A method for controlling the size of micro spring washers based on neural network prediction

Abstract

The present invention discloses a method for controlling the size of a micro spring gasket based on neural network prediction, comprising the following steps: S1, real-time monitoring of the production process of the micro spring gasket, collecting and forming a multidimensional data set; S2, preprocessing the multidimensional data set; S3, extracting key features from the preprocessed data set using a high-dimensional quantum state perturbation feature extraction method; S4, constructing a self-consistent field neural network model based on the extracted key features; S5, accurately predicting the size of the micro spring gasket based on a nonlinear chaos prediction algorithm; S6, based on the prediction results, using a multidimensional topological protection control technology to adjust the processing parameters of the production equipment in real time; S7, comparing the size of the micro spring gasket actually produced with the prediction results, and further optimizing the system using a hierarchical entropy balance mechanism. The present invention achieves high-precision control of the size of the micro spring gasket and improves the stability and consistency of the production process.

Description

Technical Field

[0001] The invention relates to the technical field of industrial control, and in particular to a method for controlling the size of a micro spring gasket based on neural network prediction. Background Art

[0002] With the continuous development of modern manufacturing, especially in the field of precision machinery and micro-components manufacturing, higher and higher requirements are placed on the precise control of product dimensions. As a common mechanical component, the dimensional accuracy of micro spring washers directly affects the overall performance and reliability of the equipment. However, in the actual production process, due to the complex process, diverse environmental variables, and the diversity of production equipment, how to achieve high-precision control of the dimensions of micro spring washers is still a technical problem.

[0003] In the existing technology, the dimensional control of micro parts usually relies on traditional quality inspection and control methods, which mainly include offline inspection, simple statistical process control, and experience-based parameter adjustment. These traditional methods have the following significant defects in the face of increasingly complex production environments and higher precision requirements:

[0004] 1. Poor real-time performance: Traditional dimensional control methods mostly rely on offline detection, that is, dimensional measurement and quality inspection are performed after the parts are produced. Although this method can detect problems in the production process, due to the lag in detection results, it is impossible to provide timely feedback and adjust the production process, resulting in a high scrap rate and low production efficiency.

[0005] 2. Lack of dynamic adaptability: Traditional statistical process control methods usually adjust parameters based on historical data and fixed control rules. However, in actual production, environmental conditions and equipment status often change dynamically. Fixed control rules are difficult to adapt to these changes, resulting in low control accuracy and unable to meet the higher requirements of modern manufacturing for product quality.

[0006] 3. Difficulty in controlling in complex environments: In the production process of micro-components, various environmental factors such as temperature, humidity, vibration, etc. will affect the size of the product. Traditional methods lack effective countermeasures when facing these complex environmental variables, resulting in poor stability of dimensional control.

[0007] 4. Insufficient predictive capabilities: Existing technologies usually rely on predictions based on experience or simple statistical models. These methods have difficulty capturing potential nonlinear dynamic behaviors in the production process and cannot accurately predict the trend of dimensional changes, especially when faced with complex production processes and multivariate coupling effects. This problem is particularly prominent.

[0008] 5. Lack of intelligence and adaptability: Most current dimensional control methods rely on manual experience and simple automated control systems, lacking intelligence and adaptability. With the increase in the amount and complexity of data in the production process, traditional methods are unable to cope with data processing and analysis, and it is difficult to extract useful information from massive data and conduct effective control.

[0009] Therefore, how to provide a method for controlling the size of micro spring washers based on neural network prediction is an urgent problem that technicians in this field need to solve. Summary of the invention

[0010] One purpose of the present invention is to propose a method for controlling the size of a micro spring gasket based on neural network prediction. The present invention makes full use of neural network algorithms, quantum state perturbation analysis and nonlinear chaos prediction technology, and describes in detail the process of intelligently realizing the size control of a micro spring gasket. It has the advantages of high precision, strong real-time performance and adaptability to complex production environments.

[0011] A method for controlling the size of a micro spring washer based on neural network prediction according to an embodiment of the present invention comprises the following steps:

[0012] S1. Monitor the production process of micro spring washers in real time, collect and form multidimensional data sets;

[0013] S2, preprocessing the collected multidimensional data sets, eliminating noise and redundant information, and making the data adapt to the input requirements of the self-consistent field neural network through standardization and normalization;

[0014] S3, using a high-dimensional quantum state perturbation feature extraction method to represent the preprocessed data set as a high-dimensional quantum state, analyzing the change of the quantum state by applying a small perturbation, and extracting key features related to the size of the micro spring gasket;

[0015] S4. Based on the key features extracted by the high-dimensional quantum state perturbation algorithm, a self-consistent field neural network model is constructed. Through multiple iterations of calculating the self-consistent field distribution of the system, the network parameters are adjusted to make the data propagation in the network reach a balanced state, thereby extracting stable features and making accurate predictions;

