Data-driven investigation of microstructure and surface integrity in additively manufactured multi-principal-element alloys

Abstract

The great variability of the additively manufactured quality is becoming a major challenge to limit its wide application. Here, we combine the high-throughput molecular dynamics simulations with the artificial neural network to search for the high surface quality and damage minimization of the selective laser melted (SLMed) FeCrNi multi-principal-element alloy (MPEA), based on the complex microstructure obtained from experiment characterization. By adjusting the laser scanning speed and laser power density, a large amount of data related to the surface roughness, dislocation characteristics, and residual stress in the SLMed MPEAs is obtained for optimizing the SLM process. Atomic simulation results show that there are various degrees of Cr segregation under different processing parameters, which leads to the uncertainty of residual stress distribution and dislocation characteristics. Thus, the surface roughness and subsurface damage is highly sensitive to the SLM parameters. According to the training results of the artificial neural network, the laser scanning speed plays a more key role in regulating the surface integrity and subsurface damage in the SLMed MPEAs. The overall prediction results exhibit that the relatively slow laser scanning speed effectively reduces the residual stress. Importantly, a complex relationship between the SLMed parameters and properties is built. The current work provides a predictive recommendation to optimize the quality and process design of the SLMed MPEAs with high surface integrity and low subsurface damage.