Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Reference Dataset and Benchmark for Reconstructing Laser Parameters from On-axis Video in Powder Bed Fusion of Bulk Stainless Steel

Cyril Blanc, Ayyoub Ahar, Kurt De Grave

TL;DR

This work tackles the lack of public data and benchmarks for machine learning tasks in LPBF by introducing raise-lpbf, a terabyte-scale dataset that randomizes laser power and speed per scan line and records on-axis high-speed video at 20,000 FPS. Ground-truth laser parameters are aligned with the recorded controller data, and a public benchmark Makebench evaluates models on reconstructing power, speed, or linear power density from video frames, using RMSE on randomized layers. Baseline evaluations show 3D CNN-based models, particularly SlowFast, achieving the best performance among tested architectures, with Transformers underperforming under current input configurations. The dataset and benchmark enable ML-driven LPBF monitoring and anomaly detection, with planned future work including CT-based porosity prediction to further advance automated, low-defect AM manufacturing of stainless steel components.

Abstract

We present RAISE-LPBF, a large dataset on the effect of laser power and laser dot speed in powder bed fusion (LPBF) of 316L stainless steel bulk material, monitored by on-axis 20k FPS video. Both process parameters are independently sampled for each scan line from a continuous distribution, so interactions of different parameter choices can be investigated. The data can be used to derive statistical properties of LPBF, as well as to build anomaly detectors. We provide example source code for loading the data, baseline machine learning models and results, and a public benchmark to evaluate predictive models.

Reference Dataset and Benchmark for Reconstructing Laser Parameters from On-axis Video in Powder Bed Fusion of Bulk Stainless Steel

TL;DR

This work tackles the lack of public data and benchmarks for machine learning tasks in LPBF by introducing raise-lpbf, a terabyte-scale dataset that randomizes laser power and speed per scan line and records on-axis high-speed video at 20,000 FPS. Ground-truth laser parameters are aligned with the recorded controller data, and a public benchmark Makebench evaluates models on reconstructing power, speed, or linear power density from video frames, using RMSE on randomized layers. Baseline evaluations show 3D CNN-based models, particularly SlowFast, achieving the best performance among tested architectures, with Transformers underperforming under current input configurations. The dataset and benchmark enable ML-driven LPBF monitoring and anomaly detection, with planned future work including CT-based porosity prediction to further advance automated, low-defect AM manufacturing of stainless steel components.

Abstract

We present RAISE-LPBF, a large dataset on the effect of laser power and laser dot speed in powder bed fusion (LPBF) of 316L stainless steel bulk material, monitored by on-axis 20k FPS video. Both process parameters are independently sampled for each scan line from a continuous distribution, so interactions of different parameter choices can be investigated. The data can be used to derive statistical properties of LPBF, as well as to build anomaly detectors. We provide example source code for loading the data, baseline machine learning models and results, and a public benchmark to evaluate predictive models.
Paper Structure (7 sections, 7 figures, 4 tables)

This paper contains 7 sections, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. LPBF setup
  4. Design of Experiment
  5. Training data
  6. Benchmark and baseline models
  7. Conclusions and future work

Figures (7)

  • Figure 1: Top view of the objects layout on the build plate. Fine grid size is 1mm.
  • Figure 2: Renderings of the 3D design from three viewpoints.
  • Figure 3: Histograms for laser dot speed, power, and linear power density.
  • Figure 4: Example laser path over a layer for one object. Every scanline is colored arbitrarily to symbolize the independent sampling of laser speed and power.
  • Figure 5: Example frames from random scan lines with the corresponding laser parameters.
  • ...and 2 more figures