Attention Mechanism-Incorporated Deep Learning for AM Part Quality Prediction
Jianjing Zhang, Peng Wang, Robert X. Gao
Abstract
To improve the consistency of part quality in Additive Manufacturing, it is critical to understand the relationship between the mechanisms underlying the layer-by-layer printing process and the resulting product quality. This paper investigates this relationship by incorporating attention mechanism into a Long Short-term Memory network, using Fused Deposition Modeling as a case study. In-process thermal variations, as reflected in the in-situ temperature measurement, are fused with machine settings to establish a data-driven model for part tensile strength prediction. Analysis using attention mechanism quantified the relative influence of each printed layer on the predictive result, providing insight into the network operation.