Thermal error compensation method for five-axis numerical control machine tool

Abstract

The invention relates to the field of numerical control machine tools, and particularly discloses a thermal error compensation method for a five-axis numerical control machine tool, which comprises the following steps of: S1, acquiring model data of a finished workpiece, and establishing a simulation thermal expansion model of the workpiece; s2, the simulated thermal expansion model of the workpiece is analyzed, and the relation between the temperature and the size change of the workpiece is determined; s3, in the operation process of the machine tool, temperature data of the machining part of the workpiece at the next time are collected; and S4, according to the relation between the temperature of the workpiece and the size change, tool compensation of the machining part at the next time is determined. On the basis of a thermal error compensation model for machine tool temperature change and tool wear, the tool is comprehensively compensated in combination with size change caused by cooling of the heated workpiece, the limitation of an existing thermal error compensation model purely considering influence factors of the machine tool and the tool is avoided, the tool compensation precision is improved, and the tool compensation precision is improved. And the machining precision of the high-precision workpiece is further improved.

Description

Technical Field

[0001] The present invention relates to the field of numerical control machine tools, and particularly to a thermal error compensation method for a five-axis numerical control machine tool. Background Art

[0002] A five-axis numerical control machine tool is a machine tool with high technological content, high precision, and specifically used for machining complex curved surfaces. Due to its high precision, the thermal error during machining has become one of the important factors leading to a reduction in the machining accuracy of workpieces.

[0003] The frictional heat, cutting heat, spindle running heat, tool wear, and ambient temperature of the machine tool generated during the long-term operation of a five-axis numerical control machine tool may all cause small but cumulative deformations in the overall structure of the machine tool, resulting in thermal errors. For workpieces with high precision requirements, these deformations will significantly affect the machining dimensions and surface quality of the workpieces.

[0004] In order to reduce the influence of thermal errors and improve machining accuracy, thermal error compensation technology is often used. However, most of them focus on the thermal error compensation on the machine tool side, and the frictional heat and cutting heat generated by the tool do not act on the tool unidirectionally, but also act on the workpiece. For example, the Chinese patent document with the publication number CN119148621A established a thermal error model by combining the additional thermal influence of tool wear on the temperature field, but it ignored the influence of the workpiece being heated, resulting in room for improvement in machining accuracy after compensation. Summary of the Invention

[0005] To solve the problem in the prior art that there is a lack of compensation for the tool in combination with the cooling of the workpiece after being heated during machining, the present invention provides a thermal error compensation method for a five-axis numerical control machine tool.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A thermal error compensation method for a five-axis numerical control machine tool includes the following steps:

[0008] S1. Obtain the model data of the finished workpiece and establish a simulation thermal expansion model of the workpiece;

[0009] S2. Analyze the simulation thermal expansion model of the workpiece to determine the relationship between the temperature of the workpiece and the dimensional change;

[0010] S3. During the operation of the machine tool, collect the temperature data of the machining part of the workpiece at the next time;

[0011] S4. According to the relationship between the temperature of the workpiece and the dimensional change, and in combination with the thermal error compensation model, determine the tool compensation for the machining part at the next time.

[0012] Preferably, in step S3, collecting the temperature data of the machining part of the workpiece at the next time includes the following steps:

[0013] S31. Collect the machining data of the machine tool to obtain the current posture of the workpiece;

[0014] S32. Collect the current infrared thermal image of the workpiece;

[0015] S33. Adjust the simulation thermal expansion model of the workpiece according to the current posture of the workpiece, so that the simulation thermal expansion model corresponds to the current posture of the workpiece;

[0016] S34. Determine the machining part at the next time according to the machining data of the machine tool, map the machining part to the simulation thermal expansion model and the infrared thermal image, and collect the temperature of the machining part in the current infrared thermal image.

