Adaptive control method, device and computer equipment for nano-grinding process

Abstract

The present invention relates to an adaptive control method, device and computer equipment for a nano-grinding process, which aims to solve the uncertainty problems caused by material heterogeneity, tool wear and environmental changes. The method first obtains nano-level feature data of a workpiece material through nano-indentation technology, and determines the initial process parameters based on these data. Then, these parameters are used for processing and the grinding and vibration data in the process are monitored in real time. Through feature analysis and spectrum analysis, process state data is extracted and integrated, process anomalies and trend changes are identified, and then target processing parameters are generated. The adaptive control algorithm adjusts the initial process parameters according to the target parameters to achieve real-time optimization of the processing process.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to an adaptive control method, device and computer equipment for a nano-grinding process. Background Art

[0002] Nano-grinding processes play a vital role in modern manufacturing, especially in areas that require high precision and ultra-smooth surfaces, such as the manufacture of optical components, semiconductor wafers, and advanced medical devices. As product requirements for surface quality and processing accuracy continue to increase, traditional grinding processes face difficult challenges, especially when controlling process parameters and dealing with complex surface processing. In order to achieve precise material removal and surface modification at the nanoscale, the grinding process must have higher flexibility and adaptability, and traditional control methods often cannot meet this requirement.

[0003] However, most current control methods for nano-grinding processes rely on fixed parameter settings, which are difficult to deal with various uncertainties that arise during actual processing. These uncertainties may come from material heterogeneity, wear of grinding tools, and changes in environmental factors. If there is not enough adaptive capability, these uncontrollable factors may lead to process instability, which in turn affects processing quality and production efficiency.

[0004] Therefore, it is particularly urgent to study an adaptive control method that can adjust processing parameters in real time to cope with dynamic changes in the nano-grinding process. This method not only requires accurate perception and analysis of various variables in the process, but also requires rapid response and optimization of control strategies to ensure nano-level processing accuracy and surface quality. The current technical difficulty lies in how to effectively integrate advanced sensing technology and control algorithms to achieve efficient management of complex process, break through the limitations of traditional control methods, and truly improve the stability and controllability of the nano-grinding process. Summary of the invention

[0005] The main purpose of the present invention is to provide an adaptive control method, device and computer equipment for nano-grinding process to solve the technical problem of how to deal with the uncertainty caused by material heterogeneity, tool wear and environmental changes during the processing.

[0006] To achieve the above-mentioned purpose, the present invention provides an adaptive control method for a nano-grinding process, comprising the following steps: using a nano-indentation material characterization method to obtain material characteristic data of a workpiece material at the nano-level, and determining the initial process parameters of the nano-grinding process based on the material characteristic data; using the initial process parameters to process the workpiece material, and monitoring the process parameters of the process in real time, wherein the process parameters at least include: grinding data and vibration data; performing feature analysis on the process parameters to extract first feature data characterizing the process state, integrating the material characteristic data of the workpiece material at the nano-level into the first feature data to obtain second feature data; using spectrum analysis technology to identify data representing process anomalies and trend changes in the second feature data to obtain target processing parameters, and inputting the target processing parameters and the initial process parameters into a preset adaptive control algorithm; obtaining the target process parameters after the adaptive control algorithm adjusts the initial process parameters based on the target processing data, and using the target process parameters to process the workpiece material in real time.

[0007] Optionally, determining the starting process parameters of the nano-grinding process based on the material characteristic data includes: normalizing the material characteristic data to convert the material characteristic data into first material characteristic data of the same dimension and magnitude; dividing the first material characteristic data into multiple categories, and based on the first material characteristic data of different categories, calculating a first scatter matrix within each category and a second scatter matrix between different categories; multiplying the first scatter matrix and the second scatter matrix to obtain eigenvalues ​​and eigenvectors for linear discrimination of the first material characteristic data; determining eigenvectors corresponding to multiple eigenvalues ​​with the largest values ​​as target projection directions, projecting the first material characteristic data onto the target projection directions to obtain second material characteristic data; inputting the second material characteristic data into a preset polynomial regression model to obtain the starting process parameters of the nano-grinding process, wherein the starting process parameters include at least: a grinding force parameter and a grinding depth parameter.

[0008] Optionally, feature analysis is performed on the processing parameters to extract first feature data characterizing the process state, including: based on a preset wavelet basis, performing wavelet analysis on the processing parameters to obtain multiple initial components, and different initial components correspond to different frequency bands; performing threshold processing on the multiple initial components to remove data above a first preset threshold and data below a second preset threshold from the multiple initial components to obtain multiple target components; performing wavelet reconstruction processing on the multiple target components, and performing Fourier transform processing on the multiple target components after the wavelet reconstruction processing to obtain analysis signals represented in the frequency domain; and performing threshold processing on the analysis signals. Perform spectrum analysis to obtain multiple frequency components and feature representations reflected by each frequency component; extract multiple target frequency components and amplitudes corresponding to the target frequency components based on the feature representation reflected by each frequency component; construct a first feature matrix based on the multiple target frequency components and amplitudes corresponding to the target frequency components; perform principal component analysis on the first feature matrix, calculate the covariance matrix of the first feature matrix, and perform eigenvalue decomposition on the covariance matrix to obtain multiple principal components; extract a preset number of target principal components from the multiple principal components, and use the target principal components as the first feature data.

[0009] Optionally, the material characteristic data of the workpiece material at the nanometer level is integrated into the first characteristic data to obtain the second characteristic data, including: normalizing the first characteristic data and the second material characteristic data to convert the first characteristic data and the second material characteristic data into the same dimension and magnitude; embedding the first characteristic data and the second material characteristic data into high-dimensional vectors in the same embedding space through multiple hidden layers of a multilayer perceptron to obtain multiple characteristic high-dimensional vectors; using the embedded characteristic high-dimensional vectors as nodes of a graph network, calculating the Euclidean distance between the characteristic high-dimensional vectors, and using the Euclidean distance between the characteristic high-dimensional vectors to establish connections between nodes to obtain a characteristic graph grid; using a graph convolutional network to update the main feature representation of each node in the characteristic graph grid to identify implicit associations between features to obtain multiple updated main feature representations; aggregating the main feature representations at different levels to obtain aggregated main feature vectors, and performing nonlinear activation processing on the main feature vectors to obtain the second characteristic data.

