Grinding disc for high-precision roller production and control method thereof

By designing an automated grinding disc system, the automatic feeding of grinding discs is achieved using motor drives and intelligent algorithms, which solves the problems of time-consuming and labor-intensive and safety hazards in the prior art, and improves the operational safety and system availability.

CN119077610AInactive Publication Date: 2024-12-06万徽精密科技(扬州)有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411275613.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-12
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing milling pans are time-consuming and labor-intensive during the discharge process, and the operators need to operate manually. The large weight of the box has safety risks, which cannot meet the needs of small-scale enterprises.

Method used

A grinding disc for high-precision roller production and its control method are designed, including support seat, grinding disc frame, guide rail and discharge plate. The transmission rod and discharge plate are driven by the motor to drive the rotary rod and discharge plate for automatic discharge, combined with intelligence The algorithm monitors and adjusts the speed and angle in real time to ensure the smooth and efficient discharge process.

Benefits of technology

The automatic discharge of grinding discs is realized, which reduces the labor intensity of operators, improves safety, avoids the weight problem during high-precision roller handling, and dynamically adjusts the transmission ratio through intelligent algorithms to ensure that the system remains under different working conditions. Optimal working condition, reducing failure rate and maintenance requirements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119077610A_ABST
    Figure CN119077610A_ABST
Patent Text Reader

Abstract

The invention discloses a high-precision roller production grinding disc and a control method thereof.The high-precision roller production grinding disc comprises a supporting seat, the top of the supporting seat is fixedly connected with a grinding disc frame, the interior of the grinding disc frame is slidably connected with a grinding disc body, the right side of the grinding disc body is fixedly connected with a supporting plate, and the top of the supporting plate is provided with a guide rail; a guide rail is fixedly connected to the left side of the grinding disc body, a discharging plate is fixedly connected to the left side of the guide rail, the grinding disc body can feed high-precision rollers into the guide rail through the discharging plate to conduct material guiding work, and the guide rail and the discharging plate are connected with a supporting plate through an adjusting structure. The discharging plate and the guide rail carry out automatic discharging work, so that the effect of facilitating discharging is achieved, meanwhile, when the high-precision rollers are automatically discharged, the safety of operators is greatly improved, and the situation that the high-precision rollers are heavy in carrying size, and high safety cannot be provided for the operators is prevented.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of high-precision rollers, and in particular to a grinding disc for producing high-precision rollers and a control method thereof. Background Art

[0002] High-precision rollers are rollers that have been precision-machined and strictly quality-controlled to achieve a higher level of precision. According to their shapes and uses, high-precision rollers can be divided into many types, such as cylindrical rollers, tapered rollers, cross rollers, etc. These rollers are widely used in high-precision mechanical equipment such as machine tools, robots, semiconductor manufacturing equipment, LCD manufacturing equipment, and printing presses.

[0003] For example, the patent number disclosed in the China Patent Network is: (CN115816194A). The present invention provides a high-precision cylindrical roller double-plane processing method based on a hard ceramic grinding disc. The method is implemented on a double-plane polishing device. During grinding: water is added to the processing area for lubrication and cooling. The lower grinding disc is connected to an external motor and rotated by the motor drive. The retaining frame is placed on the lower grinding disc, and the cylindrical roller is placed in the hole groove of the retaining frame, and the cylindrical roller is always kept in contact with the lower grinding disc; when the lower grinding disc rotates, the friction between it and the cylindrical roller drives the cylindrical roller to rotate in the hole groove. At the same time, under the action of the cylindrical roller and the lower grinding disc, the retaining frame rotates to achieve the effect of grinding the surface of the cylindrical roller; the upper grinding disc is placed on the retaining frame, and there is an eccentricity between the rotation centers of the upper grinding disc and the lower grinding disc. The technical solution in the present invention can improve the processing accuracy and consistency of the cylindrical roller, reduce processing costs, and reduce environmental pollution.

[0004] However, most grinding discs currently available on the market are generally large in size, which makes the discharging work of the grinding disc time-consuming and laborious, requiring operators to manually discharge the materials. The large size of the grinding disc brings inconvenience to the operators for manual discharging, and cannot bring convenience to users.

[0005] Secondly, when discharging materials, the grinding disc requires the operator to climb up to pick up the materials, send the high-precision rollers into the box to hold them, and lift them out after they are full. However, the heavy weight of the box may easily lead to safety hazards, and it cannot meet the needs of some small-scale enterprises. Summary of the invention

[0006] The purpose of the present invention is to solve the problem that the discharging operation of the grinding disc is time-consuming and labor-intensive and requires operators to manually discharge the materials, and to propose a high-precision grinding disc for roller production and a control method thereof.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A grinding disc for high-precision roller production and a control method thereof include a support seat, a grinding disc frame fixedly connected to the top of the support seat, a grinding disc body slidably connected to the inside of the grinding disc frame, a support plate fixedly connected to the right side of the grinding disc body, a guide rail provided on the top of the support plate, a discharge plate fixedly connected to the left side of the guide rail, the grinding disc body can send the high-precision roller into the guide rail through the discharge plate for guiding work, the guide rail and the discharge plate are connected to the support plate through an adjusting structure, the discharge plate can contact with the top of the grinding disc body through the adjusting structure, an auxiliary discharge structure is provided inside the adjusting structure, and the auxiliary discharge structure can assist the guide rail in discharging materials when the guide rail is discharging materials.

[0009] As a preferred technical solution of the present application, the adjustment structure includes a support frame, which is fixedly connected to the top of the support plate, the front of the support frame is fixedly connected to a motor, the output end of the motor is fixedly connected to a rotating rod, the back of the rotating rod is movably connected to the rear side of the inner wall of the support frame through a bearing, the surface of the rotating rod is fixedly connected to a transmission rod, the opposite side of the transmission rod is fixedly connected to the surface of the guide rail, the back of the guide rail is fixedly connected to a transmission structure, and the transmission structure can be transmitted by swinging the guide rail and fitting with the surface of the grinding disc.

[0010] As a preferred technical solution of the present application, the transmission structure includes a connecting frame, a moving wheel is fixedly connected to the bottom of the connecting frame, a transmission column is fixedly connected to the right side of the moving wheel, the transmission column can follow the rotation when the moving wheel rotates, a first bevel gear is fixedly connected to the right side of the transmission column, a second bevel gear is meshed with the top of the first bevel gear, and a coupling is fixedly connected to the top of the second bevel gear.

[0011] As a preferred technical solution of the present application, the surface of the second bevel gear is movably connected to a support block via a bearing, and the left side of the support block is fixedly connected to the right side of the moving wheel.

[0012] As a preferred technical solution of the present application, the auxiliary discharging structure includes a connecting block, a side of the connecting block close to the grinding disc body is movably connected to a first transmission disc through a rotating shaft, the surface of the first transmission disc is slidably connected to a first belt, the inside of the first belt is slidably connected to a second transmission disc, the surface of the second transmission disc is movably connected to the surface of the connecting block through a rotating shaft, a connecting rod is fixedly connected to the right side of the first transmission disc, a third transmission disc is fixedly connected to the right side of the connecting rod, the surface of the third transmission disc is slidably connected to a second belt, the inside of the second belt is slidably connected to a fourth transmission disc, the left side of the fourth transmission disc is fixedly connected to the right side of the coupling, the outer side of the fourth transmission disc is movably connected to an outer shell through a bearing, and the front side of the outer shell is fixedly connected to the back side of the guide rail.

[0013] As a preferred technical solution of the present application, an opening is provided at the top of the guide rail, the width of the opening is greater than the width of the first belt, and the opening is used in conjunction with the first belt.

[0014] As a preferred technical solution of the present application, a guide plate is fixedly connected to the top of the discharge plate, and the number of the guide plates is several and the guide plates are evenly arranged. A connecting block is fixedly connected to the right side of the discharge plate, and the connecting block penetrates into the interior of the guide rail and is fixed by fixing bolts on the surface of the guide rail.

[0015] As the preferred technical solution of this application, the following steps are included:

[0016] S1: Start the motor, the motor drives the rotating rod to rotate through the output end, the rotating rod drives the transmission rod fixed on the surface to rotate when rotating, the transmission rod drives the guide rail and the discharge plate to rotate, so that the discharge plate enters the interior of the grinding disc frame and contacts the top of the grinding disc body, when the grinding disc body needs to collect materials, the grinding disc body rotates and sends the high-precision roller to the top of the discharge plate and guides the materials through the discharge plate;

[0017] In S1, an intelligent algorithm that combines graph neural architecture search, variational autoencoders under multiple divergence theory and generalized nonlinear Hamilton-Jacobi equations is introduced to monitor the rotation speed of the grinding disc, the position of the discharge plate and the movement status of the roller in real time, and dynamically adjust the rotation speed of the rotating rod and the angle of the guide rail according to real-time data to ensure the smoothness and efficiency of the discharge process. In addition, the transmission system adaptively adjusts the transmission ratio to ensure that the grinding disc and the discharge plate can maintain the best working condition under any load conditions. At the same time, it can also predict faults and issue maintenance reminders in advance, thereby reducing downtime and improving the overall availability of the equipment. While introducing this intelligent algorithm, the improved neural network optimization algorithm is used to ensure the precise control and adaptability of the system in a highly complex environment. The specific process is:

[0018] Step 1: Data collection and preprocessing

[0019] Data collection: After the motor is started, the motor drives the rotating rod to rotate through the output end; at this time, multiple sensors in the system begin to collect key parameter data closely related to the system operation in real time; the purpose of collecting this data is to monitor the operating status of the system, ensure the synchronization and stability of each component, and thus ensure the efficiency and accuracy of the entire grinding process; the following are the main collected data and their detailed description:

[0020] Rotating rod speed (R) sensor

[0021] Use photoelectric sensors and Hall effect sensors, installed near the motor output end, close to the bearing of the rotating rod; the sensor is close to the rotating rod to reduce the delay or error caused by the mechanical transmission part; the position of the sensor should avoid interference from the mechanical structure to ensure that the sensor can directly sense the rotation speed of the rotating rod;

[0022] The sensor head should be aligned with the marking point of the rotating rod; a rotary encoder is connected to the rotating rod, and the number of revolutions per minute of the rotating rod is calculated by measuring the number of pulses of the encoder;

[0023] Rotary rod speed (R): The rotational speed of the rotary rod is expressed in revolutions per minute. The speed is an important indicator to describe the power transmission efficiency of the system. Accurately collecting and monitoring the speed change is crucial for the stable operation of the system. Unstable speed changes will cause the guide rail and the discharge plate to work in an uncoordinated manner, thus affecting the working efficiency of the grinding disc and the quality of the final product.

