Thermal error analysis and prediction method for digital twin-driven heavy numerical control machine tool
Through the digital twin-driven method, a thermal distribution factor mechanism model and deep learning algorithm are constructed, and the temperature sensor deployment and data processing are optimized, which solves the problem of poor temperature sensor deployment in thermal error modeling of CNC machine tools and improves the accuracy of thermal error prediction. performance and machining accuracy.
Patent Information
- Application Number
- CN202411193469.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-28
- Publication Date
- 2024-11-22
AI Technical Summary
The lack of thermal distribution mechanism modeling guidance in thermal error modeling of existing CNC machine tools leads to poor deployment of temperature sensors and is difficult to efficiently capture the key heat source characteristics of heavy-duty CNC machine tools, affecting machining accuracy.
Using a digital twin-driven method, a thermal distribution factor mechanism model is constructed, a temperature sensor deployment is optimized, a thermal error prediction model is established through finite element analysis and deep learning algorithms, a correlation of temperature and thermal error data, optimizing temperature data acquisition and processing, and eliminating collinearity Data, improve data quality.
It realizes effective capture of key thermal distribution factors and high-quality acquisition of temperature data, improves the training accuracy of deep learning models, and ensures the accuracy and processing accuracy of thermal error prediction.
Smart Images

Figure CN119002398A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of CNC machine tool processing, and in particular to a thermal error analysis and prediction method for heavy-duty CNC machine tools driven by a digital twin. Background Art
[0002] With the vigorous development of industries such as aviation, aerospace, nuclear power and military industry, heavy-duty CNC machine tools are in great demand in the manufacturing of large-scale high-end equipment such as steam turbines, large nuclear pumps, marine propellers and large wings. As the core equipment of intelligent manufacturing, heavy-duty CNC machine tools play a vital role in modern manufacturing. ISO international standards require that the repeat positioning accuracy of high-precision machine tool processing reach 0.005mm. However, the general positioning accuracy of most CNC machine tools is currently between 0.008mm and 0.010mm, which is far behind the international advanced level of 0.002mm and 0.003mm, and this gap is more obvious in heavy-duty CNC machine tools.
[0003] Thermal error is the error caused by thermal deformation during the machining process. It is the main factor causing the low precision of heavy-duty CNC machine tools, accounting for 40%-70% of the total error. Temperature changes have a great influence on the machining accuracy of precision parts, which leads to a low quality compliance rate of workpieces under actual machining conditions. At present, in the machining process of heavy-duty machine tools in my country, there is little consideration of using thermal distribution for modeling to achieve efficient temperature measurement. Traditional temperature measurement of heavy-duty CNC machine tools still relies on the experience of engineers, and it is difficult to evaluate the validity of the data, which leads to low robustness of thermal error modeling.
[0004] In response to the urgent need to expand CNC machining measurement and thermal error modeling methods, domestic and foreign scholars have carried out extensive research work, mainly including the following three aspects: (1) Based on CNC machining experience, a large number of temperature sensors are deployed to fully cover effective measurement points, and effective key data features are captured through intelligent sensor measurement; (2) The correlation between temperature sensors is analyzed, redundant sensors are removed, and the cost and efficiency of temperature measurement are optimized; (3) Deep learning algorithms are designed, thermal error models are constructed, and accurate prediction of thermal errors is achieved. Data-driven thermal error models can effectively get rid of the high dependence of traditional thermal error modeling on machining experience, and provide a theoretical and technical basis for related research on further improving machining accuracy. However, the current deployment of temperature sensors still lacks the guidance of thermal distribution mechanism models, which makes it difficult to efficiently capture the key heat source characteristics of heavy CNC machine tools and apply them to deep learning modeling.
[0005] In the context of interdisciplinary studies, the use of emerging technologies such as sensor systems based on digital twins, big data analysis, deep learning algorithms, data visualization, edge computing, etc., combined with traditional mechanical processing mechanisms, can effectively achieve optimization of processing accuracy, efficiency and other aspects. Summary of the invention
[0006] The purpose of the present invention is to provide a digital twin driven heavy CNC machine tool thermal error analysis and prediction method to solve the above problems.