[0010] Optionally, spectrum analysis technology is used to identify data in the second characteristic data that represent process anomalies and trend changes to obtain target processing parameters, including: performing Fourier transform on the second characteristic data to obtain spectrum data; using a preset support vector machine model to analyze each frequency component in the spectrum data to identify abnormal frequency components in the spectrum data; organizing the spectrum data according to time series, performing seasonal decomposition on each frequency component in the spectrum data, and extracting long-term trend information in the spectrum data; integrating the abnormal frequency components and the long-term trend information to obtain the target processing parameters.

[0011] Optionally, before using a preset support vector machine model to analyze each frequency component in the spectral data and identify abnormal frequency components in the spectral data, the adaptive control method also includes: using a radial basis function kernel to construct the support vector machine model, and setting the C value and gamma value of the support vector machine model; obtaining training data, and using the training data to train the support vector machine model to optimize the minimum classification error rate of the support vector machine model, wherein the training data includes: pre-prepared spectral data and process status labels corresponding to the frequency components in the pre-prepared spectral data.

[0012] Optionally, the spectral data is organized according to a time series, and each frequency component in the spectral data is seasonally decomposed to extract long-term trend information in the spectral data, including: arranging the spectral data in chronological order to form time series data, wherein the spectral data at each time point in the time series data includes the amplitude values ​​of all frequency components; combining the timestamp of the spectral data in the time series data and the frequency component data corresponding to the timestamp to generate a time series data frame, wherein the time series data frame includes a timestamp column and each frequency component column; using an additive seasonal decomposition model to analyze the time series data frame to obtain seasonal analysis data, wherein the seasonal analysis data includes: a long-term trend component, a seasonal component and random noise; drawing the long-term trend component as a trend graph, and identifying the long-term trend of the time series data based on the trend graph, wherein the long-term trend includes: long-term growth or long-term decline; evaluating the impact of the long-term trend component and the seasonal component on the long-term trend to obtain evaluation data, and integrating the evaluation data into the long-term trend information in the spectral data.

[0013] Optionally, after the target processing parameters and the initial process parameters are input into a preset adaptive control algorithm, the adaptive control method further includes: comparing and calculating the target processing parameters with the ideal processing parameters through the adaptive control algorithm to obtain error data, wherein the error data includes the error value and the error change rate of each parameter; converting the error data into a fuzzy set, and constructing a membership function for each fuzzy set, wherein the fuzzy set is represented by a linguistic variable, and the membership function is used to express the degree of fuzziness of the fuzzy set; calculating the degree of matching between the membership function of each fuzzy set and a plurality of preset fuzzy rules, and merging the fuzzy sets with a matching degree higher than a third preset threshold to obtain a fuzzy control output; using the maximum membership method to defuzzify the fuzzy control output to obtain a control signal output, and using the control signal output to adjust the initial process parameters to obtain the target process parameters.

[0014] The present invention also provides an adaptive control device for a nano-grinding process, comprising: an acquisition unit, for acquiring material characteristic data of a workpiece material at the nanometer level by adopting a nano-indentation material characterization method, and determining the initial process parameters of the nano-grinding process based on the material characteristic data; a monitoring unit, for processing the workpiece material by adopting the initial process parameters, and monitoring the processing parameters of the processing process in real time, wherein the processing parameters at least include: grinding data and vibration data; an extraction unit, for performing characteristic analysis on the processing parameters, extracting first characteristic data characterizing the process state, integrating the material characteristic data of the workpiece material at the nanometer level into the first characteristic data, and obtaining second characteristic data; an identification unit, for identifying data representing process anomalies and trend changes in the second characteristic data by adopting a spectrum analysis technique, obtaining target processing parameters, and inputting the target processing parameters and the initial process parameters into a preset adaptive control algorithm; a processing unit, for obtaining the target process parameters after the adaptive control algorithm adjusts the initial process parameters based on the target processing data, and processing the workpiece material by adopting the target process parameters in real time.

[0015] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0016] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0017] The adaptive control method, device and computer equipment of the nano-grinding process provided by the present invention can effectively cope with uncertain factors such as material heterogeneity, tool wear and environmental changes by adjusting the processing parameters in real time, and maintain the stability and accuracy of the processing technology. This real-time adaptability enables the process to maintain a high level of processing quality and efficiency in the face of constantly changing processing conditions.

[0018] Secondly, the present invention uses nanoindentation material characterization technology and spectrum analysis technology to achieve accurate acquisition and adjustment of process parameters. This precise control method optimizes the nanoscale material removal and surface modification process, and can achieve high-quality processing on an ultra-fine scale.

[0019] In addition, the present invention improves the overall understanding of process status and anomaly detection capabilities by fusing material feature data, processing parameters and spectrum data. This data fusion and analysis method can more comprehensively monitor and evaluate the processing process and identify and deal with potential problems in a timely manner.

[0020] Furthermore, the present invention breaks through the limitations of traditional fixed parameter settings and significantly improves the management capabilities of complex process through adaptive control algorithms and fuzzy control methods. Compared with traditional methods, this adaptive control method can better adapt to complex process requirements and changes.

[0021] Finally, through spectrum analysis and fuzzy control, the present invention can quickly generate and adjust the target processing parameters, shorten the optimization process time, and improve production efficiency. This efficient adjustment mechanism enables the nano-grinding process to achieve the best processing effect in a shorter time, thereby improving the economic benefits of the overall production. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 Schematic diagram of the steps of an adaptive control method for a nano-grinding process in one embodiment of the present invention;

[0023] Figure 2 is a structural block diagram of an adaptive control device for a nano-grinding process in one embodiment of the present invention;

[0024] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0025] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0027] Reference Figure 1 The embodiment of the present invention further provides an adaptive control method for a nano-grinding process, comprising the following steps:

[0028] S1, using a nano-indentation material characterization method to obtain material characteristic data of a workpiece material at the nano-level, and determining the starting process parameters of a nano-grinding process based on the material characteristic data.