[0024] During operation, the sensor records speed changes at a high frequency; capturing small changes in speed allows the system to make quick adjustments when necessary, ensuring that all components work together;

[0025] Guide rail position (P) sensor

[0026] Use a linear position sensor or laser distance sensor installed on the fixed or moving end of the guide rail to directly sense the linear displacement of the guide rail. The sensor should be installed within the travel range of the guide rail to ensure that the sensor can accurately track the position changes of the guide rail during the entire operation process. The position of the sensor should ensure that it always maintains effective sensing throughout the entire movement of the guide rail.

[0027] The sensor is installed parallel to the guide rail to ensure that the sensor head can cover the moving path of the guide rail; for long-distance guide rails, a grating ruler high-precision sensor is used to ensure that every millimeter of displacement can be accurately recorded;

[0028] Position of the guide rail (P): The position of the guide rail is measured in millimeters and describes the linear displacement of the guide rail relative to the initial position. The position of the guide rail directly affects the material conveying process between the grinding disc and the discharge plate. Therefore, accurate control of the position of the guide rail is essential to ensure smooth material conveying and the final discharge quality. Any slight error in position will lead to deviations in the discharge process, which in turn affects the overall working effect of the grinding disc.

[0029] The position of the guide rail is continuously collected by the sensor to obtain displacement data, recording every tiny movement of the guide rail. Through this data, the system ensures that the position adjustment of the guide rail conforms to the predetermined movement trajectory, avoiding system failure and reduced efficiency caused by position errors.

[0030] Angle sensor of the discharge plate (A)

[0031] Use an angle encoder installed on the rotating shaft or hinge of the discharge plate to directly measure the angular change of the discharge plate relative to the fixed shaft; the sensor needs to be closely connected to the key rotating part of the discharge plate to avoid deviation during the rotation process; the sensor position should ensure that it can provide accurate data within the entire angle adjustment range;

[0032] The angle encoder should be installed coaxially with the rotating axis of the discharge plate to ensure that the sensor can directly detect the rotation angle of the discharge plate; during installation, care should be taken to avoid the influence of mechanical friction and vibration on the accuracy of the sensor;

[0033] Angle of the discharge plate (A): The angle of the discharge plate is measured in degrees, which indicates the angle change of the discharge plate relative to the grinding disc. The change of angle directly affects the stability and uniformity of the discharge. Imprecise angle control will cause material accumulation or uneven distribution, affecting the discharge efficiency of the grinding disc and the quality of the final product. Therefore, accurate acquisition and control of the angle change of the discharge plate is the key to ensuring the discharge quality.

[0034] During operation, sensors accurately record these angle changes, ensuring that each adjustment is accurately reflected in the system operation and adapts to the coordinated work of the grinding disc and guide rails;

[0035] Data preprocessing: The collected raw data is high-dimensional, which not only contains a lot of detailed information, but also contains noise. The noise comes from sensor errors, electromagnetic interference, and mechanical vibration. In order to ensure the accuracy of data analysis, these high-dimensional data must be preprocessed. The purpose of data preprocessing is to simplify the data into more core features through dimensionality reduction and noise elimination, so as to facilitate subsequent analysis and optimization.

[0036] Dimensionality reduction and noise elimination: In order to effectively process high-dimensional data and retain its core features, a variational autoencoder based on multiple divergence theory is used for dimensionality reduction and noise elimination. The variational autoencoder can not only reduce the dimension of the data, but also remove random noise in the data while reducing the dimension, making subsequent analysis more accurate and efficient.

[0037] The variational autoencoder under the multi-divergence theory is a generative model that generates new data similar to the input data by learning the distribution of the data and optimizes the model by maximizing the following evidence lower bound:

[0038]

[0039] Among them, ELBO represents the new data formed by the generative model, x represents the observed data, which corresponds to the sensor data of speed (R), position (P) and angle (A); z is the latent variable, which represents the data characteristics after dimensionality reduction; this latent variable space is simpler than the original data space, but retains the main characteristics of the data; p(x|z) is the generative model, which describes the mapping relationship from the latent space to the observed data; this process is similar to the reverse operation of reducing high-dimensional data to low-dimensional data; p(z): prior distribution, q(z|x) is the approximate posterior distribution, which is used to map the observed data to the latent space; this distribution needs to be close to the true posterior distribution to ensure the rationality of dimensionality reduction; The expected value operator represents the expectation of the posterior distribution q(z|x) of the latent variable z; this means taking a weighted average of all the latent variables z; log p(x|z): log likelihood, which represents the probability of generating the observed data x given the latent variable z; this part represents the model's ability to reconstruct the data, that is, how well the model can restore the original data given the latent variable; D KL : KL divergence, which is used to measure the difference between two distributions; here it measures the difference between the approximate posterior distribution q(z|x) and the prior distribution p(z);

[0040] In the present invention, the goal of the variational autoencoder under the multiple divergence theory is to reduce the dimension of the original data through optimization, retain the key features in the data, the speed mode, the guide rail displacement trend and the angle change of the discharge plate, and filter out the unimportant noise information; simplify the data structure and improve the efficiency of subsequent analysis and control;

[0041] After data dimension reduction and noise elimination, the data needs to be sorted and standardized to ensure its consistency and effectiveness in subsequent analysis and modeling;

[0042] Data standardization: The data after dimensionality reduction needs to be standardized to make it have a uniform scale; eliminate the scale differences between different data dimensions to avoid the data of a certain dimension having too much impact on model training due to its large dimension;

[0043] Standardization adjusts the data to a mean of 0 and a variance of 1. This process helps improve the efficiency and stability of model training and ensures that the model does not favor a certain feature during learning.

[0044] Standardization formula:

[0045]

[0046] Among them, x is the original data value, μ is the mean of the data, σ is the standard deviation of the data; x′ is the standardized data;

[0047] The standardized data x′ will be converted to a standard normal distribution with a mean of 0 and a variance of 1, ensuring that each feature is at the same level; thus, the model has the same weight when processing this data;

[0048] Data sorting: The data after dimensionality reduction and standardization will be organized into a format suitable for subsequent analysis and modeling; each data point will be further integrated to form a feature vector for the input model; through sorting, the system ensures that all input data structures are consistent, providing a solid foundation for subsequent modeling and optimization; the sorted data is represented in matrix form, with each row representing the system state at a moment and each column representing a standardized feature;

[0049] Step 2: Graph Neural Architecture Search

[0050] In the system of the present invention, the combination of graph neural network and graph neural architecture search provides the system with powerful modeling and optimization capabilities; the graph neural network models the data after dimensionality reduction and captures the nonlinear dependencies between key parameters, while the graph neural architecture search searches and optimizes the structure of the graph neural network in an automated manner to ensure that the system achieves optimal performance; this process provides reliable technical support for the efficiency and stability of the entire system in actual operation;

[0051] Graph Neural Network Modeling

[0052] The reduced-dimensional data is used to build the graph neural network model: In the data preprocessing stage, the original high-dimensional data has been reduced in dimension through the variational autoencoder, which retains the key features of speed, position, and angle. These features are used to build the graph neural network model, which effectively captures the complex dependencies between the key parameters of the system and generates a comprehensive state representation.

[0053] Node representation: In the graph neural network, the speed, position and angle parameters are represented as nodes; there are complex nonlinear relationships between these nodes; changes in speed will affect the position of the guide rail, and the position of the guide rail will affect the angle of the discharge plate; the task of the graph neural network is to capture and reflect these dependencies by modeling these nodes and their relationships;

[0054] Adjacency matrix: The relationship between nodes is represented by the adjacency matrix; the adjacency matrix is ​​a second-order tensor matrix, in which each element indicates whether there is a connection between two nodes and the strength of the connection;

[0055]

[0056] ω RP It represents the connection weight between speed (R) and position (P), reflecting the influence of speed on the guide rail position;

[0057] ω RA It represents the connection weight between the rotation speed (R) and the angle (A), reflecting the influence of the rotation speed on the angle of the discharge plate;

[0058] ω AP It represents the connection weight between position (P) and angle (A), reflecting the influence of the guide rail position on the angle of the discharge plate;

[0059] These weights reflect the strength of the dependencies between different parameters. The larger the value, the closer the relationship between the two. Through these relationship matrices, graph neural networks capture the complex dependencies between various parameters, thereby providing a global state representation for the system.

