Experimental investigation and mathematical modelling of dry sliding wear of nitinol reinforced aluminium matrix composite using machine learning

Dry sliding wear is a complex phenomenon, an understanding of which needs careful experimentation and observation of the wear event. The present study focuses on the dry sliding wear analysis of the nitinol particulate-reinforced Al5083 matrix composite fabricated using a vacuum-assisted stir casting route. Nitinol alloys have excellent mechanical, thermo-mechanic and tribological performance along with the unique feature of shape recovery and superelasticity in the austenitic phase. The effect of varying wear control parameters, load (5-20 N), sliding distance (0-3000), sliding speed (150-600 RPM), humidity (20% - 80%) and temperature (25-100 degrees C) on the wear resistance of the fabricated composites were exhaustively examined. Steady-state, as well as Taguchi-based experimentation, were performed to investigate the wear phenomenon in the fabricated composite using a pin-on-disc tribometer. Modified Archard equation with power exponent was utilized to develop four new wear equations, which were both trained and tested using a machine learning algorithm. The results depict that among all four models, the DSW Model-4 has provided the best fit between the actual and the predicted wear volume with a coefficient of determination value of 0.9722. An infrared thermal imager was used to observe temperature variation during the wear test. Scanning electron microscopy (SEM) and a profilometer were used to carefully examine the worn surface.