[0029] S2, processing the workpiece material using the initial process parameters, and monitoring the processing parameters of the processing in real time, wherein the processing parameters at least include: grinding data and vibration data.

[0030] S3, performing feature analysis on the machining process parameters, extracting first feature data characterizing the process state, integrating the material feature data of the workpiece material at the nanometer level into the first feature data, and obtaining second feature data.

[0031] S4, using spectrum analysis technology to identify the data representing process anomalies and trend changes in the second characteristic data, obtain target processing parameters, and input the target processing parameters and the initial process parameters into a preset adaptive control algorithm.

[0032] S5, obtaining target process parameters after the adaptive control algorithm adjusts the initial process parameters based on the target processing data, and using the target process parameters to process the workpiece material in real time.

[0033] That is, in the field of modern manufacturing, precise control of the nano-grinding process is crucial to the production of high-quality optical components, semiconductor wafers and advanced medical devices. However, due to the limitations of traditional control methods, the main challenge is how to maintain the stability and accuracy of the process when processing nano-scale materials. In order to solve this problem, the present invention proposes an adaptive control method, which improves the stability and controllability of the process by adjusting the processing parameters in real time to cope with the dynamic changes in the nano-grinding process.

[0034] In this adaptive control method, the first step is to obtain nanoscale feature data of the workpiece material. This can be achieved through nanoindentation technology, which can accurately measure the properties of materials at the nanoscale, such as hardness and elastic modulus. These data are used to determine the initial process parameters so that they can be used as a benchmark in actual processing. For example, if the hardness of the material is found to be high, the initial process parameters may need to be set more conservatively to avoid excessive wear on the material.

[0035] Next, the workpiece is ground using these starting process parameters, and various parameters generated during the process, such as grinding data and vibration data, are monitored in real time. These data provide real-time feedback for understanding the actual situation during the process. By monitoring these parameters, any abnormalities in the process can be captured, such as abnormal changes in vibration amplitude that may indicate wear of the grinding tool or non-uniformity of the material.

[0036] Characterizing these process parameters is a key step. By extracting and analyzing the features in the grinding data and vibration data, the first characteristic data of the process state can be identified. Then, the nanoscale characteristic data of the material is integrated with these first characteristic data to obtain the second characteristic data. This second characteristic data includes more comprehensive process state information, which helps to more accurately identify and predict the changing trends in the process.

[0037] Through spectrum analysis technology, the second characteristic data can be further analyzed to identify data indicating process anomalies and trend changes. These anomaly and trend data help determine the target processing parameters that need to be adjusted. The target processing parameters reflect the best adjustment plan under the current process state and can effectively improve processing accuracy and surface quality.

[0038] The target processing parameters and the starting process parameters are input into the preset adaptive control algorithm, and the algorithm calculates the adjusted target process parameters based on these input data. The adaptive control algorithm adjusts the process parameters in real time based on the target processing data to optimize the processing process. Finally, using these optimized target process parameters for actual processing can significantly improve the stability and accuracy of the nano-grinding process.

[0039] For example, in actual applications, if the initial process parameters are set to a certain grinding depth and feed rate, but abnormal vibration is detected during the process, the spectrum analysis identifies that this abnormality is related to tool wear, and the algorithm will calculate the adjusted process parameters, such as reducing the feed rate or replacing the worn tool, to achieve a more stable grinding effect. This adaptive control method improves the accuracy and efficiency of the nano-grinding process through real-time adjustment and optimization, meeting the requirements of modern manufacturing for high-quality products.

[0040] In one example, determining the starting process parameters of a nano-grinding process based on the material characteristic data includes: normalizing the material characteristic data to convert the material characteristic data into first material characteristic data of the same dimension and magnitude; dividing the first material characteristic data into multiple categories, and calculating a first scatter matrix within each category and a second scatter matrix between different categories based on the first material characteristic data of different categories; multiplying the first scatter matrix and the second scatter matrix to obtain eigenvalues ​​and eigenvectors of linear discrimination of the first material characteristic data; determining eigenvectors corresponding to multiple eigenvalues ​​with the largest values ​​as target projection directions, projecting the first material characteristic data onto the target projection directions to obtain second material characteristic data; and inputting the second material characteristic data into a preset polynomial regression model to obtain the starting process parameters of the nano-grinding process, wherein the starting process parameters include at least a grinding force parameter and a grinding depth parameter.

[0041] That is, in the process of determining the initial process parameters of the nano-grinding process, the material characteristic data must first be normalized. This step aims to convert the material characteristic data of various dimensions and magnitudes into a unified standard for subsequent analysis and processing. Through normalization, it can be ensured that all characteristic data are at the same scale, thereby avoiding analysis errors caused by differences in data scales.

[0042] Next, the normalized material feature data is divided into multiple categories. The data in each category is similar in some aspects, so when calculating the first scatter matrix within each category, the scatter of the feature data in that category can be obtained. The first scatter matrix reflects the degree of variation of the data within the category. These matrices help to understand the distribution of material features in a particular category.

[0043] In addition, it is necessary to calculate the second scatter matrix between different categories, which describes the distribution differences between categories. The calculation of the second scatter matrix can reveal the distance and separation between different categories, providing support for subsequent classification. By multiplying the first scatter matrix and the second scatter matrix, the eigenvalues ​​and eigenvectors for linear discrimination of material characteristic data can be obtained. This step aims to find the main characteristic directions in the data so that the most effective projection and classification can be performed in these directions.

[0044] From the calculated eigenvalues, the eigenvectors corresponding to the eigenvalues ​​with the largest values ​​are selected as the target projection directions. These eigenvectors represent the most significant directions in the data and can retain the main information of the data to the greatest extent. The first material characteristic data is projected onto these target projection directions, and the result is the second material characteristic data, which is easier to analyze and process in the new characteristic space.