[0007] To achieve the above object, the present invention adopts the following technical solutions:
[0008] A digital twin-driven thermal error analysis and prediction method for heavy-duty CNC machine tools comprises the following steps:
[0009] S1. Construct a temperature measurement point selection model: establish a heat distribution factor mechanism model, divide the temperature measurement point deployment area based on the heat distribution factor mechanism model and the physical structure of the heavy-duty CNC machine tool, densely deploy temperature sensors in each deployment area and conduct heavy-duty CNC machine tool processing experiments under different processing parameters, obtain processing parameter information, temperature measurement point position data, temperature data and thermal error data, the temperature measurement point position data at least includes deployment area information and temperature measurement point coordinate information, the temperature data at least includes temperature value and corresponding time series, the thermal error data at least includes thermal error value and corresponding time series, analyze the influence of different processing parameters on heat distribution through correlation and make predictions, and establish a temperature measurement point selection model under different processing parameters based on finite element analysis;
[0010] S2, temperature sensor measurement optimization: based on the current processing parameters, determine the corresponding temperature measurement point selection model, use the temperature measurement point selection model to determine the temperature measurement point with the largest thermal error correlation value in each deployment area as the regional center temperature measurement point, the thermal error correlation value is the correlation between the temperature data corresponding to the temperature measurement point and the thermal error data, select the regional center temperature measurement point as the key temperature measurement point, and determine whether the thermal error correlation value of the regional center temperature measurement point is greater than the preset correlation threshold. If not, no other key temperature measurement points will be selected in the deployment area. If so, select other temperature measurement points as key temperature measurement points by the following method: normalize the temperature data and thermal error data of all temperature measurement points in the deployment area to obtain a temperature thermal error data set, construct a covariance matrix based on the temperature thermal error data set, and solve the covariance matrix eigenvalue of each temperature measurement point. The eigenvalue represents the variance contribution rate of the temperature measurement point. Combined with the temperature measurement point location information, the expected value of the temperature measurement point selection is calculated according to the following formula:
[0011]
[0012] Among them, Q k Select the expected value for the temperature measurement point, F k is the characteristic value of the kth temperature measurement point, D k is the distance from the kth temperature measurement point to the central temperature measurement point of the area, D 0is the average distance from all temperature measurement points in the deployment area to the central temperature measurement point, where a and b are constants;
[0013] Select key temperature measurement points in descending order of expected values of temperature measurement points, until the expected value of cumulative temperature measurement point selection reaches a preset threshold, traverse all deployment areas, and obtain a final set of key temperature measurement points. Deploy temperature sensors corresponding to the set of key temperature measurement points and collect temperature data. Perform collinearity processing on the collected temperature data to obtain low collinearity temperature data, and verify the collinearity processing results.
[0014] S3. Based on the time series and nonlinear characteristics of temperature data, an algorithm model for thermal error prediction is constructed, the temperature distribution and the thermal error value are associated, and the algorithm model is trained to obtain a thermal error prediction model;
[0015] S4. Obtain a thermal error prediction value using the thermal error prediction model.
[0016] Preferably, the step S1 comprises the following sub-steps:
[0017] S11. Establish a heat distribution factor mechanism model, which is used to characterize the influence of different heat sources on heat distribution. The heat sources include internal heat sources and external heat sources. The internal heat sources include the heat generated by the motor, the heat generated by the workpiece processing, the cooling capacity of the cooling system, and the heat generated by the friction between the ball screw, the spindle, the gear box, the guide rail and other components. The external heat sources include solar radiation, local heat source radiation in the environment and ambient temperature.
