ScRtO-Bi2TM: Temperature Profile Prediction Using Bidirectional Cross-Attention-Based Deep Learning Model in Additive Manufacturing

Temperature profile prediction is an emerging topic in additive manufacturing systems, used to control the printing process and provide quality products. Most of the existing techniques are found to have inefficient predictions due to the dependence on a large amount of training data, require high-intensity cost, and calibrate unknown parameters while computing. In addition to that, inefficient techniques and a lack of optimization algorithms cause complex training and can distort the evaluation with the presence of unwanted data points. To address the aforementioned challenges, the research proposes an effective search and rest optimization-enabled bidirectional cross-attention-long short-term memory (ScRtO-Bi2TM) model for temperature profile prediction. The proposed ScRtO-Bi2TM model depends on observing long-range sequential data points and only focuses on the most relevant parts of the input by applying the bidirectional cross-attention mechanism, which enables an effective training process for better prediction. Furthermore, the development of the ScRtO algorithm enhances the convergence efficiency of the proposed model by effectively fine-tuning the hyperparameters, which reduces the computational complexity as well as error loss. The ScRtO-Bi2TM model’s prediction is equipped with high-quality techniques and algorithms, resulting in improved overall efficiency and achieving a low mean squared error (MSE) of 4.84 and root mean square error (RMSE) of 0.25 compared to other conventional approaches.