[0045] Finally, the second material characteristic data is input into the preset polynomial regression model. Through the calculation of the model, the starting process parameters of the nano-grinding process, such as grinding force parameters and grinding depth parameters, can be obtained. The polynomial regression model can capture the complex relationship between characteristic data and process parameters, and then predict the optimal starting process parameters suitable for the current material characteristics. These parameters will be used as the benchmark settings in the subsequent grinding process to ensure the stability and accuracy of the processing process.

[0046] For example, if the material characteristic data contains information such as hardness, elastic modulus and surface roughness, these characteristic data can be converted into values ​​of a uniform scale after normalization. Assuming that the materials are divided into three categories: high hardness, medium hardness and low hardness, the first scatter matrix and the second scatter matrix are calculated for the characteristic data of each category. After multiplying these matrices, the eigenvalues ​​and eigenvectors obtained can help identify the best projection direction. The data is projected onto these directions, and after prediction by the polynomial regression model, the grinding force and grinding depth parameters suitable for high hardness materials are finally determined. In this way, accurate starting process parameters can be provided for different materials, thereby improving the efficiency and effectiveness of the nano-grinding process.

[0047] In one example, feature analysis is performed on the processing parameters to extract first feature data characterizing the process state, including: based on a preset wavelet basis, performing wavelet analysis on the processing parameters to obtain multiple initial components, and different initial components correspond to different frequency bands; performing threshold processing on the multiple initial components to remove data higher than a first preset threshold and data lower than a second preset threshold in the multiple initial components to obtain multiple target components; performing wavelet reconstruction processing on the multiple target components, and performing Fourier transform processing on the multiple target components after the wavelet reconstruction processing to obtain an analysis signal represented in the frequency domain; performing spectrum analysis processing on the analysis signal to obtain multiple frequency components and feature representations reflected by each frequency component; based on the feature representation reflected by each frequency component, extracting multiple target frequency components and amplitudes corresponding to the target frequency components; constructing a first feature matrix based on the multiple target frequency components and amplitudes corresponding to the target frequency components; performing principal component analysis on the first feature matrix to calculate the covariance matrix of the first feature matrix, and performing eigenvalue decomposition processing on the covariance matrix to obtain multiple principal components; extracting a preset number of target principal components from the multiple principal components, and using the target principal components as the first feature data.

[0048] When performing feature analysis on machining process parameters, we first need to perform wavelet analysis on these parameters. This step decomposes the machining process parameters into multiple initial components based on the preset wavelet basis. Each initial component corresponds to a different frequency band. This decomposition method can convert the signal from the time domain to the frequency domain, allowing us to analyze the contribution of different frequency components.

[0049] After completing the wavelet analysis, multiple initial components will be obtained. These components have different frequency ranges, but some components may contain noise or irrelevant frequency components. Therefore, these initial components need to be thresholded. This means that we will set two thresholds: one is the first preset threshold, which is used to remove data with frequency components greater than this value, and the other is the second preset threshold, which is used to remove data with frequency components less than this value. In this way, the components after thresholding will be concentrated in the frequency range of interest, and multiple target components will be obtained.

[0050] Next, these target components are processed by wavelet reconstruction, with the goal of reconstructing these target components from the frequency domain signal back to the time domain signal. Subsequently, the reconstructed signal is transformed into the frequency domain by Fourier transform, so that its frequency components can be further analyzed. Fourier transform will help us identify the different frequency components and their intensities in the signal, which reflect the different dynamic characteristics of the processing process.

[0051] The analysis signal is spectrally processed to obtain multiple frequency components and characteristic representations of each frequency component. These frequency components and their characteristic representations reveal the main frequency information in the signal, such as the main frequency of vibration or periodic changes. Based on these characteristic representations, we can extract multiple target frequency components and calculate the amplitude corresponding to each target frequency component. These target frequency components and amplitude values ​​are important characteristic data for characterizing the process status.

[0052] When constructing the first feature matrix, it will contain all the extracted target frequency components and the corresponding amplitudes. This matrix reflects the frequency characteristics and their intensity during the processing and is the basis for subsequent analysis. By performing principal component analysis on the first feature matrix, multiple principal components can be obtained by calculating the covariance matrix and performing eigenvalue decomposition. Principal component analysis is a dimensionality reduction technique that helps extract the most representative features and reduce the complexity of the data.

[0053] From the obtained multiple principal components, a preset number of target principal components are selected, which represent the most important characteristic directions in the data. Using these target principal components as the first characteristic data, the state of the processing process can be accurately characterized. These characteristic data can help identify key parameters in the process, optimize the processing process and improve the stability and efficiency of the overall process.

[0054] For example, suppose we are monitoring the vibration data of a grinding process, and use wavelet analysis to decompose the signal into initial components of different frequency bands. After thresholding these components, we retain the valid signals within the frequency range. After wavelet reconstruction and Fourier transform of these valid signals, we can obtain the representative signals in the frequency domain. After spectrum processing, the main frequency components and corresponding amplitudes are extracted, and the constructed feature matrix can be analyzed by principal components to select the principal components that best represent the process status. These principal components provide an accurate basis for subsequent process control and optimization.

[0055] In one example, the material feature data of the workpiece material at the nanometer level is integrated into the first feature data to obtain the second feature data, including: normalizing the first feature data and the second material feature data to convert the first feature data and the second material feature data to the same dimension and magnitude; embedding the first feature data and the second material feature data into high-dimensional vectors in the same embedding space through multiple hidden layers of a multilayer perceptron to obtain multiple feature high-dimensional vectors; using the embedded feature high-dimensional vectors as nodes of a graph network, calculating the Euclidean distance between the feature high-dimensional vectors, and using the Euclidean distance between the feature high-dimensional vectors to establish connections between nodes to obtain a feature graph grid; using a graph convolutional network to update the main feature representation of each node in the feature graph grid to identify implicit associations between features to obtain multiple updated main feature representations; aggregating the main feature representations at different levels to obtain aggregated main feature vectors, and performing nonlinear activation processing on the main feature vectors to obtain the second feature data.

