Micro spring washer error correction method using fuzzy logic algorithm

Abstract

The invention discloses a micro spring washer error correction method using a fuzzy logic algorithm. The method comprises the following steps: S1, acquiring a geometric dimension error, a stress distribution error and a material characteristic error; s2, generating a preliminary error correction strategy by using a nonlinear fuzzy residual transfer learning algorithm; s3, real-time entropy change is monitored; s4, adjusting a correction strategy by adopting an adaptive entropy regulation dynamic mapping algorithm; s5, carrying out correction operation, and adjusting the geometrical shape and stress distribution; s6, performing error evaluation and recording data; s7, updating algorithm parameters; and S8, repeating the steps to improve the correction precision and consistency. According to the invention, the manufacturing precision and reliability of the micro mechanical element are effectively improved.

Description

Technical Field

[0001] The invention relates to the technical field of error correction, and in particular to a micro spring washer error correction method using a fuzzy logic algorithm. Background Art

[0002] In the field of precision manufacturing, micromechanical elements such as micro spring washers have become key components in many high-precision devices due to their small size and high functional requirements. However, the manufacturing and assembly process of micro spring washers is extremely complex, which places strict requirements on manufacturing accuracy. In actual production, micro spring washers are often affected by multiple factors such as geometry, material properties, and stress distribution, which makes them prone to errors during the production process. Therefore, how to efficiently and accurately correct errors has become a major challenge in the current production of micro spring washers.

[0003] First, traditional error correction methods usually rely on deterministic mathematical models that perform corrections based on fixed rules and preset error characteristics. Although these methods are relatively effective in dealing with simple errors, they are often unable to cope with complex, multi-dimensional errors. Since micro spring washers are affected by multiple variables during the production process, the error characteristics often show high nonlinearity and uncertainty. Traditional methods are difficult to adapt to such complex situations, resulting in unsatisfactory correction results.

[0004] Secondly, current error correction technologies often lack adaptability when facing dynamic changes in the production environment. The production environment of micro spring washers is complex and changeable, and various environmental factors (such as temperature, pressure, stress, etc.) have a profound impact on the error correction process. Traditional correction methods cannot effectively adjust the correction strategy to adapt to environmental changes due to the lack of monitoring and analysis of real-time environmental data. This limitation leads to the accumulation of errors in the production process, which seriously affects the manufacturing accuracy and product consistency of micro spring washers.

[0005] Third, traditional error correction methods usually ignore the dynamic monitoring and analysis of environmental parameters. Changes in temperature, humidity, pressure, etc. in the production environment will affect the production process of micro spring washers, and these factors are often not fully considered by traditional methods. In addition, the traditional linear error correction model cannot effectively capture the complex nonlinear relationship between the error and environmental parameters, which makes the correction result deviate greatly and it is difficult to achieve the expected effect.

[0006] In summary, the existing technology has many deficiencies in error correction of micro spring washers: first, traditional correction methods lack the ability to handle complex errors; second, it is difficult to achieve adaptive correction in a dynamic environment; third, it fails to fully utilize real-time environmental data for error analysis and adjustment. These problems make it difficult to ensure the production accuracy and consistency of micro spring washers, limiting their performance in high-precision applications. Therefore, how to provide a micro spring washer error correction method using fuzzy logic algorithm is a problem that technicians in this field urgently need to solve. Summary of the invention

[0007] One object of the present invention is to propose a micro spring washer error correction method based on nonlinear fuzzy residual transfer learning and adaptive entropy control dynamic mapping algorithm. The present invention makes full use of the nonlinear fuzzy residual transfer learning algorithm and the adaptive entropy control dynamic mapping algorithm, and dynamically adjusts the error correction strategy by analyzing multidimensional error information and monitoring real-time entropy values, thereby achieving high-precision correction of the geometric shape, stress distribution and material properties of the micro spring washer, and has the advantages of high correction accuracy, strong adaptability and high production stability.

[0008] A micro spring washer error correction method using a fuzzy logic algorithm according to an embodiment of the present invention comprises the following steps:

[0009] S1. Collect multi-dimensional error information, including geometric size error, stress distribution error and material property error, and generate a multi-dimensional error data matrix;

[0010] S2, using nonlinear fuzzy residual transfer learning algorithm, based on the historical residual model and current multidimensional error data, to generate a preliminary error correction strategy;

[0011] S3, monitor the real-time entropy value changes in the manufacturing system, and provide a basis for evaluating the stability of the production environment and error fluctuations;

[0012] S4. Based on the change of entropy value, an adaptive entropy control dynamic mapping algorithm is used to adjust the execution path of the error correction strategy to adapt to the dynamic changes in the production environment;

[0013] S5. According to the adjusted error correction strategy, a correction operation is performed to adjust the geometry, stress distribution and material properties of the micro spring washer;

[0014] S6. After the correction operation, perform error evaluation and record the corrected error data;

[0015] S7. Based on the evaluation results, update the parameters of the nonlinear fuzzy residual transfer learning algorithm and the adaptive entropy control dynamic mapping algorithm to provide an optimization basis for the error correction strategy;

[0016] S8. Repeat steps S1 to S7 to achieve error correction in each production batch.

