Infrared and Laser Engineering, Volume. 52, Issue 12, 20230348(2023)
Research on surface roughness modeling based on multiple feature parameters of laser speckle image
Figures & Tables(14)
Fig. 1. The modeling process of surface roughness measurement model
Fig. 2. Schematic diagram of laser speckle image acquisition device
Fig. 3. Preprocessed laser speckle images (Plane grinding specimens). (a)
Fig. 4. Standard and predicted values of surface roughness for the test set. (a) Plane grinding; (b) Horizontal milling; (c) Vertical milling; (d) Grinding polishing
Fig. 5. Relative error graph of surface roughness prediction. (a) Plane grinding; (b) Horizontal milling; (c) Vertical milling; (d) Grinding polishing
Table 1. Material and surface roughness parameters
View tableView in ArticleTable 1. Material and surface roughness parameters
Set Process Material Standard value of surface roughness Ra/μm PG Plane grinding 45#steel 0.1 0.2 0.4 0.8 HM Horizontal milling 45#steel 0.4 0.8 1.6 3.2 VM Vertical milling 45#steel 0.4 0.8 1.6 3.2 GP Grinding polishing GCr15 0.025 0.05 0.1 -
Table 2. Dataset sample division
View tableView in ArticleTable 2. Dataset sample division
Data set PG HM VM GP Training set 160 160 160 120 Test set 24 24 24 18
Table 3. The absolute value of Spearman's correlation coefficient between feature parameters and surface roughness parameter
View tableView in ArticleTable 3. The absolute value of Spearman's correlation coefficient between feature parameters and surface roughness parameter
Process E S I L H Mmean PG 0.968 0.968 0.968 0.790 0.968 0.968 HM 0.968 0.968 0.968 0.968 0.968 0.799 VM 0.968 0.968 0.968 0.809 0.968 0.252 GP 0.943 0.943 0.935 0.928 0.943 0.943 Process Acon Bent D κ σ υ PG 0.968 0.968 0.968 0.968 0.968 0.968 HM 0.770 0.968 0.657 0.968 0.968 0.968 VM 0.284 0.945 0.177 0.959 0.968 0.968 GP 0.943 0.943 0.807 0.943 0.935 0.943
Table 4. Optimal
C andg of the SVR model with 8 input feature parametersView tableView in ArticleTable 4. Optimal
C andg of the SVR model with 8 input feature parametersHyper parameter PG HM VM GP C 3.7321 48.5029 0.8706 4.5948 g 0.0544 0.0718 8.0000 1.7411
Table 5. Prediction error of the SVR model with 8 input feature parameters
View tableView in ArticleTable 5. Prediction error of the SVR model with 8 input feature parameters
Process RMSE of the test set/μm MAPE of the test set PG 0.0151 4.77% HM 0.0375 3.72% VM 0.1979 8.16% GP 0.0022 3.88%
Table 6. Optimal
C andg of the SVR model with 5 input feature parametersView tableView in ArticleTable 6. Optimal
C andg of the SVR model with 5 input feature parametersHyper parameter PG HM VM GP C 2.8284 48.5029 1.6245 4.2871 g 0.0769 0.1015 1.6245 2.2974
Table 7. Prediction error of the SVR model with 5 input feature parameters
View tableView in ArticleTable 7. Prediction error of the SVR model with 5 input feature parameters
Process RMSE of the test set/μm MAPE of the test set PG 0.0148 3.55% HM 0.0371 3.10% VM 0.0532 3.17% GP 0.0016 2.27%
Table 8. The confusion matrix of the prediction results by the SVM classifier
View tableView in ArticleTable 8. The confusion matrix of the prediction results by the SVM classifier
Process PG HM VM GP PG 24 0 0 0 HM 0 24 0 0 VM 0 0 24 0 GP 0 0 0 18
Table 9. The confusion matrix of the prediction results by the KNN classifier
View tableView in ArticleTable 9. The confusion matrix of the prediction results by the KNN classifier
Process PG HM VM GP PG 24 0 0 0 HM 0 23 0 1 VM 0 0 24 0 GP 0 1 0 17
Tools
Get Citation
Copy Citation Text
Pengfei Wu, Zhizhong Deng, Sichen Lei, Zhenkun Tan, Jiao Wang. Research on surface roughness modeling based on multiple feature parameters of laser speckle image[J]. Infrared and Laser Engineering, 2023, 52(12): 20230348
Paper Information
Category:
Received: May. 20, 2023
Accepted: --
Published Online: Feb. 23, 2024
The Author Email:
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20