[0056] That is, in the process of integrating the nano-level material characteristic data of the workpiece material into the first characteristic data to obtain the second characteristic data, the first characteristic data and the second material characteristic data need to be normalized first. The purpose of this process is to convert two different data sets into the same dimension and magnitude so that they can be compared and analyzed under the same standard. Normalization ensures data consistency and makes subsequent feature embedding more accurate and effective.

[0057] Next, the normalized first feature data and the second material feature data are feature embedded through multiple hidden layers of a multilayer perceptron. A multilayer perceptron (MLP) is a neural network model whose hidden layers can perform complex nonlinear transformations on input data. Through these hidden layers, the input data is converted into high-dimensional vectors that represent the deep features of the data and are embedded into a common embedding space. In this way, each feature data is represented as a feature high-dimensional vector that can capture important information and relationships in the data.

[0058] These feature high-dimensional vectors are used as nodes of the graph network, and the Euclidean distance between them is calculated. Euclidean distance is a common method for measuring the distance between points, which can help us understand the similarities and differences between different nodes. By establishing connections between nodes through Euclidean distance, a feature graph grid is obtained, which reflects the relationship and similarity between nodes.

[0059] Then, a graph convolutional network (GCN) is used to update the main feature representation of each node in the feature map grid. A graph convolutional network is a neural network that processes graph data and can aggregate and update the feature representation of nodes through the connection relationship between nodes. In this way, GCN can identify implicit associations between features, for example, some features may have similar patterns or trends in neighboring nodes of the graph. The updated main feature representation more comprehensively reflects the mutual relationship and feature association of nodes.

[0060] After the feature update is completed, the main feature representations at different levels are aggregated. This process integrates the feature information at each level to generate a comprehensive main feature vector. Then, the main feature vector is nonlinearly activated to introduce nonlinear transformations and enhance the expressiveness and complexity of the model. The final main feature vector is the second feature data, which integrates the key information in the first feature data and the second material feature data, and provides an in-depth analysis and understanding of the process status.

[0061] In one example, spectrum analysis technology is used to identify data in the second feature data that represent process anomalies and trend changes to obtain target processing parameters, including: performing Fourier transform on the second feature data to obtain spectrum data; using a preset support vector machine model to analyze each frequency component in the spectrum data to identify abnormal frequency components in the spectrum data; organizing the spectrum data according to time series, performing seasonal decomposition on each frequency component in the spectrum data, and extracting long-term trend information in the spectrum data; integrating the abnormal frequency components and the long-term trend information to obtain the target processing parameters.

[0062] That is, in the process of identifying process anomalies and trend change data in the second feature data to obtain the target processing parameters, the second feature data needs to be Fourier transformed first. This step converts the time domain data into frequency domain data, allowing us to analyze the performance of different frequency components. The spectrum data generated by Fourier transform reflects the intensity and frequency characteristics of each frequency component in the signal, which is crucial for detecting anomalies and identifying trend changes.

[0063] Next, each frequency component in the spectrum data is analyzed using a preset support vector machine (SVM) model. Support vector machine is an effective classification and regression tool that can help us identify abnormal frequency components in spectrum data. These abnormal frequency components may indicate problems in the process, such as equipment failure or unstable processing conditions. The trained SVM model can accurately distinguish between normal frequency components and abnormal frequency components, thereby identifying potential process anomalies.

[0064] Subsequently, the spectral data is processed by seasonal decomposition. The purpose of seasonal decomposition is to decompose the time series information in the spectral data into several components, including long-term trends, seasonal fluctuations, and random noise. Through this decomposition, we can extract long-term trend information in the spectral data, which reveals the overall change trend of the process, such as gradually increasing processing errors or stable improvements.

[0065] In the integration process, the identified abnormal frequency components and the extracted long-term trend information are combined. The abnormal frequency components provide direct clues about process abnormalities, while the long-term trend information helps us understand how these abnormalities evolve over time. By combining the two, a comprehensive target processing parameter can be obtained, which not only takes into account the current abnormal situation, but also integrates the long-term process trend.

[0066] For example, suppose that during the processing, we perform Fourier transform on the vibration data, obtain the spectrum data, and use the SVM model to identify several abnormal frequency components. These abnormalities may correspond to mechanical failures or abnormal wear of the equipment. At the same time, the spectrum data is seasonally decomposed to extract the long-term trend of the processing process, such as the trend of gradually increasing vibration. By combining these abnormal frequency components with the long-term trend information, target processing parameters can be formulated, such as adjusting the processing speed or optimizing the equipment maintenance plan to respond to the identified anomalies and improve process stability. This integrated processing method can ensure the stability of the processing process and product quality, and improve production efficiency.

[0067] In one example, before using a preset support vector machine model to analyze each frequency component in the spectral data and identify abnormal frequency components in the spectral data, the adaptive control method also includes: using a radial basis function kernel to construct the support vector machine model, and setting the C value and gamma value of the support vector machine model; obtaining training data, and using the training data to train the support vector machine model to optimize the minimum classification error rate of the support vector machine model, wherein the training data includes: pre-prepared spectral data and process status labels corresponding to the frequency components in the pre-prepared spectral data.

[0068] Before identifying abnormal frequency components in spectral data, the support vector machine (SVM) model needs to be accurately constructed and trained. This process first involves using the radial basis function (RBF) kernel to build the SVM model. The radial basis function kernel is a commonly used kernel function that can map the input data to a high-dimensional space so that it is easier to find a decision boundary in the space that can separate different categories of data. The RBF kernel is particularly suitable for handling nonlinear classification problems because it can capture complex relationships in the data.

[0069] When building a support vector machine model, two key parameters need to be set: C value and gamma value. C value is a regularization parameter that controls how well the model fits the training data. A larger C value will make the model focus more on the accurate classification of the training data, which may lead to overfitting; a smaller C value will make the model focus more on generalization ability. Gamma value controls the influence range of the RBF kernel, that is, the influence distance of a single sample on other samples. A larger gamma value means that the influence range of each training sample is smaller, which may lead to overfitting; a smaller gamma value will make the influence range of each sample larger, which may lead to underfitting. Therefore, the settings of C value and gamma value need to be carefully adjusted to optimize the performance of the model.