[0017] Optionally, S2 includes the following steps:

[0018] S21. Collect various error data in historical production batches, such as geometric dimension error R Gi , stress distribution error R Si , Material property error R Mi , construct the historical residual matrix R hist :

[0019]

[0020] Calculate the error for the i-th batch: in, represents the actual measurement value of the i-th batch; M target represents the target design value; ξ i Indicates measurement noise or environmental disturbance, affecting measurement accuracy;

[0021] S22. Collect geometric dimension errors in real time in the current production batch Stress distribution error Material property error Construct a multidimensional error matrix E curr :

[0022]

[0023] Calculate the error value: in, represents the actual measurement value of the current batch, ζ j Represents noise or interference in real-time measurements, caused by equipment or environmental changes;

[0024] S23, establish a fuzzy logic system to handle complex error relationships through fuzzy membership functions, fuzzy rule bases and reasoning mechanisms;

[0025] Fuzzy membership function:

[0026]

[0027] Among them, c i Represents the center of the membership function and determines the peak position of the membership function; σ i It indicates the width of the membership function, controls the extension range of the membership function, and affects the degree of fuzzification;

[0028] Fuzzy rule base: Build a fuzzy rule base based on the relationship between historical data and current errors For reasoning;

[0029] S24. Based on the fuzzy logic system, a nonlinear fuzzy residual transfer function F is constructed. NLFRT ; Combining the historical residual matrix and the current error matrix, the residual transfer matrix T is obtained through fuzzy reasoning res :

[0030]

[0031] Among them, μ k Represents the fuzzy membership coefficient, dynamically adjusting the influence weight of the fuzzy rule in error transmission; Q k Represents the weight matrix, which adjusts the importance of residuals in different dimensions; σ k is the width of the Gaussian kernel function, which adjusts the blur degree of the kernel function to adapt to different error scales; A k It is a nonlinear adaptive matrix used to capture the complex dynamic behavior between residual and error, and is processed in combination with the nonlinear activation function tanh;

[0032] S25, based on the nonlinear fuzzy residual transfer function, calculate the adaptive weight matrix W dyn :

[0033]

[0034] Among them, I is the unit matrix, which is used to maintain the matrix reversibility and avoid numerical instability; α is the regularization parameter, which balances the flexibility and stability of the weight matrix to prevent overfitting;

[0035] S26. Generate a preliminary error correction strategy C using an adaptive weight matrix init , and adjusted by nonlinear optimization:

[0036]

[0037] Among them, Δ is the optimization correction vector, which is used to further adjust the correction strategy; λ is the regularization parameter, which controls the smoothness of the optimization process; Δ i is the i-th component of the correction vector Δ;

[0038] S27, generate final correction vector C final and applied to production;

[0039] S28. Based on production feedback, dynamically update parameters in the fuzzy logic system and the nonlinear fuzzy residual transfer function to optimize the error correction strategy for future production batches.

[0040] Optionally, S3 includes the following steps:

[0041] S31. During the production process, the multi-dimensional physical parameters in the manufacturing system are collected through a real-time sensor network to build a real-time data matrix D real (t), and preprocess to remove high-frequency noise to generate the processed data matrix D proc (t);

[0042] S32, based on the processed data matrix D proc (t), calculate the real-time entropy value H(t) of the manufacturing system to quantify the uncertainty and stability in the production process:

[0043]

[0044] Among them, p i (t) represents the processed data matrix D proc The probability distribution of the i-th data dimension in (t) at time t;

[0045] S33. Perform time series analysis on the entropy value H(t) and calculate the rate of change To identify dynamic fluctuations in the manufacturing process;

[0046] S34, the real-time entropy value H(t) and the preliminary error correction strategy C init Coupled, a dynamic feedback mechanism is constructed to adjust the error correction strategy of the current batch to adapt to the real-time changing production environment;

[0047] S35. Setting the warning threshold H of entropy value change th , when the real-time entropy value When the change exceeds this threshold, the error fluctuation alarm is triggered, indicating instability in the manufacturing system. At this time, the error correction strategy will be automatically adjusted to avoid production deviations caused by environmental fluctuations.

[0048] S36. Feedback the information of real-time entropy value changes to the nonlinear fuzzy residual transfer learning model, dynamically adjust the parameters of the model, and optimize the real-time error correction strategy to make it better adapt to changes in the environment and state during the manufacturing process.

[0049] Optionally, S4 includes the following steps:

[0050] S41. During the production process, the key environmental parameters in the manufacturing system are monitored through a real-time sensor network, multimodal environmental data is generated, the environmental state matrix S(t) is constructed, and the corresponding real-time entropy value H(t) and change rate are calculated.

[0051] S42. Use multimodal environmental data to construct a multi-level environmental state matrix S(t), and perform high-dimensional tensor decomposition to capture the characteristics and relationships of different levels:

[0052]

[0053] Among them, S k (t) represents the k-th level feature matrix of the environment state, which is a multi-dimensional feature extracted by tensor decomposition; is the decomposition coefficient in tensor decomposition, which represents the relationship weight between the k-th layer feature and the original data; Φ i It is a feature matrix representing the state of the environment, which contains the core features of multimodal data;

[0054] S43, couple the real-time entropy value H(t) with the multimodal environment state tensor S(t) to construct a composite mapping model Used to capture complex interactive effects in multi-dimensional dynamic environments:

[0055]

[0056] Among them, θ ijk Coupling coefficient reflects the interaction between different modal characteristics; U i ,V j ,W k They represent the three-dimensional feature matrices obtained by tensor decomposition, reflecting the coupling relationship between different environmental states; It represents the contribution to the entropy value under different modes, combining the temporal dynamic information of each modal feature;

[0057] S44, based on composite mapping model Introducing adaptive learning mechanism By combining multidimensional convolution and recurrent networks, nonlinear modeling of the complex relationship between environmental state and entropy value changes is performed:

[0058]

[0059] Among them, W p is a multidimensional convolution kernel tensor used to extract feature patterns of different scales; * is a convolution operator, which represents the operation between the convolution kernel and the environment state tensor; σ is a nonlinear activation function; RNN is a recurrent neural network layer used to process the time dependency in sequence data; ρ q is the adaptive learning weight, which represents the learning contribution at different time steps;