[0060] Graph neural networks use a message passing mechanism to propagate and aggregate the information of each node through the adjacency matrix to generate a comprehensive state representation; this process includes the following steps:

[0061] Message passing: Each node transmits information to its neighboring nodes according to the weight in the adjacency matrix; this information includes the node's own characteristics, speed value, position value, angle value and information received from neighboring nodes; this process ensures that each node can be understood in context, not just an isolated value;

[0062] Node information transmission: Each node transmits its own information to its adjacent nodes through the weight of the adjacency matrix. This information transmission process enables each node to generate a more accurate state representation based not only on its own state, but also on the relationship with other nodes.

[0063] Information aggregation: Each node aggregates the information received from neighboring nodes with its own information and updates its state representation; this process is completed through a learnable function;

[0064] Aggregate formula:

[0065]

[0066] Among them, i and j: different nodes, represents the hidden state of node i in the kth layer; this hidden state is the updated state of node i after aggregating information, which contains the comprehensive information from neighboring nodes; represents the hidden state of node j in layer k-1, represents the neighbor set of node i, that is, other nodes directly connected to node i; d i and d jare the degrees of nodes i and j, respectively, indicating the number of connections of each node; the inverse of the degree is used to normalize message passing to prevent some nodes from causing bias in information aggregation due to large degrees; W(k) is the learnable weight matrix of the kth layer, which is used for information transmission between nodes; by learning weights, the model decides which information is more important; It is a learnable weight matrix used to update the node's own information; σ is an activation function, using Sigmoid to introduce nonlinearity and enhance the model's expressiveness;

[0067] State update: Through multiple iterations, the graph neural network gradually updates the state of each node, and finally forms a comprehensive global state representation; this comprehensive representation can capture the complex dependencies between various parameters in the system and provide a comprehensive understanding of the state of the entire system;

[0068] Architecture Search and Optimization

[0069] Graph neural architecture search is an automated search algorithm designed to optimize the structure of graph neural networks to ensure optimal performance. Unlike traditional neural network design, graph neural architecture search automatically explores different network architecture combinations to find the structure that best suits a specific task.

[0070] Search space definition: Graph neural architecture search first defines a search space that contains all graph neural network architecture combinations; the dimensions of the search space include the number of network layers, the number of nodes in each layer, the choice of activation function, and the initial weight of the adjacency matrix;

[0071] Evaluation of candidate architectures: For each candidate architecture, Graph Neural Architecture Search uses an evaluation function to measure its performance. The evaluation function is based on the mean square error loss function. The purpose of the evaluation is to quantify the pros and cons of each candidate architecture and then select the optimal architecture.

[0072] Loss function:

[0073]

[0074] in, represents the value of the loss function, which measures the gap between the model prediction and the actual target; f(x i ; θ) represents the output of the graph neural network model, θ is the model parameter, x i is the input data; y i is the actual target value, indicating the true state of each node in the system; Loss(f(x i ;θ),y i ) is the error between the model prediction value and the true value, which is used to evaluate the accuracy of the model, and N is the number of models of the output of the graph neural network model;

[0075] Optimization and model selection

[0076] Through continuous search and evaluation, graph neural architecture search can find a graph neural network architecture that performs best on the evaluation function;

[0077] Optimization process: The optimization process of NAS uses the gradient descent method to reduce the value of the loss function by gradually adjusting the parameters of the candidate architecture; this process is similar to finding the global optimal solution in a complex search space;

[0078] Gradient descent method: The gradient descent method gradually adjusts the parameters to reduce the loss value by calculating the derivative of the loss function with respect to the model parameters; the formula is:

[0079]

[0080] Among them, θ is the parameter of the model; η is the learning rate, which controls the step size of parameter update; is the gradient of the loss function with respect to the parameters;

[0081] Selection of the final architecture: When the graph neural architecture search completes the search, the system will select the architecture with the smallest loss function as the final graph neural network model; this model can not only accurately capture the complex relationship between various parameters in the system, but also provide efficient prediction and control capabilities in practical applications;

[0082] Step 3: Nonlinear dynamic modeling and optimization

[0083] In the system of the present invention, nonlinear dynamic modeling and optimization are key steps to achieve precise control and stable operation of the system; by introducing the generalized nonlinear Hamilton-Jacobi equation, the system captures the complex nonlinear relationship between the transmission rod, guide rail and discharge plate, thereby formulating a real-time adjustment strategy to ensure smooth transportation and material guiding of high-precision rollers;

[0084] Modeling the Generalized Nonlinear Hamilton-Jacobi Equations

[0085] In the process of the transmission rod driving the guide rail and the discharge plate to rotate, the rotation speed of the transmission rod, the position of the guide rail, the angle of the discharge plate and other parameters in the system affect each other, forming a complex dynamic system; in order to accurately describe the nonlinear relationship between these parameters, the generalized nonlinear Hamilton-Jacobi equation is introduced into the present invention;

[0086] Nonlinear relationships: In industrial automation systems, the relationship between the drive rod, guide rail and discharge plate is not a simple linear relationship; their interaction is affected by many factors, such as mechanical resistance, friction and changes in system load; these factors make the system behavior more complex and difficult to effectively describe with traditional linear models; generalized nonlinear Hamilton-Jacobi equations provide a mathematical tool that can capture these complex relationships in a nonlinear framework, ensuring accurate modeling of system behavior;

[0087] Dynamic system control: The generalized nonlinear Hamilton-Jacobi equation is a powerful mathematical tool that can describe the dynamic characteristics of the system and generate the optimal control strategy for real-time adjustment. By solving the generalized nonlinear Hamilton-Jacobi equation, the system finds the optimal control path under different conditions, thereby achieving precise control of the system. This is of great significance to ensure stable material guidance of high-precision rollers.

[0088] The basic form of the generalized nonlinear Hamilton-Jacobi equation is as follows:

[0089]

[0090] Among them, V(x, t) is the value function, which represents the minimum cost or maximum benefit from the current moment t to the final moment under state x; the value function reflects the optimal control effect of the system under different states; in the present invention, the value function is understood as the minimum energy consumption or time required from the current moment to the completion of the task when the system is in a certain state, a certain speed, a guide rail position and a discharge plate angle; this is the goal that the system hopes to minimize. By minimizing the value function, the system can select the most energy-saving and efficient control path; represents the derivative of the value function with respect to time;

[0091] is the Hamiltonian, which represents the energy of the system in its current state. It combines the dynamic characteristics and control strategy of the system. The Hamiltonian represents the current state energy of the system and the control strategy derived from this state. It combines all the influencing factors of the current state and uses the gradient V(x,t) predicts and optimizes future state changes; Hamiltonian plays a bridging role here, closely combining the system state with the control strategy;

[0092] x represents the state vector of the system, which includes the rotation speed of the transmission rod, the position of the guide rail and the angle parameters of the discharge plate; these parameters constitute the overall state of the system and are used to describe the current working condition of the system;

[0093] t represents time, which is an independent variable in a dynamic system. Time is a key variable in the dynamic changes of the system, and the control strategy is constantly adjusted as time goes by. The time variable in the generalized nonlinear Hamilton-Jacobi equation enables the system to handle dynamic control problems that change in real time, ensuring that the system can maintain the optimal state at different time points.

[0094] It represents the gradient of the value function to the state, reflecting the impact of the state change on the overall control effect of the system. The size and direction of the gradient determine how the system needs to adjust various parameters in the current state to achieve optimal control. The larger the gradient, the greater the adjustment requirement for the current state. Through the information of the gradient, the system can determine the best adjustment direction and amplitude to ensure that each step of adjustment is towards the optimal control target.

[0095] By minimizing the value function V(x,t), we can find the control path that minimizes the system energy and maximizes efficiency. This means that under certain conditions, by adjusting the speed of the transmission rod, the position of the guide rail, and the angle of the discharge plate, we can achieve the optimal roller guide effect. Minimizing the value function is the ultimate goal of system control. By continuously adjusting various parameters, the system can complete the task with the lowest energy consumption. This not only improves the efficiency of the system, but also extends the service life of the equipment and reduces maintenance costs.

[0096] Dynamic adjustment and optimization

[0097] Control strategies generated by generalized nonlinear Hamilton-Jacobi equations: By solving the generalized nonlinear Hamilton-Jacobi equations, the system generates optimal control strategies for real-time adjustments; these strategies can guide the system to make dynamic adjustments under different conditions to ensure that the rollers can accurately and smoothly guide the material when entering the discharge plate;

[0098] Control strategy: The strategy generated by the generalized nonlinear Hamilton-Jacobi equations includes dynamic adjustments to the rotation speed of the drive rod, the position of the guide rails and the angle of the discharge plate; these adjustments are made in real time during system operation to ensure that the rollers always move on the optimal path; the control strategy is dynamic and is adjusted based on real-time data input and system status; this means that the system adaptively responds to different loads and operating conditions to maintain efficient operation;

[0099] Step 4: Model Feedback and Adaptive Control

[0100] In the system of the present invention, model feedback and adaptive control are key steps to ensure the long-term stable operation of the system and gradually improve its performance; through feedback control and adaptive optimization, the system can adjust the operating parameters in real time, and through continuous learning and optimization, adapt to different working conditions and load changes, and finally achieve the best performance of the system;

[0101] Feedback Control

[0102] After the grinding disc rotates and completes the material collection work, the system performs feedback control through real-time data; feedback control is a closed-loop process. The system continuously adjusts the operating parameters based on the real-time data of the transmission rod speed, guide rail position, and discharge plate angle obtained by the sensor to ensure the accuracy and stability of the system during operation; as part of the closed-loop control, feedback control can timely correct the system's operating deviations and maintain system stability by real-time monitoring of changes in key parameters; in industrial automation systems, environmental and load conditions change at any time, so the real-time and accuracy of feedback control are crucial; through the closed-loop feedback mechanism, the system continuously compares the current state with the expected target and makes timely adjustments to cope with external changes; this process helps to ensure that the system maintains the best operating state under various conditions, thereby avoiding operating errors caused by external changes;

[0103] In feedback control, the improved neural network optimization algorithm plays an important role; this algorithm can not only handle complex nonlinear relationships, but also has strong adaptive capabilities; by continuously learning from feedback data, the algorithm can dynamically adjust control parameters and gradually optimize the system's operating results;

[0104] The improved neural network optimization algorithm absorbs information from feedback data and automatically adjusts the weights and parameters of the model, gradually improving the control performance of the system. The core of the algorithm lies in its adaptability, that is, it can dynamically adjust itself according to the real-time feedback of the system to cope with complex nonlinear relationships. The algorithm gradually optimizes the control strategy, so that the system can still maintain efficient and stable operation under changing working conditions.

