Long-short-term memory (LSTM)-based modeling of the stiffness of 3D-printed PLA parts

This study applied a computationally efficient Taguchi-based long-short-term memory (LSTM) algorithm to predict the elastic modulus of 3D-printed polylactic acid (PLA) specimens. 128 data points were collected from the literature, and 80% were allocated for training and the rest for the validation of the LSTM algorithm. The results suggested that the LSTM algorithm, configured with 25 units in the first memory cell, 100 units in the second memory cell, the “selu” activation function in the first memory cell, the “elu” activation function in the second memory cell, the RMSprop optimizer, and a learning rate of 0.01, was precisely able to predict the elastic modulus of 3D-printed PLA parts.