[0060] S45. Combine the composite mapping model and adaptive learning with the real-time entropy value change analysis results to construct an adaptive entropy control dynamic mapping function Used to dynamically adjust the error correction strategy in real time C adj (t):

[0061]

[0062] in, represents the composite mapping result of the rth layer, capturing the complex interaction relationship in a multimodal environment; C adj (t) is the adjusted error correction strategy; γ l is a dynamic adjustment coefficient that controls the adjustment strength under different entropy value deviations; R l It is the multidimensional entropy value change moment, which is used to measure the entropy value change under different conditions;

[0063] S46, high-dimensional dynamic correction vector C generated by adaptive entropy control dynamic mapping function dyn (t), used to adjust key parameters in the manufacturing process in real time;

[0064] S47. Apply high-dimensional dynamic correction vector C dyn (t), dynamically adjust key parameters in the production process;

[0065] S48, comparing the high-dimensional error difference ΔE(t) before and after correction, feeding it back into the adaptive entropy control dynamic mapping function, and optimizing the tensor weight and learning rate;

[0066] S49. Regularly conduct multimodal analysis on entropy value changes and correction effects, and update adaptive entropy control.

[0067] Optionally, S5 includes the following steps:

[0068] S51, according to the generated high-dimensional dynamic correction vector C dyn (t), adjust the geometry of the micro spring washer;

[0069] S52. Using adaptive entropy to control dynamic mapping function The stress distribution of the micro spring washer is optimized based on the results, and the stress adjustment is performed based on the stress matrix monitored in real time.

[0070] S53, according to the generated high-dimensional dynamic correction vector C dyn (t), optimizing the material properties of the micro spring washers, especially the elastic modulus and hardness;

[0071] S54, integrate the adjustment results of geometry, stress distribution and material properties, and use the high-dimensional dynamic correction vector C dyn (t) carry out comprehensive adjustments;

[0072] S55. During the adjustment process, the geometry, stress distribution and material properties of the micro spring washer are monitored in real time, and the monitoring results are fed back to the adaptive entropy control dynamic mapping function. In order to further optimize the correction strategy in future production batches;

[0073] S56. Conduct a comprehensive performance evaluation on the adjusted micro spring washers, and further optimize the nonlinear fuzzy residual transfer learning model and the adaptive entropy control dynamic mapping function based on the evaluation results.

[0074] Optionally, the S6 comprises the following steps:

[0075] S61. After the correction operation is completed, collect the corrected multi-dimensional error data, including geometric size error Stress distribution error and material property errors Generate the corrected error data matrix E corr ;

[0076] S62, the corrected error data matrix E corr and the target design value matrix M target Compare and calculate the corrected error deviation matrix ΔE corr And the error deviation matrix ΔE is calculated by fuzzy logic inference system corr Conduct preliminary analysis;

[0077] S63, the error deviation matrix ΔE corr Input the dynamically adjusted nonlinear fuzzy logic system, and generate the adjusted correction strategy C through fuzzification processing and nonlinear activation function. corr ;

[0078] S64, compare the error deviation before and after correction, and generate the feedback matrix F corr , used to update the membership function and rule base in the fuzzy logic system;

[0079] S65, based on the feedback matrix F corr The results are optimized and the final correction strategy C is generated. fi , adjust key parameters through adaptive learning mechanisms to improve the accuracy and adaptability of the correction strategy;

[0080] S66, the final correction strategy C fi Applied to the production process, it monitors the correction effect in real time and records the corrected error data.

[0081] Optionally, the S7 includes the following steps:

[0082] S71, according to the corrected error data matrix E corr and error deviation ΔE corr , the core parameter θ of the nonlinear fuzzy residual transfer learning algorithm NLFRT To update:

[0083]

[0084] in, is the previous parameter set of the nonlinear fuzzy residual transfer learning algorithm, η is the learning rate, which is used to control the step size of parameter update, is the loss function The gradient of the parameter θ is used to minimize the error deviation ΔE corr And the corrected error matrix E corr The error between

[0085] S72, based on updated parameters Generate a new fuzzy residual matrix R new :

[0086]

[0087] Among them, R hist is the historical residual matrix, μ k is the fuzzy membership coefficient, which indicates the weight of the current fuzzy rule, σ k is the width of the Gaussian kernel function, which controls the fuzziness of the kernel function to the error. k is a nonlinear adaptive matrix used to capture the dynamic nonlinear relationship between residual and error. represents the residual matrix generated by the updated nonlinear fuzzy residual transfer learning algorithm;

[0088] S73, the generated fuzzy residual matrix R new As input, dynamically adjust the parameters γ of the adaptive entropy control dynamic mapping algorithm DM :

[0089]

[0090] in, is the previous entropy control parameter, β is the adjustment coefficient, which is used to control the amplitude of parameter adjustment, H(t) is the real-time entropy value at time t, reflecting the uncertainty in the production process, R new (t) is the fuzzy residual matrix value at time t, ∈ is a small positive number to prevent the denominator from being zero;

[0091] S74. Using the updated parameters and Generate optimized calibration strategy

[0092] S75, the optimized correction strategy It is applied to error correction operations in production batches, and the correction effect is continuously monitored through a real-time data feedback system. The parameters of the nonlinear fuzzy residual transfer learning algorithm and the adaptive entropy control dynamic mapping algorithm are further adjusted according to the feedback results.

