Estimation of mechanical properties of friction stir processed Al 6061/Al2O3-Tib2 hybrid metal matrix composite layer via artificial neural network and response surface methodology
| Autor: | Hamed H Pourasl, Arian Bahrami, Vahid M Khojastehnezhad |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Friction stir processing
Materials science Artificial neural network Mechanical Engineering 020208 electrical & electronic engineering Metal matrix composite 02 engineering and technology 021001 nanoscience & nanotechnology 0202 electrical engineering electronic engineering information engineering General Materials Science Response surface methodology Composite material 0210 nano-technology Layer (electronics) |
| Zdroj: | Proceedings of the Institution of Mechanical Engineers, Part L: Journal of Materials: Design and Applications. 235:2720-2736 |
| ISSN: | 2041-3076 1464-4207 |
| Popis: | Friction stir processing is one of the solid-state processes which can be used to modify the structure and properties of alloys. In addition, it has become one of the most promising techniques for the preparation of the surface layer composites. To pursue cost savings and a time-efficient design, the mathematical model and optimization of the process can represent a valid choice for engineers. Friction stir processing was employed to generate an Al 6061/Al2O3-TiB2 hybrid composite layer, and mechanical properties such as the hardness and wear behavior were also measured. The relationship between the hardness and wear behavior, process parameters of friction stir processing were evaluated using an artificial neural network and response surface methodology. The rotational speed (1500–1800 rpm), traverse speeds (25, 50, 100 mm/min), and the number of passes (1–4) with constant axial force (2.61 kN) were used as the input, while the hardness and weight loss values were the output. Experimentally, the results showed that the process parameters have significant effect on hardness and wear behavior of Al 6061/Al2O3-TiB2. In addition, the developed artificial neural network and response surface methodology models can be employed as alternative methods to compute the hardness and weight loss for given process parameters. The results of both models showed that the estimated values for the hardness and wear behavior of the processed zone had an error less than 0.60%, which indicated reliability, and an evaluation of the estimated values of both models and the experimental values confirmed that the artificial neural network is a better model than response surface methodology. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|