Real-time error control system for numerical control machining of hardware parts

Abstract

The invention discloses a hardware part numerical control machining real-time error control system which comprises the steps that historical machining information of a target numerical control machine tool is collected, error coupling relation analysis is conducted on the basis of the data, and therefore a coupling physical error model is constructed. Meanwhile, a processing error prediction model is established by using an LSTM network, and training is carried out by using historical data, so that the trained prediction model is associated with a physical model to form a composite error analysis model. Processing monitoring information is obtained through a machine tool sensor array, and after preprocessing, the processing monitoring information is input into the composite model for error prediction; and on the basis of a prediction result, an advanced compensation amount is generated by adopting an improved GA algorithm to control the machine tool, and whether self-correction needs to be performed on the composite model is judged according to a correction effect, so that efficient and accurate error control is realized. And processing errors and machine tool part maintenance are associated through historical maintenance information to form a maintenance knowledge graph, so that the state of the numerical control machine tool can be known, and maintenance early warning is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of real-time error control, and particularly to a real-time error control system for numerical control machining of hardware components. Background Art

[0002] With the continuous upgrading of the manufacturing industry and the improvement of the intelligent level, numerical control machining technology has become a key process in the production of hardware components. However, in the process of high-precision machining, due to the combined influence of factors such as tool wear, thermal deformation, cutting force fluctuation, and machine tool vibration, machining errors often exhibit complex spatio-temporal dynamic characteristics. Traditional error control methods mostly rely on offline detection and compensation, and often it is difficult to reflect the machining state in real time, resulting in error accumulation, machining quality fluctuation, and an increase in the scrap rate, thereby affecting production efficiency and the competitiveness of enterprises.

[0003] Therefore, how to better control the real-time error during numerical control machine tool machining to improve machining quality and assist in the maintenance of numerical control machine tools is an urgent problem to be solved. Summary of the Invention

[0004] The present invention overcomes the defects of the prior art and provides a real-time error control system for numerical control machining of hardware components.

[0005] To achieve the above object, the first aspect of the present invention provides a real-time error control method for numerical control machining of hardware components, including:

[0006] Obtain the historical machining information of the target numerical control machine tool, analyze the machining error coupling relationship based on the historical machining information, and construct a coupling error physical model of the target numerical control machine tool according to the analysis result;

[0007] Construct a machining error prediction model based on the LSTM network, train the machining error prediction model according to the historical machining information, and associate the trained machining error prediction model with the coupling error physical model to form a composite error analysis model;

[0008] Monitor the real-time machining of hardware components through a sensor array installed on the target numerical control machine tool to obtain real-time machining monitoring information, preprocess the real-time machining monitoring information and input it into the composite error analysis model for real-time machining error prediction;

[0009] Based on the real-time machining error prediction result, generate an advance compensation amount through an improved GA algorithm to control the target numerical control machine tool, and judge whether it is necessary to perform self-calibration on the composite error analysis model according to the correction effect;

[0010] Obtain the historical maintenance information of the target CNC machine tool, associate the machining error with the maintenance of machine tool components to form a maintenance knowledge graph, and judge whether component maintenance is required and give early warning prompts by monitoring the real-time error change of the target CNC machine tool.

[0011] In this solution, the historical machining information of the target CNC machine tool is obtained, the coupling relationship analysis of machining errors is carried out based on the historical machining information, and the coupling error physical model of the target CNC machine tool is constructed according to the analysis results, specifically including:

[0012] Obtain the historical machining information of the target CNC machine tool, extract features from the historical machining information to obtain the machining error features, machining type features and machining machine tool state features when the target CNC machine tool performs historical machining, and obtain the historical machining feature information;

[0013] Define three types of machining error components: geometric error, thermal error and dynamic error, and construct a historical machining feature data set with machining type - machining error - machining machine tool state as the association path according to the historical machining feature information;

[0014] Use a multiple linear regression model to analyze the contribution degree of each error component to the total error in combination with the historical machining feature data set. Take geometric error, thermal error and dynamic error as independent variables and the total error as the dependent variable, and use the least squares method to fit the regression coefficients to obtain the contribution weights of each error component to the total error, that is, the coupling relationship between each error component;

[0015] Use the frequency domain analysis method to perform FFT transformation on the vibration signal based on the historical machining feature data set to obtain a spectrogram and calibrate the frequency components, and introduce the Pearson algorithm to calculate the Pearson correlation coefficient between the three types of machining errors;

[0016] Based on the error component type, sub-item error models are constructed, including geometric error model, thermal error model and dynamic error model. The coupling relationship modeling is carried out for the construction of each sub-item error model through the Pearson correlation coefficient between the three types of error sub-items, and the thermal-geometric coupling model, thermal-dynamic coupling model and dynamic-geometric coupling model are obtained;

[0017] Based on the coupling relationship between each error component, the thermal-geometric coupling model, thermal-dynamic coupling model and dynamic-geometric coupling model are integrated to construct the coupling error physical model of the target CNC machine tool, and the model parameters are corrected and optimized through the historical machining feature data set.