[0093] The beneficial effects of the present invention are:

[0094] (1) The present invention proposes a micro spring washer error correction method based on nonlinear fuzzy residual transfer learning. A preliminary error correction strategy is generated by comprehensive analysis of historical residual data and current multi-dimensional error data. The correction strategy is monitored and adjusted in real time in combination with an adaptive entropy control dynamic mapping algorithm. This method effectively solves the problem of the existing technology that is difficult to handle complex and multi-dimensional errors, and greatly improves the accuracy of error correction. In particular, it can accurately control the geometric shape, stress distribution and material properties in a changing production environment, thereby improving the manufacturing accuracy and consistency of micro-mechanical components.

[0095] (2) The present invention introduces a real-time entropy value monitoring mechanism, which collects multi-dimensional environmental data in the manufacturing process in real time, calculates and analyzes the real-time entropy value changes of the system, and dynamically adjusts the error correction strategy according to the entropy value fluctuations. This ensures that the correction process can adaptively respond to the uncertainty and volatility in the production environment, thereby ensuring the production accuracy and quality stability of the micro spring washers, especially in complex and changeable manufacturing environments, showing significant advantages.

[0096] (3) The present invention couples complex environmental influencing factors with the error correction strategy through adaptive entropy control of the dynamic mapping function, so that the correction strategy can be dynamically adjusted according to the real-time changing environmental conditions during the production process, thereby avoiding the problem of correction error accumulation caused by changes in environmental factors in traditional methods, greatly improving the robustness and adaptability of error correction, and ensuring the high precision and high reliability of micro spring washers during the production process.

[0097] (4) The present invention constructs a highly adaptive error correction model by combining the collection of multi-dimensional error data with a nonlinear fuzzy residual transfer learning algorithm, which effectively solves the limitations of traditional error correction methods in processing complex error data, significantly improves the efficiency and accuracy of error correction, and provides a reliable technical guarantee for the high-precision manufacturing of micro-mechanical components. It can also adapt to various dynamically changing production environments, thus realizing the automation and intelligence of error correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] 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:

[0099] Figure 1 A flowchart of a micro spring washer error correction method using a fuzzy logic algorithm proposed by the present invention;

[0100] Figure 2A flowchart of generating a preliminary error correction strategy using a nonlinear fuzzy residual transfer learning algorithm proposed by the present invention;

[0101] Figure 3 This is a flow chart of the invention which proposes an adaptive entropy control dynamic mapping algorithm to adjust the error correction strategy based on the change of entropy value. DETAILED DESCRIPTION

[0102] 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.

[0103] refer to Figure 1-3 , a micro spring washer error correction method using fuzzy logic algorithm, comprising the following steps:

[0104] S1. Collect multi-dimensional error information, including geometric size error, stress distribution error and material property error, and generate a multi-dimensional error data matrix;

[0105] S2, using nonlinear fuzzy residual transfer learning algorithm, based on the historical residual model and current multidimensional error data, to generate a preliminary error correction strategy;

[0106] S3, monitor the real-time entropy value changes in the manufacturing system, and provide a basis for evaluating the stability of the production environment and error fluctuations;

[0107] S4. Based on the change of entropy value, an adaptive entropy control dynamic mapping algorithm is used to adjust the execution path of the error correction strategy to adapt to the dynamic changes in the production environment;

[0108] S5. According to the adjusted error correction strategy, a correction operation is performed to adjust the geometry, stress distribution and material properties of the micro spring washer;

[0109] S6. After the correction operation, perform error evaluation and record the corrected error data;

[0110] S7. Based on the evaluation results, update the parameters of the nonlinear fuzzy residual transfer learning algorithm and the adaptive entropy control dynamic mapping algorithm to provide an optimization basis for the error correction strategy;

[0111] S8. Repeat steps S1 to S7 to achieve error correction in each production batch.

[0112] In this implementation, S2 includes the following steps:

[0113] S21. Collect various error data in historical production batches, such as geometric dimension errors Stress distribution error Material property error Construct the historical residual matrix R hist :

[0114]

[0115] Calculate the error for the i-th batch: in, represents the actual measurement value of the i-th batch; M target represents the target design value; ξ i Indicates measurement noise or environmental disturbance, affecting measurement accuracy;

[0116] S22. Collect geometric dimension errors in real time in the current production batch Stress distribution error Material property error Construct a multidimensional error matrix E curr :

[0117]

[0118] Calculate the error value: in, represents the actual measurement value of the current batch, ζ j Represents noise or interference in real-time measurements, caused by equipment or environmental changes;

[0119] S23, establish a fuzzy logic system to handle complex error relationships through fuzzy membership functions, fuzzy rule bases and reasoning mechanisms;

[0120] Fuzzy membership function:

[0121]

[0122] Among them, c i Represents the center of the membership function and determines the peak position of the membership function; σ i It indicates the width of the membership function, controls the extension range of the membership function, and affects the degree of fuzzification;

[0123] Fuzzy rule base: Build a fuzzy rule base based on the relationship between historical data and current errors For reasoning;

[0124] S24. Based on the fuzzy logic system, a nonlinear fuzzy residual transfer function F is constructed. NLFRT ; Combining the historical residual matrix and the current error matrix, the residual transfer matrix T is obtained through fuzzy reasoning res :

[0125]

[0126] Among them, μk Represents the fuzzy membership coefficient, dynamically adjusting the influence weight of the fuzzy rule in error transmission; Q k Represents the weight matrix, which adjusts the importance of residuals in different dimensions; σ k is the width of the Gaussian kernel function, which adjusts the blur degree of the kernel function to adapt to different error scales; A k It is a nonlinear adaptive matrix used to capture the complex dynamic behavior between residual and error, and is processed in combination with the nonlinear activation function tanh;

[0127] S25, based on the nonlinear fuzzy residual transfer function, calculate the adaptive weight matrix W dyn :

[0128]

[0129] Among them, I is the unit matrix, which is used to maintain the matrix reversibility and avoid numerical instability; α is the regularization parameter, which balances the flexibility and stability of the weight matrix to prevent overfitting;

[0130] S26. Generate a preliminary error correction strategy C using an adaptive weight matrix init , and adjusted by nonlinear optimization:

[0131]

[0132] Among them, Δ is the optimization correction vector, which is used to further adjust the correction strategy; λ is the regularization parameter, which controls the smoothness of the optimization process; Δ i is the i-th component of the correction vector Δ;

[0133] S27, generate final correction vector C final and applied to production;

[0134] S28. Based on production feedback, dynamically update parameters in the fuzzy logic system and the nonlinear fuzzy residual transfer function to optimize the error correction strategy for future production batches.

