A contour error prediction and compensation control method for high-precision complex aluminum component machining

Abstract

The present invention relates to the technical field of aluminum components, and specifically to a contour error prediction compensation control method for high-precision complex aluminum component processing. First, a digital twin model is constructed, including creating a high-fidelity simulation of the physical and control systems, integrating environmental sensor data and considering machine wear; then, an error prediction model is constructed, and multi-dimensional contour error prediction is achieved by inputting multi-dimensional data and environmental data into the error prediction model; further, an adaptive compensation control strategy is set, that is, according to the error prediction results output by the error prediction model, the control parameters are adjusted to achieve real-time compensation error control; finally, the processing quality results are collected regularly, and the relevant parameters of the error prediction model and the adaptive compensation control strategy are adjusted according to the collected processing quality results.

Description

Technical Field

[0001] The invention relates to the technical field of aluminum components, and in particular to a contour error prediction compensation control method for machining high-precision complex aluminum components. Background Art

[0002] High-precision complex aluminum component processing involves the use of advanced machining technology to manufacture aluminum parts with complex shapes and strict tolerance requirements. This type of processing is often used in industries such as aerospace, automobiles, precision equipment and medical devices, among which aluminum alloys have become the preferred material due to their light weight, high strength and good corrosion resistance. High-precision processing technologies include CNC milling, turning, EDM and laser cutting, which can achieve micron or even nanometer processing accuracy while maintaining material properties.

[0003] During the machining of complex aluminum components, contour errors often occur due to factors such as equipment wear, operating errors, and non-uniform properties of the workpiece material. If these errors are not controlled, they will directly affect the functionality and life of the components, especially in those application scenarios where high precision is required. Therefore, predicting and compensating contour errors is the key to ensuring product quality and improving production efficiency. By predicting upcoming errors and adjusting machining parameters or correcting machining trajectories in real time, machining accuracy can be significantly improved, scrap rate can be reduced, and material utilization efficiency can be improved.

[0004] At present, the prediction and compensation control of contour errors mainly rely on complex machine learning models and real-time data processing technology. First, the sensors installed on the machine tool monitor the force, temperature, vibration and other data in the processing process in real time, and then combine the motion data of the machine tool to train the prediction model with this information to estimate the size and trend of the error. Commonly used prediction technologies include neural networks, support vector machines and machine learning. Once the possible errors are predicted, the corresponding control system will automatically adjust the machine tool parameters or modify the tool path to compensate for these predicted errors. In addition, the advanced feedback system can continuously optimize the error prediction model according to the final inspection results of the finished product, forming an intelligent processing system that continuously learns and improves.

[0005] In existing research, a Chinese patent with application number CN201910354964.9 proposed a contour error pre-compensation method based on CNC machining path minimization correction. First, by analyzing the error response of the servo system to typical input signals, the time domain analytical relationship between the follow-up error of the machine tool feed system and the tool position instruction sequence was established, and a progressive follow-up error prediction model was developed. This model is used to accurately predict the actual tool position point of the machine tool. Secondly, based on the linear expression of Taylor expansion, a quadratic programming model is constructed. The model takes minimization of tool position point compensation as the optimization goal and takes zero contour error as the boundary constraint. By introducing Lagrange multipliers, the complex nonlinear contour error compensation problem is transformed into a solvable linear equation system. Finally, the progressive solution method is used to correct the theoretical tool position point sequence point by point, effectively realizing the pre-compensation of contour errors. Scholars Mei Le and others proposed a method for estimating and pre-compensating contour errors based on digital twin models in response to the impact of contour errors in the CNC machine tool feed system on machining accuracy. This method constructs a mechatronics joint simulation model and combines the SE attention mechanism (Squeeze-and-Excitation Networks, SE) convolutional recurrent neural network (CNN-BiLSTM) with cross-coupling compensation to form a data-driven compensation model. The contour error is predicted by the data-driven model, and the digital twin model is pre-compensated based on the prediction results. The obtained coupling control gain is then updated to the actual control system to improve the contour control performance.

[0006] Although the existing technology predicts and compensates contour errors by applying advanced machining technology and machine learning models, it fails to fully consider the impact of environmental and wear factors on contour errors, nor does it consider the dynamic integrated multi-axis interaction effects, which limits the accurate prediction and effective control of contour errors in the processing of high-precision complex aluminum components.

[0007] Therefore, a contour error prediction and compensation control method for high-precision machining of complex aluminum components is proposed. Summary of the invention

[0008] The purpose of the present invention is to provide a contour error prediction compensation control method for high-precision complex aluminum component processing. First, a digital twin model is constructed, including creating a high-fidelity simulation of the physical and control systems, integrating environmental sensor data and considering machine wear; then an error prediction model is constructed, and multi-dimensional contour error prediction is realized by inputting multi-dimensional data and environmental data into the error prediction model; further, an adaptive compensation control strategy is set, that is, according to the error prediction results output by the error prediction model, the control parameters are adjusted to realize real-time compensation error control; finally, the processing quality results are collected regularly, and the relevant parameters of the error prediction model and the adaptive compensation control strategy are adjusted according to the collected processing quality results.