[0070] Next, you need to obtain training data, which is used to train and optimize the SVM model. The training data includes pre-prepared spectrum data and process state labels corresponding to each frequency component in the spectrum data. Process state labels are labels associated with spectrum data that indicate the process state corresponding to a specific frequency component, such as normal operation or abnormal conditions. By combining these labels with spectrum data, the model can learn the relationship between different frequency components and process states, thereby improving the ability to classify unknown data.

[0071] The goal of training the SVM model using these training data is to optimize the model's minimum classification error rate. This means that the model minimizes classification errors when processing training data to improve its prediction accuracy for new data. The training process continuously adjusts the model parameters to find the optimal model configuration that minimizes the classification error rate. After training, the model can accurately identify abnormal frequency components when processing actual spectrum data, thus providing a basis for subsequent process adjustments.

[0072] For example, suppose we have a set of spectrum data, and the process status labels corresponding to each spectrum data include "normal" or "abnormal". First, we use the RBF kernel to build a support vector machine model, and set the C value to 1.0 and the gamma value to 0.5. Then, the model is trained with these training data to optimize the classification error rate so that the model can accurately distinguish between normal frequency components and abnormal frequency components. After training, when new spectrum data is input into the model, the model can effectively identify the abnormal frequency components therein, providing effective support for further process adjustments.

[0073] In one example, the spectrum data is organized according to a time series, each frequency component in the spectrum data is seasonally decomposed, and long-term trend information in the spectrum data is extracted, including: arranging the spectrum data in time order to form time series data, wherein the spectrum data at each time point in the time series data includes the amplitude values ​​of all frequency components; combining the timestamp of the spectrum data in the time series data and the frequency component data corresponding to the timestamp to generate a time series data frame, wherein the time series data frame includes a timestamp column and each frequency component column; using an additive seasonal decomposition model to analyze the time series data frame to obtain seasonal analysis data, wherein the seasonal analysis data includes: a long-term trend component, a seasonal component and random noise; drawing the long-term trend component as a trend graph, and identifying the long-term trend of the time series data based on the trend graph, wherein the long-term trend includes: long-term growth or long-term decline; evaluating the influence of the long-term trend component and the seasonal component on the long-term trend to obtain evaluation data, and integrating the evaluation data into the long-term trend information in the spectrum data.

[0074] That is, when extracting long-term trend information from spectrum data, the spectrum data must first be organized in time series. This means arranging the spectrum data in chronological order to form a time series data set. Each time point in the data set records the amplitude values ​​of all frequency components, forming a complete time series. This arrangement method provides a structured time series data framework for subsequent analysis, which contains the spectrum data at each time point and its corresponding timestamp.

[0075] Next, combine the timestamp and the corresponding frequency component data to generate a time series data frame. This data frame includes two columns: the timestamp column and the frequency component column. The timestamp column records the time point of the time series, and the frequency component column records the spectrum data of each time point. In this way, the time information can be matched with the amplitude information of the frequency component, providing a clear data structure for analysis.

[0076] Then, the generated time series data frame is analyzed using the additive seasonal decomposition model. The additive seasonal decomposition model decomposes the time series data into three main components: long-term trend component, seasonal component, and random noise. The long-term trend component reflects the overall trend change of the data, the seasonal component represents periodic fluctuations, and the random noise is unpredictable fluctuations. Through this decomposition, the long-term trend in the data and its changing pattern can be clearly identified.

[0077] After obtaining the seasonal analysis data, the long-term trend component is plotted as a trend graph. The trend graph shows the long-term trend of the data over time, such as long-term growth or long-term decline. By analyzing the trend graph, the long-term trend of the data can be intuitively identified. For example, if the trend graph shows a gradually rising curve, it indicates that the long-term trend of the spectrum data is increasing; if the trend graph shows a gradually falling curve, it indicates that the long-term trend is declining.

[0078] Finally, evaluate the impact of long-term trend components and seasonal components on long-term trends. During the evaluation process, it is necessary to analyze how the long-term trend components are affected by seasonal fluctuations and distinguish the effects of long-term trends and short-term seasonal changes. Integrate these evaluation results to form a comprehensive understanding of the long-term trend information of the spectrum data. This integration process can help identify and explain the main trends and changes in the data, providing a strong basis for subsequent process adjustments.

[0079] In one example, after the target processing parameters and the initial process parameters are input into a preset adaptive control algorithm, the adaptive control method further includes: comparing and calculating the target processing parameters with the ideal processing parameters through the adaptive control algorithm to obtain error data, wherein the error data includes the error value and error change rate of each parameter; converting the error data into a fuzzy set, and constructing a membership function for each fuzzy set, wherein the fuzzy set is represented by a linguistic variable, and the membership function is used to express the degree of fuzziness of the fuzzy set; calculating the degree of matching between the membership function of each fuzzy set and a plurality of preset fuzzy rules, and merging the fuzzy sets with a matching degree higher than a third preset threshold to obtain a fuzzy control output; using the maximum membership method to defuzzify the fuzzy control output to obtain a control signal output, and using the control signal output to adjust the initial process parameters to obtain the target process parameters.

[0080] After the target processing parameters and the starting process parameters are input into the preset adaptive control algorithm, the further adaptive control step involves comparing these parameters with the ideal processing parameters to calculate the error data. This process first requires comparing the target processing parameters with the ideal processing parameters to obtain the error value of each parameter and the rate of change of these error values. The error value represents the gap between the target processing parameters and the ideal parameters, while the error change rate reflects the dynamic change of the error value. These error data provide basic information for subsequent control decisions.