[0018] S12. Based on the heat distribution factor mechanism model, a heavy-duty CNC machine tool processing experiment under different processing parameters is carried out, temperature sensors and laser displacement sensors are deployed, error values of the processing spindle are collected, and the influence of different processing parameters on heat distribution is predicted through correlation analysis, wherein the processing parameters at least include spindle speed and feed speed;
[0019] S13. A temperature measurement point selection model under different processing parameters is established through finite element analysis to obtain a direct correlation between the temperature measurement point and the heat distribution.
[0020] Preferably, step S2 comprises the following sub-steps:
[0021] S21, based on the current processing parameters, determine the corresponding temperature measurement point selection model;
[0022] S22, selecting a group of temperature collection points with the greatest correlation with thermal error based on the temperature measurement point selection model, and deploying temperature sensors accordingly;
[0023] S23, using a temperature sensor to collect temperature data of each temperature measuring point;
[0024] S24, performing collinearity processing on the collected temperature data to obtain low collinearity temperature data, using a grey relationship analysis algorithm and a cross comparison analysis algorithm to obtain the correlation between the temperature data of different sensors, introducing a correlation threshold, reducing the collinearity of the temperature data between different sensors, thereby obtaining low collinearity temperature data;
[0025] S25. Select part of the collected temperature data as samples to verify the reliability of the grey relationship analysis algorithm and the selected correlation threshold.
[0026] Preferably, the step S3 comprises the following sub-steps:
[0027] S31. Based on the time series and nonlinear characteristics of temperature data, an algorithm model for thermal error prediction is constructed to associate temperature distribution with thermal error values;
[0028] S32, training the algorithm model in step S31, introducing key algorithm functions to improve the ability of fitting nonlinear data, wherein the key algorithm functions at least include an activation function and a batch normalization function;
[0029] S33, using a model parameter optimization algorithm, based on a root mean square error rule, training and optimizing the algorithm model to obtain a final thermal error prediction model.
[0030] Preferably, the step S3 further comprises the following sub-steps:
[0031] S34. Randomly extract sample temperature data, use the thermal error prediction model to obtain the prediction results, and verify the prediction accuracy and robustness of the model.
[0032] Preferably, the step S3 further comprises the following sub-steps:
[0033] S35. Implement actual processing experiments to predict the thermal error of heavy CNC machine tools, compare the predicted results with real data, and if the error is higher than a preset error threshold, optimize the parameters of the thermal error prediction model to improve the prediction accuracy.
[0034] Preferably, the algorithm model in step S31 is implemented based on a long short-term memory neural network combined with a convolutional neural network.
[0035] Preferably, the model parameter optimization algorithm is stochastic gradient descent method or Adam algorithm.
[0036] Preferably, the thermal error data is measured using a laser displacement sensor or a coordinate measuring machine.
[0037] After adopting the above technical solution, the present invention has the following beneficial effects compared with the background technology:
[0038] 1. The present invention can obtain key thermal distribution factors based on the thermal distribution mechanism of heavy-duty CNC machine tools and different processing parameter environments. By constructing a temperature measurement point selection model, the deployment position of the temperature sensor is guided and optimized based on thermal error correlation and spatial distribution, and the temperature measurement of key temperature measurement points is effectively realized. At the same time, the validity of temperature data under dynamic processing conditions is improved, and finally high-quality temperature data collection is achieved.
[0039] 2. The present invention performs collinearity processing on the collected temperature data and analyzes the collinearity of the corresponding collection points of different sensors. This can eliminate temperature data collection points with high collinearity and redundant data, and provide effective data for subsequent deep learning algorithm construction and training.
[0040] 3. The present invention constructs a thermal error prediction model driven by low collinearity data and deep learning, which improves the training accuracy of deep learning and ensures the accuracy of the prediction results, thereby enabling effective and accurate processing compensation and improving the processing accuracy of heavy-duty CNC machine tools. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] Figure 1 It is a schematic diagram of the process of the present invention.
