A Method for Predicting the Surface Roughness of Workpieces in High-Speed Dry Milling

Abstract

The present invention discloses a method for predicting the surface roughness of a workpiece in high-speed dry milling, which comprises the following steps: collecting the cutting parameters of high-speed dry milling and the cutting force data received by the workpiece during the milling process; calculating and obtaining significant features related to the surface roughness of the workpiece according to the cutting force data; fusing the significant features and the cutting parameters and inputting them into a surface roughness model to obtain the prediction result of the surface roughness of the workpiece. By fusing the cutting parameters with the cutting force in the high-speed dry milling process and combining Gaussian process regression to construct a high-precision surface roughness model, the present invention can well predict the surface roughness, greatly reduce the workpiece detection cost, and at the same time can provide timely feedback in batch processing to reduce the processing cost of the workpiece.

Description

Technical Field

[0001] The present invention relates to the technical field of milling processes, and in particular to a method for predicting the surface roughness of workpieces in high-speed dry milling. Background Art

[0002] High-speed dry milling is different from traditional cutting fluid pouring cutting. High-speed dry milling uses high-speed cutting and does not use any cutting fluid during the cutting process, which is a green and advanced manufacturing technology. However, its extremely high cutting speed and the lack of cutting fluid bring about unstable machining quality. Especially under different cutting parameters, it is difficult to ensure the surface roughness quality of the workpiece. Therefore, how to optimize the cutting parameters to obtain the required surface roughness quality of the workpiece has become a hot topic in this field. Existing prediction technologies for high-speed dry milling usually only select cutting parameters and the corresponding surface roughness to construct a surface roughness model, ignoring dynamic machining signals such as cutting forces generated during the machining process, resulting in low prediction accuracy and unable to accurately predict the surface roughness well. In addition, the determination of the surface roughness of the workpiece after traditional high-speed dry milling is mostly carried out by comparing with standard blocks after machining. Whether the workpiece is qualified depends on the experience of workers, which will greatly increase production costs and time, and is not conducive to timely detecting the phenomenon of surface quality deterioration caused by tool wear and other factors.

[0003] Therefore, how to ensure machining quality has become a hot topic in high-speed dry milling, and an accurate machining quality prediction method has become a major requirement in the field of high-speed dry milling. Summary of the Invention

[0004] Aiming at the deficiencies of the above-mentioned existing technologies, the technical problem to be solved by the present invention is: to provide a method for predicting the surface roughness of workpieces in high-speed dry milling, which solves the problems of difficult accurate prediction of machining quality and high production cost in existing high-speed dry milling.

[0005] To solve the above technical problems, the present invention adopts the following technical solutions:

[0006] A method for predicting the surface roughness of workpieces in high-speed dry milling includes the following steps,

[0007] a. Collect the cutting parameters of high-speed dry milling and the cutting force data received by the workpiece during the milling process;

[0008] b. Calculate and obtain significant features related to the surface roughness of the workpiece according to the cutting force data;

[0009] c. Fuse the significant features and cutting parameters and input them into a surface roughness model to obtain the prediction result of the surface roughness of the workpiece.

[0010] As an optimization, the cutting parameters include the spindle speed, feed per tooth, cutting depth, and cutting width of high-speed dry milling.

[0011] As an optimization, the cutting force data includes the three-direction cutting force data of the workpiece during high-speed dry milling.

[0012] As an optimization, in step b, the cutting force data is preprocessed first. The cutting force magnitude data in three directions is truncated according to the acting time sequence, and the cutting force magnitude data corresponding to the first 25% and the last 25% of the acting time is truncated to obtain the effective cutting force data.

[0013] As an optimization, the significant features are multiple features in the time-domain statistical features of the cutting force data that are significantly correlated with the surface roughness. Among them, the significant correlation means that the absolute value of the Pearson correlation coefficient is ≥ 0.5.

[0014] As an optimization, the time-domain statistical features include root mean square value, standard deviation, skewness, kurtosis, peak-to-peak value, and coefficient of variation.