[0009] To achieve the above object, the present invention provides the following technical solutions:

[0010] A contour error prediction and compensation control method for high-precision complex aluminum component processing, comprising:

[0011] Constructing a digital twin model, including creating a digital twin model of a high-fidelity simulation of a physical and control system; the digital twin model integrates environmental sensor data and takes into account machine wear to obtain an environmental adjustment factor and a wear adjustment factor;

[0012] Perform multi-dimensional contour error prediction, including constructing an error prediction model, and obtaining an error prediction result by inputting multi-dimensional data and environmental data into the error prediction model; the multi-dimensional data includes The position, speed, acceleration and rotation angle of each axis; the environmental data includes tool load data and machine tool temperature and humidity data;

[0013] Setting an adaptive compensation strategy, including implementing the adaptive compensation control strategy according to the error prediction result output by the error prediction model; the adaptive compensation control strategy includes adjusting control parameters to achieve real-time compensation error control;

[0014] A feedback loop and model updating is performed; the feedback loop includes regular collection of machining quality results.

[0015] The construction of the digital twin model includes: building a detailed simulation model including all mechanical, electronic and control components; simulating mechanical stress and wear using ANSYS simulation software; and simulating the control system using Matlab.

[0016] The integrating environmental sensor data includes integrating the environmental sensor data into the digital twin model and adjusting the model parameters in real time to reflect environmental changes; the environmental sensors include temperature sensors and humidity sensors.

[0017] The formula for the environmental adjustment coefficient is:

[0018] ;

[0019] in, It's time The environmental adjustment factor, It is The influence coefficient of environmental parameters is Represents the total number of environmental parameters, It is Real-time values ​​of environmental parameters, It is reference values ​​of environmental parameters.

[0020] The taking into account machine wear includes regularly updating historical operation data; and adjusting wear adjustment parameters in real time based on the historical operation data to adjust the mechanical part of the simulation model.

[0021] The formula for the wear adjustment factor is:

[0022] ;

[0023] in, It's time Wear adjustment factor; is the attenuation coefficient, which characterizes the degree of influence of wear on performance; It is the weight coefficient of the effect of load intensity on wear, reflecting the change of wear rate under different load conditions; is the current load intensity; It is a measure of accumulated wear and is calculated from machine operating time and load conditions.

[0024] Said The axes include active motion axes, additional motion axes and characteristic motion axes; the multi-dimensional data are collected by the CNC control system.

[0025] The training process of the error prediction model includes:

[0026] Collect data: Collect 5,000 processing data from the CNC control system; the processing data includes The position, speed, acceleration and rotation angle of each axis; record the factors affecting the processing quality; the factors affecting the processing quality include tool load and machine temperature and humidity;

[0027] Data marking: marking the processing quality result corresponding to the processing data; the processing quality result includes processing error and surface roughness;

[0028] Data preprocessing: Check the processing data and process missing values ​​and abnormal values; perform periodic conversion on the rotation angle, that is, convert the angle data into sine and cosine values ​​to process its periodicity; extract analysis features from the processing data; the analysis features include the moving average of the speed and acceleration of each axis, the position coordinates of each axis, the angle data of the sine and cosine conversion of each axis, the tool load data and the temperature and humidity data of the machine tool; normalize all features;

[0029] The 5000 processed data are divided into a training set and a validation set in a ratio of 7:3;

[0030] The error prediction model is trained by the training set; the error prediction model is verified and optimized by the verification set;

[0031] The prediction process of the error prediction model includes:

[0032] Input layer: input real-time processing data; the real-time processing data includes The position, speed, acceleration and rotation angle of each axis as well as tool load data and machine temperature and humidity data;

[0033] Feature acquisition layer: Acquisition prediction feature sets; each prediction feature set corresponds to data of an axis; the prediction feature set includes position coordinates, moving averages of velocity and acceleration, converted rotation angle data, tool load data, and machine tool temperature and humidity data;

[0034] LSTM layer: including LSTM network structures; each of the LSTM network structures is used to process one of the prediction feature sets to obtain a corresponding feature analysis result;

[0035] Feature fusion layer: receives the The feature analysis results of the LSTM network structures are integrated to obtain comprehensive feature analysis results.

[0036] Output layer: obtaining an error prediction result according to the comprehensive feature analysis result; the error prediction result is a contour error value at a certain moment in the future;

[0037] The calculation formula of the contour error value is:

[0038] ;

[0039] in, It's time The predicted contour error value is It's time The environmental adjustment factor, It's time The wear adjustment factor, It is Axis in time The difference between the theoretical and actual positions, is the number of machining axes.

[0040] The adaptive compensation control strategy includes establishing an adaptive control algorithm based on the prediction result control of the error prediction model, dynamically adjusting the control parameters according to the contour error value predicted by the error prediction model, and realizing the real-time compensation error control; the adjustment formula of the control parameters is:

[0041] ;

[0042] in, Control input, that is, at time The control actions that need to be applied; is the proportional gain; is the current error; is the integral gain; is the integral of the error, which means the sum of the errors accumulated over time; is the differential gain, is the differential of the error, indicating the rate at which the error changes. To predict the gain, is the output of the error prediction model, i.e. the prediction time The contour error value, is the axis influence coefficient, indicating the Axis pair The influence of the axis, and Respectively Real-time and reference status of the axis.