[0017] Preferably, in step S4, according to the relationship between the temperature of the workpiece and the dimensional change, combined with the thermal error compensation model, determine the tool compensation amount of the machining part at the next time, including the following steps:

[0018] S41. During the operation of the machine tool, measure the position change between the tool cutting point and the workpiece to obtain thermal error data;

[0019] S42. Use the SVM model to establish a non-linear mapping relationship between the temperature-sensitive features related to the machine tool and the thermal error;

[0020] S43. Train the tool wear monitoring model based on multi-scale GRU;

[0021] S44. Add a correction function to the SVM model to obtain a thermal error compensation model; the correction function dynamically adjusts the thermal error compensation according to the real-time tool wear value output by the tool wear monitoring model;

[0022] S45. According to the relationship between the temperature of the workpiece and the dimensional change, determine the tool compensation caused by the temperature change of the machining part at the next time;

[0023] S46. According to the tool compensation caused by the temperature change of the machining part at the next time, combined with the thermal error compensation model, obtain the final tool compensation.

[0024] Preferably, in step S42, the screening of the temperature-sensitive features includes the following steps:

[0025] S421. During the operation of the machine tool, collect the temperature data of each relevant monitoring point through multiple temperature sensors;

[0026] S422. Standardize the collected temperature data, perform a correlation analysis on the temperature data and the thermal error data, and screen out the temperature-sensitive features highly correlated with the thermal error to form a temperature-sensitive feature set.

[0027] Preferably, in step S43, training the tool wear monitoring model based on multi-scale GRU includes the following steps:

[0028] S431. During the operation of the machine tool, collect the acceleration signal through a three-axis acceleration sensor, and measure the wear value of the flank face of the tool with a hand-held microscope;

[0029] S432. Preprocess the acceleration signal, use the preprocessed acceleration signal as the input, and the wear value of the flank face of the tool as the label to train the tool wear monitoring model based on multi-scale GRU.

[0030] Preferably, in step S44, the following steps are further included:

[0031] Adopt the Bayesian optimization algorithm to optimize the hyperparameters in the SVM model and the parameters in the correction function, and optimize the thermal error compensation model.

[0032] Preferably, in step S432, preprocessing the acceleration signal includes the following steps:

[0033] Remove the signal cut-in, cut-out and idling data through the change point detection algorithm, and perform data downsampling through the moving average method.

[0034] Preferably, in step S43, the multi-scale GRU includes four parallel branches, and 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 layers are 128, 64, 32, and 16 neurons respectively; then the four branches pass through two fully connected layers in sequence, and the two fully connected layers are 512 and 128 neurons respectively, and the prediction result is output after passing through the fully connected layer.

[0035] The beneficial effects of the present invention are:

[0036] Based on the thermal error compensation model of the machine tool temperature change and tool wear, combined with the dimensional change caused by the workpiece cooling after heating, comprehensive compensation for the tool is carried out, avoiding the limitations of the existing thermal error compensation model that only considers the influencing factors of the machine tool and the tool, improving the accuracy of tool compensation, and further improving the machining accuracy of high-precision workpieces. Description of the Drawings

[0037] Figure 1 It is a schematic flow chart of the thermal error compensation method for a five-axis numerical control machine tool in an embodiment of the present invention. Detailed Embodiments

[0038] The technical solutions of the present invention will be described in detail below with reference to the embodiments.

[0039] Embodiment

[0040] A thermal error compensation method for a five-axis CNC machine tool, comprising the following steps:

[0041] S1. Obtain the model data of the finished workpiece and establish a simulation thermal expansion model of the workpiece;

[0042] S2. Analyze the simulation thermal expansion model of the workpiece to determine the relationship between the temperature of the workpiece and the dimensional change;

[0043] S3. During the operation of the machine tool, collect the temperature data of the next machining part of the workpiece, specifically as follows:

[0044] S31. During the operation of the machine tool, collect the machining data of the machine tool to obtain the current posture of the workpiece;

[0045] S32. Collect the current infrared thermal image of the workpiece;

[0046] S33. Adjust the simulation thermal expansion model of the workpiece according to the current posture of the workpiece to make the simulation thermal expansion model correspond to the current posture of the workpiece;

[0047] S34. Determine the next machining part according to the machining data of the machine tool, and collect the temperature of this machining part in the current infrared thermal image;

