perform action a at time step t t Immediate rewards after

[0091] Step 5: Calculate the target Q value Q according to the Bellman equation target (s t ,a t ), used to update the network parameters θ;

[0092] Q target (s t ,a t )=r t +γmaxa t+1 Q(s t+1 ,a t+1 θ - )

[0093] Among them, r t Perform action a at time step t t The immediate reward obtained after, γ represents the discount factor, which is used to balance the importance of immediate rewards and future rewards, and θ - Represents the target network parameter, maxa t+1 Q(s t+1 ,a t+1 θ - ) means in state s t+1 Take the best action a t+1 The maximum Q value that can be obtained;

[0094] Define the mean square error as the loss function L(θ) to measure the difference between the Q value output by the network and the target Q value;

[0095]

[0096] in, represents the expected operation, D represents the experience set stored in the experience playback, Q target (s t ,a t ) represents the target Q value, Q(s t ,a t ; θ) represents the current Q value, which is estimated by the deep reinforcement learning network, and θ is the network parameter;

[0097] Use stochastic gradient descent SGD to minimize the loss function and update the network parameters θ:

[0098] θ - =θ-α▽ θ L(θ)

[0099] Among them, θ - represents the updated parameter vector; θ represents the current parameter vector; α is the learning rate, which is used to control the step size of parameter update; It represents the gradient of the loss function L with respect to the parameter θ, that is, the rate of change of the loss function with respect to the parameter.

[0100] Numerical case analysis:

[0101] The speed range of the crushing roller in the zinc powder production device is 100RPM to 1000RPM, the humidity range is 10% to 90%, the temperature range is 20℃ to 80℃, the grinding pressure range is 100kPa to 500kPa, and the particle diameter is 1mm.

[0102] Install high-resolution cameras and sensors to capture images and relevant data of raw materials in real time, rotation speed = 500RPM, humidity = 50%, temperature = 30℃, grinding pressure = 300kPa, raw material particle size 10mm;

[0103] According to the control method of this patent, the action vector generated by the deep reinforcement learning network can be: increase the speed by 70RPM, reduce the humidity by 15%, increase the temperature by 20°C, increase the grinding pressure by 100kPa,

[0104] The action vectors generated by parameter optimization are: increase the rotation speed by 66RPM, reduce the humidity by 12%, increase the temperature by 21°C, and increase the grinding pressure by 110kPa; ultimately achieving the production target of a particle diameter of 1mm.

[0105] What is described above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A zinc powder production and manufacturing device, characterized in that: include: A crushing box (1), wherein a feed hopper (2) is fixedly arranged above the crushing box (1), a discharge hopper (3) is fixedly arranged below the crushing box (1), a crushing box support rod (4) is fixedly arranged at the other end below the crushing box (1), a crushing roller (5) is rotatably connected inside the crushing box (1), a motor base (6) is fixedly arranged on the side of the crushing box (1), and a driving motor (7) is fixedly arranged above the motor base (6).

2. A zinc powder production and manufacturing device according to claim 1, characterized in that: A water washing box (8) is fixedly connected below the crushing box support rod (4); a sub-screen plate support rod (9) is fixedly arranged at one end above the water washing box (8); a sub-screen plate (10) is fixedly arranged on the side of the sub-screen plate support rod (9); a collecting box (11) is fixedly arranged on the other side of the water washing box (8); a collecting box support bottom plate (12) is fixedly arranged below the collecting box (11); the collecting box support bottom plate (12) is fixedly connected to the side of the water washing box (8); and a cleaning door (13) is arranged on one side of the collecting box (11).

3. A zinc powder production and manufacturing device according to claim 2, characterized in that: A water washing box inlet (14) is provided above the water washing box (8), a waste water storage tank (15) is fixedly provided below the water washing box (8), a cleaning inclined plate (16) is fixedly provided inside the water washing box (8), a nozzle (17) is fixedly provided at the other end of the inside of the water washing box (8), a water delivery pipe (18) is fixedly connected to one end of the outside of the water washing box (8), the other end of the water delivery pipe (18) is fixedly connected to a water tank (19), the water tank (19) is fixed to the other end of the outside of the water washing box (8), a water replenishment hole (20) is provided on one side of the water washing box (8), a drainage hole (21) is provided on one side of the waste water storage tank (15), a water washing box outlet (22) is provided on the other side of the water washing box (8), a drying box (23) is fixedly connected to the other side of the drying box (23), and a docking pipe (24) is fixedly connected to the other side of the drying box (23).