[0016] S5. Input the real-time collected production data into the trained self-consistent field neural network model, predict the size of the micro spring gasket based on the nonlinear chaos prediction algorithm, identify the chaotic characteristics in the system and establish a corresponding prediction model by analyzing the nonlinear dynamic behavior in the production process, capture the sensitivity of small initial condition changes to the system output, and use the sensitive characteristics to accurately predict the future changes in size;

[0017] S6. Based on the prediction results of self-consistent field neural network and nonlinear chaos prediction algorithm, multi-dimensional topological protection control technology is used to adjust the processing parameters of production equipment in real time;

[0018] S7. Compare the actual size of the micro spring gasket produced with the predicted result, and use the hierarchical entropy balance mechanism to further optimize the system. By analyzing and adjusting the information entropy of each level, the system can reach the optimal balance state.

[0019] Optionally, S3 includes the following steps:

[0020] S31, using a high-dimensional quantum state perturbation feature extraction method to represent the preprocessed multidimensional data set as a high-dimensional quantum state;

[0021] S32. Apply a small disturbance to the constructed high-dimensional quantum state, where the disturbance is expressed as a change in the quantum state, and its magnitude depends on the range of changes of various parameters in the production data;

[0022] S33. Introduce adaptive quantum state pruning technology to dynamically screen and prune the high-dimensional quantum state after the disturbance. By analyzing the contribution of each component in the quantum state to the final characteristics, remove the components that contribute less to the quantum state change and retain the quantum state components with high influence.

[0023] S34, analyzing the quantum state change after the adaptive quantum state pruning, calculating the amplitude and phase difference of the quantum state change, and identifying the influence of each input parameter on the quantum state change:

[0024]

[0025] Among them, Δψ total represents the overall change of the quantum state after pruning, α i is the sensitivity coefficient of the quantum state to the i-th parameter, represents the quantum state ψ(X) relative to the i-th input parameter x i The partial derivative of δx i is the perturbation of the ith parameter, represents the quantum state ψ(X) relative to the jth preserved quantum state component ψ j The change in j is the jth high-influence quantum state component retained after pruning, Prune(ψ j ) represents the quantum state component ψ retained after pruning j , β is the pruning adjustment coefficient;

[0026] S35. Based on the quantum state change results after pruning and analysis, extract the key eigenvectors related to the size of the micro-spring gasket, where each eigenvalue corresponds to a parameter that has a significant contribution to the quantum state change, and the selection criteria are based on those input parameters that cause large changes or significant phase differences;

[0027] S36. Standardize the extracted key feature vectors so that each feature value meets a specific range standard. The standardized feature vectors are used in subsequent model training and prediction processes.

[0028] Optionally, S4 includes the following steps:

[0029] S41, inputting the key feature vector extracted based on the high-dimensional quantum state perturbation algorithm into the input layer of the self-consistent field neural network model, and setting the initial weights and bias items of the network by random initialization;

[0030] S42, the feature vector is linearly transformed in each layer of neurons through the set weight matrix and bias term, and processed by the nonlinear activation function:

[0031]

[0032] Among them, a l is the output vector of the lth layer, a l-1 is the output of the previous layer, W l and b l are the weight matrix and bias term of the lth layer respectively, is a nonlinear activation function, is the quantum state change after pruning, V l is the specific adjustment vector in each layer;

[0033] S43. During the training of the network, the Hessian-Free optimization algorithm is used to perform multiple iterative calculations. In each iteration, the Hessian-Free optimization algorithm dynamically adjusts the network parameters by approximately calculating the Hessian matrix and combining the quantum state change:

[0034]

[0035] in, is the updated weight, H l is the approximation of the Hessian matrix of the lth layer, λ is the regularization parameter, I is the identity matrix, is the gradient of the loss function, γ is the quantum state adjustment coefficient;

[0036] S44. Through multiple iterative calculations, the Hessian-Free optimization algorithm gradually adjusts the network parameters so that the self-consistent field distribution of the system reaches a balanced state, extracts features with high stability, and is used to accurately predict the size of the micro spring gasket.