[0135] In this implementation, S3 includes the following steps:

[0136] S31. During the production process, the multi-dimensional physical parameters in the manufacturing system are collected through a real-time sensor network to build a real-time data matrix D real (t), and preprocess to remove high-frequency noise to generate the processed data matrix D proc (t);

[0137] S32, based on the processed data matrix D proc (t), calculate the real-time entropy value H(t) of the manufacturing system to quantify the uncertainty and stability in the production process:

[0138]

[0139] Among them, p i (t) represents the processed data matrix D proc The probability distribution of the i-th data dimension in (t) at time t;

[0140] S33. Perform time series analysis on the entropy value H(t) and calculate the rate of change To identify dynamic fluctuations in the manufacturing process;

[0141] S34, the real-time entropy value H(t) and the preliminary error correction strategy C init Coupled, a dynamic feedback mechanism is constructed to adjust the error correction strategy of the current batch to adapt to the real-time changing production environment;

[0142] S35. Setting the warning threshold H of entropy value change th , when the real-time entropy value When the change exceeds this threshold, the error fluctuation alarm is triggered, indicating instability in the manufacturing system. At this time, the error correction strategy will be automatically adjusted to avoid production deviations caused by environmental fluctuations.

[0143] S36. Feedback the information of real-time entropy value changes to the nonlinear fuzzy residual transfer learning model, dynamically adjust the parameters of the model, and optimize the real-time error correction strategy to make it better adapt to changes in the environment and state during the manufacturing process.

[0144] In this implementation, S4 includes the following steps:

[0145] S41. During the production process, the key environmental parameters in the manufacturing system are monitored through a real-time sensor network, multimodal environmental data is generated, the environmental state matrix S(t) is constructed, and the corresponding real-time entropy value H(t) and change rate are calculated.

[0146] S42. Use multimodal environmental data to construct a multi-level environmental state matrix S(t), and perform high-dimensional tensor decomposition to capture the characteristics and relationships of different levels:

[0147]

[0148] Among them, S k (t) represents the k-th level feature matrix of the environment state, which is a multi-dimensional feature extracted by tensor decomposition; is the decomposition coefficient in tensor decomposition, which represents the relationship weight between the k-th layer feature and the original data; Φ i It is a feature matrix representing the state of the environment, which contains the core features of multimodal data;

[0149] S43, couple the real-time entropy value H(t) with the multimodal environment state tensor S(t) to construct a composite mapping model Used to capture complex interactive effects in multi-dimensional dynamic environments:

[0150]

[0151] Among them, θ ijk Coupling coefficient reflects the interaction between different modal characteristics; U i ,V j ,W k They represent the three-dimensional feature matrices obtained by tensor decomposition, reflecting the coupling relationship between different environmental states; It represents the contribution to the entropy value under different modes, combining the temporal dynamic information of each modal feature;

[0152] S44, based on composite mapping model Introducing adaptive learning mechanism By combining multidimensional convolution and recurrent networks, nonlinear modeling of the complex relationship between environmental state and entropy value changes is performed:

[0153]

[0154] Among them, W p is a multidimensional convolution kernel tensor used to extract feature patterns of different scales; * is a convolution operator, which represents the operation between the convolution kernel and the environment state tensor; σ is a nonlinear activation function; RNN is a recurrent neural network layer used to process the time dependency in sequence data; ρ q is the adaptive learning weight, which represents the learning contribution at different time steps;

[0155] S45. Combine the composite mapping model and adaptive learning with the real-time entropy value change analysis results to construct an adaptive entropy control dynamic mapping function Used to dynamically adjust the error correction strategy in real time C adj (t):

[0156]

[0157] in, represents the composite mapping result of the rth layer, capturing the complex interaction relationship in a multimodal environment; C adj (t) is the adjusted error correction strategy; γ l is a dynamic adjustment coefficient that controls the adjustment strength under different entropy value deviations; R l It is the multidimensional entropy value change moment, which is used to measure the entropy value change under different conditions;

[0158] S46, high-dimensional dynamic correction vector C generated by adaptive entropy control dynamic mapping function dyn (t), used to adjust key parameters in the manufacturing process in real time;

[0159] S47. Apply high-dimensional dynamic correction vector C dyn (t), dynamically adjust key parameters in the production process;

[0160] S48, comparing the high-dimensional error difference ΔE(t) before and after correction, feeding it back into the adaptive entropy control dynamic mapping function, and optimizing the tensor weight and learning rate;

[0161] S49. Regularly conduct multimodal analysis on entropy value changes and correction effects, and update adaptive entropy control.

[0162] In this implementation, S5 includes the following steps:

[0163] S51, according to the generated high-dimensional dynamic correction vector C dyn (t), adjust the geometry of the micro spring washer;

[0164] S52. Using adaptive entropy to control dynamic mapping function The stress distribution of the micro spring washer is optimized based on the results, and the stress adjustment is performed based on the stress matrix monitored in real time.