[0018] In this solution, a processing error prediction model is constructed based on the LSTM network, and the processing error prediction model is trained according to the historical processing information. The trained processing error prediction model is associated with the coupled error physical model to form a composite error analysis model, which specifically includes:

[0019] Obtain historical processing information, extract historical processing error features according to the historical processing information, and perform temporal correlation on the extracted historical processing error features and the remaining features to form a first data set;

[0020] Construct a processing error prediction model based on the LSTM network, initialize the input layer, hidden layer, and output layer of the long short-term memory network, and import the first data set as input features into the processing error prediction model for training;

[0021] Set a learnable embedding dictionary in the hidden layer of the long short-term memory network to represent the time points when processing errors occur, and extract the corresponding embedding representations from the embedding dictionary according to the time stamp features in the historical processing data;

[0022] Obtain the periodic embedding representation of the processing error through the extracted embedding representation, set the information at different time steps through the multi-head attention mechanism to analyze the influence degree of different features on the processing error, learn the spatio-temporal correlation relationship between the processing error and the remaining processing features, and use the backpropagation method to perform iterative training and hyperparameter adjustment on the network parameters;

[0023] Obtain the coupled error physical model of the target CNC machine tool, use the coupled error physical model as the physical model branch, and use the trained processing error prediction model as the LSTM branch;

[0024] Construct a second data set based on the historical processing information. The second data set is composed of processing error data, processing type data when errors occur, and processing machine tool status data. Obtain the difference between the error prediction results of the coupled error physical model and the processing error prediction model and the actual error through the second data set to obtain difference analysis information;

[0025] Set an adaptive weighting mechanism according to the difference analysis information, and use a parallel fusion structure to associate the coupled error physical model and the processing error prediction model to construct a composite error analysis model.

[0026] In this solution, the real-time processing monitoring information is obtained by monitoring the real-time processing of hardware components through a sensor array installed on the target CNC machine tool. After preprocessing the real-time processing monitoring information, it is input into the composite error analysis model for real-time processing error prediction, which specifically includes:

[0027] Install a sensor array on the target CNC machine tool to monitor the processing of real-time hardware parts, and obtain real-time processing monitoring information through the installed sensor array;

[0028] Performing time sequence alignment on the real-time processing monitoring information, performing time sequence alignment on various types of collected real-time processing monitoring data according to corresponding time sequence features, and obtaining time sequence alignment data;

[0029] De-noising and filtering the time series alignment data, processing the collected vibration signal by wavelet transformation method, processing the acquired temperature signal by Kalman filter, and processing the cutting force data by fourth-order Butterworth low-pass filter to obtain preliminary pre-processed data;

[0030] Calculate the data standard deviation of the preprocessed data, mark the data that are not within the preset data standard deviation range as abnormal values ​​and remove them, and introduce linear interpolation method to perform data interpolation compensation to obtain the final preprocessed data;

[0031] A composite error analysis model is obtained, and the final preprocessed data is input into the composite error analysis model for error prediction. The error prediction results of the physical model branch and the LSTM branch are obtained respectively. The weighted weights are obtained based on the adaptive weighting mechanism to perform weighted fusion on the error prediction results of each branch, and real-time processing error prediction information is output.

[0032] In this solution, the target CNC machine tool is controlled by generating an advance compensation amount based on the real-time machining error prediction result through an improved GA algorithm, and judging whether the composite error analysis model needs to be self-corrected according to the correction effect, specifically including:

[0033] Acquire real-time machining error prediction information, introduce an improved GA algorithm to analyze the compensation amount of the current prediction error, preset the objective function and constraint conditions, initialize the population through the real-time machining error prediction information, calculate the individual fitness in the initial population, perform selection, mutation and crossover operations according to the individual fitness, and obtain an initial solution;

[0034] The taboo search algorithm is introduced for improvement, the initial solution is used as the input of the taboo search algorithm, the parameters are initialized and the taboo table is set, the neighborhood solution is generated based on the neighborhood movement rule, the optimal solution is selected according to the generated neighborhood solution and the corresponding fitness value and neighborhood movement are obtained, and the taboo table is updated;

[0035] Perform iterative search and update the taboo table to obtain the optimized individuals after the search to update the population, perform iterative optimization based on the updated population, output the optimal solution, and generate the optimal advance compensation amount of the current predicted processing error based on the optimal solution;

[0036] Generate a control plan for the target CNC machine tool according to the optimal early compensation amount of the current predicted machining error, perform error compensation according to the control plan, and detect the surface error of the hardware parts in real time through the installed laser displacement sensor;

[0037] Calculate the residual error by calculating the surface error of the real-time hardware parts detected and the real-time machining error prediction information, and judge the calculated residual error with a preset threshold;

[0038] If the residual error is greater than the preset threshold, it means that the current correction effect does not meet the prediction, then trigger the model self-correction mechanism, and use the real-time machining monitoring data to correct the parameters of the composite error prediction model.

[0039] In this solution, obtain the historical maintenance information of the target CNC machine tool, associate the machining error with the maintenance of the machine tool components to form a maintenance knowledge graph, and judge whether component maintenance is required and give an early warning prompt by monitoring the real-time error change of the target CNC machine tool, specifically including:

[0040] Obtain the historical maintenance information of the target CNC machine tool, extract the maintenance type characteristics, maintenance time characteristics and maintenance status characteristics of the target CNC machine tool during maintenance from the historical maintenance information, and obtain the historical maintenance characteristic information;

[0041] Obtain the historical machining information of the target CNC machine tool, extract the historical machining error characteristics of the target CNC machine tool based on the historical machining information of the target CNC machine tool, perform time series processing to obtain the historical machining error time series, and analyze the change trend of the machining error to obtain the change trend analysis information;

[0042] Align the historical maintenance characteristic information and the change trend analysis information in time series, obtain the historical machining error characteristics corresponding to each historical maintenance event time interval based on the time series alignment result and perform association to generate a third data set;

[0043] Use the K-means clustering algorithm to classify the associated data stored in the third data set according to the maintenance type to obtain the classification information; construct a maintenance knowledge graph with the change of machining error and the maintenance of machine tool components as the association path based on the classification information;

[0044] When the target CNC machine tool performs the daily hardware parts processing task, extract the machining error characteristics of the target CNC machine tool in real time, input them into the maintenance knowledge graph for matching analysis, and obtain the matching analysis information;

[0045] Extract the historical machining error feature curve that matches the current real-time machining error feature according to the matching analysis information, and conduct an overlap degree analysis. If the overlap degree between the current real-time machining error feature and the historical machining error feature curve is within the preset standard range, it means that the current CNC machine tool needs to be repaired, and a maintenance warning prompt is generated.