[0037] Optionally, S5 includes the following steps:

[0038] S51, inputting the real-time collected production data into a trained and optimized self-consistent field neural network model, wherein the input data includes pre-processed multi-dimensional feature vectors, which are used as initial inputs in the model to predict the size of the micro spring gasket in real time;

[0039] S52, after the input data enters the network, the feature vector passes through the weight matrix W of the current network l and the bias term b l Perform linear transformation and pass nonlinear activation function Process and output the result a l As the input of the next layer, it propagates layer by layer;

[0040] S53. During the data propagation process, the size of the micro spring gasket is predicted in real time based on the nonlinear chaotic prediction algorithm:

[0041]

[0042] Among them, x t+1 is the state of the system at the next moment, x t is the current state, f(x t ,W l ) is the state transfer function, η is the chaos intensity adjustment factor, γ is the adjustment coefficient, α i is the i-th input feature x i The sensitivity coefficient, represents the sensitivity of the quantum state to the i-th feature, is the quantum state change associated with the ith feature, V l is a specific regulation vector in the network, Indicates the effect of the phase difference between different quantum state components on the state change;

[0043] S54. In the nonlinear chaos prediction process, the enhanced sensitivity detection technology is used to analyze the system state in detail. Through a more sophisticated detection mechanism, the potential sensitivity changes in the production process are identified and amplified, thereby capturing the sensitivity of small initial condition changes to the system output and further optimizing the accuracy of size prediction;

[0044] S55. Based on the sensitivity changes captured by the enhanced sensitivity detection technology, the future changes of the size are accurately predicted and the prediction results are generated.

[0045] Optionally, the S6 comprises the following steps:

[0046] S61, constructing a multi-dimensional topological structure diagram according to the size prediction results generated by the self-consistent field neural network model and the nonlinear chaos prediction algorithm;

[0047] S62. The nodes and connection relationships in the multidimensional topological structure diagram are converted into a topological characteristic matrix T. By considering the interaction between nodes and the quantum state change, the elements of the topological characteristic matrix are:

[0048]

[0049] in, For node n i and n j The feature similarity measurement function between ij For node n i and n j The connection weight between ij is the adjustment coefficient, controlling the impact of nonlinear terms, β ij is the attenuation coefficient related to the node distance, γ ij is the regulation factor related to the quantum state change, Δψ ij For node n i and n j The quantum state change between

[0050] S63. Introduce quantum topological interference control mechanism into topological protection control technology, optimize the generated topological characteristic matrix T, simulate quantum interference effect, perform parallel calculations between multiple possible adjustment paths, and select the optimal path based on quantum state superposition:

[0051]

[0052] Among them, α k is the adjustment coefficient of each path, Interference(T k ,Q k ) is the quantum interference function, simulating the topological characteristic matrix T corresponding to path k k With the quantum state matrix Q k The interference effect, β is the adjustment coefficient of gradient adjustment, Represents the gradient calculation of the topological feature matrix T, represents the total amount of quantum state change after pruning, Represents the contribution of the phase difference of quantum state change between nodes i and j to the system impact;

[0053] S64, based on the optimized adjustment vector P adjust , multi-dimensional topology protection control technology is used to adjust the processing parameters of production equipment in real time.

[0054] Optionally, the S7 includes the following steps:

[0055] S71, comparing the size of the micro spring gasket actually produced with the prediction results generated by the self-consistent field neural network model and the nonlinear chaotic prediction algorithm to generate a size error matrix D, where D represents the difference between the actual size and the predicted size under different production conditions;

[0056] S72. Perform hierarchical analysis on the dimension error matrix D, construct a multi-level information entropy model, and divide the various parameters in the production process into different levels. Each level corresponds to an information entropy value H. l ,The information entropy value represents the uncertainty and variation range of the parameters at this level;

[0057] S73. Introduce a real-time anomaly detection function in the hierarchical entropy balance mechanism to monitor the changes in information entropy at each level in real time, identify and locate abnormal fluctuations in the system, and immediately trigger the entropy correction mechanism when an abnormality is detected in the information entropy of a certain level to correct the parameters of that level, so that the system can quickly restore balance under abnormal circumstances;

[0058] S74. Correct and optimize the abnormal information entropy based on real-time abnormality detection and entropy correction mechanism:

[0059]

[0060] Among them, H l ′ is the corrected information entropy value, α l is the adjustment factor, is the gradient calculation of the error matrix D, γ is the adjustment factor related to the quantum state change, D i is the i-th element in the error matrix, Δψ ij is the quantum state change between nodes i and j, β l To correct the adjustment factor, is the quantum interference function;

[0061] S75. Further adjust the production system through the hierarchical information entropy model after real-time anomaly detection and entropy correction, so that the information entropy of each level reaches the optimal balance state, especially in the case of abnormal fluctuations, so that the system can recover quickly and maintain stability;

[0062] S76. Apply the optimized hierarchical entropy balance mechanism to the production system, adjust the processing parameters of the production equipment in real time, and optimize the difference between the actual production size and the predicted size.

[0063] The beneficial effects of the present invention are:

[0064] (1) The present invention significantly improves the accuracy and stability of the size control of micro spring washers by combining a high-dimensional quantum state disturbance feature extraction method, a nonlinear chaos prediction algorithm, and a self-consistent field neural network model. In particular, in terms of data analysis and real-time adjustment in complex production environments, the present invention overcomes the limitations of traditional methods, can accurately extract key features related to size, and achieve high-precision control of the production process.