[0043] The processing quality results include error data and surface roughness;

[0044] The model update includes adjusting relevant parameters of the error prediction model and the adaptive compensation control strategy according to the collected processing quality results; the model update uses an online learning method to adapt to changes in operating conditions to ensure that the error prediction model is always synchronized with the current processing status.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. The present invention integrates environmental sensor data including temperature sensors and humidity sensors, and considers wear factors such as machine usage frequency, load intensity and usage time to accurately adjust the environmental adjustment coefficient and the wear adjustment coefficient. Such comprehensive considerations can more accurately optimize possible deviations in the manufacturing process, achieve higher accuracy and lower scrap rate, and thus improve production efficiency.

[0047] 2. The present invention constructs an error prediction model, which uses multi-axis position, velocity, acceleration and rotation angle data, and combines tool load and machine tool environment data, so that the error prediction model can more comprehensively understand and predict dynamic changes in the machining process. This comprehensive analysis of multi-dimensional data not only improves the accuracy of predicting contour errors, but also optimizes the entire machining process, improving machining efficiency and product quality.

[0048] 3. The present invention comprehensively considers the prediction results of the error prediction model and the interaction between multiple axes to dynamically adjust the control parameters and realize real-time compensation error control. This adaptive control strategy not only reduces the error caused by inter-axis interference, but also enhances the response ability of the processing system to uncertain factors, thereby ensuring high stability and high precision in complex processing tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 An example diagram of a high-precision complex aluminum component provided by an embodiment of the present invention;

[0050] Figure 2 A flowchart of a contour error prediction and compensation control method for high-precision complex aluminum component machining provided by an embodiment of the present invention;

[0051] Figure 3 A comparison diagram of the predicted contour error and the actual contour error provided by an embodiment of the present invention;

[0052] Figure 4 A comparison diagram of contour errors before and after implementing the adaptive compensation control strategy provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0054] High-precision complex aluminum components are aluminum parts that meet extremely high precision standards in size, shape and surface quality. For specific examples of high-precision complex aluminum components, see Figure 1. Such components are mainly used in industries such as aerospace, automobile manufacturing, precision machinery and medical equipment, where the accuracy and reliability of the components are very strict. The processing of high-precision complex aluminum components usually involves advanced CNC machine tool technology, precision casting or forging, and subsequent fine processing technologies such as CNC milling, turning and grinding. These processes require a high degree of technical integration and precise control systems to ensure that each step of processing can meet the predetermined design specifications.

[0055] It is crucial to predict and control the contour errors of machining, as this is directly related to the quality and performance of the final product. The existence of contour errors may cause the part dimensions to deviate from the design specifications, affecting the assembly quality and functionality of the parts. In high-end applications such as aerospace or medical devices, this deviation may cause serious safety issues. Therefore, by predicting and controlling the contour errors during machining, potential machining deviations can be identified and corrected in advance, thereby ensuring that the product meets strict quality standards.

[0056] Existing approaches include using advanced measuring instruments and techniques for real-time monitoring, as well as applying statistical process control (SPC) and machine learning techniques to optimize processing parameters and improve the predictability of the production process.

[0057] However, the existing technology does not fully consider environmental factors and wear factors. In real-time dynamic adjustment, the changes of these factors directly affect the processing accuracy. In addition, the dynamic integration of multi-axis interaction effects is particularly critical for high-precision processing, because the movement of each axis may affect the performance of other axes, and the existing technology fails to effectively solve this complex dynamic relationship. These defects lead to the fact that in the actual processing process, despite the support of advanced simulation and prediction models, the accuracy of the prediction results may still be insufficient, thus affecting the overall processing quality and efficiency.

[0058] The purpose of the present invention is to provide a contour error prediction and compensation control method for high-precision complex aluminum component processing, which is used to predict and compensate for contour errors that may exist during the processing. For a specific method flow chart, see Figure 2 . Embodiment 1

[0059] In this embodiment, a high-fidelity simulation of the physical and control systems is achieved by constructing a digital twin model, and environmental sensor data is integrated and machine wear is considered to optimize the machining process. By inputting multidimensional data and environmental data into the error prediction model, the method proposed in the present invention can accurately predict the multidimensional contour error, and adjust the control parameters in real time according to the prediction results of the error prediction model through an adaptive compensation control strategy to achieve error control. In addition, by regularly collecting machining quality results and updating model parameters, the present invention ensures continuous optimization of the machining process.

[0060] The method proposed by the present invention and its specific operation and effects in practical applications will be described in detail below.

[0061] Reference Figure 1 In S10, a digital twin model is first constructed, which realizes high-fidelity simulation of physical and control systems and integrates environmental sensor data and machine wear information; the integrated environmental sensor data and machine wear information are used to obtain an environmental adjustment coefficient and a wear adjustment coefficient;

[0062] Furthermore, the construction of the digital twin model includes: constructing a detailed simulation model including all mechanical, electronic and control components; simulating mechanical stress and wear using ANSYS simulation software; and simulating the control system using Matlab.

[0063] ANSYS's advanced finite element analysis function can accurately simulate mechanical stress and wear, optimize components to resist stress and fatigue under extreme working conditions, thereby improving product durability and safety; Matlab's control system simulation environment allows accurate evaluation and optimization of system control strategies to ensure the stability and efficiency of system operation. This comprehensive simulation method not only speeds up the development cycle, but also significantly reduces the cost of trial and error, providing solid technical support for the manufacture of high-performance complex mechanical systems.

[0064] In this embodiment, the details of the mechanical structure, electronic control and software behavior of the machine tool are comprehensively considered, ANSYS simulation software is used to simulate physical stress, thermal effect and wear, and Matlab is used to simulate the control system to ensure the coordinated work of the control strategy and the physical model.