[0043] Figure 2 Schematic diagram of the deep learning neural network structure. DETAILED DESCRIPTION
[0044] 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 embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0045] Example
[0046] The embodiment of the present invention discloses a thermal error analysis and prediction method for a heavy-duty CNC machine tool driven by a digital twin, comprising the following steps:
[0047] S1. Construct a temperature measurement point selection model: establish a heat distribution factor mechanism model, and divide the temperature measurement point deployment area based on the heat distribution factor mechanism model and the physical structure of the heavy-duty CNC machine tool. Deploy temperature sensors at dense intervals in each deployment area and conduct heavy-duty CNC machine tool processing experiments under different processing parameters to obtain processing parameter information, temperature measurement point position data, temperature data, and thermal error data. The temperature measurement point position data at least includes deployment area information and temperature measurement point coordinate information, the temperature data at least includes temperature values and corresponding time series, and the thermal error data at least includes thermal error values and corresponding time series. Through correlation analysis and prediction of the degree of influence of different processing parameters on heat distribution, a temperature measurement point selection model under different processing parameters is established based on finite element analysis. Step S1 includes the following sub-steps:
[0048] S11. Establish a heat distribution factor mechanism model, which is used to characterize the influence of different heat sources on heat distribution. The heat sources include internal heat sources and external heat sources. The internal heat sources include the heat generated by the motor, the heat generated by the workpiece processing, the cooling capacity of the cooling system, and the heat generated by the friction between the ball screw, spindle, gear box, guide rail and other components. The external heat sources include solar radiation, local heat source radiation in the environment and ambient temperature. The establishment of the heat distribution factor mechanism model is based on the principles of thermodynamics, heat conduction and heat transfer, and characterizes the key generation factors of heat distribution, including:
[0049] The heat generated by the current in the stator and rotor coils of the spindle motor and servo motor is calculated as P e =I·R 2 , P e is the current thermal power, I is the current flowing through the stator and rotor coils, and R is the resistance of the stator and rotor coils; the heat generated by the friction between the spindle and servo bearings of the spindle motor and servo motor is calculated as P f =0.0001047·n·M,P f is the heat energy generated by friction, M is the total friction torque; the heat generated by friction on the guide rail is calculated as P m =μ f ·N·V,P m is the heat generated by friction on the guide rail, μ f is the rail friction coefficient, N is the normal force, and V is the translational travel velocity.
[0050] The heat transfer process between the external structure-spindle-motion axis-tool-tool position is based on Fourier's law. The heat transfer model is expressed as:
[0051]
[0052] Where T(x, t) is the temperature at position x and time t, α is the thermal diffusivity of the material, is the second-order derivative Laplace operator, and Q(x, t) is the heat source term per unit volume.
[0053] S12. Based on the mechanism model of heat distribution factors, heavy-duty CNC machine tool processing experiments are carried out under different processing parameters. Temperature sensors and laser displacement sensors are deployed to collect the error values of the processing spindle. The influence of different processing parameters on heat distribution is predicted through correlation analysis. The processing parameters include at least spindle speed and feed speed.
[0054] S13. Through finite element analysis, a temperature measurement point selection model under different processing parameters is established to obtain a direct correlation between the temperature measurement point and the heat distribution, thereby obtaining effective key temperature measurement points. Finite element analysis is used to divide the heat distribution solution domain of heavy CNC machine tools into a finite number of small elements (such as quadrilaterals), and local approximate functions are used on each element to approximate the true solution. Then, a global equation system is constructed by assembling the contributions of all elements, and this system is solved to obtain an approximate solution on the entire domain.
[0055] S2, temperature sensor measurement optimization: Based on the current processing parameters, determine the corresponding temperature measurement point selection model, and use the temperature measurement point selection model to determine the temperature measurement point with the largest thermal error correlation value in each deployment area as the regional center temperature measurement point. The thermal error correlation value is the correlation between the temperature data corresponding to the temperature measurement point and the thermal error data. The regional center temperature measurement point is selected as the key temperature measurement point, and then other key temperature measurement points are selected based on the thermal error impact and spatial distribution. Deploy temperature sensors accordingly and collect temperature data, perform collinearity processing on the collected temperature data to obtain low collinearity temperature data, and verify the collinearity processing results. The step S2 includes the following sub-steps:
[0056] S21. Based on the current processing parameters, determine the corresponding temperature measurement point selection model.