[0065] (2) The present invention optimizes the real-time adjustment of production equipment processing parameters and the dynamic balance control of the system by introducing multi-dimensional topological protection control technology and hierarchical entropy balance mechanism, and improves the response speed and adaptability of the system in the face of variable production conditions. At the same time, combined with the real-time anomaly detection function, the present invention can quickly identify and correct abnormal situations in the production process, ensuring the continuous and stable operation of the production system.

[0066] (3) The present invention provides an efficient, accurate and adaptive micro spring gasket size control solution by comprehensively applying a variety of advanced analysis and control technologies, enabling the system to maintain high-precision size control in a complex production environment, automatically handle size deviations, and optimize parameter adjustments during the production process, thereby achieving efficient production and quality assurance of micro components. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0068] Figure 1 A flowchart of a method for controlling the size of a micro spring gasket based on neural network prediction proposed by the present invention;

[0069] Figure 2 A schematic diagram of the high-dimensional quantum state disturbance feature extraction method proposed in the present invention;

[0070] Figure 3 Schematic diagram of the hierarchical entropy balancing mechanism and real-time anomaly detection and correction proposed in the present invention. DETAILED DESCRIPTION

[0071] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0072] refer to Figure 1-3 , a method for controlling the size of a micro spring gasket based on neural network prediction, comprising the following steps:

[0073] S1. Monitor the production process of micro spring washers in real time, collect and form multidimensional data sets;

[0074] S2, preprocessing the collected multidimensional data sets, eliminating noise and redundant information, and making the data adapt to the input requirements of the self-consistent field neural network through standardization and normalization;

[0075] S3, using a high-dimensional quantum state perturbation feature extraction method to represent the preprocessed data set as a high-dimensional quantum state, analyzing the change of the quantum state by applying a small perturbation, and extracting key features related to the size of the micro spring gasket;

[0076] S4. Based on the key features extracted by the high-dimensional quantum state perturbation algorithm, a self-consistent field neural network model is constructed. Through multiple iterations of calculating the self-consistent field distribution of the system, the network parameters are adjusted to make the data propagation in the network reach a balanced state, thereby extracting stable features and making accurate predictions;

[0077] S5. Input the real-time collected production data into the trained self-consistent field neural network model, predict the size of the micro spring gasket based on the nonlinear chaos prediction algorithm, identify the chaotic characteristics in the system and establish a corresponding prediction model by analyzing the nonlinear dynamic behavior in the production process, capture the sensitivity of small initial condition changes to the system output, and use the sensitive characteristics to accurately predict the future changes in size;

[0078] S6. Based on the prediction results of self-consistent field neural network and nonlinear chaos prediction algorithm, multi-dimensional topological protection control technology is used to adjust the processing parameters of production equipment in real time;

[0079] S7. Compare the actual size of the micro spring gasket produced with the predicted result, and use the hierarchical entropy balance mechanism to further optimize the system. By analyzing and adjusting the information entropy of each level, the system can reach the optimal balance state.

[0080] In this implementation, S3 includes the following steps:

[0081] S31, using a high-dimensional quantum state perturbation feature extraction method to represent the preprocessed multidimensional data set as a high-dimensional quantum state;

[0082] S32. Apply a small disturbance to the constructed high-dimensional quantum state, where the disturbance is expressed as a change in the quantum state, and its magnitude depends on the range of changes of various parameters in the production data;

[0083] S33. Introduce adaptive quantum state pruning technology to dynamically screen and prune the high-dimensional quantum state after the disturbance. By analyzing the contribution of each component in the quantum state to the final characteristics, remove the components that contribute less to the quantum state change and retain the quantum state components with high influence.

[0084] S34, analyzing the quantum state change after the adaptive quantum state pruning, calculating the amplitude and phase difference of the quantum state change, and identifying the influence of each input parameter on the quantum state change:

[0085]

[0086] Among them, Δψ total represents the overall change of the quantum state after pruning, α i is the sensitivity coefficient of the quantum state to the i-th parameter, represents the quantum state ψ(X) relative to the i-th input parameter x i The partial derivative of δx i is the perturbation of the ith parameter, represents the quantum state ψ(X) relative to the jth preserved quantum state component ψ j The change in j is the jth high-influence quantum state component retained after pruning, Prune(ψ j ) represents the quantum state component ψ retained after pruning j , β is the pruning adjustment coefficient;

[0087] S35. Based on the quantum state change results after pruning and analysis, extract the key eigenvectors related to the size of the micro-spring gasket, where each eigenvalue corresponds to a parameter that has a significant contribution to the quantum state change, and the selection criteria are based on those input parameters that cause large changes or significant phase differences;

[0088] S36. Standardize the extracted key feature vectors so that each feature value meets a specific range standard. The standardized feature vectors are used in subsequent model training and prediction processes.

