Abstract:The advancement of machine learning (ML) provides possibilities for in-situ monitoring and feedback control during directed energy deposition (DED) additive manufacturing (AM) process. However, obtaining sufficient and high-quality training samples experimentally is expensive and infeasible. The scarcity of samples and practical features is a critical barrier to the performance of ML models on DED monitoring tasks. Physics-based high-fidelity simulation can effectively investigate the mechanisms of metal additive manufacturing (MAM) process in detail. The physics-based model incorporates the physical laws, governing equations, material constitutive models and practical experience which are already verified, the above-mentioned physics-based knowledge can be beneficial in improving the performance of ML models. In this work, to rapidly and accurately determine the molten pool morphologies and internal defects, the experimentally validated physics-based simulation and empirical formulas are adopted to calculate the physics-informed model inputs. The physics-based simulation and empirical formulas enable the acquisition of sufficient training samples and high-quality input features in DED cases. The results show that the physics-informed inputs reduce the mean absolute errors (MAEs) of the predicted outputs in the molten pool morphology prediction task, improve the classification accuracy in the defect identification task, and significantly enhance the robustness and generalization capability of ML models. More importantly, with the help of feature importance analysis, the physics-informed features allow for more comprehensive explorations and assessments of the multivariate effects on DED process, leading to a better explanation of the process-property mechanisms. Finally, the DED process windows of 12CrNi2 are obtained for the first time based on the decision boundaries given by ML models. In summary, this work demonstrates the feasibility of constructing the physics-informed model inputs, and the significant improvement of the physics-informed features on the performance of ML models in MAM-related cases.

A Robust Recurrent Neural Networks-Based Surrogate Model for Thermal History and Melt Pool Characteristics in Directed Energy Deposition

A Robust Autoregressive Long-Term Spatiotemporal Forecasting Framework for Surrogate-Based Turbulent Combustion Modeling Via Deep Learning

Prediction of Melt Pool Geometry Using Deep Neural Networks

Temperature evolution prediction for laser directed energy deposition enabled by finite element modelling and bi-directional gated recurrent unit

Prediction of melt pool shape in additive manufacturing based on machine learning methods

Surrogate Modeling of Melt Pool Thermal Field using Deep Learning

Physics-informed machine learning approach for molten pool morphology prediction and process evaluation in directed energy deposition of 12CrNi2 alloy steel

Predictions of in-situ melt pool geometric signatures via machine learning techniques for laser metal deposition

Machine learning analysis for melt pool geometry prediction of direct energy deposited SS316L single tracks

Online melt pool depth estimation during directed energy deposition using coaxial infrared camera, laser line scanner, and artificial neural network

In Situ Monitoring and Innovative Feature Fusion Neural Network for Enhanced Laser-Directed Energy Deposition Track Geometry Prediction and Control

Inconel-718 processing windows by directed energy deposition: a framework combining computational fluid dynamics and machine learning models with experimental validation

Traditional machine learning and deep learning for predicting melt-pool cross-sectional morphology of laser powder bed fusion additive manufacturing with thermographic monitoring

Deep Learning for Melt Pool Depth Contour Prediction From Surface Thermal Images via Vision Transformers

Prediction of molten pool temperature and processing quality in laser metal deposition based on back propagation neural network algorithm

Model Predictive Synchronous Control of Barrel Temperature for Injection Molding Machine Based on Diagonal Recurrent Neural Networks

An Augmented Surprise-guided Sequential Learning Framework for Predicting the Melt Pool Geometry

Self-consistency Reinforced minimal Gated Recurrent Unit for surrogate modeling of history-dependent non-linear problems: Application to history-dependent homogenized response of heterogeneous materials

A Physics-Informed Neural Network Framework to Predict 3D Temperature Field Without Labeled Data in Process of Laser Metal Deposition

Monitoring process stability in robotic wire-laser directed energy deposition based on multi-modal deep learning

Fast and accurate prediction of temperature evolutions in additive manufacturing process using deep learning