[0015] As an optimization, the surface roughness model is constructed based on the Gaussian process regression model, and it includes the following steps:

[0016] c1. Determine the hyperparameters of the Gaussian process regression model using historical process data:

[0017] ξ(x) ∼ GP(μ(x), k(x, z)) (1)

[0018] where μ(x) = E(ξ(x)) represents the mean function, and k(x, z) = E((ξ(x) - μ(x))(ξ(z) - μ(z))) represents the covariance function;

[0019] c2. Use the hyperparameters to construct the surface roughness model:

[0020]

[0021] where Force feature = [feature1, feature2,..., feature d are the significant features extracted from the cutting force dataset, d is the number of cutting force features, Y is the prior distribution of historical process data, n0 is the spindle speed, f z is the feed per tooth, a P is the cutting depth, a e is the cutting width. x represents all samples x in the dataset S i and x * represents a new sample to be predicted, R ais the surface roughness value corresponding to the new sample, σ is the standard deviation of the noise, I is the identity matrix, and K(*, *) is the covariance matrix.

[0022] As an optimization, the hyperparameters of the Gaussian process regression are the hyperparameters of the mean function and the covariance function; among them, the mean function is a linear mean function, and the covariance function is a Matern 5 / 2 function with automatic relevance determination function.

[0023] As an optimization, the historical working condition data includes the three-dimensional cutting force data, cutting parameters, and surface roughness data collected during the historical high-speed dry milling process of the workpiece.

[0024] The present application has the following beneficial effects compared with the prior art:

[0025] In the present invention, by first collecting the three-dimensional cutting force data during the high-speed dry milling process, intercepting the stable cutting area of the three-dimensional cutting force data, then calculating the significant features of the three-dimensional cutting force data, and finally fusing the three-dimensional cutting force data with the cutting parameters as feature vectors and inputting them into the constructed surface roughness model, the surface roughness value corresponding to this set of cutting parameters is obtained. By fusing the cutting parameters with the cutting force during the high-speed dry milling process and constructing a high-precision surface roughness model in combination with Gaussian process regression, the present invention can well predict the surface roughness, greatly reduce the workpiece detection cost, and at the same time can provide timely feedback during batch processing and reduce the workpiece processing cost. Description of the Drawings

[0026] Figure 1 is the prediction flow chart of the present invention;

[0027] Figure 2 is the cutting force data in the X direction of the workpiece during high-speed dry milling in the present invention;

[0028] Figure 3 is the cutting force data in the Y direction of the workpiece during high-speed dry milling in the present invention;

[0029] Figure 4 is the cutting force data in the Z direction of the workpiece during high-speed dry milling in the present invention;

[0030] Figure 5 is the cutting force data in the X direction of the stable area of the workpiece in the present invention;

[0031] Figure 6 is the milling experiment factor level table of the present invention;

[0032] Figure 7 is the cutting force feature and R a Pearson correlation coefficient;

[0033] Figure 8 This is a comparison between the predicted value and the observed value of the surface roughness of the workpiece after high-speed dry milling by the surface roughness model in the present invention. Specific embodiments

[0034] The present invention will be further described in detail below with reference to the accompanying drawings.

[0035] During specific implementation: Refer to Figure 1-8 ,

[0036] A method for predicting the surface roughness of a workpiece in high-speed dry milling includes the following steps.

[0037] a. Collect the cutting parameters of high-speed dry milling and the cutting force data received by the workpiece during the milling process.

[0038] Specifically, the cutting parameters include the spindle speed, feed per tooth, cutting depth, and cutting width of high-speed dry milling. The cutting force data includes the X, Y, and Z three-direction cutting force data received by the workpiece during high-speed dry milling. For example, Figures 2-4 , and after truncating the first 25% and the last 25% of the three-direction cutting force data, the cutting force data in the stable region is obtained, which is the effective cutting force data. For example, Figure 5 As shown, and the time-domain statistical characteristics are calculated based on this effective cutting force data. The time-domain statistical characteristics include root mean square value, standard deviation, skewness, kurtosis, peak-to-peak value, and coefficient of variation.

[0039] b. Calculate and obtain the significant features related to the surface roughness of the workpiece based on the cutting force data.

