Selective laser melting of LM25 aluminium alloy reinforced with boron carbide and graphite for machine learning based wear prediction and microstructural characterization

Abstract

The Selective Laser Melting (SLM) is a modern and innovative aluminium alloy metals additive synthesis technique for fabricating lightweight aluminium matrix composites with complex geometries and superior engineering performance. In the present investigation, LM25 aluminium alloy reinforced with boron carbide (B 4 C) and graphite (Gr) particles was successfully fabricated using the SLM process to enhance its tribological properties. Pin-on-disc wear experiments were conducted under different practical loads, sliding distances, sliding velocities, and sliding times to evaluate the specific wear rate (SWR), Coefficient of Friction (CoF), and Frictional Force (FF). Optical Microstructural analysis using optical microscopy showed refined α-Al dendrites, more uniform distribution of eutectic Si, and decreased porosity with increase in the amount of hybrid reinforcement, whereas XRD analysis showed that α-Al, Si, Mg₂Si, B 4 C, and graphite phases were present without any unwanted reaction products. The worn surface morphology and wear mechanisms were investigated using Field Emission Scanning Electron Microscopy (FESEM), which revealed reduced wear grooves, limited plastic deformation, lower material removal, and the formation of a protective lubricating tribo-layer in the B 4 C/Gr-reinforced hybrid composites, confirming their superior wear resistance compared with the unreinforced LM25 alloy. The four types of machine learning (ML) models, Artificial Neural Network (ANN), Random Forest (RF), Support Vector Machine (SVM), and Gaussian Process Regression (GPR) were trained on experimental wear data to predict the tribological responses accurately. Prediction performances of the models were assessed using the coefficient of determination (R 2 ), root mean square error (RMSE), mean absolute error (MAE), mean squared error (MSE), and mean absolute percentage error (MAPE). Among the four evaluated modelling architectures, the RF model had the highest accuracy with R 2 of 0.999, RMSE of 0.0013, MAE of 0.0010, MSE of 1.7 × 10 –6 , and MAPE of 0.30%. In addition, the LM25/2 wt.% B 4 C/2 wt.% Gr hybrid composite showed the lowest specific wear rate and coefficient of friction values, indicating highly improved tribological properties compared to the LM25 alloy without any reinforcements. The developed experimental-mechanical learning framework provides approach for predicting and optimizing the tribological behaviour of SLM-fabricated LM25/ B 4 C/Gr hybrid composites under investigated wear testing conditions.