[0165] S53, according to the generated high-dimensional dynamic correction vector C dyn (t), optimizing the material properties of the micro spring washers, especially the elastic modulus and hardness;

[0166] S54, integrate the adjustment results of geometry, stress distribution and material properties, and use the high-dimensional dynamic correction vector C dyn (t) carry out comprehensive adjustments;

[0167] S55. During the adjustment process, the geometry, stress distribution and material properties of the micro spring washer are monitored in real time, and the monitoring results are fed back to the adaptive entropy control dynamic mapping function. In order to further optimize the correction strategy in future production batches;

[0168] S56. Conduct a comprehensive performance evaluation on the adjusted micro spring washers, and further optimize the nonlinear fuzzy residual transfer learning model and the adaptive entropy control dynamic mapping function based on the evaluation results.

[0169] In this implementation, S6 includes the following steps:

[0170] S61. After the correction operation is completed, collect the corrected multi-dimensional error data, including geometric size error Stress distribution error and material property errors Generate the corrected error data matrix E corr ;

[0171] S62, the corrected error data matrix E corr and the target design value matrix M target Compare and calculate the corrected error deviation matrix ΔE corr And the error deviation matrix ΔE is calculated by fuzzy logic inference system corr Conduct preliminary analysis;

[0172] S63, the error deviation matrix ΔE corr Input the dynamically adjusted nonlinear fuzzy logic system, and generate the adjusted correction strategy C through fuzzification processing and nonlinear activation function. corr ;

[0173] S64, compare the error deviation before and after correction, and generate the feedback matrix F corr , used to update the membership function and rule base in the fuzzy logic system;

[0174] S65, based on the feedback matrix F corr The results are optimized and the final correction strategy C is generated. fi , adjust key parameters through adaptive learning mechanisms to improve the accuracy and adaptability of the correction strategy;

[0175] S66, the final correction strategy C fi Applied to the production process, it monitors the correction effect in real time and records the corrected error data.

[0176] In this implementation, S7 includes the following steps:

[0177] S71, according to the corrected error data matrix E corr and error deviation ΔE corr , the core parameter θ of the nonlinear fuzzy residual transfer learning algorithm NLFRT To update:

[0178]

[0179] in, is the previous parameter set of the nonlinear fuzzy residual transfer learning algorithm, η is the learning rate, which is used to control the step size of parameter update, is the loss function The gradient of the parameter θ is used to minimize the error deviation ΔE corr And the corrected error matrix E corr The error between

[0180] S72, based on updated parameters Generate a new fuzzy residual matrix R new :

[0181]

[0182] Among them, R hist is the historical residual matrix, μ k is the fuzzy membership coefficient, which indicates the weight of the current fuzzy rule, σ k is the width of the Gaussian kernel function, which controls the fuzziness of the kernel function to the error. k is a nonlinear adaptive matrix used to capture the dynamic nonlinear relationship between residual and error. represents the residual matrix generated by the updated nonlinear fuzzy residual transfer learning algorithm;

[0183] S73, the generated fuzzy residual matrix R new As input, dynamically adjust the parameters γ of the adaptive entropy control dynamic mapping algorithm DM :

[0184]

[0185] in, is the previous entropy control parameter, β is the adjustment coefficient, which is used to control the amplitude of parameter adjustment, H(t) is the real-time entropy value at time t, reflecting the uncertainty in the production process, R new (t) is the fuzzy residual matrix value at time t, ∈ is a small positive number to prevent the denominator from being zero;

[0186] S74. Using the updated parameters and Generate optimized calibration strategy

[0187] S75, the optimized correction strategy It is applied to error correction operations in production batches, and the correction effect is continuously monitored through a real-time data feedback system. The parameters of the nonlinear fuzzy residual transfer learning algorithm and the adaptive entropy control dynamic mapping algorithm are further adjusted according to the feedback results.

[0188] Embodiment 1:

[0189] In the production line of a high-end precision equipment manufacturer, micro spring washers are a key component, which are used to support and disperse pressure while maintaining stable operation of the equipment under high load conditions. The micro spring washers produced by the factory are used in internal components of spacecraft and need to have extremely high precision and consistency. Any slight error may cause the failure of the entire equipment. However, in the actual production process, due to differences in material properties, geometric dimensions and stress distribution, these micro spring washers often have unforeseen errors in manufacturing. These errors not only lead to high scrap rates of products, but also affect production efficiency and quality consistency.

[0190] In order to deal with these problems, the factory decided to introduce the micro spring washer error correction method based on nonlinear fuzzy residual transfer learning and adaptive entropy control dynamic mapping algorithm proposed in this invention. The specific application is as follows:

[0191] First, during the production process, the manufacturing system collects multi-dimensional error information of micro spring washers in real time, including geometric size error, stress distribution error and material property error. Through high-precision sensors, these error data are converted into a multi-dimensional error data matrix and input into the system in real time.

[0192] Then, the system compares and analyzes historical production data and current real-time data through a nonlinear fuzzy residual transfer learning algorithm to generate a preliminary error correction strategy. For example, for the geometric dimension error of a batch of products, the system identifies that the deviation from historical data is large, which may affect the assembly accuracy of the gasket, so it recommends fine-tuning the size.

[0193] At the same time, the system will continuously monitor the real-time entropy changes in the manufacturing system and evaluate the stability of the production environment and the error fluctuations. Assuming that the entropy value increases significantly within a certain period of time, the system will use an adaptive entropy control dynamic mapping algorithm based on this change to dynamically adjust the error correction strategy to ensure that the fluctuations in the production environment will not have a negative impact on product accuracy.

[0194] In actual operation, when the system detects that the errors in the outer diameter and thickness of a batch of micro spring washers exceed the preset threshold, the correction procedure is immediately initiated. Guided by the high-dimensional dynamic correction vector, the system makes online corrections to the product, adjusting its geometry, stress distribution and material properties to meet the design requirements.