[0046] The second aspect of the present invention provides a real-time error control system for the numerical control machining of hardware components. The system includes: a memory and a processor. The memory contains a program for the real-time error control method of the numerical control machining of hardware components. When the program for the real-time error control method of the numerical control machining of hardware components is executed by the processor, the following steps are implemented:

[0047] Obtain the historical machining information of the target CNC machine tool, conduct an analysis of the machining error coupling relationship based on the historical machining information, and construct a coupling error physical model of the target CNC machine tool according to the analysis result;

[0048] Construct a machining error prediction model based on the LSTM network, train the machining error prediction model according to the historical machining information, and associate the trained machining error prediction model with the coupling error physical model to form a composite error analysis model;

[0049] Monitor the real-time machining of hardware components through a sensor array installed on the target CNC machine tool to obtain real-time machining monitoring information. After preprocessing the real-time machining monitoring information, input it into the composite error analysis model for real-time machining error prediction;

[0050] Based on the real-time machining error prediction result, generate an advance compensation amount through an improved GA algorithm to control the target CNC machine tool, and judge whether it is necessary to perform self-calibration on the composite error analysis model according to the correction effect;

[0051] Obtain the historical maintenance information of the target CNC machine tool, associate the machining error with the maintenance of machine tool components to form a maintenance knowledge graph, and judge whether component maintenance is required by monitoring the real-time error change of the target CNC machine tool and give a warning prompt.

[0052] The present invention discloses a real-time error control system for numerical control machining of hardware components, including: collecting historical machining information of a target numerically controlled machine tool, analyzing the error coupling relationship based on this data, and thus constructing a coupled physical error model. At the same time, an LSTM network is used to establish a machining error prediction model, which is trained with historical data, and then the trained prediction model is associated with the physical model to form a composite error analysis model. Machining monitoring information is obtained through a machine tool sensor array, preprocessed and then input into the composite model for error prediction; based on the prediction results, an improved GA algorithm is used to generate an advance compensation amount to control the machine tool, and it is judged whether it is necessary to perform self-calibration on the composite model according to the correction effect, so as to achieve efficient and accurate error control. And a maintenance knowledge graph is constructed by associating machining errors with machine tool component maintenance through historical maintenance information to help understand the state of the numerically controlled machine tool and achieve maintenance warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments or exemplary examples of the present invention, the following will briefly introduce the drawings required for use in the embodiments or exemplary descriptions. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the drawings shown.

[0054] Figure 1 It is a flowchart of a real-time error control method for numerical control machining of hardware components provided by an embodiment of the present invention;

[0055] Figure 2 It is a flowchart of a maintenance warning method based on real-time error analysis provided by an embodiment of the present invention;

[0056] Figure 3 It is a block diagram of a real-time error control system for numerical control machining of hardware components provided by an embodiment of the present invention;

[0057] The realization, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] In order to be able to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below with reference to the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0059] Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0060] Figure 1 A flowchart of a real-time error control method for numerically controlled machining of hardware components provided by an embodiment of the present invention;

[0061] As Figure 1 shown, the present invention provides a flowchart of a real-time error control method for numerically controlled machining of hardware components, including:

[0062] S102, obtaining the historical machining information of the target numerically controlled machine tool, analyzing the coupling relationship of machining errors based on the historical machining information, and constructing a coupling error physical model of the target numerically controlled machine tool according to the analysis result;

[0063] S104, constructing a machining error prediction model based on the LSTM network, training the machining error prediction model according to the historical machining information, and associating the trained machining error prediction model with the coupling error physical model to form a composite error analysis model;

[0064] S106, monitoring the real-time machining of hardware components through a sensor array installed on the target numerically controlled machine tool to obtain real-time machining monitoring information, preprocessing the real-time machining monitoring information and inputting it into the composite error analysis model for real-time machining error prediction;

[0065] S108, generating an advance compensation amount based on the real-time machining error prediction result through an improved GA algorithm to control the target numerically controlled machine tool, and judging whether it is necessary to perform self-calibration on the composite error analysis model according to the correction effect;

[0066] S110, obtaining the historical maintenance information of the target numerically controlled machine tool, associating the machining error with the maintenance of machine tool components to form a maintenance knowledge graph, and judging whether it is necessary to perform component maintenance and giving an early warning prompt by monitoring the real-time error change of the target numerically controlled machine tool.

[0067] Further, in a preferred embodiment of the present invention, the obtaining the historical machining information of the target numerically controlled machine tool, analyzing the coupling relationship of machining errors based on the historical machining information, and constructing a coupling error physical model of the target numerically controlled machine tool specifically includes:

[0068] Obtaining the historical machining information of the target numerically controlled machine tool, extracting features from the historical machining information to obtain the machining error features, machining type features and machining machine tool state features when the target numerically controlled machine tool performs historical machining, and obtaining historical machining feature information;

[0069] Defining three types of machining error components: geometric error, thermal error and dynamic error, and constructing a historical machining feature data set with the machining type - machining error - machining machine tool state as the association path according to the historical machining feature information;

[0070] Using a multiple linear regression model and combining with the historical processing feature dataset, analyze the contribution degree of each error component to the total error. Take geometric error, thermal error, and dynamic error as independent variables, and the total error as the dependent variable. Use the least squares method to fit the regression coefficients to obtain the contribution weights of each error component to the total error, that is, the coupling relationship between each error component;

[0071] Using frequency domain analysis method, perform FFT transformation on the vibration signal based on the historical processing feature dataset to obtain a spectrogram and calibrate the frequency components, and introduce the Pearson algorithm to calculate the Pearson correlation coefficient between the three types of processing errors;

[0072] Based on the error component type, construct sub - item error models including geometric error model, thermal error model, and dynamic error model. Through the Pearson correlation coefficient between the three types of error sub - items, perform coupling relationship modeling for constructing each sub - item error model to obtain a thermal - geometric coupling model, a thermal - dynamic coupling model, and a dynamic - geometric coupling model;

[0073] Based on the coupling relationship between each error component, integrate the thermal - geometric coupling model, the thermal - dynamic coupling model, and the dynamic - geometric coupling model to construct a coupling error physical model of the target CNC machine tool, and perform model parameter correction and optimization through the historical processing feature dataset.

