A thermal error compensation method for five-axis CNC machine tools considering tool wear

Abstract

The present invention discloses a thermal error compensation method for a five-axis CNC machine tool considering tool wear, comprising: collecting temperature data of each monitoring point, measuring the position change between the tool cutting point and the workpiece to obtain thermal error data, collecting acceleration signals, and measuring tool wear values; (2) standardizing the collected temperature data to screen out temperature sensitive features that are highly correlated with thermal errors; (3) using an SVM model to establish a nonlinear mapping relationship between temperature sensitive features and thermal errors; (4) preprocessing the acceleration signal to train a tool wear monitoring model; (5) introducing a Sigmoid correction function based on the SVM model to obtain a thermal error compensation model; (6) using a Bayesian optimization algorithm to optimize the parameters in the SVM model and the Sigmoid correction function; and using the optimized thermal error compensation model to compensate for the thermal error of the machine tool in real time. The present invention can meet the demand for high-precision thermal error compensation in a long-term processing environment.

Description

Technical Field

[0001] The invention relates to the field of numerical control machining process monitoring and thermal error compensation, and in particular to a thermal error compensation method for a five-axis numerical control machine tool taking tool wear into consideration. Background Art

[0002] In modern high-precision machine tool processing, thermal error is one of the important factors that lead to reduced workpiece processing accuracy. The friction heat, cutting heat, spindle operation heat, etc. generated by the machine tool during long-term operation will cause thermal expansion of the components, resulting in small but cumulative deformation of the overall structure of the machine tool, which in turn affects the processing size and surface quality of the workpiece. In order to maintain the processing accuracy, the existing thermal error compensation technology usually collects the temperature data of the machine tool in real time through sensors and compensates based on the thermal error model. However, in actual production, as the tool wears, the friction heat between the tool and the workpiece gradually increases, exacerbating the change in the overall temperature of the machine tool. This process is crucial to the impact of thermal error.

[0003] In the prior art, a Chinese patent document with publication number CN118519391A discloses a thermal error compensation method based on a temperature sensor, which collects machine tool temperature information and performs real-time compensation. Although this method can reduce the processing error caused by temperature to a certain extent, it ignores the additional thermal impact of tool wear on the temperature field. As the tool gradually wears, frictional heat gradually increases, and this additional heat is not included in the compensation model, resulting in a gradual decrease in compensation accuracy after long-term processing, and the consistency and high precision of processing cannot be maintained. Another Chinese patent document with publication number CN114895627A uses a thermal analysis model to predict and control thermal errors, and predicts and compensates thermal errors by establishing a mathematical model of temperature field and error. However, this patent also ignores the significant contribution of tool wear to frictional heat.

[0004] Existing thermal error compensation technologies do not fully consider the interaction between tool wear and thermal error. In particular, in the later stages of tool wear, friction heat increases significantly, causing the overall temperature of the machine tool to accelerate. Since this additional heat source is not included in the compensation model, traditional compensation methods are difficult to maintain high precision during the machining process, and the compensation accuracy gradually decreases, affecting machining quality and efficiency. In addition, the wear state will have a significant impact on the overall temperature distribution of the machine tool, and the heat source superposition effect caused by tool wear cannot be adjusted in real time, limiting the application of existing thermal error compensation methods. Therefore, there is an urgent need for a new thermal error compensation method that can simultaneously consider the coupling effect of tool wear state and thermal error to improve the stability and accuracy of long-term machining. Summary of the invention

[0005] The present invention provides a thermal error compensation method for a five-axis CNC machine tool taking tool wear into consideration, which can meet the demand for high-precision thermal error compensation in a long-term machining environment, and can effectively improve the compensation effect especially when tool wear increases.