[0040] Specifically, the significant features are multiple features in the time-domain statistical characteristics of the cutting force data that are significantly correlated with the surface roughness. Among them, the significant correlation means that the absolute value of the Pearson correlation coefficient ≥ 0.5, and the Pearson correlation coefficient is calculated according to the following formula:

[0041]

[0042] Among them, cov(x1, x2) represents the covariance of variables x1 and x2. respectively represent the standard deviations of variables x1 and x2.

[0043] c. Fuse the significant features and the cutting parameters and input them into a surface roughness model to obtain the predicted result of the surface roughness of the workpiece.

[0044] Specifically, after determining the significant features of the three-direction cutting force and fusing them with the cutting parameters, an input feature vector is formed, and then it is standardized by the Z-score method as the input of the model.

[0045] Specifically, the surface roughness model is constructed based on Gaussian process regression, which includes the following steps: obtaining historical process data collected under the same working conditions. The historical process data includes three-dimensional cutting force data, cutting parameters, and corresponding surface roughness data. The same working conditions refer to the same machine tool, workpiece material, and tool. The prediction of the surface roughness model is only suitable for predicting the surface roughness of high-speed dry milling workpieces under the same working conditions.

[0046] c1. Determine the hyperparameters of the Gaussian process regression. The hyperparameters of the Gaussian process regression consist of the hyperparameters of the mean function and the covariance function.

[0047] Specifically, the Gaussian process is defined as follows:

[0048] ξ(x)~GP(μ(x), k(x, z)) (1)

[0049] where μ(x) = E(ξ(x)) represents the mean function, and k(x, z) = E((ξ(x) - μ(x))(ξ(z) - μ(z))) represents the covariance function. According to the characteristics of the high-speed dry milling surface roughness data, the mean function is selected as the linear mean function μ(x) = a T x, and the covariance function is selected as the Matern 5 / 2 function with automatic relevance determination function, which can be expressed as:

[0050]

[0051]

[0052] where Λ = diag(λ), D is the dimension of the model input. The hyperparameters can be expressed as θ = (σ f , a, λ1,..., λ D ), which can be obtained by maximizing the log marginal likelihood. Among them, σ f is one of the hyperparameters, which needs to be calculated using the maximum log marginal likelihood method according to the historical process data. After all the hyperparameters are calculated, substitute the hyperparameters into the above covariance function and mean function to obtain the determined covariance function and mean function. The determined covariance function and mean function can be used to calculate the following R a .

[0053] c2. Construct a surface roughness model with the hyperparameters and historical working condition data.

[0054] Specifically, based on the calculated hyperparameters and historical process data, assume that the historical process data set can be expressed as S = {(x i , y i) | i = 1, 2, …, n}, where n is the number of samples. Thus, the workpiece surface roughness model for high-speed dry milling can be obtained:

[0055]

[0056] Among them, Force feature = [feature1, feature2, …, feature d are the significant features extracted from the cutting force dataset, d is the number of cutting force features, Y is the prior distribution of historical process data, n0 is the spindle speed, f z is the feed per tooth, a P is the cutting depth, a e is the cutting width. x represents all the x i in the dataset S, and x * represents the new sample to be predicted. R a is the corresponding surface roughness value under the new sample, σ is the standard deviation of the noise, I is the identity matrix, and K(*, *) is the covariance matrix.

[0057] Specifically, taking the milling of 30CrMnSiNiA material as an example. The machine tool is a three-axis vertical machining center, model BV8H, with a spindle power of 7.5 kW. The cutting tool is an insert type carbide end mill, with two cutting edges, a diameter of 20 mm, and the blade surface is coated with a PVD coating. Data is collected according to the full factor experiment, and the experimental scheme is as shown in Figure 6 .

[0058] Collect and analyze the three-way cutting force data, and perform statistical feature extraction including root mean square value (RMS), standard deviation (STD), skewness, kurtosis, peak-to-peak value, and coefficient of variation (CV). The calculated Pearson correlation coefficients are as shown in Figure 7 .

[0059] It can be seen that the statistical features RMS, STD, and PP of the cutting force in the Y direction are significant features and are used for the construction of the surface roughness model.