[0048] S4. According to the relationship between the temperature of the workpiece and the dimensional change, and in combination with the thermal error compensation model, determine the tool compensation for the next machining part, specifically as follows:

[0049] S41. During the operation of the machine tool, measure the position change between the cutting point of the tool and the workpiece to obtain thermal error data;

[0050] S42. Use the SVM model to establish a non-linear mapping relationship between the temperature-sensitive characteristics related to the machine tool and the thermal error. Among them, the screening of the temperature-sensitive characteristics includes the following steps:

[0051] S421. During the operation of the machine tool, collect the temperature data of each relevant monitoring point through multiple temperature sensors. Taking the input characteristics at each moment as the temperature values of each temperature sensor, a temperature data matrix is formed, and the matrix form is:

[0052] x(t) = [T 1 (t), T 2 (t), …, T l (t)];

[0053] In the formula, T l (t) is the temperature value of the l-th temperature sensor T l at the moment t;

[0054] S422. Standardize the collected temperature data to eliminate the influence of data dimension and unit. The specific standardization formula is as follows:

[0055]

[0056] In the formula, T i (t) is the temperature value of the i-th temperature sensor at time t, is the mean value of the temperature values of the temperature sensor T i , and is its standard deviation;

[0057] Perform a correlation analysis on the temperature data and the thermal error data, and use the Pearson correlation coefficient to analyze and screen the temperature-sensitive features that are highly correlated with the thermal error E thermal (t). The calculation formula of the Pearson correlation coefficient is as follows:

[0058]

[0059] In the formula, E thermal (t) is the thermal error value at time t, and are the mean values of the temperature sensor T i and the thermal error respectively, and N is the number of temperature sensors;

[0060] Set the correlation threshold to 0.85, and retain the temperature features with a correlation coefficient greater than 0.85 to form a temperature-sensitive feature set:

[0061]

[0062] In the formula, M is the number of selected sensitive temperature features;

[0063] The SVM adopts the radial basis function (RBF) kernel, and the form of the kernel function is:

[0064] k(x i , x j ) = exp(-γ||x i - x j || 2 );

[0065] In the formula, k(x i , x j ) is the kernel function between two vectors x i and x j , and γ is the hyperparameter of the radial basis function (RBF), which controls the similarity between input features;

[0066] Use the historical temperature data and thermal error data to train the SVM model to generate a non-linear mapping relationship between the temperature-sensitive features and the thermal error. The model is:

[0067] E thermal E(t) = fSVM(x selcted (t));

[0068] In the formula, f SVM is a trained SVM regression model that can predict thermal error based on temperature-sensitive feature data.

[0069] S43. Train a tool wear monitoring model based on multi-scale GRU. The multi-scale GRU includes four parallel branches, and each branch contains two layers of GRU units. The first layer of GRU units in the four branches all have 128 neurons, and the second layer has 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. After passing through the fully connected layers, the prediction results are output, as follows:

[0070] S431. During the operation of the machine tool, collect acceleration signals through a three-axis acceleration sensor, and measure the wear value of the flank face of the tool with a hand-held microscope.

[0071] S432. Preprocess the acceleration signals. Remove the signal cut-in, cut-out, and idle data through the change point detection algorithm, and perform data downsampling through the moving average method. Use the preprocessed acceleration signals as the input and the wear value of the flank face of the tool as the label to train the tool wear monitoring model based on multi-scale GRU. The process of signal preprocessing is as follows:

[0072] Divide the acceleration signal S of a single cutting into segments, and set each segment as S 1 , S 2 , …, S i , and each segment contains m data points:

[0073] S i = [S i1 , S i2 , …, S im ;

[0074] For each segment of signal S i calculate its mean μ i and variance The formulas are as follows:

[0075]

[0076] Compare the means μ and variances σ of adjacent data segments S i and S i+1 to determine whether there is a significant change between the two segments of data. Calculate Δ and Δ threshold respectively through the following formulas:

[0077] Δ = |μ i+1 - μ i | + |σ i+1 - σ i |;

[0078] Δ threshold = p·(max(S) - min(S)), p ∈ [0.05, 0.1];

