Process parameter optimisation method based on data-driven prediction model and multi-objective optimisation for the laser metal deposition manufacturing process monitoring
•A data-driven prediction model based on random forest was developed. The model takes into account internal defects porosity and cracks, and establishes a non-explicit function between process parameters and quality as the objective optimisation function.•A multi-objective optimisation algorithm for...
Gespeichert in:
| Veröffentlicht in: | Computers & industrial engineering Jg. 204; S. 111108 |
|---|---|
| Hauptverfasser: | , , , , , , |
| Format: | Journal Article |
| Sprache: | Englisch |
| Veröffentlicht: |
Elsevier Ltd
01.06.2025
|
| Schlagworte: | |
| ISSN: | 0360-8352 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Zusammenfassung: | •A data-driven prediction model based on random forest was developed. The model takes into account internal defects porosity and cracks, and establishes a non-explicit function between process parameters and quality as the objective optimisation function.•A multi-objective optimisation algorithm for process parameters suitable for manufacturing processes is constructed. The algorithm combines random forest and NSGA-II for the first time, eliminates the need for expensive and time-consuming multiple testing experiments, and improves the reliability and applicability of the optimisation model.•The optimal solution search strategy is established. The strategy enables the optimisation algorithm can automatically search the best process parameters from the set of Pareto solutions and obtain the optimal solution, which meets the actual requirements of industrial manufacturing.
Process parameter optimisation is essential for laser metal deposition manufacturing process monitoring, which can minimize internal defects and enhance component quality. However, existing process parameter optimisation mainly focuses on experimental design and curve fitting, which are time-consuming, labour-intensive, and expensive to test, thus, they are not able to effectively ensure accuracy. In addition, the dynamic changes in quality under the same process parameters in the manufacturing processes make it more difficult to analyse the optimal process parameter combination through experimentation. In this regard, a process parameter optimisation method based on a data-driven prediction model and the multi-objective optimisation algorithm is proposed in this paper to obtain the optimal process parameter combination. This method carries out multi-process parameter deposition experiments to count the number of each quality level based on the established quality evaluation standard. Then, a data-driven prediction model by random forest was used to automatically develop a non-explicit prediction function, which establishes the relationship between process parameters and different quality levels. Subsequently, the NSGA-II (Non-dominated Sorting Genetic Algorithm II) multi-objective optimisation algorithm was utilised to generate the optimal set of Pareto solutions for the process parameters. Finally, the optimal process parameter combinations are automatically searched based on the proposed search strategy. Experimental results show that the components under the optimal process parameters have the least internal defects and the best quality, which indicates that the proposed method can provide effective guidance for the manufacturing process monitoring. |
|---|---|
| ISSN: | 0360-8352 |
| DOI: | 10.1016/j.cie.2025.111108 |