[0089] In this implementation manner, S4 includes the following steps:

[0090] S41, inputting the key feature vector extracted based on the high-dimensional quantum state perturbation algorithm into the input layer of the self-consistent field neural network model, and setting the initial weights and bias items of the network by random initialization;

[0091] S42, the feature vector is linearly transformed in each layer of neurons through the set weight matrix and bias term, and processed by the nonlinear activation function:

[0092]

[0093] Among them, a l is the output vector of the lth layer, a l-1 is the output of the previous layer, W l and b l are the weight matrix and bias term of the lth layer respectively, is a nonlinear activation function, is the quantum state change after pruning, V l is the specific adjustment vector in each layer;

[0094] S43. During the training of the network, the Hessian-Free optimization algorithm is used to perform multiple iterative calculations. In each iteration, the Hessian-Free optimization algorithm dynamically adjusts the network parameters by approximately calculating the Hessian matrix and combining the quantum state change:

[0095]

[0096] in, is the updated weight, H l is the approximation of the Hessian matrix of the lth layer, λ is the regularization parameter, I is the identity matrix, is the gradient of the loss function, γ is the quantum state adjustment coefficient;

[0097] S44. Through multiple iterative calculations, the Hessian-Free optimization algorithm gradually adjusts the network parameters so that the self-consistent field distribution of the system reaches a balanced state, extracts features with high stability, and is used to accurately predict the size of the micro spring gasket.

[0098] In this implementation, S5 includes the following steps:

[0099] S51, inputting the real-time collected production data into a trained and optimized self-consistent field neural network model, wherein the input data includes pre-processed multi-dimensional feature vectors, which are used as initial inputs in the model to predict the size of the micro spring gasket in real time;

[0100] S52, after the input data enters the network, the feature vector passes through the weight matrix W of the current network l and the bias term b l Perform linear transformation and pass nonlinear activation function Process and output the result a l As the input of the next layer, it propagates layer by layer;

[0101] S53. During the data propagation process, the size of the micro spring gasket is predicted in real time based on the nonlinear chaotic prediction algorithm:

[0102]

[0103] Among them, x t+1 is the state of the system at the next moment, x t is the current state, f(x t ,W l ) is the state transfer function, η is the chaos intensity adjustment factor, γ is the adjustment coefficient, α i is the i-th input feature x i The sensitivity coefficient, represents the sensitivity of the quantum state to the i-th feature, is the quantum state change associated with the ith feature, V l is a specific regulation vector in the network, Indicates the effect of the phase difference between different quantum state components on the state change;

[0104] S54. In the nonlinear chaos prediction process, the enhanced sensitivity detection technology is used to analyze the system state in detail. Through a more sophisticated detection mechanism, the potential sensitivity changes in the production process are identified and amplified, thereby capturing the sensitivity of small initial condition changes to the system output and further optimizing the accuracy of size prediction;

[0105] S55. Based on the sensitivity changes captured by the enhanced sensitivity detection technology, the future changes of the size are accurately predicted and the prediction results are generated.

[0106] In this implementation manner, S6 includes the following steps:

[0107] S61, constructing a multi-dimensional topological structure diagram according to the size prediction results generated by the self-consistent field neural network model and the nonlinear chaos prediction algorithm;

[0108] S62. The nodes and connection relationships in the multidimensional topological structure diagram are converted into a topological characteristic matrix T. By considering the interaction between nodes and the quantum state change, the elements of the topological characteristic matrix are:

[0109]

[0110] in, For node n i and n j The feature similarity measurement function between ij For node n i and n j The connection weight between ij is the adjustment coefficient, controlling the impact of nonlinear terms, β ij is the attenuation coefficient related to the node distance, γ ij is the regulation factor related to the quantum state change, Δψ ijFor node n i and n j The quantum state change between

[0111] S63. Introduce quantum topological interference control mechanism into topological protection control technology, optimize the generated topological characteristic matrix T, simulate quantum interference effect, perform parallel calculations between multiple possible adjustment paths, and select the optimal path based on quantum state superposition:

[0112]

[0113] Among them, α k is the adjustment coefficient of each path, Interference(T k ,Q k ) is the quantum interference function, simulating the topological characteristic matrix T corresponding to path k k With the quantum state matrix Q k The interference effect, β is the adjustment coefficient of gradient adjustment, Represents the gradient calculation of the topological feature matrix T, represents the total amount of quantum state change after pruning, Represents the contribution of the phase difference of quantum state change between nodes i and j to the system impact;

[0114] S64, based on the optimized adjustment vector P adjust , multi-dimensional topology protection control technology is used to adjust the processing parameters of production equipment in real time.

