Multi-objective optimisation of machining process parameters using deep learning-based data-driven genetic algorithm and TOPSIS
Machining process is currently widely employed in mechanical manufacturing systems. Optimum selection of machining process parameters can improve the environmental impact and production efficiency of the machining process effectively. However, existing studies toward machining process parameters opt...
Uloženo v:
| Vydáno v: | Journal of manufacturing systems Ročník 64; s. 40 - 52 |
|---|---|
| Hlavní autoři: | , , , , , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
Elsevier Ltd
01.07.2022
|
| Témata: | |
| ISSN: | 0278-6125, 1878-6642 |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Shrnutí: | Machining process is currently widely employed in mechanical manufacturing systems. Optimum selection of machining process parameters can improve the environmental impact and production efficiency of the machining process effectively. However, existing studies toward machining process parameters optimisation are focusing on computationally expensive numerical simulations and costly physical models, which are inefficient and labor-expensive. Moreover, the numerical simulations and physical models often show an unsatisfactory accuracy in the actual exploitation stage, which would make the final optimisation solution cannot achieve the best optimum results. Therefore, this paper proposes a deep learning based data-driven genetic algorithm and TOPSIS for multi objective optimisation of machining process parameters and searching the final solutions. First, deep learning is employed in this paper to automatically develop the data-driven prediction function of different optimized objectives. Then the developed optimized objective prediction function is converted into the surrogate model and integrated with the genetic algorithm for generating the Pareto set. Finally, the TOPSIS is employed to automatically search the best optimum processing parameter from the generated Pareto set. The experiments conducted on a milling machine and the experimental results show that the proposed parameters selection method is feasible and effective, and it can effectively and adjustably help operators to realize a balance among the multiple different conflicting objectives.
•Data-driven genetic algorithm is proposed and employed for machining process parameters optimization.•Deep learning is employed to automatically develop the data-driven prediction function of different objectives.•This work provides a cost-efficient way for the operators to select the process parameters of machining process.•Little effort and expert knowledge are required during the optimal selection of process parameters.•The results can help manufacturers to select best optimal energy-efficient machining process in a flexible way. |
|---|---|
| ISSN: | 0278-6125 1878-6642 |
| DOI: | 10.1016/j.jmsy.2022.05.016 |