[0105] Input data processing: After each operation, the system collects feedback data provided by sensors, including transmission rod speed (R), guide rail position (P), and discharge plate angle (A); these feedback data reflect the performance of the system in actual operation. Any deviation is captured through these data, and the system uses these data to make the next round of parameter adjustments; feedback data is the basis for algorithm adjustment. By collecting and processing these data in real time, the system can accurately perceive deficiencies in operation and optimize control strategies accordingly; these data provide the system with real operational feedback, helping the system identify potential problems and make necessary adjustments;

[0106] Error calculation: The system calculates the error in each round of operation based on the set target value and the feedback data from the actual operation;

[0107]

[0108] Among them, n means there are n rounds of operation, i means the i-th batch in a round of operation, and the target value iIndicates the target value for the i-th time, the actual value i Indicates the actual value of the i-th time

[0109] Parameter adjustment: Based on the error, the system uses a neural network optimization algorithm to adjust the parameters; the algorithm uses a back-propagation mechanism to adjust the weight of each parameter based on the size and direction of the error, thereby reducing the error in the next round of operations;

[0110] Formula derivation:

[0111]

[0112] Among them, θ t represents the current parameter value, θ t+1 represents the updated parameter value, η is the learning rate, and represents the step size of parameter adjustment. It represents the derivative of the error with respect to the current reference value;

[0113] Through this formula, the system can gradually adjust the parameters so that the next round of operations is closer to the target value. The core of the back-propagation mechanism is the gradient descent method, which enables the system to optimize the control effect by gradually reducing the error. Through continuous adjustment, the system parameters gradually tend to the optimal value, and finally achieve efficient and stable control effects.

[0114] Adaptive Optimization

[0115] Adaptive optimization means that after each operation, the system automatically adjusts the parameters of the neural network model according to the results of the operation feedback, so that it gradually adapts to different working conditions and load changes; this process is dynamic, and as the operation proceeds, the control ability of the system will continue to improve, and eventually achieve the best performance;

[0116] Adaptive optimization is a dynamic learning process. The system gradually optimizes the model through feedback data after each operation, so that it can better adapt to different operating environments and load changes. This capability enables the system to maintain high efficiency and stability in long-term operation and to cope with changes in various operating conditions. Adaptive optimization accumulates operating experience, so that the system can still maintain efficient operation in the face of complex working conditions.

[0117] Working principle of adaptive optimization: After each operation, the system will adjust the parameters of the model according to the feedback data, so that the next operation is closer to the expected goal; this process is similar to online learning in machine learning. The system gradually accumulates experience during operation and continuously improves its adaptability to different working conditions; specific steps:

[0118] Data collection and analysis: The system collects relevant data after each operation and analyzes the difference between the operation effect and the expected goal;

[0119] Through data analysis, the system can identify deficiencies in operations and provide a basis for the next round of adjustments; data collection and analysis are the basis of adaptive optimization. By analyzing deviations in operations, the system determines the next optimization direction; the analysis results provide important guidance for subsequent adjustments to the system, ensuring that each optimization is carried out in the right direction;

[0120] Model update: Based on the analysis results, the system automatically adjusts the parameters of the neural network model; this includes weight updates and learning rate adjustments to ensure that the model can better adapt to the current working conditions in the next round of operation;

[0121]

[0122] Among them, θ t represents the current parameter value, θ t+1 represents the updated parameter value, η is the learning rate, and represents the step size of parameter adjustment. It represents the derivative of the error with respect to the current reference value;

[0123] Ultimately, by continuously updating the model parameters, the system gradually reduces the operating error and improves the control accuracy. Model updating is the core of adaptive optimization. Through the feedback information after each operation, the model parameters are continuously adjusted to gradually approach the optimal state. Each update enables the system to cope with the ever-changing operating conditions and improve the overall efficiency.

[0124] Through feedback control and adaptive optimization, the system is able to continuously improve its performance and adapt to different operating conditions, ultimately achieving the goals of precise control and stable operation; this process not only improves the system's production efficiency, but also significantly reduces operational errors caused by load changes.

[0125] S2: At the same time, when the grinding disc body rotates, the moving wheel rubs against the grinding disc body, which can drive the moving wheel to rotate. When the moving wheel rotates, it can drive the transmission column fixed on the surface to rotate. The transmission column drives the first bevel gear to rotate. The first bevel gear meshes with the second bevel gear and drives the second bevel gear to rotate. The second bevel gear drives the coupling to rotate. The coupling is fixed on the surface of the fourth transmission disc, driving the fourth transmission disc to rotate, thereby achieving the purpose of transmission;

[0126] S3: When the fourth transmission disc rotates, it can drive the third transmission disc to rotate through the second belt, the third transmission disc drives the connecting rod to rotate, the connecting rod drives the second transmission disc to rotate, the second transmission disc rotates at the same time through the first transmission disc, thereby driving the first belt to rotate, the first belt is used in conjunction with the guide rail, when the high-precision roller moves to the connection between the guide rail and the first belt, the first belt can be used to assist in discharging.

[0127] Compared with the prior art, the present invention provides a grinding disc for high-precision roller production and a control method thereof, which has the following beneficial effects:

[0128] 1. The grinding disc for high-precision roller production and the control method thereof make the guide rail and the discharge plate contact with the grinding disc body by adjusting the structure. When the grinding disc body rotates, the discharge plate and the guide rail automatically discharge the material, thereby achieving the effect of facilitating the discharge of the material. At the same time, when the high-precision roller is automatically discharged, the safety of the operator is greatly increased, preventing the high-precision roller from being heavy to carry and failing to provide high safety for the operator.

[0129] 2. The high-precision roller production grinding disc and its control method use a motor to make the rotating rod rotate in a circle, and the rotating rod drives the transmission rod to rotate, so that the guide rail and the discharge plate fit the surface of the grinding disc body to perform the discharge work, which can facilitate the operator to perform automatic discharge work. When the grinding disc body is working, the guide rail and the discharge plate can be lifted to prevent affecting the normal operation of the grinding disc body;

[0130] 3. The grinding disc for high-precision roller production and the control method thereof, when the moving wheel rotates following the grinding disc body, the moving wheel rotates by itself to drive the transmission column to rotate, and the transmission column drives the coupling to rotate through the first bevel gear and the second bevel gear, and the coupling rotates to drive the fourth transmission disc in different positions to rotate, so that the moving wheel can assist in generating transmission force, thereby driving the auxiliary discharging structure to work, saving the power consumption of the auxiliary discharging structure;

[0131] 4. The grinding disc for high-precision roller production and the control method thereof can provide auxiliary support to the second bevel gear through the provided support block, thereby preventing the second bevel gear from being unstable due to single-point support and causing deviation during operation, thereby improving the stability of the second bevel gear.

[0132] 5. The grinding disc for high-precision roller production and its control method are connected and driven by the fourth transmission disc, etc. to drive the first belt to rotate, so that the high-precision roller on the surface of the guide rail can assist in discharging when it moves to the position. The guide rail is an arc-shaped rising type, which can easily cause the high-precision roller to get stuck. It can effectively prevent the high-precision roller from being blocked during discharging, thereby improving the work efficiency of the guide rail discharging work.

[0133] 6. The grinding disc for high-precision roller production and the control method thereof can make the first belt stably rotate inside the guide rail through the provided opening, thereby preventing the first belt from getting stuck during rotation.

[0134] 7. The grinding disc for high-precision roller production and the control method thereof can guide the high-precision roller through the setting of the guide plate and the connecting block, prevent the high-precision roller from getting stuck on the surface of the discharge plate, enable the high-precision roller to discharge materials stably, install and fix the discharge plate and the guide rail, and have the function of easy disassembly, so that the discharge plate can be quickly replaced.

[0135] 8. The grinding disc for high-precision roller production and its control method, by introducing graph neural architecture search and variational autoencoder under multiple divergence theory, the present invention can more accurately monitor the rotation speed of the grinding disc, the position of the discharge plate and the movement status of the roller in real time. The intelligent algorithm can dynamically adjust the rotation speed of the rotating rod and the angle of the guide rail according to these real-time data, thereby ensuring that the discharge process is smoother and more efficient. In traditional systems, the adjustment of the transmission ratio usually depends on preset parameters or simple feedback control. Through the introduction of intelligent algorithms, the system can automatically adjust the transmission ratio according to the actual load conditions during operation, ensuring that the grinding disc and the discharge plate can maintain the best working state under different working conditions. This dynamic adaptive capability significantly improves the response speed and adjustment accuracy of the system.