[0065] Furthermore, the integrated environmental sensor data includes integrating the environmental sensor data into the digital twin model and adjusting the model parameters in real time to reflect environmental changes; the environmental sensors include temperature sensors and humidity sensors.

[0066] By integrating environmental sensor data such as temperature sensors and humidity sensors into the simulation model, the present invention can adjust model parameters in real time to accurately reflect environmental changes. This real-time data integration improves the responsiveness and accuracy of the model, allowing the processing process to be optimized according to actual environmental conditions, thereby ensuring the consistency of processing quality in different environments, reducing processing errors caused by environmental fluctuations, and significantly improving production efficiency and product reliability.

[0067] The formula for the environmental adjustment coefficient is:

[0068] ;

[0069] in, It's time The environmental adjustment factor, It is The influence coefficient of environmental parameters is Represents the total number of environmental parameters, It is Real-time values ​​of environmental parameters, It is reference values ​​of environmental parameters.

[0070] The environmental adjustment coefficient formula proposed in this embodiment systematically considers the specific impact of various environmental parameters on the processing process, and dynamically calculates the environmental adjustment coefficient by real-time monitoring and comparing the actual value of the environmental parameter with the reference value. This method allows the processing equipment to automatically adapt to environmental changes and optimize operating conditions, thereby ensuring that the processing accuracy is not negatively affected by temperature fluctuations or humidity changes.

[0071] Further, the taking into account machine wear includes regularly updating historical operation data, and adjusting wear adjustment parameters in real time based on the historical operation data to adjust the mechanical part of the simulation model.

[0072] This method can effectively reflect the actual wear of the machine, dynamically optimize processing parameters, extend the service life of the equipment, and ensure the continuous stability of processing accuracy. By timely updating and adjusting wear parameters, the processing error caused by wear can be significantly reduced, and production efficiency and product quality can be improved.

[0073] The formula for the wear adjustment factor is:

[0074] ;

[0075] in, It's time Wear adjustment factor; is the attenuation coefficient, which characterizes the degree of influence of wear on performance; It is the weight coefficient of the effect of load intensity on wear, reflecting the change of wear rate under different load conditions; is the current load intensity; It is a measure of accumulated wear and is calculated from machine operating time and load conditions.

[0076] In this embodiment, This parameter represents the accumulated wear of the machine tool. It is calculated based on the historical operation data of the machine tool. Usually this value increases with the increase of usage time, indicating that the wear degree gradually deepens with the continuous use of the machine tool. This value can be calculated based on the total time of machine tool operation or the accumulated time of specific operation, thus reflecting the impact of usage time on machine tool performance.

[0077] This parameter represents the current load intensity of the machine tool. In machine tool operation, load intensity can be regarded as an indicator of frequency of use, because it not only reflects the workload of the machine tool at a certain point in time, but also implies the degree to which the machine tool is frequently used. High load intensity usually means that the machine tool has been subjected to a high workload in a short period of time, which is equivalent to a high frequency of use to a certain extent.

[0078] is a weighting factor used to adjust the calculation of the effect of load intensity on wear. It indicates that the wear rate may vary significantly under different load conditions. High frequency of use conditions (such as frequent high speed or high load operation) may lead to faster wear accumulation, while The value of is used to quantify this effect.

[0079] By comprehensively considering factors such as the frequency of use, load intensity, and cumulative operating time of the machine tool, the degree of wear of the machine tool can be effectively evaluated. In this formula, the attenuation coefficient reveals the specific impact of wear on the performance of the machine tool, while the weight coefficient of the impact of load intensity on wear reflects the change in wear rate under different working conditions. Through this dynamic evaluation and adjustment, potential faults and wear problems can be discovered in time, so that preventive measures can be taken.

[0080] Reference Figure 1 In S20, an error prediction model is then constructed to implement multi-dimensional contour error prediction by inputting multi-dimensional data and environmental data into the error prediction model;

[0081] The multidimensional data includes The position, speed, acceleration and rotation angle of each axis; the environmental data includes tool load data and machine tool temperature and humidity data;

[0082] Said The axes include active motion axes, additional motion axes and characteristic motion axes; the multi-dimensional data are collected by the CNC control system.

[0083] In this embodiment, It is 7, including 3 active motion axes, 3 additional motion axes and 1 characteristic motion axis; the specific description is as follows:

[0084] The specific description is shown in Table 1:

[0085]

[0086] Among them, 3 active motion axes constitute the basic motion capability of the machine tool, allowing the tool to be accurately positioned in three-dimensional space. 3 additional motion axes allow the tool to rotate, thereby achieving more complex processing operations; 1 special motion axis is specially configured to meet the specific construction requirements of aluminum components, enabling it to perform customized processing applications.

[0087] Due to the high-precision and complex structure of aluminum components, special motion axes are usually required to meet delicate processing requirements, such as controlling the tool to move along the designed nonlinear path to adapt to the complex geometry of the blade and ensure that each part can meet the designed accuracy requirements while maintaining efficient cutting speed and optimized cutting depth.

[0088] In the process of machining high-precision complex aluminum components, due to the complexity of the task, the machine tool needs to be equipped with multiple types of motion axes. In the analysis process of this embodiment, the application of active motion axes, additional motion axes and characteristic motion axes is considered to ensure the accuracy and flexibility of the machining process.

