Training-Validation-Testing Dataset for "Real-Time Prediction of Thermal History and Hardness in Laser Powder Bed Fusion Using Deep Learning"

The dataset contains 2 .zip files (V14_FEMData__TrainingValidation and V14_FEMData__Testing) with 3 different process parameter combinations (RW006, RW030, RW076), three different build plate preheating temperatures (BPT025, BPT100, BPT200) and different cross sections (sqaure 3 mm x 3 mm up to 12 mm x12 mm) for the training and validation. As well data for the testing with modified process parameter combinations and cross sections.

The FEM simulations were based on the multiscale framework published for the quenched and tempered steel 42CrMo4 / AISI 4140 (Schüßler, P., Nouri, N., Schulze, V., & Dietrich, S. (2023). A novel multiscale process simulation to predict the impact of intrinsic heat treatment on local microstructure gradients and bulk hardness of AISI 4140 manufactured by laser powder bed fusion. Virtual and Physical Prototyping, 18(1). https://doi.org/10.1080/17452759.2023.2271455).

Each dataset is saved in the hdf5 format with 20k rows and 25 columns. Each row represents a time step. The columns are listed below:

• 0: step index [-]
• 1: total frame [-]
• 2: step frame [-]
• 3: total time [s]
• 4: step time [s]
• 5: export interval [s]
• 6: step percentage [%]
• 7: laser spot diameter [mm]
• 8: laser power [W]
• 9: laser speed [mms/s]
• 10: scan vector start x coordinate [mm]
• 11: scan vector start y coordinate [mm]
• 12: scan vector start z coordinate [mm]
• 13: scan vector end x coordinate [mm]
• 14: scan vector end y coordinate [mm]
• 15: scan vector end z coordinate [mm]
• 16: measurement point x coordinate [mm]
• 17: measurement point y coordinate [mm]
• 18: measurement point z coordinate [mm]
• 19: hatch_angle [°]
• 20: RESULT temperature [K]
• 21: RESULT austenite fraction [-]
• 22: RESULT martensite fraction [-]
• 23: RESULT Hollomon-Jaffe tempering parameter PHJ [-]
• 24: RESULT Vickers hardness [HV]