[0006] A thermal error compensation method for a five-axis CNC machine tool considering tool wear comprises the following steps:

[0007] (1) During the operation of the machine tool, multiple temperature sensors are used to collect temperature data at each monitoring point; the position change between the tool cutting point and the workpiece is measured to obtain thermal error data; acceleration signals are collected through a three-axis acceleration sensor; and the wear value of the tool back face is measured;

[0008] (2) The collected temperature data is standardized and then subjected to correlation analysis to screen out temperature-sensitive features that are highly correlated with thermal errors to form a temperature-sensitive feature set;

[0009] (3) Based on the screened temperature-sensitive features, the SVM model is used to establish the nonlinear mapping relationship between the temperature-sensitive features and the thermal error;

[0010] (4) Preprocessing the acceleration signal, using the preprocessed acceleration signal as input and the tool wear value measured by the handheld microscope as a label, training a tool wear monitoring model based on a multi-scale GRU; using the tool wear monitoring model to monitor and output the real-time wear value of the tool;

[0011] (5) Based on the SVM model, the Sigmoid correction function is introduced to obtain the thermal error compensation model; the Sigmoid correction function dynamically adjusts the thermal error compensation according to the real-time wear status of the tool output by the tool wear monitoring model;

[0012] (6) The Bayesian optimization algorithm is used to optimize the hyperparameters in the SVM model and the parameters in the Sigmoid correction function. The thermal error of the machine tool is compensated in real time through the optimized thermal error compensation model.

[0013] The present invention improves the accuracy of thermal error compensation by screening temperature characteristics and using Sigmoid correction based on tool wear. This method combines the influence of tool wear on machine tool temperature change, establishes a coupling model of tool wear and thermal error, and uses Sigmoid activation function to dynamically adjust the compensation intensity, which can effectively improve the compensation effect, especially when tool wear increases.

[0014] In step (2), the temperature sensitive features that are highly correlated with thermal error are screened out, and the formula is:

[0015] ;

[0016] In the formula, It is A temperature sensor at time The temperature value, It's time The thermal error value, and Temperature sensor and the mean of the thermal errors, Indicates the number of temperature sensors.

[0017] In step (3), the kernel function of the SVM model uses the radial basis function RBF, which is in the form of:

[0018] ;

[0019] In the formula, are two vectors and The kernel function between It is a hyperparameter of the radial basis function RBF, which controls the similarity between input features.

[0020] In step (4), the acceleration signal is preprocessed, including: removing signal cut-in, cut-out and idling data by using a change point detection algorithm, and downsampling the data by using a moving average method.

[0021] In step (4), the multi-scale GRU includes four parallel branches, each branch contains two layers of GRU units; the first layer of GRU units in the four branches are all 128 neurons, and the second layer are 128, 64, 32, and 16 neurons respectively; then the four branches pass through two fully connected layers in sequence, the two fully connected layers have 512 and 128 neurons respectively, and the prediction results are output after passing through the fully connected layers.

[0022] In step (5), the formula of the Sigmoid correction function is:

[0023] ;

[0024] In the formula, yes t Tool wear value at time, is the parameter that controls the steepness of the sigmoid curve. is the threshold at which tool wear begins to significantly affect thermal errors.

[0025] In step (5), the formula of the thermal error compensation model is:

[0026] ;

[0027] In the formula, express Temperature sensitive characteristics of the moment, is the trained SVM model, Represents the SVM model prediction Thermal error at time, Indicates that the Sigmoid correction function is Thermal error correction value at time.

[0028] In step (6), the objective function of the Bayesian optimization algorithm is:

[0029] ;

[0030] In the formula, is the hyperparameter of the radial basis function RBF, is the penalty factor, is the actual thermal error at time t, is the total thermal error compensation.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] 1. The present invention fully considers the influence of tool wear on the temperature change of machine tools by establishing a coupled compensation model of tool wear and thermal error. The nonlinear relationship between temperature characteristics and thermal errors is established by support vector machine (SVM), and combined with the real-time monitoring of tool wear status, the compensation model can dynamically reflect the changes in thermal errors during machine tool processing. This coupling method solves the limitation of existing compensation models that do not consider the influence of wear, and effectively improves the accuracy of thermal error compensation.

[0033] 2. The present invention introduces a tool wear correction strategy based on the Sigmoid function and utilizes the S-curve characteristics of the Sigmoid function to simulate the temperature change trend. In particular, when the temperature change tends to be stable, the compensation strength gradually weakens, and when the tool wear intensifies, the compensation strength rapidly increases. This correction strategy can adaptively adjust the thermal error compensation, significantly improve the compensation effect in the later stage of wear, and ensure the high precision and stability of machine tool processing.