[0079] When Δ > Δ threshold , this part of the data is excluded; the remaining stable signal S is downsampled by a sliding window. If the window size is too small at w = 5, then the mean value in each sliding window can be expressed as:

[0080]

[0081] In the formula, ave w is the mean value of w sample points (from k to k + w - 1) within the window starting from position k;

[0082] The entire signal is processed by means of a sliding window to generate the downsampled signal S down = [ave 1 , ave 2 , …, ave l ;

[0083] S44. A correction function is added to the SVM model to obtain a thermal error compensation model; the correction function dynamically adjusts the thermal error compensation according to the real-time tool wear value output by the tool wear monitoring model. Among them, the Bayesian optimization algorithm is also used to optimize the hyperparameters in the SVM model and the parameters in the correction function to optimize the thermal error compensation model;

[0084] The multi-scale GRU model is used to monitor the tool wear state in real time. The input is the preprocessed acceleration signal S down (t). The GRU model captures the wear state changes at different time scales and outputs the real-time tool wear state W(t). The recurrence formula for each layer of GRU is:

[0085] z t = σ(W r a t + U r h t-1 );

[0086] r t = σ(W z a t + U z h t-1 );

[0087] h t = (1 - zt ) ⊙ h t-1 z t + tanh ⊙ (W h a t + U h (r t ⊙ h t-1 ));

[0088] Where z t is the update gate, r t is the reset gate, h t is the hidden state output at the current time t, h t-1 is the hidden state output at the previous time, a t is the input vector at the current time, W z , U z are the weight matrices of the update gate, W r , U r are the weight matrices of the reset gate, W h , U h are the weight matrices of the candidate hidden state, σ is the sigmoid activation function, tanh is the hyperbolic tangent activation function, and ⊙ represents element-wise multiplication;

[0089] Considering the influence of tool wear on the thermal error, a sigmoid correction function is introduced to dynamically adjust the thermal error compensation. The formula of the correction function is:

[0090]

[0091] Where W(t) is the tool wear at time t, β is the parameter controlling the steepness of the sigmoid curve, and W threshold is the threshold at which tool wear begins to significantly affect the thermal error;

[0092] The formula for the corrected thermal error compensation model is:

[0093] E total (t) = f SVM (x selected (t) + f(W(t)));

[0094] Using Bayesian optimization to jointly optimize the hyperparameters in the model. The optimized parameters include the hyperparameters γ and C in SVM, and β and W threshold in the tool wear correction function. The objective function of Bayesian optimization is:

[0095]

[0096] Where C is the penalty factor, E actual (t) is the actual thermal error at time t, and E total (t) is the total thermal error compensation;

[0097] Through Bayesian optimization, the optimal combination γ of the hyperparameters and correction function parameters in the SVM is finally obtained * , C * , β * , The formula for the final thermal error compensation model is as follows:

[0098]

[0099] S45. Determine the tool compensation generated by the temperature change of the workpiece for the machining part at the next time according to the relationship between the temperature and dimensional change of the workpiece

[0100] S46. Obtain the final tool compensation by combining the tool compensation generated by the temperature change of the workpiece for the machining part at the next time with the final thermal error compensation model

[0101] In the present invention, a simulation thermal expansion model of the workpiece is constructed, and the simulation thermal expansion model of the workpiece is adjusted in real time according to the current posture of the workpiece at preset time intervals to match the infrared thermal imaging of the workpiece at the corresponding time, so as to cool and predict the error amount required for the machining point at the next moment after the workpiece is cooled and formed after machining. Finally, the error amount is combined with the output of the thermal error compensation model to determine the final tool compensation amount. On the basis of the thermal error compensation model for the temperature change of the machine tool and tool wear, combined with the dimensional change caused by the cooling of the workpiece after heating, comprehensive compensation for the tool is carried out, avoiding the limitations of the existing thermal error compensation model that only considers the influencing factors of the machine tool and tool, improving the accuracy of tool compensation, and further improving the machining accuracy of high-precision workpieces

[0102] The above embodiments only represent the specific implementation manners of the present invention, and the description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several deformations and improvements can still be made, and these all belong to the protection scope of the present invention

