A credal decision tree classifier approach for surface condition monitoring of friction stir weldment through vibration patterns
| Autor: | Aravinth Sivakumar, Kuppan Chetty Ramanathan, Meenakshi Prabhakar, Joshuva Arockia Dhanraj, Balachandar Krishnamurthy, Bhavya Lingampalli |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
010302 applied physics
Materials science business.industry Decision tree learning Process (computing) Condition monitoring 02 engineering and technology Structural engineering Welding 021001 nanoscience & nanotechnology 01 natural sciences law.invention Vibration Transverse plane law 0103 physical sciences Friction stir welding 0210 nano-technology business Spinning |
| Zdroj: | Materials Today: Proceedings. 46:1127-1133 |
| ISSN: | 2214-7853 |
| DOI: | 10.1016/j.matpr.2021.02.054 |
| Popis: | Solid-state welding that uses a non-consumable method is Friction Stir Welding (FSW). Because of the friction produced between the workpiece and the tool, the metal becomes plasticized and, with a particular feed rate and spinning speed, the tool is transverse in the desired direction. Present work was carried out with vibration data for the condition monitoring of the FSW weldment. The welding was done using an H13 hexagonal pin profile for two dissimilar metals (AA6063 and AA5052). The test was carried out with the process parameters such as 1600 rpm spindle speed, 1 mm/s feed rate, and a 5.9 mm penetrating depth. The requisite descriptive statistical characteristics have been extracted from the vibration data and feature selection through C4.5. The credal decision tree classifier (CDT) used to perform the classification of the faults like tunnel defect (TD), kissing bonds (KB), root sticking (RS), incomplete fusion (IF), flash (FL), weld root-flaw crack (WC), oxidation (OX) and lack of fill defects (LF) and the prediction accuracy was achieved to be 97.44% with the computational time of 0.46 s. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect