Machine learning approach to model and predict the shear strength of TZM-graphite joint bonded by spark plasma sintering

These days, the study of titanium alloys has grown into a sizable topic of study. Due to its remarkable high-temperature applications, titanium zirconium molybdenum (TZM) is one such alloy with considerable research potential. In this study, the shear strength of TZM-graphite alloy, which was bonded using the spark plasma sintering (SPS) process with titanium foil as an interlayer, has been predicted using a machine learning approach. The shear strength of the TZM-graphite alloy depends on various process parameters such as sintering temperature, sintering pressure, holding time, and intermediate layer thickness. Since the correlation between input variables and output variables is intricate and non-linear, an artificial neural network (ANN) model was developed in this research to investigate the relationship between bonding parameters and the shear strength of the TZM-graphite joint. The feed-forward backpropagation neural network was utilized for training the model and predicting the shear strength. By computing the Mean square error (MSE) and Average error (AE), the optimum number of neurons in the hidden layers was determined. Consequently, the model with 4–9-9–1 architecture was constructed, and its accuracy was assessed by contrasting the values obtained by the neural network with actual experimental data. The model was then validated using a variety of performance measuring indicators such as mean absolute percentage error (MAPE) and root mean square error (RMSE). The achieved correlation coefficient (R-value) of 0.99614% demonstrates that the proposed ANN model is an excellent fit for the experimental data to predict the accurate shear strength of the TZM-graphite joint. For further understanding of the influence of input parameters on the shear strength of the TZM-graphite joint, 2D and 3D surface graphs were plotted.