[0074] It should be noted that first, historical machining data of the target numerically controlled machine tool is collected, and features of these data are extracted to obtain error features, machining type features, and machine tool state features generated during the machining process, forming the basis of historical machining feature information. Then, three machining error components, namely geometric error, thermal error, and dynamic error, are defined, and a historical machining feature dataset with machining type, machining error, and machine tool state as the associated paths is constructed. Subsequently, by establishing a multiple linear regression model, with geometric, thermal, and dynamic errors as independent variables and total error as the dependent variable, and using the least squares method to fit the regression coefficients, the contribution of each error component to the total error is analyzed, and the coupling relationship between the components is obtained. At the same time, the vibration signal in the historical machining feature data is subjected to FFT transformation using the frequency domain analysis method, the frequency spectrum diagram is obtained and the frequency components are calibrated, and then the correlation coefficients between the three types of machining errors are calculated in combination with the Pearson algorithm to quantify their correlation. On this basis, sub-error models for geometric, thermal, and dynamic errors are constructed based on the error component types, and the coupling relationship of the sub-models is modeled using the Pearson correlation coefficients between the sub-items, obtaining the thermal-geometric coupling model, thermal-dynamic coupling model, and dynamic-geometric coupling model respectively. The geometric-thermal-force coupling error model can be established using the multi-body system theory, finite element analysis method, and cutting force-deformation coupling theory, covering the geometric error, thermal deformation, and cutting force-deformation coupling effect of the machine tool. Finally, integration is carried out to construct the coupling error physical model of the target numerically controlled machine tool, and the model parameters are corrected and optimized using the historical machining feature dataset to achieve more accurate error control.

[0075] Further, in a preferred embodiment of the present invention, the machining error prediction model is constructed based on the LSTM network, the machining error prediction model is trained according to the historical machining information, and the trained machining error prediction model is associated with the coupling error physical model to form a composite error analysis model, specifically including:

[0076] Obtain historical machining information, extract historical machining error features according to the historical machining information, perform temporal correlation on the extracted historical machining error features and the remaining features to form a first dataset;

[0077] Construct a machining error prediction model based on the LSTM network, initialize the input layer, hidden layer, and output layer of the long short-term memory network, and import the first dataset as input features into the machining error prediction model for training;

[0078] Set a learnable embedding dictionary in the hidden layer of the long short-term memory network to represent the time points when machining errors occur, and extract the corresponding embedding representations from the embedding dictionary according to the time stamp features in the historical machining data;

[0079] Obtain the periodic embedding representation of the processing error through the extracted embedding representation. Set the information at different time steps through the multi-head attention mechanism to analyze the influence degree of different features on the processing error, learn the spatio-temporal correlation relationship between the processing error and other processing features, and use the backpropagation method to iteratively train the network parameters and adjust the hyperparameters;

[0080] Obtain the physical model of the coupling error of the target CNC machine tool. Take the physical model of the coupling error as the physical model branch, and take the trained processing error prediction model as the LSTM branch;

[0081] Construct a second data set based on the historical processing information. The second data set is composed of processing error data, processing type data and processing machine tool state data when the error occurs. Obtain the difference between the error prediction results of the coupling error physical model and the processing error prediction model and the actual error through the second data set to obtain the difference analysis information;

[0082] Set an adaptive weighting mechanism according to the difference analysis information, and use a parallel fusion structure to associate the coupling error physical model and the processing error prediction model to construct a composite error analysis model.

[0083] It should be noted that the machining error features are extracted from the historical machining data of the target CNC machine tool, and time series correlation is performed with the remaining features to form a first data set, so as to comprehensively capture the evolution law of machining errors. Among them, the remaining features are the machining type features and the machining machine tool state features when errors occur. Then, a machining error prediction model is constructed based on the LSTM network, and the input layer, hidden layer and output layer of the LSTM network are initialized so that it can effectively learn the error change pattern in time series data. During the training process, the first data set is imported as input features into the LSTM model to establish the prediction ability for machining errors. A learnable embedding dictionary is introduced into the hidden layer of the LSTM network to represent the time points when machining errors occur, and the corresponding embedding representations are extracted from this embedding dictionary using the timestamp information in the historical machining data to enhance the model's understanding of time dependence. The periodic information of machining errors is obtained through the extracted embedding representations, and the multi-head attention mechanism is used to analyze the influence degree of data features at different time steps on machining errors, so as to learn the spatio-temporal correlation relationship between machining errors and other machining features. At the same time, the backpropagation method is used to iteratively optimize the network parameters of the LSTM model, and combined with hyperparameter adjustment to improve the prediction accuracy. On this basis, the physical model of the coupling error of the target CNC machine tool is obtained and used as the physical model branch, and the trained machining error prediction model is used as the LSTM branch to integrate the advantages of physical modeling and deep learning methods. Subsequently, a second data set is constructed based on the historical machining information, which includes machining error data, machining type information when errors occur, and machining machine tool state information. The difference analysis information is obtained by calculating the difference between the error prediction values and the actual errors of the coupling error physical model and the machining error prediction model using the second data set. An adaptive weighting mechanism is set according to the difference analysis information, and the coupling error physical model and the machining error prediction model are fused using a parallel fusion structure to construct a composite error analysis model, so as to achieve complementary advantages between physical modeling and data-driven methods and improve the accuracy of error prediction and compensation.