[0089] Furthermore, the training process of the error prediction model includes:

[0090] Collect data: Collect 5,000 processing data from the CNC control system; the processing data includes The position, speed, acceleration and rotation angle of each axis; record the factors affecting the processing quality; the factors affecting the processing quality include tool load and machine temperature and humidity;

[0091] Data marking: marking the processing quality result corresponding to the processing data; the processing quality result includes processing error and surface roughness;

[0092] Data preprocessing: Check the processing data and process missing values ​​and abnormal values; perform periodic conversion on the rotation angle, that is, convert the angle data into sine and cosine values ​​to process its periodicity; extract analysis features from the processing data; the analysis features include the moving average of the speed and acceleration of each axis, the position coordinates of each axis, the angle data of the sine and cosine conversion of each axis, the tool load data and the temperature and humidity data of the machine tool; normalize all features;

[0093] The 5000 processed data are divided into a training set and a validation set in a ratio of 7:3;

[0094] The error prediction model is trained by the training set; the error prediction model is verified and optimized by the verification set;

[0095] Since angle data is cyclical, this means that the values ​​360 degrees and 0 degrees are actually the same position. Machine learning models that use raw angle values ​​directly may not be able to effectively recognize this cyclical nature, leading to inaccurate predictions. For example, the model may mistakenly believe that going from 359 degrees to 1 degree is a huge change, when in fact it is a very small change.

[0096] In order to enable the loss prediction model to correctly understand the periodicity of the angle, the angle data is converted in this embodiment. Specifically, the sine function and cosine function are used to convert each angle value into two values ​​(sine value and cosine value). This conversion can maintain the periodicity of the angle while providing the model with more continuous and smooth data.

[0097] The prediction process of the error prediction model includes:

[0098] Input layer: input real-time processing data; the real-time processing data includes The position, speed, acceleration and rotation angle of each axis as well as tool load data and machine temperature and humidity data;

[0099] Feature acquisition layer: Acquisition prediction feature sets; each prediction feature set corresponds to data of an axis; the prediction feature set includes position coordinates, moving averages of velocity and acceleration, converted rotation angle data, tool load data, and machine tool temperature and humidity data;

[0100] LSTM layer: including LSTM network structures; each of the LSTM network structures is used to process one of the prediction feature sets to obtain a corresponding feature analysis result;

[0101] Feature fusion layer: receives the The feature analysis results of the LSTM network structures are integrated to obtain comprehensive feature analysis results.

[0102] Output layer: obtaining an error prediction result according to the comprehensive feature analysis result; the error prediction result is a contour error value at a certain moment in the future;

[0103] The calculation formula of the contour error value is:

[0104] ;

[0105] in, It's time The predicted contour error value is It's time The environmental adjustment factor, It's time The wear adjustment factor, It is Axis in time The difference between the theoretical and actual positions, is the number of machining axes.

[0106] In this embodiment, the comparison diagram of the predicted contour error and the actual contour error is shown in Figure 3 .

[0107] The error prediction model proposed in this embodiment integrates the interactive data of multiple axes, including the temperature and humidity, position, speed, acceleration and rotation angle of the machine tool and the tool load. The model uses the advanced LSTM network structure to perform feature analysis on the dynamic data of each axis, so that it can comprehensively consider the interaction and influence between the axes, thereby accurately adjusting the error compensation during the processing. This integration and intelligent analysis of multi-dimensional data significantly improves the processing accuracy, reduces the possibility of error generation, and ensures the quality and efficiency of the processing of complex parts.

[0108] Reference Figure 1 In S30, further, setting an adaptive compensation strategy includes implementing the adaptive compensation control strategy according to the error prediction result output by the error prediction model; the adaptive compensation control strategy includes adjusting control parameters to achieve real-time compensation error control;

[0109] Furthermore, the adaptive compensation control strategy includes establishing an adaptive control algorithm based on the prediction result control of the error prediction model, dynamically adjusting the control parameters according to the contour error value predicted by the error prediction model, and realizing the real-time compensation error control; the adjustment formula of the control parameters is:

[0110] ;

[0111] in, Control input, that is, at time The control actions that need to be applied include position, velocity or acceleration adjustment inputs of the machine tool; is the proportional gain, which is used to adjust the current error The response strength of proportional control is to directly adjust the control output according to the error size. The larger the error, the stronger the adjustment. is the current error, which refers to the time The difference between the target profile and the actual profile, i.e. the deviation between the set point and the actual output; is the integral gain, which responds to the accumulation of error and helps to eliminate the steady-state error. Integral control takes into account the cumulative effect of error and can continue to act on the system until the error is zero; It is the integral of the error, which means the sum of the accumulated errors over time, and has a good adjustment effect on the persistent small errors; The differential gain responds to the speed of error change, helps predict the future behavior of the system, and resists sudden changes in error. The differential control focuses on the error change rate, which helps to suppress overshoot and oscillation of the system. It is the differential of the error, indicating the rate of error change. By monitoring the rate of error change, the differential control can make preventive adjustments before the error changes significantly. Forecast gain is used to enhance the response to future errors. By incorporating predictions of future errors into current control decisions, control measures can be adjusted in advance to mitigate or avoid predicted errors; is the output of the error prediction model, i.e. the prediction time The contour error value can help the control system foresee future changes and respond in advance, enhancing the adaptability and robustness of the system. is the axis influence coefficient, indicating the Axis pair The influence of the axis, and Respectively Real-time and reference status of the axis.