[0034] 3. The present invention uses Bayesian optimization technology to optimize and adjust the key parameters in the SVM model and the tool wear correction function to ensure that the compensation model can achieve the best compensation effect under different working conditions. By dynamically adjusting the hyperparameters through Bayesian optimization, the accuracy of thermal error compensation is further improved, overcoming the problem of insufficient compensation caused by fixed parameters in the prior art, especially in long-term processing and the late stage of tool wear. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is the work flow chart of the present invention.

[0036] Figure 2 This is a flow chart of a thermal error compensation method for a five-axis CNC machine tool taking tool wear into consideration according to an embodiment of the present invention.

[0037] Figure 3 4 is a structural diagram of a tool wear monitoring model in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.

[0039] like Figure 1 As shown, a thermal error compensation method for a five-axis CNC machine tool considering tool wear includes:

[0040] During the operation of the machine tool, temperature data of each monitoring point is collected through multiple temperature sensors to form a temperature data sequence. The position change between the tool cutting point and the workpiece is measured online using a laser interferometer to obtain thermal error data. Acceleration signals are collected through a three-axis acceleration sensor. A handheld microscope is used to measure the tool back face wear value on the machine.

[0041] The collected temperature data are standardized to eliminate dimensional differences. Subsequently, correlation analysis is performed to screen out temperature-sensitive features that are highly correlated with thermal errors to form a temperature-sensitive feature set.

[0042] Based on the screened temperature-sensitive features, a support vector machine (SVM) model is used to establish a nonlinear mapping relationship between temperature-sensitive features and thermal errors.

[0043] The acceleration signal is preprocessed, and a tool wear monitoring model based on a multi-scale gated recurrent unit (GRU) is trained to monitor the tool wear status in real time and output the tool wear data.

[0044] Construct a thermal error compensation model combined with tool wear status. Dynamically adjust the compensation strength through Sigmoid correction function, and perform thermal error correction based on tool wear data.

[0045] The Bayesian optimization algorithm is used to optimize the hyperparameters in the SVM model and the parameters in the Sigmoid correction function. The thermal error of the machine tool is compensated in real time through the optimized thermal error compensation model.

[0046] Specifically, Figure 2 As shown, the following steps are included:

[0047] S1: During the operation of the machine tool, 24 temperature sensors are used to collect the temperature data of the machine tool. The installation positions of the temperature sensors are shown in Table 1.

[0048] Table 1

[0049]

[0050] The input feature at each moment is the temperature value of each sensor, forming a temperature data matrix. The matrix form is:

[0051] ;

[0052] S2: Standardize the collected temperature data to eliminate the influence of data dimension and dimension. The specific standardization formula is as follows:

[0053] ;

[0054] In the formula, For temperature sensor T i The mean of is its standard deviation.

[0055] S3: Analysis of screening and thermal errors using Pearson correlation coefficient Highly correlated temperature characteristics. The correlation coefficient is calculated as:

[0056] ;

[0057] In the formula, It is A temperature sensor at time The temperature value, It's time The thermal error value, and Temperature sensor and the mean of the thermal errors, Indicates the number of temperature sensors.

[0058] Set the correlation threshold to 0.85, retain the temperature features with correlation coefficients greater than 0.85, and form a temperature-sensitive feature set:

[0059] ;

[0060] In the formula, Indicates the number of sensitive temperature features screened out.

[0061] S4: Use support vector machine (SVM) to model the screened temperature sensitive feature set and establish a nonlinear mapping relationship between temperature features and thermal errors. SVM uses radial basis function (RBF) kernel, and the kernel function form is:

[0062] ;

[0063] In the formula, are two vectors and The kernel function between It is a hyperparameter of the radial basis function RBF, which controls the similarity between input features.