[0115] In this implementation manner, the S7 includes the following steps:

[0116] S71, comparing the size of the micro spring gasket actually produced with the prediction results generated by the self-consistent field neural network model and the nonlinear chaotic prediction algorithm to generate a size error matrix D, where D represents the difference between the actual size and the predicted size under different production conditions;

[0117] S72. Perform hierarchical analysis on the dimension error matrix D, construct a multi-level information entropy model, and divide the various parameters in the production process into different levels. Each level corresponds to an information entropy value H. l ,The information entropy value represents the uncertainty and variation range of the parameters at this level;

[0118] S73. Introduce a real-time anomaly detection function in the hierarchical entropy balance mechanism to monitor the changes in information entropy at each level in real time, identify and locate abnormal fluctuations in the system, and immediately trigger the entropy correction mechanism when an abnormality is detected in the information entropy of a certain level to correct the parameters of that level, so that the system can quickly restore balance under abnormal circumstances;

[0119] S74. Correct and optimize the abnormal information entropy based on real-time abnormality detection and entropy correction mechanism:

[0120]

[0121] Among them, H l ′ is the corrected information entropy value, α l is the adjustment factor, is the gradient calculation of the error matrix D, γ is the adjustment factor related to the quantum state change, D i is the i-th element in the error matrix, Δψ ij is the quantum state change between nodes i and j, β l To correct the adjustment factor, is the quantum interference function;

[0122] S75. Further adjust the production system through the hierarchical information entropy model after real-time anomaly detection and entropy correction, so that the information entropy of each level reaches the optimal balance state, especially in the case of abnormal fluctuations, so that the system can recover quickly and maintain stability;

[0123] S76. Apply the optimized hierarchical entropy balance mechanism to the production system, adjust the processing parameters of the production equipment in real time, and optimize the difference between the actual production size and the predicted size.

[0124] Embodiment 1:

[0125] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a precision manufacturing plant located in Munich, Germany, which mainly produces miniature spring washers for high-end mechanical equipment. Since the dimensional accuracy requirements of these washers are extremely high, any dimensional deviation will directly affect the overall performance of the mechanical equipment. In the past, the factory used traditional processes for production, relying on fixed process parameter settings, and controlling product quality through simple inspections during the production process. However, this method exposes many problems in practical applications, especially under high-precision production requirements, the defective rate is high, resulting in serious impact on production efficiency and product quality. In this embodiment, we selected a miniature spring washer size control method based on neural network prediction of the present invention, and applied it to the production line of the factory, in the hope of improving the product qualification rate and production efficiency by real-time monitoring and adjustment of production parameters.

[0126] First, the production line is equipped with multiple high-precision sensors, which are responsible for real-time collection of multi-dimensional data of micro spring washers during the production process, including temperature, pressure, raw material thickness, and cutting speed. In the key processing link, we also arranged a laser measuring instrument to accurately measure the size of each washer. The collected multi-dimensional data is input into the system, and the system pre-processes the data, including noise removal, standardization, and normalization, to ensure the accuracy and consistency of the data.

[0127] Next, the system used a high-dimensional quantum state perturbation feature extraction method to extract key features closely related to the gasket size from the preprocessed data set. These features were input into the self-consistent field neural network model, which generated a real-time prediction of the gasket size during the production process through multiple iterative calculations. During the prediction process, the system applied a nonlinear chaos prediction algorithm to analyze the nonlinear dynamic behavior that may occur in the production process in real time, and adjusted the processing parameters of the production equipment according to the prediction results.

[0128] For example, when the system predicts that a batch of gaskets may have dimensional deviations, the system immediately adjusts the cutting speed or material feed rate to ensure that the final product meets the design requirements. In addition, through the layered entropy balance mechanism, the system further optimizes the entire production process to ensure that each production link is in the best state.

[0129] In the three months from June to August 2023, we conducted a detailed comparative analysis of the production data before and after the implementation of the invention. Before implementation, the factory produced about 1 million micro spring washers per month, with a pass rate of 94%, and about 60,000 washers were scrapped each month due to unqualified dimensions. The main reason for defective products is excessive dimensional deviation, especially when producing high-precision washers, the pass rate is only 90%.

[0130] Table 1: Comparison of data before and after implementation

[0131]

[0132] After implementing the dimension control method of the present invention, the performance of the production line improved significantly. In the first month, the product qualification rate increased from 94% to 96%, and the number of defective products decreased to 40,000 pieces. In the second month, the qualification rate further increased to 97.5%, and the number of defective products decreased to 25,000 pieces. In the third month, the qualification rate stabilized at more than 98%, and the number of defective products decreased to less than 20,000 pieces.

[0133] The effect is particularly significant when producing gaskets with high precision requirements. After the implementation of the present invention, the qualified rate of high-precision gaskets increased from 90% to 96%, and the number of defective products decreased from 100,000 pieces to 40,000 pieces. At the same time, production efficiency was also improved, and the average production time of each batch of gaskets was shortened by about 15%.