[0084] Furthermore, in a preferred embodiment of the present invention, the real-time machining monitoring information is obtained by monitoring the real-time machining of hardware components through a sensor array installed on the target CNC machine tool, and after preprocessing the real-time machining monitoring information, it is input into the composite error analysis model for real-time machining error prediction, which specifically includes:

[0085] Install a sensor array on the target CNC machine tool to monitor the real-time machining of hardware components, and obtain real-time machining monitoring information through the installed sensor array;

[0086] Performing time sequence alignment on the real-time processing monitoring information, performing time sequence alignment on various types of collected real-time processing monitoring data according to corresponding time sequence features, and obtaining time sequence alignment data;

[0087] De-noising and filtering the time series alignment data, processing the collected vibration signal by wavelet transformation method, processing the acquired temperature signal by Kalman filter, and processing the cutting force data by fourth-order Butterworth low-pass filter to obtain preliminary pre-processed data;

[0088] Calculate the data standard deviation of the preprocessed data, mark the data that are not within the preset data standard deviation range as abnormal values ​​and remove them, and introduce linear interpolation method to perform data interpolation compensation to obtain the final preprocessed data;

[0089] A composite error analysis model is obtained, and the final preprocessed data is input into the composite error analysis model for error prediction. The error prediction results of the physical model branch and the LSTM branch are obtained respectively. The weighted weights are obtained based on the adaptive weighting mechanism to perform weighted fusion on the error prediction results of each branch, and real-time processing error prediction information is output.

[0090] It should be noted that a sensor array is installed on the target numerically controlled machine tool to monitor the machining state of the hardware parts in real time and collect information such as vibration, temperature, and cutting force during the machining process, so as to obtain real-time machining monitoring data. During the time series alignment process, the real-time machining monitoring data collected by different sensors are synchronously processed according to the corresponding time series characteristics to generate time-consistent time series aligned data. After obtaining the time series aligned data, it is necessary to perform denoising and filtering processing on the data to remove the interference components in the signal and improve the data quality. First, the wavelet transform method is used to perform multi-scale analysis on the vibration signal to separate the noise components and retain the effective signal characteristics; second, the Kalman filter is used to smooth the temperature signal to eliminate the measurement error; in addition, the fourth-order Butterworth low-pass filter is used to process the cutting force data to filter out the high-frequency interference and obtain the preliminary preprocessed data. Then, the data standard deviation is calculated for the preliminary preprocessed data, and the outliers are removed according to the preset data standard deviation range. The data exceeding the range are marked as outliers and removed to ensure the stability of the data. For the missing data that appears after the outlier removal, the linear interpolation method is introduced for data interpolation compensation to make the data continuous and complete, so as to obtain the final preprocessed data. After completing the data preprocessing, the final preprocessed data is input into the composite error analysis model to perform real-time machining error prediction. The composite error analysis model includes a physical model branch and an LSTM branch, which respectively predict the error and combine the adaptive weighting mechanism to perform weighted fusion on the two prediction results. The adaptive weighting mechanism assigns weights to the prediction results of each branch according to the current machining state to optimize the accuracy of the error prediction. Finally, the real-time machining error prediction information is output through weighted fusion, providing accurate reference for machining error compensation and machine tool state optimization.

[0091] Furthermore, in a preferred embodiment of the present invention, the target numerically controlled machine tool is controlled by generating an advance compensation amount through an improved GA algorithm based on the real-time machining error prediction result, and it is judged whether it is necessary to perform self-correction on the composite error analysis model according to the correction effect, specifically including:

[0092] Obtain the real-time machining error prediction information, introduce the improved GA algorithm to analyze the compensation amount of the current prediction error, preset the objective function and constraints, perform population initialization through the real-time machining error prediction information, calculate the individual fitness in the initial population, and perform selection, mutation, and crossover operations according to the individual fitness to obtain the initial solution;

[0093] Introduce the tabu search algorithm for improvement, use the initial solution as the input of the tabu search algorithm, perform parameter initialization and set the tabu list, generate neighborhood solutions based on the neighborhood movement rule, select the optimal solution according to the generated neighborhood solutions and obtain the corresponding fitness value and neighborhood movement, and update the tabu list;

[0094] Perform iterative search and update the taboo list to obtain the optimized individual after search, update the population, perform iterative optimization based on the updated population, output the optimal solution, and generate the optimal early compensation amount for the current predicted machining error based on the optimal solution;

[0095] Generate a control scheme for the target CNC machine tool according to the optimal early compensation amount of the current predicted machining error, perform error compensation according to the control scheme, and detect the surface error of the hardware parts in real time through the installed laser displacement sensor;

[0096] Calculate the residual error by calculating the residual error between the detected real-time surface error of the hardware parts and the real-time machining error prediction information, and judge the calculated residual error with a preset threshold;

[0097] If the residual error is greater than the preset threshold, it means that the current correction effect does not conform to the prediction, then trigger the model self-correction mechanism, and use the real-time machining monitoring data to correct the parameters of the composite error prediction model.

[0098] It should be noted that after obtaining the real-time machining error prediction information, an improved genetic algorithm is introduced to analyze the compensation amount of the current prediction error, and the objective function and constraint conditions are set to optimize the accuracy and stability of the compensation amount calculation. Based on the real-time machining error prediction information, population initialization is performed to generate the initial population, and the fitness value of each individual is calculated to measure the advantages and disadvantages of different compensation schemes. On the basis of fitness evaluation, selection, mutation, and crossover operations are performed to improve the diversity of the population and enhance the search ability, thereby obtaining the initial solution. The tabu search algorithm is introduced for improvement, and the initial solution is used as the input of the tabu search algorithm, and parameter initialization and tabu list setting are performed to avoid falling into the local optimum. Multiple neighborhood solutions are generated based on the neighborhood movement rule, and the optimal solution is selected from them, and its fitness value is calculated at the same time. By continuously updating the tabu list, neighborhood search is iteratively executed to improve the global optimization ability. After multiple rounds of iterative optimization, the optimized individual after search is finally obtained, and the population is updated. Based on the optimized population, iterative optimization is continued, and finally the optimal solution is output, and the optimal advance compensation amount of the current predicted machining error is generated based on the optimal solution, so as to provide a more accurate error compensation strategy. According to the calculated optimal advance compensation amount, the control scheme of the target numerical control machine tool is generated and compensation control is executed to reduce the real-time machining error. After the error compensation is executed, the error on the surface of the hardware part is detected in real time by the installed laser displacement sensor to verify the compensation effect. The detected real-time error on the surface of the hardware part is compared with the real-time machining error prediction information, and the residual error is calculated. If the residual error exceeds the preset threshold, it indicates that the current correction effect does not conform to the prediction result, and the model self-correction mechanism needs to be triggered. The parameters of the composite error prediction model are updated and optimized by using the real-time machining monitoring data to improve the error prediction accuracy of the model, and the error compensation strategy is dynamically adjusted to ensure that the machining accuracy of the numerical control machine tool is always in the optimal state.