[0064] The SVM model is trained using historical temperature and thermal error data to learn the nonlinear mapping relationship between input features and output thermal errors. The model form is:

[0065] ;

[0066] In the formula, It is a trained SVM regression model that can predict thermal errors based on temperature data.

[0067] S5: Remove invalid data from the cutting acceleration signal, including data from the machine tool idling and cutting-in and cutting-out stages. Then downsample the remaining stable signals to reduce the amount of data and retain the necessary change information. The signal preprocessing process is achieved through the following formula:

[0068] The acceleration signal of a cutting S Segmentation is performed, assuming that each segment is , each segment contains m Data points:

[0069] ;

[0070] For each signal Calculate its mean and variance , the formula is as follows:

[0071] ;

[0072] ;

[0073] Comparing adjacent data segments and The mean and variance , to determine whether there is a significant change between the data at both ends. When this part of data is removed, the change The calculation formula is as follows:

[0074] ;

[0075] In the formula, .

[0076] The remaining stable signal S is downsampled by sliding window, and the window size is w=5, then the mean in each sliding window can be expressed as:

[0077] ;

[0078] In the formula, Indicates from position In the start window Sample points (from arrive ) is the mean of .

[0079] The entire signal is processed by sliding window to generate a downsampled signal .

[0080] S6: Use the multi-scale GRU model to monitor the tool wear status in real time. The tool wear monitoring process is as follows: Figure 3 As shown in the figure. The input is the preprocessed acceleration signal, and the GRU model captures the wear state changes at different time scales. The input of GRU is the acceleration signal S down (t), through the recursive structure of the GRU network, the real-time wear status of the tool is finally output W ( t ), the recursive formula of each layer of GRU is:

[0081] ;

[0082] ;

[0083] ;

[0084] In the formula, It is the update gate. It is the reset gate. It is the current moment t The hidden state output of is the hidden state output of the previous moment, is the input vector at the current moment, , is the weight matrix of the update gate, , is the weight matrix of the reset gate, , is the weight matrix of the candidate hidden state, is the sigmoid activation function, tanh is the hyperbolic tangent activation function, It is element-wise multiplication.

[0085] S7: In order to consider the influence of tool wear on thermal error, the present invention introduces a Sigmoid correction function to dynamically adjust the thermal error compensation according to the tool wear state. The correction function is in the form of:

[0086] ;

[0087] In the formula, yes t Tool wear at all times, is the parameter that controls the steepness of the sigmoid curve. is the threshold at which tool wear begins to significantly affect thermal errors.

[0088] The corrected thermal error compensation model is:

[0089] ;

[0090] S8: Use Bayesian optimization to jointly optimize the hyperparameters in the model. The optimization parameters include the hyperparameters in SVM and And the tool wear correction function and , the objective function of Bayesian optimization is:

[0091] ;

[0092] In the formula, is the hyperparameter of the radial basis function RBF, is the penalty factor, yes t The actual thermal error at the moment, is the total thermal error compensation.

[0093] Through Bayesian optimization, the optimal combination of SVM hyperparameters and correction function parameters was finally obtained. , , , The final thermal error compensation model is:

[0094] ;

[0095] S9: Through the optimized thermal error compensation model The thermal error of the machine tool is compensated in real time. In the later stage of tool wear, the model can adaptively enhance the compensation strength to ensure machining accuracy.

[0096] This embodiment proposes a thermal error compensation method combined with the tool wear state based on the coupling relationship between tool wear and machine tool thermal error. A dynamic thermal error compensation model is constructed by screening temperature characteristics and combining the tool wear state monitored in real time. This method uses SVM to perform nonlinear modeling between temperature and thermal error, and dynamically adjusts the compensation intensity according to the tool wear state through the Sigmoid correction function, which significantly improves the accuracy of thermal error compensation. Compared with the traditional single compensation method, the present invention can more comprehensively capture the impact of tool wear on machine tool thermal error, and ensure the real-time and accuracy of thermal error compensation during processing.