[0060] Using the historical process dataset, after calculating the hyperparameters by the maximum log marginal likelihood method, the workpiece surface roughness model for high-speed dry milling can be obtained. The prediction performance of the constructed high-speed dry milling surface roughness model (the collected dataset is divided into a training set and a test set, and here the test set is used for evaluation) is as shown in Figure 8 . It can be seen that the proposed high-speed dry milling surface roughness prediction method has a high prediction accuracy, and the coefficient of determination R 2 is 0.9736.

[0061] The present invention first collects the three-direction cutting force data during high-speed dry milling, intercepts the stable cutting area of the three-direction cutting force data, then calculates the significant features of the three-direction cutting force data, and finally fuses the three-direction cutting force data with the cutting parameters as feature vectors and inputs them into the constructed surface roughness model to obtain the corresponding surface roughness value under this set of cutting parameters. By fusing the cutting parameters with the cutting force during high-speed dry milling and combining Gaussian process regression to construct a high-precision surface roughness model, the present invention can well predict the surface roughness, greatly reduce the workpiece detection cost, and at the same time can provide timely feedback during batch processing to reduce the processing cost of workpieces.

[0062] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and basis of the present invention. The scope of the present invention is defined by the appended claims and their equivalents. Therefore, the embodiments of the present invention are only illustrative examples of the present invention, and in no way constitute a limitation to the present invention from any point of view.

Claims

1. A method for predicting the surface roughness of a workpiece in high-speed dry milling, characterized in that, It includes the following steps: a. Collect the cutting parameters of high-speed dry milling and the cutting force data exerted on the workpiece during the milling process. b. Calculate and obtain the significant features related to the surface roughness of the workpiece based on the cutting force data; the significant features are multiple features in the time-domain statistical features of the cutting force data that have a significant correlation with the surface roughness, where the significant correlation means that the absolute value of the Pearson correlation coefficient ≥ 0.5; the time-domain statistical features include root mean square value, standard deviation, skewness, kurtosis, peak-to-peak value, and coefficient of variation. c. Fusion the significant features and the cutting parameters and input them into the surface roughness model to obtain the prediction result of the workpiece surface roughness; specifically, The surface roughness model is constructed based on the Gaussian process regression model, and it includes the following steps: c1. Determine the hyperparameters of the Gaussian process regression model using historical process data: ξ(x)~GP(μ(x),k(x,z)) (1) where μ(x) = E(ξ(x)) represents the mean function, and k(x,z) = E((ξ(x) - μ(x))(ξ(z) - μ(z))) represents the covariance function. c2. Use the hyperparameters to construct the surface roughness model: Among them, Force feature = [feature1, feature2, …, feature d are the significant features extracted from the cutting force dataset, d is the number of cutting force features, Y is the prior distribution of historical process data, n0 is the spindle speed, f z is the feed per tooth, a P is the cutting depth, a e is the cutting width; x represents the x i of all samples in the dataset S, x * represents the new sample to be predicted, R a is the corresponding surface roughness value under the new sample, σ is the standard deviation of the noise, I is the identity matrix, K(*, *) is the covariance matrix; The Gaussian process regression model is composed of a mean function and a covariance function. The hyperparameters are the hyperparameters of the mean function and the covariance function, and the hyperparameters are obtained by using the historical process data with the maximum log marginal likelihood function method; among them, the mean function is a linear mean function, and the covariance function is a Matern 5 / 2 function with automatic relevance determination function.

2. The workpiece surface roughness prediction method for high-speed dry milling according to claim 1, characterized in that, The cutting parameters include the spindle speed, feed per tooth, cutting depth, and cutting width of high-speed dry milling.

3. A method for predicting the surface roughness of a workpiece in high-speed dry milling according to claim 1, characterized in that, The cutting force data includes the magnitude data of the cutting forces in three directions exerted on the workpiece during high-speed dry milling.

4. A workpiece surface roughness prediction method for high-speed dry milling according to claim 3, characterized in that, In step b, first preprocess the cutting force data. Cut off the cutting force magnitude data corresponding to the first 25% and the last 25% of the action time in chronological order of the cutting force magnitude data in three directions to obtain the effective cutting force data.

5. A workpiece surface roughness prediction method for high-speed dry milling according to claim 1, characterized in that The historical process data includes the three-direction cutting force data, cutting parameters, and surface roughness data collected during the historical high-speed dry milling process.

Images