[0099] Figure 2 Flowchart of a maintenance warning method based on real-time error analysis provided by an embodiment of the present invention;

[0100] As Figure 2 shown, the present invention provides a flowchart of a maintenance warning method based on real-time error analysis, including:

[0101] S202, obtaining the historical maintenance information of the target numerical control machine tool, extracting features from the historical maintenance information to obtain the maintenance type feature, maintenance time feature, and maintenance status feature when the target numerical control machine tool is maintained, and obtaining the historical maintenance feature information;

[0102] S204. Obtain the historical processing information of the target CNC machine tool, extract the historical processing error characteristics of the target CNC machine tool based on the historical processing information of the target CNC machine tool, perform time series processing to obtain the historical processing error time series, and analyze the change trend of the processing error to obtain the change trend analysis information;

[0103] S206. Align the historical maintenance feature information and the change trend analysis information in time series, obtain the historical processing error characteristics corresponding to each historical maintenance event time interval based on the time series alignment result, and perform association to generate a third data set;

[0104] S208. Use the K-means clustering algorithm to classify the associated data stored in the third data set based on the maintenance type to obtain the classification information; construct a maintenance knowledge graph with the change of processing error and the maintenance of machine tool components as the associated path based on the classification information;

[0105] S210. When the target CNC machine tool performs the daily processing task of hardware parts, extract the processing error characteristics of the target CNC machine tool in real time, input them into the maintenance knowledge graph for matching analysis, and obtain the matching analysis information;

[0106] S212. Extract the historical processing error characteristic curve that matches the current real-time processing error characteristic according to the matching analysis information, and perform the coincidence degree analysis. If the coincidence degree between the current real-time processing error characteristic and the historical processing error characteristic curve is within the preset standard range, it means that the current CNC machine tool needs to be repaired, and then generate a maintenance warning prompt.

[0107] It should be noted that when the CNC machine tool is working, with the increase in the number of machined parts, there will inevitably be varying degrees of damage to the CNC machine tool, such as tool damage, overheating and structural deformation, etc. When such situations occur, it is necessary to promptly repair the machine tool components to ensure the production quality. At the same time, there is also a great connection between the damage of the CNC machine tool and the change of errors. For example, when the tool wear changes, the machining may cause the error to become larger, and it represents that the error compensation amount becomes larger. Thus, it can be known that the change of the error can reflect the condition of the corresponding CNC machine tool and judge whether maintenance is required. This solution extracts features such as the maintenance type, maintenance time, and maintenance status from the historical maintenance information of the target CNC machine tool to form historical maintenance feature information; at the same time, it collects historical machining information, extracts machining error features to form a historical machining error time series, and analyzes the error change trend to obtain change trend analysis information. Subsequently, the historical maintenance feature information and the error change trend are aligned in time series, so as to obtain the historical machining error features corresponding to the time intervals of each historical maintenance event, and generate a third data set. Then, using the K-means clustering algorithm, the data in the third data set is classified according to the maintenance type, and a maintenance knowledge graph with the change of machining error and the maintenance of machine tool components as the associated path is constructed. During the daily machining process, the machining error features of the CNC machine tool are collected in real time and input into the maintenance knowledge graph for matching analysis, so as to extract the historical error feature curve that matches the current real-time error feature and perform the coincidence degree analysis; if the coincidence degree between the current error feature and the historical curve meets the preset standard, it indicates that the machine tool needs to be repaired, thus generating a maintenance warning prompt, so as to provide better quality assurance for the processing of hardware components, reduce the production unqualified rate and improve the operation safety of the CNC machine tool.

[0108] Figure 3 A real-time error control system 3 for the numerical control machining of hardware components provided by an embodiment of the present invention, the system includes: a memory 31 and a processor 32. The memory 31 contains a program for the real-time error control method for the numerical control machining of hardware components. When the program for the real-time error control method for the numerical control machining of hardware components is executed by the processor 32, the following steps are implemented:

[0109] Obtain the historical machining information of the target CNC machine tool, analyze the coupling relationship of machining errors based on the historical machining information, and construct a coupling error physical model of the target CNC machine tool according to the analysis results;

[0110] Construct a machining error prediction model based on the LSTM network, train the machining error prediction model according to the historical machining information, and associate the trained machining error prediction model with the coupling error physical model to form a composite error analysis model;

[0111] Monitor the real-time machining of hardware components through the sensor array installed on the target CNC machine tool to obtain real-time machining monitoring information, preprocess the real-time machining monitoring information, and then input it into the composite error analysis model for real-time machining error prediction;

[0112] Based on the real-time machining error prediction result, generate an advance compensation amount through the improved GA algorithm to control the target CNC machine tool, and judge whether it is necessary to perform self-calibration on the composite error analysis model according to the correction effect;

[0113] Obtain the historical maintenance information of the target CNC machine tool, associate the machining error with the machine tool component maintenance to form a maintenance knowledge graph, and judge whether component maintenance is required and give a warning prompt by monitoring the real-time error change of the target CNC machine tool.