[0081] Next, the error data is converted into fuzzy sets for easier processing and control. Fuzzy sets are represented by linguistic variables, such as "high", "medium", and "low", which help convert the quantitative values ​​of the error data into qualitative descriptions, thereby simplifying the control process. For each fuzzy set, it is necessary to construct corresponding membership functions, which are used to describe the degree to which the error data belongs to a certain fuzzy set. The values ​​of the membership functions are between 0 and 1, indicating the degree of membership of the error data in a specific fuzzy set.

[0082] After constructing the fuzzy sets and their membership functions, it is necessary to calculate the degree of match between the membership of each fuzzy set and the preset multiple fuzzy rules. Fuzzy rules exist in the form of "if-then", such as "if the error is large and the error rate of change is fast, then the control strength needs to be increased." By evaluating the degree of match between the membership of each fuzzy set and these rules, the fuzzy sets that meet the rules can be identified. Fuzzy sets with a matching degree above a third preset threshold are merged to generate a fuzzy control output. This merging process helps to integrate the control information of multiple fuzzy sets, thereby obtaining a more comprehensive control strategy.

[0083] Finally, the maximum membership method is used to defuzzify the fuzzy control output. The maximum membership method determines the most suitable control signal by selecting the fuzzy control output with the highest membership. This control signal represents the specific measures of adjustment, such as increasing or decreasing the value of a process parameter. The obtained control signal output is used to adjust the starting process parameters to make them close to the target process parameters, thereby achieving the desired processing effect.

[0084] For example, suppose that in a nano-grinding process, the target processing parameters are grinding force and grinding depth, while the ideal processing parameters are grinding force of 100N and grinding depth of 0.01mm. After calculation, the error data is obtained as grinding force error of 10N and grinding depth error of 0.002mm. After these error data are converted into fuzzy sets, such as "high error", "medium error", etc., membership functions are constructed. According to the preset fuzzy rules, such as "if the grinding force error is high and the grinding depth error is medium, then increase the grinding force", the membership of the fuzzy set is calculated to match the degree of matching the rule, and the fuzzy sets that meet the rules are merged to obtain the fuzzy control output. Finally, the maximum membership method is used to defuzzify and obtain the final control signal, such as "increase the grinding force by 10N", and the initial process parameters are adjusted according to this control signal so that the grinding force and grinding depth are closer to the target value.

[0085] Reference Figure 3 In an embodiment of the present invention, there is also provided an adaptive control device for a nano-grinding process, comprising:

[0086] An acquisition unit 1 is used to acquire material characteristic data of a workpiece material at a nanometer level by adopting a nano-indentation material characterization method, and determine the starting process parameters of a nano-grinding process based on the material characteristic data;

[0087] A monitoring unit 2, used for processing the workpiece material using the initial process parameters and monitoring the processing parameters of the processing process in real time, wherein the processing parameters at least include: grinding data and vibration data;

[0088] An extraction unit 3 is used to perform feature analysis on the machining process parameters, extract first feature data characterizing the process state, and integrate the material feature data of the workpiece material at the nanometer level into the first feature data to obtain second feature data;

[0089] An identification unit 4 is used to identify data representing process anomalies and trend changes in the second characteristic data by using spectrum analysis technology, obtain target processing parameters, and input the target processing parameters and the initial process parameters into a preset adaptive control algorithm;

[0090] The processing unit 5 is used to obtain the target process parameters after the adaptive control algorithm adjusts the initial process parameters based on the target processing data, and to process the workpiece material using the target process parameters in real time.

[0091] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0092] Reference Figure 3In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0093] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0094] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0095] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0096] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0097] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An adaptive control method for a nano-grinding process, characterized in that: The following steps are involved: A nano-indentation material characterization method is adopted to obtain material characteristic data of the workpiece material at the nanometer level, and the material characteristic data is normalized to convert the material characteristic data into first material characteristic data of the same dimension and magnitude; the first material characteristic data is divided into multiple categories, and based on the first material characteristic data of different categories, a first scatter matrix within each category and a second scatter matrix between different categories are calculated; the first scatter matrix and the second scatter matrix are multiplied to obtain eigenvalues ​​and eigenvectors of linear discrimination of the first material characteristic data; the eigenvectors corresponding to the multiple eigenvalues ​​with the largest values ​​are determined as the target projection direction, and the first material characteristic data is projected onto the target projection direction to obtain the second material characteristic data; Inputting the second material characteristic data into a preset polynomial regression model to obtain the initial process parameters of the nano-grinding process, wherein the initial process parameters at least include: a grinding force parameter and a grinding depth parameter; The workpiece material is processed using the initial process parameters, and the process parameters of the processing are monitored in real time, wherein the process parameters at least include: grinding data and vibration data; Performing feature analysis on the machining process parameters to extract first feature data characterizing the process state, integrating the material feature data of the workpiece material at the nanometer level into the first feature data to obtain second feature data; Using spectrum analysis technology, identifying data in the second characteristic data indicating process anomalies and trend changes, obtaining target processing parameters, and inputting the target processing parameters and the initial process parameters into a preset adaptive control algorithm; The target process parameters are obtained after the adaptive control algorithm adjusts the initial process parameters based on the target processing data, and the workpiece material is processed using the target process parameters in real time.

2. The adaptive control method according to claim 1, characterized in that: Performing feature analysis on the processing parameters to extract first feature data representing the process state includes: Based on a preset wavelet basis, the processing parameters are subjected to wavelet analysis to obtain a plurality of initial components, wherein different initial components correspond to different frequency bands; Performing threshold processing on the multiple initial components, removing data above a first preset threshold and data below a second preset threshold from the multiple initial components, to obtain multiple target components; Performing wavelet reconstruction processing on the multiple target components, and performing Fourier transform processing on the multiple target components after the wavelet reconstruction processing to obtain analysis signals represented in the frequency domain; Performing spectrum analysis on the analysis signal to obtain a plurality of frequency components and a characteristic representation reflected by each frequency component; Extracting a plurality of target frequency components and amplitudes corresponding to the target frequency components based on the feature representation reflected by each frequency component; constructing a first characteristic matrix based on the multiple target frequency components and the amplitudes corresponding to the target frequency components; Performing principal component analysis on the first feature matrix, calculating a covariance matrix of the first feature matrix, and performing eigenvalue decomposition processing on the covariance matrix to obtain a plurality of principal components; A preset number of target principal components are extracted from the plurality of principal components, and the target principal components are used as the first feature data.