[0134] Through data analysis, we found that the present invention effectively solved the problem of high defective rate and low qualified rate caused by the difficulty in real-time adjustment of production parameters in traditional processes, and significantly improved the overall performance of the production line. This not only improves product quality, but also provides strong technical support for the field of high-precision manufacturing and ensures the market competitiveness of products.

[0135] 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 method for controlling the size of a micro spring washer based on neural network prediction, characterized in that: The steps include: S1. Monitor the production process of micro spring washers in real time, collect and form multidimensional data sets; S2, preprocessing the collected multidimensional data sets, eliminating noise and redundant information, and making the data adapt to the input requirements of the self-consistent field neural network through standardization and normalization; S3, using a high-dimensional quantum state perturbation feature extraction method to represent the preprocessed data set as a high-dimensional quantum state, analyzing the change of the quantum state by applying a small perturbation, and extracting key features related to the size of the micro spring gasket; S4. Based on the key features extracted by the high-dimensional quantum state perturbation algorithm, a self-consistent field neural network model is constructed. Through multiple iterations of calculating the self-consistent field distribution of the system, the network parameters are adjusted to make the data propagation in the network reach a balanced state, thereby extracting stable features and making accurate predictions; S5, inputting the real-time collected production data into the trained self-consistent field neural network model, and accurately predicting the size of the micro spring gasket based on the nonlinear chaos prediction algorithm; S6. Based on the prediction results of self-consistent field neural network and nonlinear chaos prediction algorithm, multi-dimensional topological protection control technology is used to adjust the processing parameters of production equipment in real time; S7. Compare the actual size of the micro spring gasket produced with the predicted result, and use the hierarchical entropy balance mechanism to further optimize the system. By analyzing and adjusting the information entropy of each level, the system can reach the optimal balance state.

2. The method for controlling the size of a micro spring washer based on neural network prediction according to claim 1, characterized in that: The S3 specifically includes: S31, using a high-dimensional quantum state perturbation feature extraction method to represent the preprocessed multidimensional data set as a high-dimensional quantum state; S32. Apply a small disturbance to the constructed high-dimensional quantum state, where the disturbance is expressed as a change in the quantum state, and its magnitude depends on the range of changes of various parameters in the production data; S33. Introduce adaptive quantum state pruning technology to dynamically screen and prune the high-dimensional quantum state after the disturbance. By analyzing the contribution of each component in the quantum state to the final characteristics, remove the components that contribute less to the quantum state change and retain the quantum state components with high influence. S34, analyzing the quantum state change after the adaptive quantum state pruning, calculating the amplitude and phase difference of the quantum state change, and identifying the influence of each input parameter on the quantum state change: Among them, Δψ total represents the overall change of the quantum state after pruning, α i is the sensitivity coefficient of the quantum state to the i-th parameter, represents the quantum state ψ(X) relative to the i-th input parameter x i The partial derivative of δx i is the perturbation of the ith parameter, represents the quantum state ψ(X) relative to the jth preserved quantum state component ψ j The change in j is the jth high-influence quantum state component retained after pruning, Prune(ψ j ) represents the quantum state component ψ retained after pruning j , β is the pruning adjustment coefficient; S35. Based on the quantum state change results after pruning and analysis, the key feature vectors related to the size of the micro spring gasket are extracted; S36. Standardize the extracted key feature vectors so that each feature value meets a specific range standard. The standardized feature vectors are used in subsequent model training and prediction processes.

3. The method for controlling the size of a micro spring washer based on neural network prediction according to claim 1, characterized in that: The S4 specifically includes: S41, inputting the key feature vector extracted based on the high-dimensional quantum state perturbation algorithm into the input layer of the self-consistent field neural network model, and setting the initial weights and bias items of the network by random initialization; S42, the feature vector is linearly transformed in each layer of neurons through the set weight matrix and bias term, and processed by the nonlinear activation function: Among them, a l is the output vector of the lth layer, a l-1 is the output of the previous layer, W l and b l are the weight matrix and bias term of the lth layer respectively, is a nonlinear activation function, is the quantum state change after pruning, V l is the specific adjustment vector in each layer; S43. During the training of the network, the Hessian-Free optimization algorithm is used to perform multiple iterative calculations. In each iteration, the Hessian-Free optimization algorithm dynamically adjusts the network parameters by approximately calculating the Hessian matrix and combining the quantum state change: in, is the updated weight, H l is the approximation of the Hessian matrix of the lth layer, λ is the regularization parameter, I is the identity matrix, is the gradient of the loss function, γ is the quantum state adjustment coefficient; S44. Through multiple iterative calculations, the Hessian-Free optimization algorithm gradually adjusts the network parameters, extracts features with high stability, and is used to accurately predict the size of micro spring washers.