[0136] 9. The grinding disc for high-precision roller production and its control method. The maintenance of traditional systems is usually based on a predetermined schedule or after a failure occurs. The introduced intelligent algorithm can predict possible failures in advance and automatically issue maintenance reminders through modeling and optimization of the generalized nonlinear Hamilton-Jacobi equation. This can reduce the occurrence of unexpected downtime and improve the overall availability and reliability of the equipment.

[0137] 10. The grinding disc for high-precision roller production and its control method, through the combination of graph neural architecture search and variational autoencoder under multiple divergence theory, the system can search for the global optimal path and formulate control strategies in high-dimensional, complex geometric and dynamic spaces. This capability ensures that the system can still maintain efficient operation in the face of complex and changing production environments. Through the generalized nonlinear Hamilton-Jacobi equation, the system can optimize the control of complex nonlinear dynamic behaviors. This capability is particularly important when dealing with a wide range of operating conditions and load fluctuations, ensuring the stability and efficiency of the system in various complex environments.

[0138] 11. The grinding disc for high-precision roller production and its control method, through the improvement of optimization methods such as neural network optimization algorithm and multi-objective gray wolf algorithm, the system can continuously self-learn and evolve during operation to adapt to new working conditions and operating requirements. Over time, the system will become more and more intelligent and can complete tasks with higher efficiency and accuracy. During the optimization process, the system can simultaneously consider multiple goals, such as energy efficiency, production speed, equipment life, etc., and achieve the best balance between these goals. This multi-objective optimization capability ensures that the system can still maintain excellent performance under changing production needs.

[0139] 12. The grinding disc for high-precision roller production and its control method, by introducing advanced optimization methods such as Gaussian quantum improved particle swarm algorithm, difference creative search algorithm, improved whale optimization algorithm, etc., the system can maintain high robustness and stability when facing complex environments and high-dimensional data. Even when high load or emergencies occur, the system can respond quickly and adjust to the optimal state.

[0140] 13. The grinding disc and control method for high-precision roller production, due to the high adaptability and scalability of the introduced algorithm, the system can be easily adjusted and applied to different production environments and process requirements. Whether in high-precision roller production or other mechanical processing fields, this combination can provide strong support to ensure the scalability and versatility of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0141] Figure 1 It is a structural schematic diagram of the present invention;

[0142] Figure 2 This is a flip diagram of the discharge plate of the present invention;

[0143] Figure 3 It is a front view of the guide rail of the present invention;

[0144] Figure 4 An exploded view of the mobile wheel of the present invention;

[0145] Figure 5 It is a front view of the motor of the present invention;

[0146] Figure 6 This is a front view of the connecting card block of the present invention.

[0147] Figure 7 : Flow chart of the intelligent algorithm introduced in step S1 of the present invention

[0148] In the figure: 1. support seat; 2. grinding disc frame; 3. grinding disc body; 4. support plate; 5. guide rail; 6. discharge plate; 7. adjustment structure; 8. auxiliary discharge structure; 9. support frame; 10. motor; 11. rotating rod; 12. transmission rod; 13. transmission structure; 14. connecting frame; 15. moving wheel; 16. transmission column; 17. first bevel gear; 18. second bevel gear; 19. coupling; 20. support block; 21. connecting block; 22. first transmission disc; 23. first belt; 24. second transmission disc; 25. connecting rod; 26. third transmission disc; 27. second belt; 28. fourth transmission disc; 29. ​​outer shell; 30. opening; 31. guide plate; 32. connecting block. DETAILED DESCRIPTION

[0149] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention; it is obvious that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments, and all other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making creative work are within the scope of protection of the present invention.

[0150] Embodiment 1:

[0151] Reference Figure 1-6A grinding disc for high-precision roller production and a control method thereof include a support seat 1, a grinding disc frame 2 is fixedly connected to the top of the support seat 1, a grinding disc body 3 is slidably connected inside the grinding disc frame 2, a support plate 4 is fixedly connected to the right side of the grinding disc body 3, a guide rail 5 is arranged on the top of the support plate 4, a discharge plate 6 is fixedly connected to the left side of the guide rail 5, the grinding disc body 3 can send the high-precision roller into the guide rail 5 through the discharge plate 6 for guiding work, the guide rail 5 and the discharge plate 6 are connected to the support plate 4 through an adjusting structure 7, the discharge plate 6 can contact with the top of the grinding disc body 3 through the adjusting structure 7, an auxiliary discharge structure 8 is arranged inside the adjusting structure 7, the auxiliary discharge structure 8 can assist the guide rail 5 in discharging when the guide rail 5 is discharging, and when working, the adjusting structure 7 The motor 10 is controlled to rotate, the grinding disc 3 rotates and sends the high-precision roller to the top of the discharge plate 6 and guides the material through the discharge plate 6. The guide rail 5 and the discharge plate 6 are brought into contact with the grinding disc 3 through the adjusting structure 7. When the grinding disc 3 rotates, the discharge plate 6 and the guide rail 5 automatically discharge the material, thereby achieving the effect of facilitating the discharge of the material. At the same time, when the high-precision roller automatically discharges the material, the safety of the operator is greatly increased to prevent the high-precision roller from being heavy in carrying and failing to provide high safety for the operator. The adjusting structure 7 includes a support frame 9, the support frame 9 is fixedly connected to the top of the support plate 4, the front side of the support frame 9 is fixedly connected to the motor 10, the output end of the motor 10 is fixedly connected to the rotating rod 11, and the back side of the rotating rod 11 is connected by the shaft The bearing is movably connected to the rear side of the inner wall of the support frame 9, and a transmission rod 12 is fixedly connected to the surface of the rotating rod 11, and the opposite side of the transmission rod 12 is fixedly connected to the surface of the guide rail 5. The back of the guide rail 5 is fixedly connected to a transmission structure 13, and the transmission structure 13 can transmit through the swing of the guide rail 5 and the surface of the grinding disc 3. The rotating rod 11 is rotated in a circle by the motor 10, and the rotating rod 11 drives the transmission rod 12 to rotate, so that the guide rail 5 and the discharge plate 6 are in contact with the surface of the grinding disc 3 to perform the discharge work, which can facilitate the operator to perform automatic discharge work. When the grinding disc 3 is working, the guide rail 5 and the discharge plate 6 can be lifted to prevent affecting the normal operation of the grinding disc 3. The transmission structure 13 includes a connecting frame 14, and the bottom of the connecting frame 14 A moving wheel 15 is fixedly connected, and a transmission column 16 is fixedly connected to the right side of the moving wheel 15. The transmission column 16 can rotate along with the moving wheel 15 when it rotates. A first bevel gear 17 is fixedly connected to the right side of the transmission column 16. A second bevel gear 18 is meshed on the top of the first bevel gear 17. A coupling 19 is fixedly connected to the top of the second bevel gear 18. When the moving wheel 15 rotates along with the grinding disc body 3, the moving wheel 15 rotates by itself to drive the transmission column 16 to rotate. The transmission column 16 drives the coupling 19 to rotate through the first bevel gear 17 and the second bevel gear 18. The coupling 19 rotates to drive the eccentric fourth transmission disc 28 to rotate, so that the moving wheel 15 can assist in generating transmission force, thereby driving the auxiliary discharging structure 8 to work.The power consumption of the auxiliary discharging structure 8 is saved. The surface of the second bevel gear 18 is movably connected with a support block 20 through a bearing. The left side of the support block 20 is fixedly connected to the right side of the moving wheel 15. The second bevel gear 18 can be auxiliary supported by the support block 20 to prevent the second bevel gear 18 from being unstable at a single point and being offset during operation, thereby improving the stability of the second bevel gear 18. The auxiliary discharging structure 8 includes a connecting block 21. The side of the connecting block 21 close to the grinding disc body 3 is movably connected with a first transmission disc 22 through a rotating shaft. The surface of the first transmission disc 22 is slidably connected with a first belt 23. The first belt The second transmission disc 24 is slidably connected to the inside of the first transmission disc 23, and the surface of the second transmission disc 24 is movably connected to the surface of the connecting block 21 through a rotating shaft. The right side of the first transmission disc 22 is fixedly connected to a connecting rod 25, and the right side of the connecting rod 25 is fixedly connected to a third transmission disc 26. The surface of the third transmission disc 26 is slidably connected to a second belt 27, and the inside of the second belt 27 is slidably connected to a fourth transmission disc 28. The left side of the fourth transmission disc 28 is fixedly connected to the right side of the coupling 19, and the outer side of the fourth transmission disc 28 is movably connected to a housing 29 through a bearing. The front side of the housing 29 is fixedly connected to the back side of the guide rail 5, and the fourth transmission disc 28 is connected through the fourth transmission disc 28. The transmission drives the first belt 23 to rotate, so that the high-precision rollers on the surface of the guide rail 5 can assist in discharging when they move to the position. The guide rail 5 is an arc-shaped rising type, which can easily cause the high-precision rollers to get stuck, and can effectively prevent the high-precision rollers from getting stuck during discharging, thereby improving the discharging efficiency of the guide rail 5. An opening 30 is provided on the top of the guide rail 5, and the width of the opening 30 is greater than the width of the first belt 23. The opening 30 is used in conjunction with the first belt 23. Through the setting of the opening 30, the first belt 23 can be stably rotated inside the guide rail 5, thereby preventing the first belt 23 from getting stuck during rotation, and the discharging plate The top of the discharge plate 6 is fixedly connected with a guide plate 31, the number of guide plates 31 is several, and the several guide plates 31 are evenly arranged. The right side of the discharge plate 6 is fixedly connected with a connecting block 32, which penetrates into the interior of the guide rail 5 and is fixed by a fixing bolt on the surface of the guide rail 5. By setting the guide plate 31 and the connecting block 32, the high-precision roller can be guided to prevent the high-precision roller from getting stuck on the surface of the discharge plate 6, so that the high-precision roller can discharge stably, and the discharge plate 6 and the guide rail 5 are installed and fixed, and it is easy to disassemble, and the discharge plate 6 can be quickly replaced.