[0057] S22. Use the temperature measurement point selection model to determine the temperature measurement point with the largest thermal error correlation value in each deployment area as the central temperature measurement point of the area. The thermal error correlation value is the correlation between the temperature data corresponding to the temperature measurement point and the thermal error data. The central temperature measurement point of the area is selected as the key temperature measurement point.
[0058] Determine whether the thermal error correlation value of the temperature measurement point in the center of the area is greater than the preset correlation threshold. If not, no other key temperature measurement points are selected in the deployment area. If yes, other temperature measurement points are selected as key temperature measurement points by the following method: normalize the temperature data and thermal error data of all temperature measurement points in the deployment area to obtain a temperature thermal error data set, build a covariance matrix based on the temperature thermal error data set, and solve the eigenvalue of the covariance matrix of each temperature measurement point. The eigenvalue represents the variance contribution rate of the temperature measurement point. Combined with the location information of the temperature measurement point, the expected value of the temperature measurement point selection is calculated according to the following formula:
[0059]
[0060] Among them, Q k Select the expected value for the temperature measurement point, F k is the characteristic value of the kth temperature measurement point, D k is the distance from the kth temperature measurement point to the central temperature measurement point of the area, D 0 is the average distance from all temperature measurement points in the deployment area to the central temperature measurement point, a and b are constants, 0.8 <a<1,0<b<0.3。
[0061] According to the order of expected values of temperature measurement points from high to low, key temperature measurement points are selected in sequence until the expected value of cumulative temperature measurement point selection reaches the preset threshold, and all deployment areas are traversed to obtain a final set of key temperature measurement points. Temperature sensors are deployed according to this set of key temperature measurement points. With the help of the temperature measurement point selection model, temperature sensors can be reasonably deployed, and the core temperature data of heavy CNC machine tools in dynamic processing environments can be efficiently collected, eliminating invalid measurement data.
[0062] S23, using a temperature sensor to collect temperature data at each temperature measuring point.
[0063] S24, collinearity processing is performed on the collected temperature data to obtain low collinearity temperature data, and the gray relationship analysis algorithm and the cross comparison analysis algorithm are used to obtain the correlation of temperature data between different sensors, and the correlation threshold is introduced to reduce the collinearity of temperature data between different sensors, so as to obtain low collinearity temperature data. The gray relationship analysis algorithm is a multi-factor statistical analysis method, which first normalizes the temperature data, then calculates the gray correlation coefficient, then calculates the correlation, and finally sorts according to the size of the correlation. The implementation method of the cross comparison analysis algorithm is to first normalize the temperature data, then pair the temperature data sets collected by different sensors and calculate the similarity (such as calculating the similarity using the Euclidean distance), and finally obtain the correlation of temperature data between different sensors. The temperature data is preprocessed by step S24, and the collinearity of the corresponding collection points of different sensors is analyzed, so that the temperature data collection points with high collinearity and redundant data can be eliminated, and effective data can be provided for the subsequent deep learning algorithm construction and training. The present invention can effectively improve the collection efficiency and data quality of temperature data by combining step S22 with step S24.
[0064] S25. Select part of the collected temperature data as samples to verify the reliability of the grey relationship analysis algorithm and the selected correlation threshold. Apply the sample to the deep learning test model to obtain the model training efficiency and accuracy, which can verify the reliability of collinearity processing on the one hand and provide a basis for sensor optimization on the other.
