CN-116663674-B - Milling surface integrity prediction method

CN116663674BCN 116663674 BCN116663674 BCN 116663674BCN-116663674-B

Abstract

The invention discloses a milling surface integrity prediction method which comprises the steps of S1, collecting milling test data sets, S2, preprocessing the milling test data sets to obtain processed test data sets, S3, inputting the processed test data sets into a milling surface integrity prediction model to conduct model training to obtain a trained prediction model, S4, inputting milling data to be detected into the trained prediction model, and outputting a prediction result of milling surface integrity. The method can remarkably improve the accuracy of predicting the integrity of the milling surface, and is beneficial to improving the efficiency and quality of the milling process.

Inventors

LIU ZONGMIN
XING BIN
HE ZHENYU
LIU LANHUI
GUAN TING

Assignees

重庆工业大数据创新中心有限公司
重庆工商大学

Dates

Publication Date: 20260505
Application Date: 20230524

Claims (8)

1. A milling surface integrity prediction method is characterized by comprising the following steps: s1, collecting a milling test data set; S2, preprocessing a milling test data set to obtain a processed test data set; S3, inputting the processed test data set into a milling surface integrity prediction model for model training to obtain a trained prediction model, wherein the method specifically comprises the following steps of: Selecting by using the processed test data set Regression model and pair Respectively optimizing the selected regression models to obtain Selecting optimized regression model The regression model specifically comprises: Respectively to using training set Training the regression models, wherein each regression model respectively carries out 3 times of training by taking the surface roughness, the surface residual stress in the feeding direction and the transverse surface residual stress as label values, Greater than ; After training is completed, predicting the test set by using the trained model; Scoring each regression model by using the evaluation index, and taking the regression model with the scoring value meeting the set condition as the selected regression model; For a pair of Different combinations are carried out on the optimized regression models to obtain base learners with different combinations, the same meta learner is used for carrying out Stacking model fusion on the base learners with different combinations, and the optimal base learner combination is selected from the base learner combinations with the multiple model fusion; The optimal basic learner combination is used as a first-layer learner, different regression models are selected on the basis of the first-layer learner to be used as element learners for model fusion comparison, and the regression model with the highest prediction precision in model fusion is selected to be used as the element learner in model fusion; s4, inputting milling data to be detected into the trained prediction model, and outputting a prediction result of the integrity of the milling surface.
2. The method of claim 1, wherein the milling test data set comprises feature variable data and processed workpiece surface data; the characteristic variable data comprise machining parameters, tool posture parameters and tool geometric parameters; the processed workpiece surface data comprises surface roughness and surface residual stress, wherein the surface residual stress comprises surface residual stress in the feeding direction and transverse surface residual stress.
3. The method of claim 2, wherein the machining parameters include spindle speed, feed rate, and depth of cut; the tool posture parameters comprise a rake angle and an inclination angle; The tool geometry parameters include tool diameter, number of edges and helix angle.
4. The method of milling surface integrity prediction according to claim 1, wherein: the pretreatment comprises the following steps: Performing dimensionless treatment on the milling test data set to obtain a dimensionless test data set; the test data set after dimensionless is divided into a training set and a testing set.
5. The method for predicting the integrity of a milled surface according to claim 1, wherein optimizing the selected regression model comprises: traversing all the super-parameter combinations in the regression model, performing cross-validation on each group of super-parameters to obtain performance indexes of each group of super-parameters, and finally selecting a group of super-parameters with optimal performance as final model parameters.
6. The method of claim 1, wherein the milling surface integrity prediction model comprises a surface roughness prediction model, a feed direction surface residual stress prediction model, and a lateral surface residual stress prediction model.
7. The method of claim 6, wherein the optimal combination of basis learners in the surface roughness prediction model comprises a Bagging regression model, a random forest regression model, a ExtraTree regression model, a gradient lifting regression model, a XGBoost regression model, and a CatBoost regression model; The optimal basic learner combination in the feeding direction surface residual stress prediction model comprises a Bagging regression model, a random forest regression model, a ExtraTree regression model, a XGBoost regression model and a CatBoost regression model; The optimal combination of the base learners in the transverse surface residual stress prediction model comprises ExtraTree regression models, XGBoost regression models and CatBoost regression models.
8. The method for predicting the integrity of a milled surface according to claim 7, wherein a gradient lifting regression model is used as a meta learner in model fusion in the surface roughness prediction model; a meta learner for taking a random forest regression model as a model fusion in the feed direction surface residual stress prediction model; And taking the gradient lifting regression model as a meta-learner when the model is fused in the transverse surface residual stress prediction model.