[0195] Data Proof

[0196] In order to verify the effectiveness of this method, we chose to monitor and collect data on the factory's production line for half a year from July 2023 to January 2024.

[0197] Table 1 Data table

[0198]

[0199]

[0200] The above data table clearly shows the error correction effect of the method of the present invention on the micro spring washers in different batches of production. By applying the method of the present invention, the scrap rate is effectively reduced, the production consistency and product quality are significantly improved, and the beneficial effects of the present invention are fully demonstrated.

[0201] 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 micro spring washer error correction method using fuzzy logic algorithm, characterized in that: The steps include: S1. Collect multi-dimensional error information, including geometric size error, stress distribution error and material property error, and generate a multi-dimensional error data matrix; S2, using nonlinear fuzzy residual transfer learning algorithm, based on the historical residual model and current multidimensional error data, to generate a preliminary error correction strategy; S3, monitor the real-time entropy value changes in the manufacturing system, and provide a basis for evaluating the stability of the production environment and error fluctuations; S4. Based on the change of entropy value, an adaptive entropy control dynamic mapping algorithm is used to adjust the execution path of the error correction strategy; S5. Based on the adjusted error correction strategy, a correction operation is performed to adjust the geometry, stress distribution and material properties of the micro spring washer; S6. After the correction operation, perform error evaluation and record the corrected error data; S7. Update the parameters of the nonlinear fuzzy residual transfer learning algorithm and the adaptive entropy control dynamic mapping algorithm according to the evaluation results; S8. Repeat steps S1 to S7 to achieve error correction in each production batch.

2. The micro spring washer error correction method using fuzzy logic algorithm according to claim 1, characterized in that: The S2 comprises the following steps: S21. Collect various error data in historical production batches, such as geometric dimension errors Stress distribution error Material property error Construct the historical residual matrix R hist : Calculate the error for the i-th batch: in, represents the actual measurement value of the i-th batch; M target represents the target design value; ξ i represents measurement noise or environmental disturbance; S22. Collect geometric dimension errors in real time in the current production batch Stress distribution error Material property error Construct a multidimensional error matrix E curr : Calculate the error value: in, represents the actual measurement value of the current batch, ζ j Represents noise or interference in real-time measurements; S23, establish a fuzzy logic system to handle complex error relationships through fuzzy membership functions, fuzzy rule bases and reasoning mechanisms; Fuzzy membership function: Among them, c i Represents the center of the membership function and determines the peak position of the membership function; σ i Indicates the width of the membership function; Fuzzy rule base: Build a fuzzy rule base based on the relationship between historical data and current errors S24. Based on the fuzzy logic system, a nonlinear fuzzy residual transfer function F is constructed. NLFRT ; Combining the historical residual matrix and the current error matrix, the residual transfer matrix T is obtained through fuzzy reasoning res : Among them, μ k represents the fuzzy membership coefficient; Q k represents the weight matrix; σ k is the width of the Gaussian kernel function; A k is the nonlinear adaptive matrix; S25, based on the nonlinear fuzzy residual transfer function, calculate the adaptive weight matrix W dyn : Where I is the identity matrix; α is the regularization parameter; S26. Generate a preliminary error correction strategy C using an adaptive weight matrix init , and adjusted by nonlinear optimization: Among them, Δ is the optimization correction vector; λ is the regularization parameter; Δ i is the i-th component of the correction vector Δ; S27, generate final correction vector C final and applied to production; S28. Based on production feedback, dynamically update parameters in the fuzzy logic system and the nonlinear fuzzy residual transfer function.

3. The micro spring washer error correction method using fuzzy logic algorithm according to claim 1, characterized in that: The S3 comprises the following steps: S31. During the production process, the multi-dimensional physical parameters in the manufacturing system are collected through a real-time sensor network to build a real-time data matrix D real (t), and preprocess to remove high-frequency noise to generate the processed data matrix D proc (t); S32, based on the processed data matrix D proc (t), calculate the real-time entropy value H(t) of the manufacturing system to quantify the uncertainty and stability in the production process: Among them, p i (t) represents the processed data matrix D proc The probability distribution of the i-th data dimension in (t) at time t; S33. Perform time series analysis on the entropy value H(t) and calculate the rate of change To identify dynamic fluctuations in the manufacturing process; S34, the real-time entropy value H(t) and the preliminary error correction strategy C init Coupled, a dynamic feedback mechanism is constructed to adjust the error correction strategy of the current batch; S35. Setting the warning threshold H of entropy value change th , when the real-time entropy value When the change exceeds this threshold, the error fluctuation alarm is triggered, indicating instability in the manufacturing system. At this time, the error correction strategy will be automatically adjusted to avoid production deviations caused by environmental fluctuations. S36. Feedback the information of real-time entropy value changes to the nonlinear fuzzy residual transfer learning model, dynamically adjust the parameters of the model, and optimize the real-time error correction strategy.

