Soft computing-based process optimization in laser metal deposition of Ti-6Al-4 V

Parameter optimization is significant to a successful laser metal deposition (LMD). While conventional optimization methods have been used, the prowess of soft computing techniques is still less explored in LMD towards ensuring reduced experimental costs and throughput. This study develops a process optimization and wear volume prediction model for Ti-6Al-4 V using soft computing techniques. The particle swarm optimization (PSO) model was used to optimize a single objective function to determine optimal process parameters. A supervised learning model using artificial neural network (ANN) was developed to predict the wear volume from known process parameters. The model hyperparameters were tuned by several trials until optimal parameters were obtained. The ANN model was trained and tested with 70% and 30% of the dataset, respectively. The ANN model was evaluated using known statistical performance metrics and user-friendly interfaces, where process optimization can be carried out within upper and lower design bounds, which were developed for the two intelligent models. From the model evaluation result, a root mean square error (RMSE) of 0.0052, mean absolute deviation (MAD) of 0.0031, coefficient of determination (R²) of 0.9733, and a mean absolute percentage error (MAPE) of 14.0152 was obtained from the model testing phase. Overall, soft computing techniques prove helpful in ensuring process integrity, efficient, and cost-effective LMD.