Machine Learning Tools to Optimize Metal Additive Manufacturing Process Parameters to Enhance Fatigue Performance of Aircraft Components

Global Engineering and Materials, Inc. (GEM) along with team members, Northwestern University (NU) and National Institute for Aviation Research (NIAR) at Wichita State University, propose a coupled tool based on machine learning and integrated computational materials engineering (ICME) to optimize metal additive manufacturing (AM) process parameters to improve fatigue performance of aircraft components. This data-driven framework will utilize a novel transfer learning approach that integrates data from multiple simulation and experimental sources, with distinct but overlapping sets of inputs and outputs, to generate fatigue predictions superior to those obtained through any individual simulation or experiment. NU’s process simulation tools, including thermal-fluid models at multiple levels of fidelity and crystal plasticity self-consistent clustering analysis (CPSCA), will be leveraged to predict thermal history, surface roughness, and fatigue performance under given parameter ranges. Existing literature fatigue data will be supplemented with new measurements, including in-situ sensor data providing thermal and melt pool details, and incorporated into the framework. Experimental specimens for training and validating the model will be fabricated by NU. Pre-test nondestructive evaluation and fatigue tests will be performed by NIAR both to build the training dataset and to demonstrate improved fatigue performance via process optimization guided by our new model.