4. The micro spring washer error correction method using fuzzy logic algorithm according to claim 1, characterized in that: The S4 comprises the following steps: S41. During the production process, the key environmental parameters in the manufacturing system are monitored through a real-time sensor network, multimodal environmental data is generated, the environmental state matrix S(t) is constructed, and the corresponding real-time entropy value H(t) and change rate are calculated. S42. Use multimodal environmental data to construct a multi-level environmental state matrix S(t), and perform high-dimensional tensor decomposition to capture the characteristics and relationships of different levels: Among them, S k (t) represents the k-th level feature matrix of the environment state, which is a multi-dimensional feature extracted by tensor decomposition; is the decomposition coefficient in tensor decomposition, which represents the relationship weight between the k-th layer feature and the original data; Φ i It is a feature matrix representing the state of the environment, containing the core features of multimodal data; S43, couple the real-time entropy value H(t) with the multimodal environment state tensor S(t) to construct a composite mapping model Used to capture complex interactive effects in multi-dimensional dynamic environments: Among them, θ ijk Coupling coefficient reflects the interaction between different modal characteristics; U i ,V j ,W k They represent the three-dimensional feature matrices obtained by tensor decomposition; It represents the contribution to the entropy value under different modes, combining the temporal dynamic information of each mode feature; S44, based on composite mapping model Introducing adaptive learning mechanism By combining multidimensional convolution and recurrent networks, nonlinear modeling of the complex relationship between environmental state and entropy value changes is performed: Among them, W p is a multidimensional convolution kernel tensor used to extract feature patterns of different scales; * is a convolution operator, which represents the operation between the convolution kernel and the environment state tensor; σ is a nonlinear activation function; RNN is a recurrent neural network layer used to process the time dependency in sequence data; ρ q is the adaptive learning weight, which represents the learning contribution at different time steps; S45. Combine the composite mapping model and adaptive learning with the real-time entropy value change analysis results to construct an adaptive entropy control dynamic mapping function Used to dynamically adjust the error correction strategy in real time C adj (t): in, represents the composite mapping result of the rth layer; C adj (t) is the adjusted error correction strategy; γ l is the dynamic adjustment coefficient; R l is the multidimensional entropy value change moment; S46, high-dimensional dynamic correction vector C generated by adaptive entropy control dynamic mapping function dyn (t), used to adjust key parameters in the manufacturing process in real time; S47. Apply high-dimensional dynamic correction vector C dyn (t), dynamically adjust key parameters in the production process; S48, comparing the high-dimensional error difference ΔE(t) before and after correction, feeding it back into the adaptive entropy control dynamic mapping function, and optimizing the tensor weight and learning rate; S49. Regularly conduct multimodal analysis on entropy value changes and correction effects, and update adaptive entropy control.

5. The micro spring washer error correction method using fuzzy logic algorithm according to claim 1, characterized in that: The S5 comprises the following steps: S51, according to the generated high-dimensional dynamic correction vector C dyn (t), adjust the geometry of the micro spring washer; S52. Using adaptive entropy to control dynamic mapping function The stress distribution of the micro spring washer is optimized based on the results, and the stress adjustment is performed based on the stress matrix monitored in real time. S53, according to the high-dimensional dynamic correction vector C dyn (t), optimizing the material properties of the micro spring washers, especially the elastic modulus and hardness; S54, integrate the adjustment results of geometry, stress distribution and material properties, and use the high-dimensional dynamic correction vector C dyn (t) carry out comprehensive adjustments; S55. During the adjustment process, the geometry, stress distribution and material properties of the micro spring washer are monitored in real time, and the monitoring results are fed back to the adaptive entropy control dynamic mapping function. In order to further optimize the correction strategy in future production batches; S56. Conduct a comprehensive performance evaluation on the adjusted micro spring washers, and further optimize the nonlinear fuzzy residual transfer learning model and the adaptive entropy control dynamic mapping function based on the evaluation results.

6. The micro spring washer error correction method using fuzzy logic algorithm according to claim 1, characterized in that: The S6 comprises the following steps: S61. After the correction operation is completed, collect the corrected multi-dimensional error data, including geometric size error Stress distribution error and material property errors Generate the corrected error data matrix E corr ; S62, the corrected error data matrix E corr and the target design value matrix M target Compare and calculate the corrected error deviation matrix ΔE corr And the error deviation matrix ΔE is calculated by fuzzy logic inference system corr Conduct preliminary analysis; S63, the error deviation matrix ΔE corr Input the dynamically adjusted nonlinear fuzzy logic system, and generate the adjusted correction strategy C through fuzzification processing and nonlinear activation function. corr ; S64, compare the error deviation before and after correction, and generate the feedback matrix F corr , used to update the membership function and rule base in the fuzzy logic system; S65, based on the feedback matrix F corr The results are optimized and the final correction strategy C is generated. fi , adjust key parameters through adaptive learning mechanisms to improve the accuracy and adaptability of the correction strategy; S66, the final correction strategy C fi Applied to the production process, it monitors the correction effect in real time and records the corrected error data.

7. The micro spring washer error correction method using fuzzy logic algorithm according to claim 1, characterized in that: The S7 comprises the following steps: S71, according to the corrected error data matrix E corr and error deviation ΔE corr , the core parameter θ of the nonlinear fuzzy residual transfer learning algorithm NLFRT To update: in, is the previous parameter set of the nonlinear fuzzy residual transfer learning algorithm; η is the learning rate; is the loss function The gradient of the parameter θ; S72, based on updated parameters Generate a new fuzzy residual matrix R new : Among them, R hist is the historical residual matrix, μ k is the fuzzy membership coefficient, which indicates the weight of the current fuzzy rule, σ k is the width of the Gaussian kernel function, which controls the fuzziness of the kernel function to the error; A k is a nonlinear adaptive matrix used to capture the dynamic nonlinear relationship between residual and error. represents the residual matrix generated by the updated nonlinear fuzzy residual transfer learning algorithm; S73, the generated fuzzy residual matrix R new As input, dynamically adjust the parameters γ of the adaptive entropy control dynamic mapping algorithm DM : in, is the previous entropy control parameter; β is the adjustment coefficient, which is used to control the amplitude of parameter adjustment; H(t) is the real-time entropy value at time t; R new (t) is the fuzzy residual matrix value at time t, ∈ is a small positive number; S74. Using the updated parameters and Generate optimized calibration strategy S75, the optimized correction strategy C initopt Applied to error correction operations in production batches.

Images