Description

Milling surface integrity prediction method Technical Field The invention relates to the field of milling, in particular to a milling surface integrity prediction method. Background Surface integrity is a technical indicator used to describe, evaluate and control the impact of surface performance in a process engineering and the various changes that may occur in its process surface layer. As the manufacturing industry evolves and the demands for part surface integrity continue to increase, the need to predict and control the integrity of the milled surface is also increasing. However, due to the complexity and variability of the milling process, it is difficult to accurately predict the mill surface integrity by empirical and experimental methods alone, requiring prediction and analysis by means of computer simulation techniques. At present, a method or a product for carrying out integrity prediction on a milling surface based on a computer simulation technology usually uses an existing model frame directly, the construction process is too simple, and effective recognition on milling surface characteristics cannot be carried out, so that the prediction effect is not ideal enough, and effective prediction on the milling surface integrity cannot be realized, and therefore, the method for predicting the milling surface integrity is needed, and the problems can be solved. Disclosure of Invention In view of the above, the present invention aims to overcome the defects in the prior art, and provide a method for predicting the integrity of a milling surface, which can significantly improve the accuracy of predicting the integrity of the milling surface and is helpful for improving the efficiency and quality of the milling process. The milling surface integrity prediction method comprises the following steps: s1, collecting a milling test data set; S2, preprocessing a milling test data set to obtain a processed test data set; S3, inputting the processed test data set into a milling surface integrity prediction model for model training to obtain a trained prediction model; s4, inputting milling data to be detected into the trained prediction model, and outputting a prediction result of the integrity of the milling surface. Further, the milling test data set comprises characteristic variable data and processed workpiece surface data; the characteristic variable data comprise machining parameters, tool posture parameters and tool geometric parameters; the processed workpiece surface data comprises surface roughness and surface residual stress, wherein the surface residual stress comprises surface residual stress in the feeding direction and transverse surface residual stress. Further, the processing parameters comprise spindle rotation speed, feed speed and cutting depth; the tool posture parameters comprise a rake angle and an inclination angle; The tool geometry parameters include tool diameter, number of edges and helix angle. Further, the preprocessing includes: Performing dimensionless treatment on the milling test data set to obtain a dimensionless test data set; the test data set after dimensionless is divided into a training set and a testing set. Further, inputting the processed test data set into a milling surface integrity prediction model for model training to obtain a trained prediction model, which specifically comprises the following steps: Selecting k regression models by using the processed test data set, and respectively optimizing the k selected regression models to obtain k optimized regression models; Different combinations are carried out on the k optimized regression models to obtain base learners with different combinations, the same meta learner is used for carrying out Stacking model fusion on the base learners with different combinations, and the optimal base learner combination is selected from the base learner combinations with the multiple model fusion; and the optimal basic learner combination is used as a first-layer learner, different regression models are selected on the basis of the first-layer learner to be used as the element learner for model fusion comparison, and the regression model with the highest prediction precision in model fusion is selected to be used as the element learner in model fusion. Further, selecting k regression models, specifically including: training m regression models by using a training set, wherein each regression model respectively carries out 3 times of training by taking surface roughness, surface residual stress in a feeding direction and transverse surface residual stress as label values, and m is larger than k; After training is completed, predicting the test set by using the trained model; and scoring each regression model by using the evaluation index, and taking the regression model with the bisection value meeting the set condition as the selected regression model. Further, optimizing the selected regression model specifically includes: traversi