Prediction of the Melt Pool Size in Single-Layer Single-Channel Selective Laser Melting Based on Neural Network

Abstract

The unstable forming quality of parts formed by selective laser melting (SLM) process has been one of the obstacles of its development and application, and the thermal process directly influences the forming quality in SLM process, such as the melt pool geometry. For the sake of studying the influence of different process parameters on the melt pool size in SLM forming process, a finite element model by ANSYS was established and single-layer single-channel temperature field imitation of the SLM 316 L stainless steel part under the combination of different laser power, scanning speed, focusing spot diameter and layer thickness was conducted in this paper. Since the neural network (NN) can fully approximate the complex nonlinear relationship, the melt pool size obtained by simulation is used as the training samples, and the NN model is trained to establish the mapping relation model between the SLM process parameters and the melt pool size, which provides the reference for the SLM process parameter optimization. The experimental results indicate that the deviation between the predicted results and the measured results is less, which indicates that the model has high prediction accuracy. A good mapping relation between the studied process parameters and the melt pool size is established.