Claims

1. A thermal error compensation method for a five-axis CNC machine tool, characterized in that: The steps include: S1. Obtain model data of the finished workpiece and establish a simulated thermal expansion model of the workpiece; S2. Analyze the simulated thermal expansion model of the workpiece to determine the relationship between the temperature and the dimensional change of the workpiece; S3. During the operation of the machine tool, the temperature data of the next processing part of the workpiece is collected; S4. According to the relationship between the temperature and the size change of the workpiece, combined with the thermal error compensation model, the tool compensation of the processing part at the next time is determined.

2. The thermal error compensation method for a five-axis CNC machine tool according to claim 1, characterized in that: In step S3, the temperature data of the next processing part of the workpiece is collected, including the following steps: S31, collecting the processing data of the machine tool to obtain the current posture of the workpiece; S32, collecting the current infrared thermal image of the workpiece; S33, adjusting the simulated thermal expansion model of the workpiece according to the current posture of the workpiece, so that the simulated thermal expansion model corresponds to the current posture of the workpiece; S34, determining the processing part at the next time according to the processing data of the machine tool, corresponding the processing part to the simulated thermal expansion model and the infrared thermal imaging, and collecting the temperature of the processing part in the current infrared thermal imaging.

3. The thermal error compensation method for a five-axis CNC machine tool according to claim 2, characterized in that: In step S4, according to the relationship between the temperature and the size change of the workpiece, combined with the thermal error compensation model, the tool compensation amount of the processing part at the next time is determined, including the following steps: S41, during the operation of the machine tool, measuring the position change between the cutting point of the tool and the workpiece to obtain thermal error data; S42, using the SVM model to establish a nonlinear mapping relationship between the temperature sensitive features related to the machine tool and the thermal error; S43, training a tool wear monitoring model based on multi-scale GRU; S44, adding a correction function to the SVM model to obtain a thermal error compensation model; The correction function dynamically adjusts the thermal error compensation according to the real-time tool wear value output by the tool wear monitoring model; S45, determining the tool compensation caused by the temperature change of the workpiece at the next processing position according to the relationship between the temperature and the size change of the workpiece; S46, according to the tool compensation caused by the temperature change of the workpiece in the next processing part combined with the thermal error compensation model, the final tool compensation is obtained.

4. The thermal error compensation method for a five-axis CNC machine tool according to claim 3, characterized in that: In step S42, the screening of temperature sensitive features includes the following steps: S421. During the operation of the machine tool, temperature data of each relevant monitoring point is collected through multiple temperature sensors; S422, standardize the collected temperature data, perform correlation analysis on the temperature data and thermal error data, screen out temperature sensitive features that are highly correlated with thermal errors, and form a temperature sensitive feature set.

5. The thermal error compensation method for a five-axis CNC machine tool according to claim 4, characterized in that: In step S43, training a tool wear monitoring model based on a multi-scale GRU includes the following steps: S431. During the operation of the machine tool, the acceleration signal is collected by a three-axis acceleration sensor, and the wear value of the tool back face is measured by a handheld microscope; S432, preprocessing the acceleration signal, taking the preprocessed acceleration signal as input and the wear value of the tool back face as a label, and training a tool wear monitoring model based on a multi-scale GRU.

6. The thermal error compensation method for a five-axis CNC machine tool according to claim 5, characterized in that: In step S44, the following steps are also included: The Bayesian optimization algorithm is used to optimize the hyperparameters in the SVM model and the parameters in the correction function, and the thermal error compensation model is optimized.

7. The thermal error compensation method for a five-axis CNC machine tool according to claim 6, characterized in that: In step S432, the acceleration signal is preprocessed, including the following steps: The signal entry and exit and idling data are removed through the change point detection algorithm, and the data is downsampled through the moving average method.

8. The thermal error compensation method for a five-axis CNC machine tool according to claim 7, characterized in that: In step S43, 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 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 turn, the two fully connected layers have 512 and 128 neurons respectively, and the prediction results are output after passing through the fully connected layers.

Images