4. The method for controlling the size of micro spring washers based on neural network prediction according to claim 1, characterized in that: The S5 specifically includes: S51, inputting the real-time collected production data into the trained and optimized self-consistent field neural network model; S52, after the input data enters the network, the feature vector passes through the weight matrix W of the current network l and the bias term b l Perform linear transformation and pass nonlinear activation function Process and output the result a l As the input of the next layer, it propagates layer by layer; S53. During the data propagation process, the size of the micro spring gasket is predicted in real time based on the nonlinear chaotic prediction algorithm: Among them, x t+1 is the state of the system at the next moment, x t is the current state, f(x t ,W l ) is the state transfer function, η is the chaos intensity adjustment factor, γ is the adjustment coefficient, α i is the i-th input feature x i The sensitivity coefficient, represents the sensitivity of the quantum state to the i-th feature, is the quantum state change associated with the i-th feature, V l is a specific regulation vector in the network, Indicates the effect of the phase difference between different quantum state components on the state change; S54. In the nonlinear chaos prediction process, the enhanced sensitivity detection technology is used to analyze the system state in detail. Through a more sophisticated detection mechanism, the potential sensitivity changes in the production process are identified and amplified, thereby capturing the sensitivity of small initial condition changes to the system output and further optimizing the accuracy of size prediction; S55. Based on the sensitivity changes captured by the enhanced sensitivity detection technology, the future changes of the size are accurately predicted and the prediction results are generated.

5. The method for controlling the size of micro spring washers based on neural network prediction according to claim 1, characterized in that: The S6 specifically includes: S61, constructing a multi-dimensional topological structure diagram according to the size prediction results generated by the self-consistent field neural network model and the nonlinear chaos prediction algorithm; S62. The nodes and connection relationships in the multidimensional topological structure diagram are converted into a topological characteristic matrix T. By considering the interaction between nodes and the quantum state change, the elements of the topological characteristic matrix are: in, For node n i and n j The feature similarity measurement function between ij For node n i and n j The connection weight between ij is the adjustment coefficient, controlling the impact of nonlinear terms, β ij is the attenuation coefficient related to the node distance, γ ij is the regulation factor related to the quantum state change, Δψ ij For node n i and n j The quantum state change between S63. Introduce quantum topological interference control mechanism into topological protection control technology, optimize the generated topological characteristic matrix T, simulate quantum interference effect, perform parallel calculations between multiple adjustment paths, and select the optimal path based on quantum state superposition: Among them, α k is the adjustment coefficient of each path, Interference(T k ,Q k ) is the quantum interference function, simulating the topological characteristic matrix T corresponding to path k k With the quantum state matrix Q k The interference effect, β is the adjustment coefficient of gradient adjustment, Represents the gradient calculation of the topological feature matrix T, represents the total amount of quantum state change after pruning, Represents the contribution of the phase difference of quantum state change between nodes i and j to the system impact; S64, based on the optimized adjustment vector P adjust , multi-dimensional topology protection control technology is used to adjust the processing parameters of production equipment in real time.

6. The method for controlling the size of micro spring washers based on neural network prediction according to claim 1, characterized in that: The S7 specifically includes: S71, comparing the size of the micro spring gasket actually produced with the prediction results generated by the self-consistent field neural network model and the nonlinear chaotic prediction algorithm to generate a size error matrix D, where D represents the difference between the actual size and the predicted size under different production conditions; S72. Perform hierarchical analysis on the dimension error matrix D, construct a multi-level information entropy model, and divide the various parameters in the production process into different levels. Each level corresponds to an information entropy value H. l ,The information entropy value represents the uncertainty and variation range of the parameters at this level; S73. Introduce a real-time anomaly detection function in the hierarchical entropy balance mechanism to monitor the changes in information entropy at each level in real time, identify and locate abnormal fluctuations in the system, and immediately trigger the entropy correction mechanism when an abnormality is detected in the information entropy of a certain level to correct the parameters of that level, so that the system can quickly restore balance under abnormal circumstances; S74. Correct and optimize the abnormal information entropy based on real-time abnormality detection and entropy correction mechanism: Among them, H l ′ is the corrected information entropy value, α l is the adjustment factor, is the gradient calculation of the error matrix D, γ is the adjustment factor related to the quantum state change, D i is the i-th element in the error matrix, Δψ ij is the quantum state change between nodes i and j, β l To correct the adjustment factor, is the quantum interference function; S75. Further adjust the production system through the hierarchical information entropy model after real-time anomaly detection and entropy correction, so that the information entropy of each level reaches the optimal balance state; S76. Apply the optimized hierarchical entropy balance mechanism to the production system, adjust the processing parameters of the production equipment in real time, and optimize the difference between the actual production size and the predicted size.

Images