4. A zinc powder production and manufacturing device according to claim 3, characterized in that: A grinding basin (25) is fixedly connected to the other side of the docking pipe (24), a bottom plate (26) is fixedly arranged below the grinding basin (25), a grinding disc support seat (27) is fixedly arranged on the side of the bottom plate (26), a telescopic rod (28) is fixedly arranged below the grinding disc support seat (27), a rotating shaft (29) is rotatably connected below the telescopic rod (28), a grinding disc (30) is rotatably connected below the rotating shaft (29), a grinding disc (31) is fixedly arranged inside the grinding basin (25), and a material extraction pump (32) is fixedly connected to the other side of the grinding basin (25).

5. A zinc powder production and manufacturing device according to claim 1, characterized in that: The driving motor (7) drives the pulverizing roller (5) to rotate through the outside of the pulverizing box (1), and the pulverizing roller (5) can preliminarily pulverize the raw materials, and the pulverized raw materials are discharged from the discharge hopper (3).

6. A zinc powder production and manufacturing device according to claim 2, characterized in that: The sub-screen plate (10) is located directly below the discharge funnel (3). The discharge funnel (3) has a certain inclination angle, so that raw material particles larger than the mesh size of the sub-screen plate (10) can slide into the collecting box (11). The cleaning door (13) on the side of the collecting box (11) facilitates the removal of separated raw materials, which are then put into the crushing box (1) again for secondary crushing. Raw material particles that meet the size fall from the sub-screen plate (10) through the water washing box inlet (14) into the water washing box (8).

7. A zinc powder production and manufacturing device according to claim 3, characterized in that: The cleaning inclined plate (16) also has a certain angle, which is convenient for the raw materials to slide out of the water washing box (8) from the water washing box outlet (22) and enter the drying box (23). The nozzle (17) is aimed at the cleaning inclined plate (16) for spraying, and the raw materials will directly fall on the cleaning inclined plate (16) from the water washing box inlet (14). The surface of the cleaning inclined plate (16) is provided with water flow holes, and the cleaned sewage will fall into the wastewater storage tank (15) through the water flow holes. The drying box (23) can dry the cleaned raw materials, and the dried raw materials are transmitted into the grinding basin (25) through the docking pipe (24).

8. A zinc powder production and manufacturing device according to claim 4, characterized in that: The grinding disc (31) divides the interior of the grinding basin (25) into two parts, the upper and lower parts. The raw materials will fall onto the grinding disc (31) through the docking pipe (24). The grinding disc (30) rotates through the rotating shaft (29) to grind and crush the raw material particles on the grinding disc (31). The ground and crushed zinc powder falls into the lower part of the grinding basin (25) through the through holes on the surface of the grinding disc (31), and then the extraction pump (32) extracts the ground and crushed zinc powder.

9. A zinc powder production process, characterized in that: The following steps are involved: Step 1: First, the raw material enters the crushing box (1) through the feed hopper (2), and the driving motor (7) drives the crushing roller (5) to rotate inside the crushing box (1) through an external connection to perform preliminary crushing on the raw material. The crushed raw material is discharged through a feed hopper (3) with a certain inclination angle and falls onto a screening plate (10). The screening plate (10) is located directly below the feed hopper (3) and can screen out raw material particles that meet the mesh size of the screen and make them fall into the water washing box (8) through the water washing box inlet (14). Raw material particles larger than the mesh size of the screen slide into the collecting box (11). These raw material particles can be conveniently taken out through the cleaning door (13) for secondary crushing. Step 2: Secondly, the raw material particles entering the washing box (8) will first fall on the cleaning inclined plate (16), and the nozzle (17) will draw water from the water tank (19) through the water pipe (18) to spray and clean the raw materials on the cleaning inclined plate (16). The cleaned sewage will flow into the wastewater storage tank (15) through the water flow holes on the surface of the cleaning inclined plate (16), and the cleaned raw materials will slide out along the inclination angle of the cleaning inclined plate (16) and enter the drying box (23) for drying. Step 3: Finally, the dried raw materials enter the grinding basin (25) through the connection between the drying box (23) and the docking pipe (24). In the grinding basin (25), the raw materials fall on the grinding disc (31). The grinding disc (31) divides the inside of the grinding basin (25) into two parts, an upper part and an lower part. The grinding disc (30) rotates through the rotating shaft (29) connected by the telescopic rod (28) to grind and crush the raw material particles on the grinding disc (31). The ground and crushed zinc powder falls into the lower part of the grinding basin (25) through the through hole on the surface of the grinding disc (31). The ground and crushed zinc powder is extracted by the extraction pump (32), completing the entire zinc powder production process.