[0065] S3. Based on the time series and nonlinear characteristics of temperature data, an algorithm model for thermal error prediction is constructed, the temperature distribution and thermal error value are associated, and the algorithm model is trained to obtain a thermal error prediction model. Figure 2 As shown, the thermal error prediction model is based on a deep learning algorithm, and the structure of the artificial neural network used is composed of multiple input, output and hidden layers. Each layer contains units that can convert input data into information for the next layer to use for specific prediction tasks. The step S3 includes the following sub-steps:
[0066] S31. Based on the time series and nonlinear characteristics of the temperature data, an algorithm model for thermal error prediction is constructed to associate the temperature distribution with the thermal error value. In this embodiment, the algorithm model is implemented by combining a long short-term memory neural network with a convolutional neural network. The algorithm model is expressed as:
[0067]
[0068] in, is the predicted thermal error value at time t+1, is the thermal error value at time t, x t is the temperature data at time t, and W is the weight parameter.
[0069] S32, training the algorithm model in step S31, introducing key algorithm functions to improve the ability to fit nonlinear data, the key algorithm functions at least including activation functions and batch normalization functions. The activation function is used in a neural network to introduce nonlinearity on the output of neurons, so that the network can learn and simulate complex functions. In this embodiment, the activation function adopts one of the ReLU function, PReLU function, Softmax function, and Swish function.
[0070] S33, using a model parameter optimization algorithm, based on a root mean square error rule, to train and optimize the algorithm model to obtain a final thermal error prediction model. The model parameter optimization algorithm may use a stochastic gradient descent method or an Adam algorithm.
[0071] The stochastic gradient descent method is used to optimize the thermal error prediction model, which is specifically achieved through the following methods:
[0072] 1) Choose an initial point θ 0 , define θ as the model parameter of the thermal error prediction model, θ 0 For initialization parameters.
[0073] 2) Select a loss function L(θ) that measures the difference between the prediction and the actual observation under the model parameters θ of the current thermal error prediction model.
[0074] 3) Determine an appropriate learning rate η, which controls the step size of updating the parameters in each iteration.
[0075] 4) Iterative update: for each temperature data training sample i, calculate the gradient of the loss function with respect to θ:
[0076]
[0077] Among them, g i is the gradient, x i is the input temperature data, y i is the corresponding predicted value of thermal error.
[0078] Using the gradient g i And the learning rate η updates the parameters:
[0079] θ=θ 0 -η·g i
[0080] 5) Repeat the above iterative update steps until a stopping condition is met, where the stopping condition is that a predetermined number of iterations is reached or the update amplitude of θ is less than a preset threshold.
[0081] 6) Use a validation set or test set of temperature data to evaluate model performance.
[0082] The Adam algorithm is used to optimize the thermal error prediction model, which is specifically achieved through the following methods:
[0083] 1) Initialize the model parameters θ of the thermal error prediction model and initialize the first moment m 0 and the second moment v 0 is the zero vector.
[0084] 2) Define η as the learning rate, β 1 , β 2 Bias correction factor, a small constant ∈ is usually used to prevent the denominator from reaching zero.
[0085] 3) Iterative update: For each temperature data training sample i, perform the following steps:
[0086] Calculate the gradient of the gradient with respect to θ:
[0087]
[0088] Among them, g i is the gradient, x i is the input temperature data, y i is the corresponding thermal error prediction;
[0089] Update first moment estimate:
[0090] m t =β 1 ·m t-1 +(1-β 1 )·g i
[0091] Among them, m t is the estimate at step t;
[0092] Update the second moment estimate:
[0093]
[0094] Among them, v t is the estimate at step t;
[0095] Compute the first and second moments of the bias correction:
[0096]
[0097] Update the parameters θ, Among them, ∈ is a small constant.
[0098] 4) Repeat the above iterative update steps until a stopping condition is met, where the stopping condition is that a predetermined number of iterations is reached or the update amplitude of θ is less than a preset threshold.
[0099] 5) Use a validation set or test set of temperature data to evaluate model performance.
[0100] S34. Randomly extract sample temperature data, use the thermal error prediction model to obtain the prediction results, and verify the prediction accuracy and robustness of the model.
