On the machine learning algorithm combined evolutionary optimization to understand different tool designs’ wear mechanisms and other machinability metrics during dry turning of D2 steel
Engineering tool designs through geometrical changes and surface texturing is gaining attention to enhance the tribological behaviour in metal processing industries. However, understanding the interactions of these designs with machining parameter selection considered time-taking process through var...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-03-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123025000866 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Engineering tool designs through geometrical changes and surface texturing is gaining attention to enhance the tribological behaviour in metal processing industries. However, understanding the interactions of these designs with machining parameter selection considered time-taking process through various trial and error experiments. In this study, three-step novel modelling approach for optimal prediction of dry turning parameters is proposed. Firstly, an economical design of experiment approach is opted to evaluate two inserts with distinct designs with machining parameters such as cutting speed (VCS), feed rate (FR), and depth of cut (DOC). Secondly, the open-source supervised machine learning architecture search as carried out to map process characteristics such as surface evolution, tool wear, and chip morphology (width and thickness). Lastly, an evolutionary optimization method is applied to democratize multi-objective process characteristics’ learnings to facilitate machinists and users within manufacturing community. An artificial neural network (ANN) based modelling has been performed to find the correlation between the predicted and the experimental values. It has been found that, an ANN based modelling gave the better results by strongly giving the R2 nearly equal to unity in training, and equal to unit for testing and validation for the response measures. According to non-dominated sorting genetic algorithm (NSGA-II), the results of turning operation of AISI D2 steel in terms of tool life (TLIFE), surface asperities (Ra, Rt, Rz), chip width (CW) and chip thickness (CTH) have been enhanced when compared to the experimental results to the NSGA-II suggested optimal settings. It has been found that the TLIFE, Ra, Rt, Rz, CW and CTH, improved by 354.05 %, 93.87 %, 81.35 %, 88.01 %, 127.01 %, and 124.71 %, respectively, when parametric combinations (VCS = 100 m/min, FR = 0.2 mm/rev, DOC = 0.5 mm, and INType = Xcel) are used. |
|---|---|
| ISSN: | 2590-1230 |