[0112] In this embodiment, the contour error comparison diagram before and after the implementation of the adaptive compensation control strategy is shown in Figure 4 .

[0113] The adaptive compensation control strategy takes into account the prediction gain and prediction time. The contour error value can be used to predict the future error development trend. This design enables the control module to adjust the control strategy more accurately to prevent impending errors. By further considering the interaction between multiple axes, the accuracy of the compensation strategy is improved, and the coordinated control between multiple axes is achieved, thus ensuring high precision and high efficiency of the machining process.

[0114] Reference Figure 1 In S40, finally, a feedback loop and a model update are performed; the feedback loop includes regularly collecting machining quality results. The machining quality results include error data and surface roughness;

[0115] The model update includes adjusting relevant parameters of the error prediction model and the adaptive compensation control strategy according to the collected processing quality results; the model update uses an online learning method to adapt to changes in operating conditions to ensure that the error prediction model is always synchronized with the current processing status.

[0116] In this embodiment, by regularly collecting machining quality results, including error data and surface roughness, the error prediction model and its adaptive compensation control strategy can be evaluated and adjusted in real time. The model update adopts an online learning method, so that the prediction model can adapt to changes in operating conditions and ensure continuous synchronization with the current machining state. This dynamic adjustment process not only improves the reaction speed and efficiency of the machining operation, but also ensures that high-quality machining results are maintained in a constantly changing production environment.

[0117] Embodiment 2

[0118] Reference Figure 1 In S10, a digital twin model is first constructed, which realizes high-fidelity simulation of physical and control systems and integrates environmental sensor data and machine wear information; the integrated environmental sensor data and machine wear information are used to obtain an environmental adjustment coefficient and a wear adjustment coefficient;

[0119] Furthermore, the construction of the digital twin model includes: constructing a detailed simulation model including all mechanical, electronic and control components; simulating mechanical stress and wear using ANSYS simulation software; and simulating the control system using Matlab.

[0120] ANSYS's advanced finite element analysis capabilities can accurately simulate mechanical stress and wear, optimize components to resist stress and fatigue under extreme working conditions, thereby improving product durability and safety;

[0121] Matlab's control system simulation environment allows accurate evaluation and optimization of system control strategies to ensure the stability and efficiency of system operation. This comprehensive simulation method not only speeds up the development cycle, but also significantly reduces the cost of trial and error, providing solid technical support for the manufacture of high-performance complex mechanical systems.

[0122] Furthermore, the integrating environmental sensor data includes integrating the environmental sensor data into a simulation model and adjusting model parameters in real time to reflect environmental changes; the environmental sensors include temperature sensors and humidity sensors.

[0123] The formula for the environmental adjustment coefficient is:

[0124] ;

[0125] in, It's time The environmental adjustment factor, It is The influence coefficient of environmental parameters is Represents the total number of environmental parameters, It is Real-time values ​​of environmental parameters, It is reference values ​​of environmental parameters.

[0126] Furthermore, the consideration of machine wear includes regularly updating a wear model based on historical operation data, and adjusting the mechanical part of the simulation model by adjusting wear adjustment parameters in real time.

[0127] The formula for the wear adjustment factor is:

[0128] ;

[0129] in, It's time Wear adjustment factor; is the attenuation coefficient, which characterizes the degree of influence of wear on performance; It is the weight coefficient of the effect of load intensity on wear, reflecting the change of wear rate under different load conditions; is the current load intensity; It is a measure of accumulated wear and is calculated from machine operating time and load conditions.

[0130] In this embodiment, This parameter represents the accumulated wear of the machine tool. It is calculated based on the historical operation data of the machine tool. Usually this value increases with the increase of usage time, indicating that the wear degree gradually deepens with the continuous use of the machine tool. This value can be calculated based on the total time of machine tool operation or the accumulated time of specific operation, thus reflecting the impact of usage time on machine tool performance.

[0131] This parameter represents the current load intensity of the machine tool. In machine tool operation, load intensity can be regarded as an indicator of frequency of use, because it not only reflects the workload of the machine tool at a certain point in time, but also implies the degree to which the machine tool is frequently used. High load intensity usually means that the machine tool has been subjected to a high workload in a short period of time, which is equivalent to a high frequency of use to a certain extent.

[0132] is a weighting factor used to adjust the calculation of the effect of load intensity on wear. It indicates that the wear rate may vary significantly under different load conditions. High frequency of use conditions (such as frequent high speed or high load operation) may lead to faster wear accumulation, while The value of is used to quantify this effect

[0133] Reference Figure 1 In S20, an error prediction model is then constructed to implement multi-dimensional contour error prediction by inputting multi-dimensional data and environmental data into the error prediction model;

[0134] The multidimensional data includes The position, speed, acceleration and rotation angle of each axis; the environmental data includes tool load data and machine tool temperature and humidity data;

[0135] Said The axes include active motion axes, additional motion axes and characteristic motion axes; the multi-dimensional data are collected by the CNC control system.

[0136] In this embodiment, It is 7, including 3 active motion axes, 3 additional motion axes and 1 characteristic motion axis; the specific description is shown in Table 2:

[0137]

[0138] Among them, 3 active motion axes constitute the basic motion capability of the machine tool, allowing the tool to be accurately positioned in three-dimensional space. 3 additional motion axes allow the tool to rotate, thereby achieving more complex processing operations; 1 special motion axis is specially configured to meet the specific construction requirements of aluminum components, enabling it to perform customized processing applications.