10. A zinc powder production process according to claim 9, characterized in that: In step 1, the intelligent control process of the initial crushing of the raw materials is as follows: Step 1: Install a high-resolution camera and sensor inside the feed hopper (2) to capture images and related data of the raw materials in real time, and use the sensor to collect information on the rotation speed, humidity, temperature, grinding pressure of the grinding roller (5) and particle size of the raw material particles in real time; Step 2: Use the image data obtained by the camera and the raw material feature data collected by the sensor to establish and annotate the data set, classify different types of raw material particles, and use the region-based fast convolutional network algorithm Faster R-CNN to generate candidate boxes through the region recommendation network RPN. Its output is a series of candidate boxes and their corresponding scores. The loss function of RPN consists of two parts: classification loss and bounding box regression loss. Among them, L cls represents the classification loss, L reg Bounding box regression loss, N cls is the total number of samples, N reg The total number of bounding boxes, C represents the number of categories, and p i c represents the probability that the i-th sample in the true label belongs to the c-th category, Indicates the probability that the i-th sample belongs to the c-th category in the model prediction, t i The actual bounding box parameters, Predicted bounding box parameters L smooth The model’s prediction error for the bounding box, λ is the weight of the bounding box regression loss; x represents the difference between the predicted value and the true value, |x| is the absolute value of x; The total loss function of RPN is the weighted sum of classification loss and bounding box regression loss: THE total =L cls +λL reg Among them, L cls represents the classification loss, L reg Bounding box regression loss, λ is the weight of the bounding box regression loss; Step 4: Use deep reinforcement learning to achieve intelligent control of the grinding roller speed and grinding force; take the speed, humidity, temperature, grinding pressure and particle size information data of the raw material particles as input and express them as state vectors s, s t represents the state vector at time step t; the action of adjusting the speed, humidity, temperature, grinding pressure and particle size information parameters of the raw material particles is represented by the action vector a, a t represents the action vector at time step t; Choose multilayer perceptron as the basic structure of deep Q network; Input layer: 5 neurons, corresponding to the dimension of the state space; Hidden layer: 128 hidden layers are set, each hidden layer contains 8 neurons; Output layer: 5 neurons, corresponding to the dimension of the action space; The Q-value function of the deep reinforcement learning network is defined as: Q(s t ,a t (i) Among them, θ is the network parameter, a t represents the action vector at time step t, s t represents the state vector at time step t; Formulate the reward function R(s t ,a t ,s t+1 ), used to measure the agent in state s t Execute action a t Then transfer to state s t+1 The degree of pros and cons; using the experience replay mechanism to store the agent’s experience in the environment (s t ,a t ,r t ,s t+1 ), and randomly sample from them for training, r t Perform action a at time step t t Immediate rewards after Step 5: Calculate the target Q value Q according to the Bellman equation target (s t ,a t ), used to update the network parameters θ; Q target (s t ,a t )=r t +γmaxa t+1 Q(s t+1 ,a t+1 ;θ - ) Among them, r t Perform action a at time step t t The immediate reward obtained after, γ represents the discount factor, which is used to balance the importance of immediate rewards and future rewards, and θ - Represents the target network parameter, maxa t+1 Q(s t+1 ,a t+1 θ - ) means in state s t+1 Take the best action a t+1 The maximum Q value that can be obtained; Define the mean square error as the loss function L(θ) to measure the difference between the Q value output by the network and the target Q value; in, represents the expected operation, D represents the experience set stored in the experience playback, Q target (s t ,a t ) represents the target Q value, Q(s t ,a t ; θ) represents the current Q value, which is estimated by the deep reinforcement learning network, and θ is the network parameter; Use stochastic gradient descent SGD to minimize the loss function and update the network parameters θ: i - =θ-α▽ θ L(θ) Among them, θ - represents the updated parameter vector; θ represents the current parameter vector; α is the learning rate, which is used to control the step size of parameter update; ▽ θ L(θ) represents the gradient of the loss function L with respect to the parameter θ, that is, the rate of change of the loss function with respect to the parameter.

Images