[0152] A grinding disc for high-precision roller production and a control method thereof are characterized in that they include the following steps:

[0153] S1: Start the motor 10, the motor 10 drives the rotating rod 11 to rotate through the output end, the rotating rod 11 drives the transmission rod 12 fixed on the surface to rotate when rotating, the transmission rod 12 drives the guide rail 5 and the discharge plate 6 to rotate, so that the discharge plate 6 enters the interior of the grinding disc frame 2 and contacts the top of the grinding disc body 3, when the grinding disc body 3 needs to collect materials, the grinding disc body 3 rotates and sends the high-precision roller to the top of the discharge plate 6 and guides the materials through the discharge plate 6;

[0154] Step 1: Data collection and preprocessing

[0155] Rotational speed of the rotating rod (R): The data collected by the sensor is 1200RPM±50RPM, and the frequency is 100Hz, that is, 100 data points are recorded per second.

[0156] Position of the guide rail (P): The initial position is 30 mm, the position changes from 28 mm to 32 mm during operation, and the frequency is 100 Hz.

[0157] Angle of the discharge plate (A): The initial angle is 10°, and the angle changes from 9° to 11° during operation, with a frequency of 100 Hz.

[0158] After data preprocessing, the original data is reduced in dimension by the variational autoencoder, retaining the main features and effectively removing the noise. After standardization, the mean of the speed, position and angle data is 0 and the variance is 1.

[0159] Step 2: Graph Neural Architecture Search (NAS)

[0160] Feature vector after dimensionality reduction: Assume that after dimensionality reduction and standardization, the system obtains the following feature vector:

[0161] R'=0.1±0.05

[0162] P'=-0.05±0.02

[0163] A'=0.02±0.01

[0164] Using graph neural networks for modeling, the captured nonlinear relationships are represented by adjacency matrices:

[0165]

[0166] Through search and optimization, NAS finally selected a 3-layer GNN structure with 50 nodes in each layer, ReLU activation function, and an initial adjacency matrix weight of 0.8.

[0167] Step 3: Nonlinear dynamic modeling and optimization

[0168] State vector: x = [R, P, A] = [1200RPM, 30mm, 10°]

[0169] Generalized nonlinear Hamilton-Jacobi equation (HJE):

[0170] Assuming the goal is to minimize energy consumption and maintain system stability, the control strategy generated by the system requires that at time t1, R be adjusted to 1180 RPM, P to 30.5 mm, and A to 10.2°.

[0171] By solving HJE, the system determines the optimal path at the next time point.

[0172] Step 4: Model Feedback and Adaptive Control

[0173] Initial feedback data:

[0174] R=1150RPM,P=32mm,A=11°

[0175] Error calculation:

[0176] Total error = 2505

[0177] Parameter adjustment:

[0178] Adjust R to 1180RPM, P to 30.5mm, A to 10.2°

[0179] Adaptive Optimization:

[0180] In the next operation, the system automatically adjusted the model parameters so that the target speed reached 1190RPM, the guide rail position was 30.2mm, and the discharge plate angle was 10.1°, gradually approaching the expected target.

[0181] Through the cycle and optimization of these four steps, the system successfully achieved precise dynamic adjustment and control in response to different working conditions and load changes. Ultimately, during the continuous optimization process, the system was able to maintain high-precision control of the discharge plate angle within the range of 10°±0.2°, demonstrating excellent adaptability and operational stability. This process significantly improved production efficiency, ensured product quality, and demonstrated the practical application value of the system in complex industrial environments.

[0182] S2: At the same time, when the grinding disc body 3 rotates, the moving wheel 15 rubs against the grinding disc body 3, which can drive the moving wheel 15 to rotate. When the moving wheel 15 rotates, it can drive the transmission column 16 fixed on the surface to rotate. The transmission column 16 drives the first bevel gear 17 to rotate. The first bevel gear 17 meshes with the second bevel gear 18 and drives the second bevel gear 18 to rotate. The second bevel gear 18 drives the coupling 19 to rotate. The coupling 19 is fixed on the surface of the fourth transmission disc 28, and drives the fourth transmission disc 28 to rotate, thereby achieving the purpose of transmission;

[0183] S3: When the fourth transmission disc 28 rotates, the third transmission disc 26 can be driven to rotate through the second belt 27, the third transmission disc 26 drives the connecting rod 25 to rotate, the connecting rod 25 drives the second transmission disc 24 to rotate, and the second transmission disc 24 rotates at the same time through the first transmission disc 22, thereby driving the first belt 23 to rotate. The first belt 23 is used in conjunction with the guide rail 5. When the high-precision roller moves to the connection between the guide rail 5 and the first belt 23, the first belt 23 can be used to assist in discharging.

[0184] Specifically, when the high-precision roller production grinding disc is working / in use: the motor 10 is started, and the motor 10 drives the rotating rod 11 to rotate through the output end. When the rotating rod 11 rotates, it drives the transmission rod 12 fixed on the surface to rotate, and the transmission rod 12 drives the guide rail 5 and the discharge plate 6 to rotate, so that the discharge plate 6 enters the interior of the grinding disc frame 2 and contacts the top of the grinding disc body 3. When the grinding disc body 3 needs to collect materials, the grinding disc body 3 rotates and sends the high-precision roller to the top of the discharge plate 6 and guides the material through the discharge plate 6. At the same time, when the grinding disc body 3 rotates, the moving wheel 15 rubs against the grinding disc body 3, which can drive the moving wheel 15 to rotate. When the moving wheel 15 rotates, it can drive the transmission column 16 fixed on the surface to rotate, and the transmission column 16 drives the first bevel gear The wheel 17 rotates, the first bevel gear 17 engages with the second bevel gear 18 and drives the second bevel gear 18 to rotate, the second bevel gear 18 drives the coupling 19 to rotate, the coupling 19 is fixed to the surface of the fourth transmission disk 28, and drives the fourth transmission disk 28 to rotate, thereby achieving the purpose of transmission. When the fourth transmission disk 28 rotates, it can drive the third transmission disk 26 to rotate through the second belt 27, the third transmission disk 26 drives the connecting rod 25 to rotate, the connecting rod 25 drives the second transmission disk 24 to rotate, and the second transmission disk 24 rotates at the same time through the first transmission disk 22, thereby driving the first belt 23 to rotate. The first belt 23 is used in conjunction with the guide rail 5. When the high-precision roller moves to the connection between the guide rail 5 and the first belt 23, the first belt 23 can be used to assist in discharging.

[0185] The above description 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 grinding disc for high-precision roller production, comprising a support seat (1), characterized in that: The top of the support seat (1) is fixedly connected to a grinding disc frame (2), the inside of the grinding disc frame (2) is slidably connected to a grinding disc body (3), the right side of the grinding disc body (3) is fixedly connected to a support plate (4), the top of the support plate (4) is provided with a guide rail (5), the left side of the guide rail (5) is fixedly connected to a discharge plate (6), the grinding disc body (3) can send a high-precision roller into the guide rail (5) through the discharge plate (6) for material guiding, the guide rail (5) and the discharge plate (6) are connected to the support plate (4) through an adjustment structure (7), the discharge plate (6) can contact the top of the grinding disc body (3) through the adjustment structure (7), the adjustment structure (7) is provided with an auxiliary discharge structure (8), the auxiliary discharge structure (8) can assist the guide rail (5) in discharging materials when the guide rail (5) is discharging materials.

2. The grinding disc for high-precision roller production according to claim 1, characterized in that: The adjustment structure (7) comprises a support frame (9), wherein the support frame (9) is fixedly connected to the top of the support plate (4), the front side of the support frame (9) is fixedly connected to a motor (10), the output end of the motor (10) is fixedly connected to a rotating rod (11), the back side of the rotating rod (11) is movably connected to the rear side of the inner wall of the support frame (9) through a bearing, the surface of the rotating rod (11) is fixedly connected to a transmission rod (12), the opposite side of the transmission rod (12) is fixedly connected to the surface of the guide rail (5), the back side of the guide rail (5) is fixedly connected to a transmission structure (13), and the transmission structure (13) can be transmitted by swinging the guide rail (5) and fitting with the surface of the grinding disc body (3).

3. The grinding disc for high-precision roller production according to claim 2, characterized in that: The transmission structure (13) comprises a connecting frame (14), a moving wheel (15) is fixedly connected to the bottom of the connecting frame (14), a transmission column (16) is fixedly connected to the right side of the moving wheel (15), and the transmission column (16) can rotate along with the moving wheel (15) when the moving wheel (15) rotates, a first bevel gear (17) is fixedly connected to the right side of the transmission column (16), a second bevel gear (18) is meshed with the top of the first bevel gear (17), and a coupling (19) is fixedly connected to the top of the second bevel gear (18).

4. The grinding disc for high-precision roller production according to claim 3, characterized in that: The surface of the second bevel gear (18) is movably connected to a support block (20) via a bearing, and the left side of the support block (20) is fixedly connected to the right side of the moving wheel (15).