[0097] The thermal error compensation model in this embodiment further adjusts key hyperparameters through Bayesian optimization, avoiding the problem of reduced accuracy of the traditional compensation model during long-term processing. Compared with the traditional fixed parameter compensation strategy, the present invention realizes an adaptive compensation strategy by dynamically optimizing the SVM hyperparameters and the parameters of the tool wear correction function, especially in the later stage of tool wear, the compensation effect is more significant. This method greatly improves the adaptability of the model to complex working conditions and provides a reliable solution for long-term high-precision processing.

[0098] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A thermal error compensation method for a five-axis CNC machine tool considering tool wear, characterized in that: The following steps are involved: (1) During the operation of the machine tool, multiple temperature sensors are used to collect temperature data at each monitoring point; the position change between the tool cutting point and the workpiece is measured to obtain thermal error data; acceleration signals are collected through a three-axis acceleration sensor; and the wear value of the tool back face is measured; (2) The collected temperature data is standardized and then subjected to correlation analysis to screen out temperature-sensitive features that are highly correlated with thermal errors to form a temperature-sensitive feature set; (3) Based on the screened temperature-sensitive features, the SVM model is used to establish the nonlinear mapping relationship between the temperature-sensitive features and the thermal error; (4) Preprocessing the acceleration signal, using the preprocessed acceleration signal as input and the tool wear value measured by the handheld microscope as a label, training a tool wear monitoring model based on a multi-scale GRU; using the tool wear monitoring model to monitor and output the real-time wear value of the tool; (5) Based on the SVM model, the Sigmoid correction function is introduced to obtain the thermal error compensation model. The Sigmoid correction function dynamically adjusts the thermal error compensation according to the real-time wear status of the tool output by the tool wear monitoring model. The formula of the Sigmoid correction function is: ； In the formula, yes Tool wear value at time, is the parameter that controls the steepness of the sigmoid curve. is the threshold at which tool wear begins to significantly affect thermal error; (6) The Bayesian optimization algorithm is used to optimize the hyperparameters in the SVM model and the parameters in the Sigmoid correction function. The thermal error of the machine tool is compensated in real time through the optimized thermal error compensation model.

2. The thermal error compensation method for a five-axis CNC machine tool considering tool wear according to claim 1, characterized in that: In step (2), the temperature sensitive features that are highly correlated with thermal error are screened out, and the formula is: ； In the formula, It is A temperature sensor at time The temperature value, It's time The thermal error value, and Temperature sensor and the mean of the thermal errors, Indicates the number of temperature sensors.

3. The thermal error compensation method for a five-axis CNC machine tool considering tool wear according to claim 1, characterized in that: In step (3), the kernel function of the SVM model uses the radial basis function RBF, which is in the form of: ； In the formula, are two vectors and The kernel function between It is a hyperparameter of the radial basis function RBF, which controls the similarity between input features.

4. The thermal error compensation method for a five-axis CNC machine tool considering tool wear according to claim 1, characterized in that: In step (4), the acceleration signal is preprocessed, including: removing signal cut-in, cut-out and idling data by using a change point detection algorithm, and downsampling the data by using a moving average method.

5. The thermal error compensation method for a five-axis CNC machine tool considering tool wear according to claim 1, characterized in that: In step (4), the multi-scale GRU includes four parallel branches, each branch contains two layers of GRU units; the first layer of GRU units in the four branches are all 128 neurons, and the second layer are 128, 64, 32, and 16 neurons respectively; then the four branches pass through two fully connected layers in sequence, the two fully connected layers have 512 and 128 neurons respectively, and the prediction results are output after passing through the fully connected layers.

6. The thermal error compensation method for a five-axis CNC machine tool considering tool wear according to claim 1, characterized in that: In step (5), the formula of the thermal error compensation model is: ； In the formula, express Temperature sensitive characteristics of the moment, is the trained SVM model, Represents the SVM model prediction Thermal error at time, Indicates that the Sigmoid correction function is Thermal error correction value at time.

7. The thermal error compensation method for a five-axis CNC machine tool considering tool wear according to claim 6, characterized in that: In step (6), the objective function of the Bayesian optimization algorithm is: ； In the formula, is the hyperparameter of the radial basis function RBF, is the penalty factor, yes t The actual thermal error at the moment, is the total thermal error compensation.

Images