In this work, machine learning (ML) models have been developed to predict the feature responses in additive manufacturing to guide the process operation to achieve higher prediction accuracy. Artificial neural network models have been trained with the help of nature-inspired metaheuristic algorithms. This work also demonstrates the power of nature-inspired algorithms for learning the complex non-linearity present inside the data set. Importantly, the proposed models serve as a foundation for Digital Twin enabled AM process optimization under uncertainty, enhancing real-time monitoring and predictive analytics in additive manufacturing systems. Several prediction models were established among the laser power, scanning speed and track width, track depth, track height, and contact angle on the formation of a single track. The performance of various models in terms of their prediction accuracy, adaptability and relevance were further evaluated by comparing each of the models outcome. The experimental data has been collected from the literature. The result shows that the iFANN gives a better performance for predicting the track width, depth, height, and contact angle. The metaheuristic algorithm-guided neural network models perform extremely well for predicting the track height and contact angle compared to the traditional regression model. The iFANN model achieved minimized RMSE prediction value with only 30 generations, which traditional backpropagation model fails to achieve the same accuracy even with higher number of generations.
Ball et al. (Thu,) studied this question.