[0101] S35, implement actual processing experiments, predict the thermal error of heavy CNC machine tools, compare the predicted results with real data, and if the error is higher than a preset error threshold, optimize the parameters of the thermal error prediction model to improve the prediction accuracy. In this embodiment, the thermal error of the processing experiment is measured using a laser displacement sensor or a three-coordinate measuring machine.
[0102] S4. Obtain a thermal error prediction value using the thermal error prediction model.
[0103] The thermal error prediction method of the present invention is based on deep learning and data-driven implementation. It can combine the thermal distribution mechanism of heavy-duty CNC machine tools and different processing parameter environments to obtain key thermal distribution factors. By constructing a temperature measurement point selection model, it guides and optimizes the deployment position of the temperature sensor, effectively realizes the temperature measurement of key temperature measurement points, and at the same time improves the effectiveness of temperature data under dynamic processing conditions, and finally achieves high-quality collection of temperature data; by constructing a thermal error prediction model driven by low collinearity data and deep learning, the training accuracy of deep learning is improved, the accuracy of the prediction results is ensured, and then effective and accurate processing compensation can be performed to improve the processing accuracy of heavy-duty CNC machine tools. The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or replacements that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.
Claims
1. A digital twin-driven heavy-duty CNC machine tool thermal error analysis and prediction method, characterized in that: The following steps are involved: S1. Construct a temperature measurement point selection model: establish a heat distribution factor mechanism model, divide the temperature measurement point deployment area based on the heat distribution factor mechanism model and the physical structure of the heavy-duty CNC machine tool, densely deploy temperature sensors in each deployment area and conduct heavy-duty CNC machine tool processing experiments under different processing parameters, obtain processing parameter information, temperature measurement point position data, temperature data and thermal error data, the temperature measurement point position data at least includes deployment area information and temperature measurement point coordinate information, the temperature data at least includes temperature value and corresponding time series, the thermal error data at least includes thermal error value and corresponding time series, analyze the influence of different processing parameters on heat distribution through correlation and make predictions, and establish a temperature measurement point selection model under different processing parameters based on finite element analysis; S2, temperature sensor measurement optimization: based on the current processing parameters, determine the corresponding temperature measurement point selection model, use the temperature measurement point selection model to determine the temperature measurement point with the largest thermal error correlation value in each deployment area as the regional center temperature measurement point, the thermal error correlation value is the correlation between the temperature data corresponding to the temperature measurement point and the thermal error data, select the regional center temperature measurement point as the key temperature measurement point, and determine whether the thermal error correlation value of the regional center temperature measurement point is greater than the preset correlation threshold. If not, no other key temperature measurement points will be selected in the deployment area. If so, select other temperature measurement points as key temperature measurement points by the following method: normalize the temperature data and thermal error data of all temperature measurement points in the deployment area to obtain a temperature thermal error data set, construct a covariance matrix based on the temperature thermal error data set, and solve the covariance matrix eigenvalue of each temperature measurement point. The eigenvalue represents the variance contribution rate of the temperature measurement point. Combined with the temperature measurement point location information, the expected value of the temperature measurement point selection is calculated according to the following formula: Among them, Q k Select the expected value for the temperature measurement point, F k is the characteristic value of the kth temperature measurement point, D k is the distance from the kth temperature measurement point to the central temperature measurement point in the area, D0 is the average distance from all temperature measurement points in the deployment area to the central temperature measurement point, and a and b are constants; Select key temperature measurement points in descending order of expected values of temperature measurement points, until the expected value of cumulative temperature measurement point selection reaches a preset threshold, traverse all deployment areas, and obtain a final set of key temperature measurement points. Deploy temperature sensors corresponding to the set of key temperature measurement points and collect temperature data. Perform collinearity processing on the collected temperature data to obtain low collinearity temperature data, and verify the collinearity processing results. S3. Based on the time series and nonlinear characteristics of temperature data and thermal error data, an algorithm model for thermal error prediction is constructed, the temperature distribution and the thermal error value are associated, and the algorithm model is trained to obtain a thermal error prediction model; S4. Obtain a thermal error prediction value using the thermal error prediction model.
2. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 1, characterized in that: The step S1 comprises the following sub-steps: S11. Establish a heat distribution factor mechanism model, which is used to characterize the influence of different heat sources on heat distribution. The heat sources include internal heat sources and external heat sources. The internal heat sources include the heat generated by the motor, the heat generated by the workpiece processing, the cooling capacity of the cooling system, and the heat generated by the friction between the ball screw, the spindle, the gear box, the guide rail and other components. The external heat sources include solar radiation, local heat source radiation in the environment and ambient temperature. S12. Based on the heat distribution factor mechanism model, a heavy-duty CNC machine tool processing experiment under different processing parameters is carried out, temperature sensors and laser displacement sensors are deployed, error values of the processing spindle are collected, and the influence of different processing parameters on heat distribution is predicted through correlation analysis, wherein the processing parameters at least include spindle speed and feed speed; S13. A temperature measurement point selection model under different processing parameters is established through finite element analysis to obtain a direct correlation between the temperature measurement point and the heat distribution.
3. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 2, characterized in that: The step S2 comprises the following sub-steps: S21, based on the current processing parameters, determine the corresponding temperature measurement point selection model; S22, selecting a group of temperature collection points with the greatest correlation with thermal error based on the temperature measurement point selection model, and deploying temperature sensors accordingly; S23, using a temperature sensor to collect temperature data of each temperature measuring point; S24, performing collinearity processing on the collected temperature data to obtain low collinearity temperature data, using a grey relationship analysis algorithm and a cross comparison analysis algorithm to obtain the correlation between the temperature data of different sensors, introducing a correlation threshold, reducing the collinearity of the temperature data between different sensors, thereby obtaining low collinearity temperature data; S25. Select part of the collected temperature data as samples to verify the reliability of the grey relationship analysis algorithm and the selected correlation threshold.
4. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 3, characterized in that: The step S3 comprises the following sub-steps: S31. Based on the time series and nonlinear characteristics of temperature data, an algorithm model for thermal error prediction is constructed to associate temperature distribution with thermal error values; S32, training the algorithm model in step S31, introducing key algorithm functions to improve the ability of fitting nonlinear data, wherein the key algorithm functions at least include an activation function and a batch normalization function; S33, using a model parameter optimization algorithm, based on a root mean square error rule, training and optimizing the algorithm model to obtain a final thermal error prediction model.
5. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 4, characterized in that: The step S3 also includes the following sub-steps: S34. Randomly extract sample temperature data, use the thermal error prediction model to obtain the prediction results, and verify the prediction accuracy and robustness of the model.
6. A digital twin driven heavy CNC machine tool thermal error analysis and prediction method as claimed in claim 5, characterized in that: The step S3 also includes the following sub-steps: S35. Implement actual processing experiments to predict the thermal error of heavy CNC machine tools, compare the predicted results with real data, and if the error is higher than a preset error threshold, optimize the parameters of the thermal error prediction model to improve the prediction accuracy.
7. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 4, characterized in that: The algorithm model in step S31 is implemented based on a long short-term memory neural network combined with a convolutional neural network.
8. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 4, characterized in that: The model parameter optimization algorithm is the stochastic gradient descent method or the Adam algorithm.
9. The method for thermal error analysis and prediction of a heavy-duty CNC machine tool driven by a digital twin as claimed in claim 1, characterized in that: The thermal error data is measured using a laser displacement sensor or a three-coordinate measuring machine.
Citation Information
Cited By
Longitudinal cutting automatic lathe thermal error modeling method based on single temperature sensor
CN120630874A
Numerical control machine tool electric spindle performance evaluation and analysis method based on deep learning
CN121209422A