[0114] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces. The indirect coupling or communication connection of the devices or units can be electrical, mechanical, or other forms.

[0115] The units described above as separate components may or may not be physically separated. The components shown as units may or may not be physical units; they can be located in one place or distributed to multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0116] In addition, each functional unit in the embodiments of the present invention can be all integrated in a processing unit, or each unit can be separately used as a unit, or two or more units can be integrated in a unit; the above integrated units can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0117] Those of ordinary skill in the art can understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including those of the above method embodiments. The aforementioned storage medium includes various media that can store program codes, such as removable storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs.

[0118] Alternatively, if the above integrated unit is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media that can store program codes, such as removable storage devices, ROM, RAM, magnetic disks, or optical discs.

[0119] As described above, the above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A real-time error control method for numerical control machining of hardware parts, characterized in that: include: Acquire historical processing information of a target CNC machine tool, perform processing error coupling relationship analysis based on the historical processing information, and construct a coupling error physical model of the target CNC machine tool according to the analysis results; Building a machining error prediction model based on an LSTM network, training the machining error prediction model according to the historical machining information, and associating the trained machining error prediction model with a coupling error physical model to form a composite error analysis model; The real-time processing of hardware parts is monitored by a sensor array installed on a target CNC machine tool to obtain real-time processing monitoring information, and the real-time processing monitoring information is pre-processed and then input into the composite error analysis model to perform real-time processing error prediction; Based on the real-time machining error prediction result, an advance compensation amount is generated by an improved GA algorithm to control the target CNC machine tool, and whether the composite error analysis model needs to be self-corrected is determined according to the correction effect; Obtain the historical maintenance information of the target CNC machine tool, associate the processing error with the machine tool component maintenance to form a maintenance knowledge graph, and determine whether component maintenance is needed and issue an early warning by monitoring the real-time error changes of the target CNC machine tool.

2. The real-time error control method for numerical control machining of hardware parts according to claim 1 is characterized in that: The acquiring of historical processing information of the target CNC machine tool, performing a processing error coupling relationship analysis based on the historical processing information, and constructing a coupling error physical model of the target CNC machine tool according to the analysis result specifically includes: Obtain historical processing information of the target CNC machine tool, perform feature extraction on the historical processing information to obtain processing error features of the target CNC machine tool during historical processing, processing type features when errors occur, and processing machine tool status features, and obtain historical processing feature information; Three types of machining error components, namely geometric error, thermal error and dynamic error, are defined, and a historical machining feature data set with machining type-machining error-machining machine tool status as the association path is constructed according to the historical machining feature information; A multivariate linear regression model is used in combination with the historical processing feature data set to analyze the contribution of each error component to the total error. The geometric error, thermal error, and dynamic error are used as independent variables, and the total error is used as the dependent variable. The least squares method is used to fit the regression coefficient to obtain the contribution weight of each error component to the total error, that is, the coupling relationship between the error components. The frequency domain analysis method is used to perform FFT changes on the vibration signal based on the historical processing feature data set to obtain the spectrum diagram and calibrate the frequency components, and the Pearson algorithm is introduced to calculate the Pearson correlation coefficients between the three types of processing errors; Based on the error component type, the sub-item error model is constructed, including geometric error model, thermal error model and dynamic error model. The coupling relationship between the sub-item error models is modeled through the Pearson correlation coefficient between the three types of error sub-items, and the thermal-geometric coupling model, thermal-dynamic coupling model and dynamic-geometric coupling model are obtained. Based on the coupling relationship between the error components, the thermal-geometry coupling model, the thermal-dynamic coupling model and the dynamic-geometry coupling model are integrated to construct a coupling error physical model of the target CNC machine tool, and the model parameters are corrected and optimized through the historical processing feature data set.

3. The real-time error control method for numerical control machining of hardware parts according to claim 1 is characterized in that: The processing error prediction model is constructed based on the LSTM network, the processing error prediction model is trained according to the historical processing information, and the trained processing error prediction model is associated with the coupling error physical model to form a composite error analysis model, specifically including: Acquire historical processing information, extract historical processing error features according to the historical processing information, and perform time series correlation between the extracted historical processing error features and the remaining features to form a first data set; Building a processing error prediction model based on the LSTM network, initializing the input layer, hidden layer, and output layer of the long short-term memory network, and importing the first data set as input features into the processing error prediction model for training; Setting a learnable embedding dictionary in a hidden layer of the long short-term memory network to represent the time point when the processing error occurs, and extracting the corresponding embedding representation from the embedding dictionary according to the timestamp feature in the historical processing data; The periodic embedding representation of the processing error is obtained through the extracted embedding representation. The information of different time steps is set through the multi-head attention mechanism to analyze the influence of different features on the processing error, and the spatiotemporal correlation between the processing error and other processing features is learned. The back-propagation method is used to iteratively train the network parameters and adjust the hyperparameters. Obtain a coupling error physical model of a target CNC machine tool, use the coupling error physical model as a physical model branch, and use the trained machining error prediction model as an LSTM branch; constructing a second data set based on the historical processing information, the second data set consisting of processing error data, processing type data when the error occurs, and processing machine status data, obtaining a difference between an error prediction result of a coupling error physical model and a processing error prediction model and an actual error through the second data set, and obtaining difference analysis information; An adaptive weighting mechanism is set according to the difference analysis information, and a parallel fusion structure is used to associate the coupling error physical model and the machining error prediction model to construct a composite error analysis model.

