Effects of process parameters on the properties of Ba𝐀𝐥𝟐𝐎𝟒 based coating deposited on steels by novel explosive spray technique and prediction of coating hardness by using machine learning technique

Abstract

In the present work, the mixture of barium nitrate (Ba(NO3)2) and aluminum (Al) powders are burned to deposit barium aluminate (BaAl2O4) composite coatings on mild steel substrate by a novel explosive spray coating setup. Various aluminium amounts (1, 1.5, 2, and 2.5 gm) and different stand-off distances (60, 80, and 100 mm) are employed during the deposition process. The X-ray diffractometer (XRD) is used to analyze the phases present in the different coatings. The significant peaks of barium aluminate phase along with some secondary phases, are observed with varying Al contents and changing stand-off distances (SOD). The hardness of the coating and substrate is estimated by using Vickers microhardness tester. The maximum hardness value (1313HV0.05) is noticed for the coating fabricated by employing 2 gm of Al at 80 mm SOD. The hardness of the substrate at 20 μm, below the interface of coating-substrate is found to be maximum for coating obtained through 1 gm of Al at 100 mm SOD. The depth of the hardened zone of coating is observed to be decreased with an increase in Al amount from 1 to 2.5 gm at all three different SODs. A stylus profilometer is engaged to measure the average surface roughness of the coatings and grit-blasted substrate. The coating fabricated by using 2 gm of Al at 80 mm SOD shows the minimum roughness value of 3.5 μm. The thickness of the coating is measured through an inverted optical microscope. The maximum coating thickness is achieved by employing 2.5 gm of Al at 60 mm of SOD among all different combinations of Al content and SOD. Using a machine learning model to predict the hardness of the coating. Random Forest model shows an accuracy of 99.22% prediction of hardness coating and its RMSE value is also 19.95, which is less than other machine learning models (Decision Tree and Linear Regression).