Tue, 17 September, 2024
A new paper, which was co-authored by representatives from TWI, the University of Birmingham and the National Structural Integrity Research Centre (NSIRC) has been published in the Scripta Materialia journal.
The paper, ‘A Novel Porosity Prediction Framework Based on Reinforcement Learning for Process Parameter Optimization in Additive Manufacturing,’ demonstrates a novel framework based on reinforcement learning (RL) for porosity prediction in metal laser-powder bed fusion (L-PBF).
This work comes as machine learning continues to influence additive manufacturing through developments including the ability to predict complex patterns and behaviours through data, with examples including design optimisation, process control, and cost minimisation.
The novelty of reinforcement learning to predict porosity in metal laser-powder bed fusion is twofold. Firstly, there is the integration of RL in L-PBF for porosity prediction where the state space consists of permutations of three parameters (laser power, scan speed, and hatch spacing) for optimal parameter combinations. In addition, the paper details how an appropriately formulated reward function allowed for physics-informed principles based on the Eagar-Tsai thermal model for training to be embedded.
The proposed framework was experimentally validated on L-PBF high-strength A205 Al alloy, with the results demonstrating high fidelity with the predicted optimal parameters, despite a few outliers.
The paper was co-authored by Ahmed M. Faizan Mohamed and Leonardo Stella from the University of Birmingham School of Computer Science, alongside Moataz M. Attallah of the University of Birmingham School of Metallurgy and Materials, Francesco Careri from the University of Birmingham and National Structural Integrity Research Centre (NSIRC), and Dr Raja Khan from TWI.
You can see the paper, in full, by following this link.