[0139] Furthermore, the training process of the error prediction model includes:

[0140] Collect data: Collect 5,000 processing data from the CNC control system; the processing data includes The position, speed, acceleration and rotation angle of each axis; record the factors affecting the processing quality; the factors affecting the processing quality include tool load and machine temperature and humidity;

[0141] Data marking: marking the processing quality result corresponding to the processing data; the processing quality result includes processing error and surface roughness;

[0142] Data preprocessing: Check the processing data and process missing values ​​and abnormal values; perform periodic conversion on the rotation angle, that is, convert the angle data into sine and cosine values ​​to process its periodicity; extract analysis features from the processing data; the analysis features include the moving average of the speed and acceleration of each axis, the position coordinates of each axis, the angle data of the sine and cosine conversion of each axis, the tool load data and the temperature and humidity data of the machine tool; normalize all features;

[0143] The 5000 processed data are divided into a training set and a validation set in a ratio of 7:3;

[0144] The error prediction model is trained by the training set; the error prediction model is verified and optimized by the verification set;

[0145] Since angle data is cyclical, this means that the values ​​360 degrees and 0 degrees are actually the same position. Machine learning models that use raw angle values ​​directly may not be able to effectively recognize this cyclical nature, leading to inaccurate predictions. For example, the model may mistakenly believe that going from 359 degrees to 1 degree is a huge change, when in fact it is a very small change.

[0146] In order to enable the loss prediction model to correctly understand the periodicity of the angle, the angle data is converted in this embodiment. Specifically, the sine function and cosine function are used to convert each angle value into two values ​​(sine value and cosine value). This conversion can maintain the periodicity of the angle while providing the model with more continuous and smooth data.

[0147] The prediction process of the error prediction model includes:

[0148] Input layer: input real-time processing data; the real-time processing data includes The position, speed, acceleration and rotation angle of each axis as well as tool load data and machine temperature and humidity data;

[0149] Feature acquisition layer: Acquisition prediction feature sets; each prediction feature set corresponds to data of an axis; the prediction feature set includes position coordinates, moving averages of velocity and acceleration, converted rotation angle data, tool load data, and machine tool temperature and humidity data;

[0150] LSTM layer: including LSTM network structures; each of the LSTM network structures is used to process one of the prediction feature sets to obtain a corresponding feature analysis result;

[0151] Feature fusion layer: receives the The feature analysis results of the LSTM network structures are integrated to obtain comprehensive feature analysis results.

[0152] Output layer: obtaining an error prediction result according to the comprehensive feature analysis result; the error prediction result is a contour error value at a certain moment in the future;

[0153] The calculation formula of the contour error value is:

[0154] ;

[0155] in, It's time The predicted contour error value is It's time The environmental adjustment factor, It's time The wear adjustment factor, It is Axis in time The difference between the theoretical and actual positions, is the number of machining axes.

[0156] Reference Figure 1 In S30, further, setting an adaptive compensation strategy includes implementing the adaptive compensation control strategy according to the error prediction result output by the error prediction model; the adaptive compensation control strategy includes adjusting control parameters to achieve real-time compensation error control;

[0157] Furthermore, the adaptive compensation control strategy includes establishing an adaptive control algorithm based on the prediction result control of the error prediction model, dynamically adjusting the control parameters according to the contour error value predicted by the error prediction model, and realizing the real-time compensation error control; the adjustment formula of the control parameters is:

[0158] ;

[0159] in, Control input, that is, at time The control actions that need to be applied include position, velocity or acceleration adjustment inputs of the machine tool; is the proportional gain, which is used to adjust the current error The response strength of proportional control is to directly adjust the control output according to the error size. The larger the error, the stronger the adjustment. is the current error, which refers to the time The difference between the target profile and the actual profile, i.e. the deviation between the set point and the actual output; is the integral gain, which responds to the accumulation of error and helps to eliminate the steady-state error. Integral control takes into account the cumulative effect of error and can continue to act on the system until the error is zero; It is the integral of the error, which means the sum of the accumulated errors over time, and has a good adjustment effect on the persistent small errors; The differential gain responds to the speed of error change, helps predict the future behavior of the system, and resists sudden changes in error. The differential control focuses on the error change rate, which helps to suppress overshoot and oscillation of the system. It is the differential of the error, indicating the rate of error change. By monitoring the rate of error change, the differential control can make preventive adjustments before the error changes significantly. Forecast gain is used to enhance the response to future errors. By incorporating predictions of future errors into current control decisions, control measures can be adjusted in advance to mitigate or avoid predicted errors; is the output of the error prediction model, i.e. the prediction time The contour error value can help the control system foresee future changes and respond in advance, enhancing the adaptability and robustness of the system. is the axis influence coefficient, indicating the Axis pair The influence of the axis, and Respectively Real-time and reference status of the axis.

[0160] Reference Figure 1 In S40, finally, a feedback loop and a model update are performed; the feedback loop includes regularly collecting machining quality results. The machining quality results include error data and surface roughness;

[0161] The model update includes adjusting relevant parameters of the error prediction model and the adaptive compensation control strategy according to the collected processing quality results; the model update uses an online learning method to adapt to changes in operating conditions to ensure that the error prediction model is always synchronized with the current processing status.

