Dataset for the manuscript Adaptive Run-Time Control of Laser-based Direct Energy Deposition using a Recurrent Neural Network Surrogate Controller

This repository provides the dataset and model‑training resources supporting the manuscript “Adaptive Run‑Time Control of Laser‑based Direct Energy Deposition using a Recurrent Neural Network Surrogate Controller.” It includes the synthetic high‑fidelity (HF) training data generated using an experimentally calibrated Finite Element (FE) thermal model for surrogate‑model development. The repository also contains data from the statistical analysis of input features, the hyperparameter‑tuning process for the recurrent models, and a historical‑sensitivity study used to determine the appropriate amount of past information required by the controller. Additionally, the repository provides test results comprising laser‑power correction targets and the corresponding thermal and melt‑pool geometric process indicators for three evaluation cases, two of which are not part of the training dataset. These files correspond to the results presented in the manuscript and enable full reproduction of the figures, analyses, and the workflow used to identify the most effective controller configuration for stabilizing the DED‑LB process.