5. The grinding disc for high-precision roller production according to claim 3, characterized in that: The auxiliary discharging structure (8) comprises a connecting block (21), a side of the connecting block (21) close to the grinding disc body (3) is movably connected to a first transmission disc (22) via a rotating shaft, a surface of the first transmission disc (22) is slidably connected to a first belt (23), the interior of the first belt (23) is slidably connected to a second transmission disc (24), the surface of the second transmission disc (24) is movably connected to the surface of the connecting block (21) via a rotating shaft, a connecting rod (25) is fixedly connected to the right side of the first transmission disc (22), a third transmission disc (26) is fixedly connected to the right side of the connecting rod (25), a surface of the third transmission disc (26) is slidably connected to a second belt (27), the interior of the second belt (27) is slidably connected to a fourth transmission disc (28), the left side of the fourth transmission disc (28) is fixedly connected to the right side of a coupling (19), the outer side of the fourth transmission disc (28) is movably connected to a housing (29) via a bearing, and the front side of the housing (29) is fixedly connected to the back side of the guide rail (5).

6. The grinding disc for high-precision roller production according to claim 3, characterized in that: An opening (30) is provided at the top of the guide rail (5); the width of the opening (30) is greater than the width of the first belt (23); the opening (30) is used in conjunction with the first belt (23).

7. The grinding disc for high-precision roller production according to claim 1, characterized in that: A guide plate (31) is fixedly connected to the top of the discharge plate (6), the number of the guide plates (31) is several, and the several guide plates (31) are evenly arranged, and a connecting block (32) is fixedly connected to the right side of the discharge plate (6), the connecting block (32) penetrates into the interior of the guide rail (5), and is fixed by fixing bolts on the surface of the guide rail (5).

8. The method for controlling a grinding disc for producing high-precision rollers according to claim 1, characterized in that: The following steps are involved: S1: starting the motor (10), the motor (10) drives the rotating rod (11) to rotate through the output end, the rotating rod (11) drives the transmission rod (12) fixed on the surface to rotate when rotating, the transmission rod (12) drives the guide rail (5) and the discharge plate (6) to rotate, so that the discharge plate (6) enters the interior of the grinding disc frame (2) and contacts the top of the grinding disc body (3), when the grinding disc body (3) needs to collect materials, the grinding disc body (3) rotates and sends the high-precision roller to the top of the discharge plate (6) and guides the materials through the discharge plate (6); S2: At the same time, when the grinding disc body (3) rotates, the moving wheel (15) rubs against the grinding disc body (3), which can drive the moving wheel (15) to rotate. When the moving wheel (15) rotates, it can drive the transmission column (16) fixed on the surface to rotate. The transmission column (16) drives the first bevel gear (17) to rotate. The first bevel gear (17) meshes with the second bevel gear (18) and drives the second bevel gear (18) to rotate. The second bevel gear (18) drives the coupling (19) to rotate. The coupling (19) is fixed on the surface of the fourth transmission disc (28) and drives the fourth transmission disc (28) to rotate, thereby achieving the purpose of transmission; S3: When the fourth transmission disc (28) rotates, the third transmission disc (26) can be driven to rotate through the second belt (27), the third transmission disc (26) drives the connecting rod (25) to rotate, the connecting rod (25) drives the second transmission disc (24) to rotate, the second transmission disc (24) rotates simultaneously through the first transmission disc (22), thereby driving the first belt (23) to rotate, the first belt (23) and the guide rail (5) are used in conjunction, when the high-precision roller moves to the connection between the guide rail (5) and the first belt (23), the first belt (23) can be used to assist in discharging.