4. The real-time error control method for numerical control machining of hardware parts according to claim 1 is characterized in that: The real-time hardware parts processing is monitored by a sensor array installed on a target CNC machine tool to obtain real-time processing monitoring information, and the real-time processing monitoring information is pre-processed and then input into the composite error analysis model to perform real-time processing error prediction, specifically including: Install a sensor array on the target CNC machine tool to monitor the processing of real-time hardware parts, and obtain real-time processing monitoring information through the installed sensor array; Performing time sequence alignment on the real-time processing monitoring information, performing time sequence alignment on various types of collected real-time processing monitoring data according to corresponding time sequence features, and obtaining time sequence alignment data; De-noising and filtering the time series alignment data, processing the collected vibration signal by wavelet transformation method, processing the acquired temperature signal by Kalman filter, and processing the cutting force data by fourth-order Butterworth low-pass filter to obtain preliminary pre-processed data; Calculate the data standard deviation of the preprocessed data, mark the data that are not within the preset data standard deviation range as abnormal values ​​and remove them, and introduce linear interpolation method to perform data interpolation compensation to obtain the final preprocessed data; A composite error analysis model is obtained, and the final preprocessed data is input into the composite error analysis model for error prediction. The error prediction results of the physical model branch and the LSTM branch are obtained respectively. The weighted weights are obtained based on the adaptive weighting mechanism to perform weighted fusion on the error prediction results of each branch, and real-time processing error prediction information is output.

5. The real-time error control method for numerical control machining of hardware parts according to claim 1 is characterized in that: The method of generating an advance compensation amount based on the real-time machining error prediction result by improving the GA algorithm to control the target CNC machine tool, and judging whether it is necessary to self-correct the composite error analysis model according to the correction effect, specifically includes: Acquire real-time machining error prediction information, introduce an improved GA algorithm to analyze the compensation amount of the current prediction error, preset the objective function and constraint conditions, initialize the population through the real-time machining error prediction information, calculate the individual fitness in the initial population, perform selection, mutation and crossover operations according to the individual fitness, and obtain an initial solution; The taboo search algorithm is introduced for improvement, the initial solution is used as the input of the taboo search algorithm, the parameters are initialized and the taboo table is set, the neighborhood solution is generated based on the neighborhood movement rule, the optimal solution is selected according to the generated neighborhood solution and the corresponding fitness value and neighborhood movement are obtained, and the taboo table is updated; Perform iterative search and update the taboo table to obtain the optimized individuals after the search to update the population, perform iterative optimization based on the updated population, output the optimal solution, and generate the optimal advance compensation amount of the current predicted processing error based on the optimal solution; Generate a control plan for the target CNC machine tool based on the optimal advance compensation amount of the current predicted processing error, perform error compensation according to the control plan, and detect the surface error of hardware parts in real time through the installed laser displacement sensor; Perform residual error calculation on the detected real-time surface error of the hardware parts and the real-time processing error prediction information to obtain the residual error, and judge the calculated residual error with a preset threshold; If the residual error is greater than a preset threshold, it means that the current correction effect does not meet the prediction, then the model self-correction mechanism is triggered, and the parameters of the composite error prediction model are corrected using real-time processing monitoring data.

6. The real-time error control method for numerical control machining of hardware parts according to claim 1 is characterized in that: The method of obtaining the historical maintenance information of the target CNC machine tool, associating the machining error with the maintenance of machine tool components to form a maintenance knowledge graph, and determining whether component maintenance is required and issuing an early warning prompt by monitoring the real-time error change of the target CNC machine tool, specifically includes: Acquire historical maintenance information of the target CNC machine tool, perform feature extraction on the historical maintenance information to obtain maintenance type features, maintenance time features, and maintenance status features of the target CNC machine tool when it is maintained, and obtain historical maintenance feature information; Acquire historical processing information of a target CNC machine tool, extract historical processing error characteristics of the target CNC machine tool based on the historical processing information of the target CNC machine tool, perform time series processing to obtain a time series sequence of the historical processing errors, and analyze a change trend of the processing errors to obtain change trend analysis information; Performing time series alignment on the historical maintenance feature information and the change trend analysis information, obtaining and associating historical processing error features corresponding to the time intervals of each historical maintenance event based on the time series alignment result, and generating a third data set; Using a K-means clustering algorithm to classify the associated data stored in the third data set based on the maintenance type, and obtain classification information; based on the classification information, constructing a maintenance knowledge graph with machining error changes and machine tool component maintenance as the associated path; When the target CNC machine tool performs daily hardware parts processing tasks, the processing error characteristics of the target CNC machine tool are extracted in real time and input into the maintenance knowledge graph for matching analysis to obtain matching analysis information; According to the matching analysis information, a historical machining error characteristic curve that matches the current real-time machining error characteristic is extracted, and an overlap analysis is performed. If the overlap degree between the current real-time machining error characteristic and the historical machining error characteristic curve is within a preset standard range, it means that the current CNC machine tool needs maintenance, and a maintenance warning prompt is generated.

7. A real-time error control system for CNC machining of hardware parts, characterized in that: The system comprises: a memory and a processor, wherein the memory contains a real-time error control method program for numerical control machining of hardware parts, and when the real-time error control method program for numerical control machining of hardware parts is executed by the processor, the following steps are implemented: Acquire historical processing information of a target CNC machine tool, perform processing error coupling relationship analysis based on the historical processing information, and construct a coupling error physical model of the target CNC machine tool according to the analysis results; Building a machining error prediction model based on an LSTM network, training the machining error prediction model according to the historical machining information, and associating the trained machining error prediction model with a coupling error physical model to form a composite error analysis model; The real-time processing of hardware parts is monitored by a sensor array installed on a target CNC machine tool to obtain real-time processing monitoring information, and the real-time processing monitoring information is pre-processed and then input into the composite error analysis model to perform real-time processing error prediction; Based on the real-time machining error prediction result, an advance compensation amount is generated by an improved GA algorithm to control the target CNC machine tool, and whether the composite error analysis model needs to be self-corrected is determined according to the correction effect; Obtain the historical maintenance information of the target CNC machine tool, associate the processing error with the machine tool component maintenance to form a maintenance knowledge graph, and determine whether component maintenance is needed and issue an early warning by monitoring the real-time error changes of the target CNC machine tool.

Images