[0162] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A contour error prediction and compensation control method for high-precision complex aluminum component processing, characterized in that: include: Constructing a digital twin model, including creating a digital twin model of a high-fidelity simulation of a physical and control system; the digital twin model integrates environmental sensor data and takes into account machine wear to obtain an environmental adjustment factor and a wear adjustment factor; Perform multi-dimensional contour error prediction, including constructing an error prediction model, and obtaining an error prediction result by inputting multi-dimensional data and environmental data into the error prediction model; the multi-dimensional data includes The position, speed, acceleration and rotation angle of each axis; the environmental data includes tool load data and machine tool temperature and humidity data; Setting an adaptive compensation control strategy, including implementing the adaptive compensation control strategy according to the error prediction result output by the error prediction model; the adaptive compensation control strategy includes adjusting control parameters to achieve real-time compensation error control; Performing a feedback loop and model updating; the feedback loop includes regularly collecting machining quality results; The formula for the wear adjustment factor is: ； in, It's time Wear adjustment factor; is the attenuation coefficient, which characterizes the degree of influence of wear on performance; It is the weight coefficient of the effect of load intensity on wear, reflecting the change of wear rate under different load conditions; is the current load intensity; It is a measure of accumulated wear and is calculated from machine operating time and load conditions.

2. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The construction of the digital twin model includes: building a detailed simulation model including all mechanical, electronic and control components; simulating mechanical stress and wear using ANSYS simulation software; and simulating the control system using Matlab.

3. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The integrating environmental sensor data includes integrating the environmental sensor data into the digital twin model and adjusting the model parameters in real time to reflect environmental changes; the environmental sensors include temperature sensors and humidity sensors.

4. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The formula for the environmental adjustment coefficient is: ； in, It's time The environmental adjustment factor, It is The influence coefficient of environmental parameters is Represents the total number of environmental parameters, It is Real-time values ​​of environmental parameters, It is reference values ​​of environmental parameters.

5. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The taking into account machine wear includes regularly updating historical operation data; and adjusting wear adjustment parameters in real time based on the historical operation data to adjust the mechanical part of the simulation model.

6. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: Said The axes include active motion axes, additional motion axes and characteristic motion axes; the multi-dimensional data are collected by the CNC control system.

7. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The training process of the error prediction model includes: Collect data: Collect 5,000 processing data from the CNC control system; the processing data includes The position, speed, acceleration and rotation angle of each axis; record the factors affecting the processing quality; the factors affecting the processing quality include tool load and machine temperature and humidity; Data marking: marking the processing quality result corresponding to the processing data; the processing quality result includes processing error and surface roughness; Data preprocessing: Check the processing data and process missing values ​​and abnormal values; perform periodic conversion on the rotation angle, that is, convert the angle data into sine and cosine values ​​to process its periodicity; extract analysis features from the processing data; the analysis features include the moving average of the speed and acceleration of each axis, the position coordinates of each axis, the angle data of the sine and cosine conversion of each axis, the tool load data and the temperature and humidity data of the machine tool; normalize all features; The 5000 processed data are divided into a training set and a validation set in a ratio of 7:3; The error prediction model is trained by the training set; the error prediction model is verified and optimized by the verification set; The prediction process of the error prediction model includes: Input layer: input real-time processing data; the real-time processing data includes The position, speed, acceleration and rotation angle of each axis as well as tool load data and machine temperature and humidity data; Feature acquisition layer: Acquisition prediction feature sets; each prediction feature set corresponds to data of an axis; the prediction feature set includes position coordinates, moving averages of velocity and acceleration, converted rotation angle data, tool load data, and machine tool temperature and humidity data; LSTM layer: including LSTM network structures; each of the LSTM network structures is used to process one of the prediction feature sets to obtain a corresponding feature analysis result; Feature fusion layer: receives the The feature analysis results of the LSTM network structures are integrated to obtain comprehensive feature analysis results. Output layer: obtaining an error prediction result according to the comprehensive feature analysis result; the error prediction result is a contour error value at a certain moment in the future; The calculation formula of the contour error value is: ； in, It's time The predicted contour error value is It's time The environmental adjustment factor, It's time The wear adjustment factor, It is Axis in time The difference between the theoretical and actual positions, is the number of machining axes.

8. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The adaptive compensation control strategy includes establishing an adaptive control algorithm based on the prediction result control of the error prediction model, dynamically adjusting the control parameters according to the contour error value predicted by the error prediction model, and realizing the real-time compensation error control; the adjustment formula of the control parameters is: ； in, Control input, that is, at time The control actions that need to be applied; is the proportional gain; is the current error; is the integral gain; is the integral of the error, which means the sum of the errors accumulated over time; is the differential gain, is the differential of the error, indicating the rate at which the error changes. To predict the gain, is the output of the error prediction model, i.e. the prediction time The contour error value, is the axis influence coefficient, indicating the Axis pair The influence of the axis, Respectively Real-time and reference status of the axis.

9. The contour error prediction and compensation control method for high-precision complex aluminum component processing according to claim 1 is characterized in that: The processing quality results include error data and surface roughness; The model update includes adjusting relevant parameters of the error prediction model and the adaptive compensation control strategy according to the collected processing quality results; the model update uses an online learning method to adapt to changes in operating conditions to ensure that the error prediction model is always synchronized with the current processing status.

Images