3. The adaptive control method according to claim 1, characterized in that: Integrating the material characteristic data of the workpiece material at the nanometer level into the first characteristic data to obtain second characteristic data, including: Normalizing the first characteristic data and the second material characteristic data so that the first characteristic data and the second material characteristic data are converted into the same dimension and magnitude; The first feature data and the second material feature data are feature-embedded into a high-dimensional vector in the same embedding space through multiple hidden layers of a multilayer perceptron to obtain multiple feature high-dimensional vectors; The embedded feature high-dimensional vectors are used as nodes of the graph network, and the Euclidean distances between the feature high-dimensional vectors are calculated. The Euclidean distances between the feature high-dimensional vectors are used to establish connections between the nodes to obtain a feature graph grid; Using a graph convolutional network to update the main feature representation of each node in the feature graph grid to identify implicit associations between features, and obtain multiple updated main feature representations; Aggregation processing is performed on main feature representations at different levels to obtain aggregated main feature vectors, and nonlinear activation processing is performed on the main feature vectors to obtain the second feature data.

4. The adaptive control method according to claim 1, characterized in that: Using spectrum analysis technology, the data representing process anomalies and trend changes in the second characteristic data are identified to obtain target processing parameters, including: Performing Fourier transformation on the second characteristic data to obtain frequency spectrum data; Using a preset support vector machine model to analyze each frequency component in the spectrum data to identify abnormal frequency components in the spectrum data; Organizing the spectrum data according to time series, performing seasonal decomposition on each frequency component in the spectrum data, and extracting long-term trend information in the spectrum data; The abnormal frequency component and the long-term trend information are integrated and processed to obtain the target processing parameter.

5. The adaptive control method according to claim 4, characterized in that: Before using a preset support vector machine model to analyze each frequency component in the spectrum data and identify abnormal frequency components in the spectrum data, the adaptive control method further includes: The support vector machine model is constructed using a radial basis function kernel, and the C value and gamma value of the support vector machine model are set; Acquire training data, and use the training data to train the support vector machine model to optimize the minimum classification error rate of the support vector machine model, wherein the training data includes: pre-prepared spectrum data and process state labels corresponding to frequency components in the pre-prepared spectrum data; and / or, Organizing the spectrum data according to time series, seasonally decomposing each frequency component in the spectrum data, and extracting long-term trend information in the spectrum data, including: Arranging the spectrum data in chronological order to form time series data, wherein the spectrum data at each time point in the time series data includes amplitude values ​​of all frequency components; Combining the timestamp of the spectrum data in the time series data with the frequency component data corresponding to the timestamp to generate a time series data frame, wherein the time series data frame includes a timestamp column and each frequency component column; Adopting an additive seasonal decomposition model to analyze the time series data frame to obtain seasonal analysis data, wherein the seasonal analysis data includes: a long-term trend component, a seasonal component and random noise; Plotting the long-term trend component as a trend graph, and identifying the long-term trend of the time series data based on the trend graph, wherein the long-term trend includes: long-term growth or long-term decline; The influence of the long-term trend component and the seasonal component on the long-term trend is evaluated to obtain evaluation data, and the evaluation data is integrated into the long-term trend information in the spectrum data.

6. The adaptive control method according to claim 1, characterized in that: After inputting the target processing parameters and the initial process parameters into a preset adaptive control algorithm, the adaptive control method further includes: By using the adaptive control algorithm, the target processing parameters are compared and calculated with the ideal processing parameters to obtain error data, wherein the error data includes the error value and error change rate of each parameter; Converting the error data into fuzzy sets, and constructing a membership function for each fuzzy set, wherein the fuzzy set is represented by a linguistic variable, and the membership function is used to express the fuzziness degree of the fuzzy set; Calculating the matching degree between the membership function of each fuzzy set and the preset multiple fuzzy rules, and merging the fuzzy sets with matching degrees higher than a third preset threshold to obtain a fuzzy control output; The fuzzy control output is defuzzified by using the maximum membership method to obtain a control signal output, and the initial process parameters are adjusted by using the control signal output to obtain the target process parameters.

7. An adaptive control device for nano-grinding process, characterized in that: include: An acquisition unit is used to acquire material characteristic data of a workpiece material at a nanometer level by adopting a nano-indentation material characterization method, and to normalize the material characteristic data to convert the material characteristic data into first material characteristic data of the same dimension and magnitude; divide the first material characteristic data into a plurality of categories, and based on the first material characteristic data of different categories, calculate a first scatter matrix within each category and a second scatter matrix between different categories; perform a product calculation on the first scatter matrix and the second scatter matrix to obtain eigenvalues ​​and eigenvectors of linear discrimination of the first material characteristic data; determine eigenvectors corresponding to a plurality of eigenvalues ​​with the largest values ​​as target projection directions, project the first material characteristic data onto the target projection directions, and obtain second material characteristic data; Inputting the second material characteristic data into a preset polynomial regression model to obtain the initial process parameters of the nano-grinding process, wherein the initial process parameters at least include: a grinding force parameter and a grinding depth parameter; A monitoring unit, used to process the workpiece material using the initial process parameters and monitor the process parameters of the process in real time, wherein the process parameters at least include: grinding data and vibration data; An extraction unit, used for performing feature analysis on the machining process parameters, extracting first feature data characterizing the process state, integrating the material feature data of the workpiece material at the nanometer level into the first feature data, and obtaining second feature data; an identification unit, configured to identify data indicative of process anomalies and trend changes in the second characteristic data by using spectrum analysis technology, obtain target processing parameters, and input the target processing parameters and the initial process parameters into a preset adaptive control algorithm; The processing unit is used to obtain the target process parameters after the adaptive control algorithm adjusts the initial process parameters based on the target processing data, and to process the workpiece material using the target process parameters in real time.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images