9. The control method of the grinding disc for high-precision roller production according to claim 8, characterized in that: In S1: an intelligent algorithm is introduced to monitor the rotation speed of the grinding disc, the position of the discharge plate and the movement status of the roller in real time, and the rotation speed of the rotating rod and the angle of the guide rail are dynamically adjusted according to the real-time data to ensure the smooth and efficient discharge process; in addition, the transmission system adaptively adjusts the transmission ratio to ensure that the grinding disc and the discharge plate can maintain the best working condition under any load condition; the specific process is as follows: Step 1: Data collection and preprocessing Data collection: After the motor is started, the motor drives the rotating rod to rotate through the output end; at this time, multiple sensors in the system begin to collect key parameter data closely related to the system operation in real time; the purpose of collecting this data is to monitor the operating status of the system, ensure the synchronization and stability of each component, and thus ensure the efficiency and accuracy of the entire grinding process; the following are the main collected data and their detailed description: Rotating rod speed (R) sensor: Use photoelectric sensors and Hall effect sensors, installed near the motor output end, close to the bearing of the rotating rod; the sensor is close to the rotating rod to reduce the delay or error caused by the mechanical transmission part; the position of the sensor should avoid interference from the mechanical structure to ensure that the sensor can directly sense the rotation speed of the rotating rod; Guide rail position (P) sensor: Use a linear position sensor or laser distance sensor installed on the fixed or moving end of the guide rail to directly sense the linear displacement of the guide rail. The sensor should be installed within the travel range of the guide rail to ensure that the sensor can accurately track the position changes of the guide rail during the entire operation process. The position of the sensor should ensure that it always maintains effective sensing throughout the entire movement of the guide rail. Angle sensor of the discharge plate (A): Use an angle encoder installed on the rotating shaft or hinge of the discharge plate to directly measure the angular change of the discharge plate relative to the fixed shaft; the sensor needs to be closely connected to the key rotating part of the discharge plate to avoid deviation during the rotation process; the sensor position should ensure that it can provide accurate data within the entire angle adjustment range; Data preprocessing: The collected raw data is high-dimensional, which not only contains a lot of detailed information but also contains noise; these high-dimensional data must be preprocessed; Dimensionality reduction and noise elimination: In order to effectively process high-dimensional data and retain its core features, a variational autoencoder based on multiple divergence theory is used for dimensionality reduction and noise elimination. The variational autoencoder can not only reduce the dimension of the data, but also remove random noise in the data while reducing the dimension, making subsequent analysis more accurate and efficient. The variational autoencoder under the multi-divergence theory is a generative model that generates new data similar to the input data by learning the distribution of the data and optimizes the model by maximizing the following evidence lower bound: Among them, ELBO represents the new data formed by the generative model, x represents the observed data, which corresponds to the sensor data of speed (R), position (P) and angle (A); z is the latent variable, which represents the data characteristics after dimensionality reduction; this latent variable space is simpler than the original data space, but retains the main characteristics of the data; p(x|z) is the generative model, which describes the mapping relationship from the latent space to the observed data; this process is similar to the reverse operation of reducing high-dimensional data to low-dimensional data; p(z): prior distribution, q(z|x) is the approximate posterior distribution, which is used to map the observed data to the latent space; this distribution needs to be close to the true posterior distribution to ensure the rationality of dimensionality reduction; The expected value operator represents the expectation of the posterior distribution q(z|x) of the latent variable z; this means taking a weighted average of all the latent variables z; log p(x|z): log likelihood, which represents the probability of generating the observed data x given the latent variable z; this part represents the model's ability to reconstruct the data, that is, how well the model can restore the original data given the latent variable; D KL : KL divergence is used to measure the difference between two distributions; here it measures the difference between the approximate posterior distribution q(z|x) and the prior distribution p(z); Data standardization: The data after dimensionality reduction needs to be standardized to make it have a uniform scale; eliminate the scale differences between different data dimensions to avoid the data of a certain dimension having too much impact on model training due to its large dimension; Standardization adjusts the data to a mean of 0 and a variance of 1. This process helps improve the efficiency and stability of model training and ensures that the model does not favor a certain feature during learning. Standardization formula: Among them, x is the original data value, μ is the mean of the data, σ is the standard deviation of the data; x′ is the standardized data; The standardized data x′ will be converted to a standard normal distribution with a mean of 0 and a variance of 1, ensuring that each feature is at the same level; thus, the model has the same weight when processing this data; Data sorting: The data after dimensionality reduction and standardization will be organized into a format suitable for subsequent analysis and modeling; each data point will be further integrated to form a feature vector for the input model; through sorting, the system ensures that all input data structures are consistent, providing a solid foundation for subsequent modeling and optimization; the sorted data is represented in matrix form, with each row representing the system state at a moment and each column representing a standardized feature; Step 2: Graph Neural Architecture Search Graph Neural Network Modeling The reduced-dimensional data is used to build the graph neural network model: In the data preprocessing stage, the original high-dimensional data has been reduced in dimension through the variational autoencoder, which retains the key features of speed, position, and angle. These features are used to build the graph neural network model, which effectively captures the complex dependencies between the key parameters of the system and generates a comprehensive state representation. Node representation: In the graph neural network, the speed, position and angle parameters are represented as nodes; there are complex nonlinear relationships between these nodes; changes in speed will affect the position of the guide rail, and the position of the guide rail will affect the angle of the discharge plate; the task of the graph neural network is to capture and reflect these dependencies by modeling these nodes and their relationships; Adjacency matrix: The relationship between nodes is represented by the adjacency matrix; the adjacency matrix is ​​a second-order tensor matrix, in which each element indicates whether there is a connection between two nodes and the strength of the connection; ω RP It represents the connection weight between speed (R) and position (P), reflecting the influence of speed on the guide rail position; ω RA It represents the connection weight between the rotation speed (R) and the angle (A), reflecting the influence of the rotation speed on the angle of the discharge plate; ω AP It represents the connection weight between position (P) and angle (A), reflecting the influence of the guide rail position on the angle of the discharge plate; These weights reflect the strength of the dependencies between different parameters. The larger the value, the closer the relationship between the two. Through these relationship matrices, graph neural networks capture the complex dependencies between various parameters, thereby providing a global state representation for the system. Graph neural networks use a message passing mechanism to propagate and aggregate the information of each node through the adjacency matrix to generate a comprehensive state representation; this process includes the following steps: Message passing: Each node transmits information to its neighboring nodes according to the weight in the adjacency matrix; this information contains the node's own characteristics, speed value, position value, angle value and information received from neighboring nodes; this process ensures that each node can be understood in context, rather than just an isolated value; Node information transmission: Each node transmits its own information to its adjacent nodes through the weight of the adjacency matrix. This information transmission process enables each node to generate a more accurate state representation based not only on its own state, but also on the relationship with other nodes. Information aggregation: Each node aggregates the information received from neighboring nodes with its own information and updates its state representation; this process is completed through a learnable function; Aggregate formula: Among them, i and j: different nodes, represents the hidden state of node i in the kth layer; this hidden state is the updated state of node i after aggregating information, which contains the comprehensive information from neighboring nodes; represents the hidden state of node j in layer k-1, represents the neighbor set of node i, that is, other nodes directly connected to node i; d i and d j are the degrees of nodes i and j, respectively, indicating the number of connections of each node; the inverse of the degree is used to normalize message passing to prevent some nodes from causing bias in information aggregation due to large degrees; W(k) is the learnable weight matrix of the kth layer, which is used for information transmission between nodes; by learning weights, the model decides which information is more important; It is a learnable weight matrix used to update the node's own information; σ is an activation function, using Sigmoid to introduce nonlinearity and enhance the model's expressiveness; State update: Through multiple iterations, the graph neural network gradually updates the state of each node, and finally forms a comprehensive global state representation; this comprehensive representation can capture the complex dependencies between various parameters in the system and provide a comprehensive understanding of the state of the entire system; Architecture Search and Optimization Search space definition: Graph neural architecture search first defines a search space that contains all graph neural network architecture combinations; the dimensions of the search space include the number of network layers, the number of nodes in each layer, the choice of activation function, and the initial weight of the adjacency matrix; Evaluation of candidate architectures: For each candidate architecture, Graph Neural Architecture Search uses an evaluation function to measure its performance. The evaluation function is based on the mean square error loss function. The purpose of the evaluation is to quantify the pros and cons of each candidate architecture and then select the optimal architecture. Loss function: in, represents the value of the loss function, which measures the gap between the model prediction and the actual target; f(x i ; θ) represents the output of the graph neural network model, θ is the model parameter, x i is the input data; y i is the actual target value, indicating the true state of each node in the system; Loss(f(x i ;θ),y i ) is the error between the model prediction value and the true value, which is used to evaluate the accuracy of the model, and N is the number of models of the output of the graph neural network model; Optimization and model selection Through continuous search and evaluation, graph neural architecture search can find a graph neural network architecture that performs best on the evaluation function; Optimization process: The optimization process of graph neural architecture search (NAS) uses the gradient descent method to gradually adjust the parameters of the candidate architecture to reduce the value of the loss function; this process is similar to finding the global optimal solution in a complex search space; Gradient descent method: The gradient descent method gradually adjusts the parameters to reduce the loss value by calculating the derivative of the loss function with respect to the model parameters; the formula is: Among them, θ is the parameter of the model; η is the learning rate, which controls the step size of parameter update; is the gradient of the loss function with respect to the parameters; Selection of the final architecture: When the graph neural architecture search completes the search, the system will select the architecture with the smallest loss function as the final graph neural network model; this model can not only accurately capture the complex relationship between various parameters in the system, but also provide efficient prediction and control capabilities in practical applications; Step 3: Nonlinear dynamic modeling and optimization Modeling the Generalized Nonlinear Hamilton-Jacobi Equations In the process of the transmission rod driving the guide rail and the discharge plate to rotate, the rotation speed of the transmission rod, the position of the guide rail, the angle of the discharge plate and other parameters in the system affect each other, forming a complex dynamic system; in order to accurately describe the nonlinear relationship between these parameters, the generalized nonlinear Hamilton-Jacobi equation is introduced into the present invention; The basic form of the generalized nonlinear Hamilton-Jacobi equation is as follows: Among them, V(x, t) is the value function, which represents the minimum cost or maximum benefit from the current moment t to the final moment under state x; the value function reflects the optimal control effect of the system under different states; in the present invention, the value function is understood as the minimum energy consumption or time required from the current moment to the completion of the task when the system is in a certain state, a certain speed, a guide rail position and a discharge plate angle; this is the goal that the system hopes to minimize. By minimizing the value function, the system can select the most energy-saving and efficient control path; represents the derivative of the value function with respect to time; is the Hamiltonian, which represents the energy of the system in its current state. It combines the dynamic characteristics and control strategy of the system. The Hamiltonian represents the current state energy of the system and the control strategy derived from this state. It combines all the influencing factors of the current state and uses the gradient V(x,t) predicts and optimizes future state changes; Hamiltonian plays a bridging role here, closely combining the system state with the control strategy; x represents the state vector of the system, which includes the rotation speed of the transmission rod, the position of the guide rail and the angle parameters of the discharge plate; these parameters constitute the overall state of the system and are used to describe the current working condition of the system; t represents time, which is an independent variable in a dynamic system. Time is a key variable in the dynamic changes of the system, and the control strategy is constantly adjusted as time goes by. The time variable in the generalized nonlinear Hamilton-Jacobi equation enables the system to handle dynamic control problems that change in real time, ensuring that the system can maintain the optimal state at different time points. It represents the gradient of the value function to the state, reflecting the impact of the state change on the overall control effect of the system. The size and direction of the gradient determine how the system needs to adjust various parameters in the current state to achieve optimal control. The larger the gradient, the greater the adjustment requirement for the current state. Through the information of the gradient, the system can determine the best adjustment direction and amplitude to ensure that each step of adjustment is towards the optimal control target. By minimizing the value function V(x,t), we can find the control path that minimizes the system energy and maximizes efficiency. This means that under certain conditions, by adjusting the speed of the transmission rod, the position of the guide rail, and the angle of the discharge plate, we can achieve the optimal roller guide effect. Minimizing the value function is the ultimate goal of system control. By continuously adjusting various parameters, the system can complete the task with the lowest energy consumption. This not only improves the efficiency of the system, but also extends the service life of the equipment and reduces maintenance costs. Dynamic adjustment and optimization Control strategy: The strategy generated by the generalized nonlinear Hamilton-Jacobi equations includes dynamic adjustments to the rotation speed of the drive rod, the position of the guide rails and the angle of the discharge plate; these adjustments are made in real time during system operation to ensure that the rollers always move on the optimal path; the control strategy is dynamic and is adjusted based on real-time data input and system status; this means that the system adaptively responds to different loads and operating conditions to maintain efficient operation; Step 4: Model Feedback and Adaptive Control Input data processing: After each operation, the system collects feedback data provided by sensors, including transmission rod speed (R), guide rail position (P), and discharge plate angle (A); these feedback data reflect the performance of the system in actual operation. Any deviation is captured through these data, and the system uses these data to make the next round of parameter adjustments; feedback data is the basis for algorithm adjustment. By collecting and processing these data in real time, the system can accurately perceive deficiencies in operation and optimize control strategies accordingly; these data provide the system with real operational feedback, helping the system identify potential problems and make necessary adjustments; Error calculation: The system calculates the error in each round of operation based on the set target value and the feedback data from the actual operation; Among them, n means there are n rounds of operation, i means the i-th batch in a round of operation, and the target value i Indicates the target value for the i-th time, the actual value i represents the actual value of the i-th time; Parameter adjustment: Based on the error, the system uses a neural network optimization algorithm to adjust the parameters; the algorithm uses a back-propagation mechanism to adjust the weight of each parameter based on the size and direction of the error, thereby reducing the error in the next round of operations; Formula derivation: Among them, θ t represents the current parameter value, θ t+1 represents the updated parameter value, η is the learning rate, and represents the step size of parameter adjustment. It represents the derivative of the error with respect to the current reference value; Through this formula, the system can gradually adjust the parameters so that the next round of operations is closer to the target value. The core of the back-propagation mechanism is the gradient descent method, which enables the system to optimize the control effect by gradually reducing the error. Through continuous adjustment, the system parameters gradually tend to the optimal value, and finally achieve efficient and stable control effects. Data collection and analysis: The system collects relevant data after each operation and analyzes the difference between the operation effect and the expected goal; Through data analysis, the system can identify deficiencies in operations and provide a basis for the next round of adjustments; data collection and analysis are the basis of adaptive optimization. By analyzing deviations in operations, the system determines the next optimization direction; the analysis results provide important guidance for subsequent adjustments to the system, ensuring that each optimization is carried out in the right direction; Model update: Based on the analysis results, the system automatically adjusts the parameters of the neural network model; this includes weight updates and learning rate adjustments to ensure that the model can better adapt to the current working conditions in the next round of operation; Among them, θ t represents the current parameter value, θ t+1 represents the updated parameter value, η is the learning rate, and represents the step size of parameter adjustment. It represents the derivative of the error with respect to the current reference value; Ultimately, by continuously updating the model parameters, the system gradually reduces operating errors and improves control accuracy; model updating is the core of adaptive optimization. Through feedback information after each operation, the model parameters are continuously adjusted to gradually approach the optimal state; each update enables the system to cope with changing operating conditions and improve overall efficiency.