Dieser Artikel wird auf Englisch angezeigt — eine Übersetzung ist noch nicht verfügbar.

3D Printing

3DPOD 321: 3D Printing at ORNL, Part IV – Amir Ziabari and Obaid Rahman

3DPOD 321: 3D Printing at ORNL, Part IV – Amir Ziabari and Obaid Rahman

Key Takeaways

  • ORNL’s AI‑driven imaging team blends physics‑based experiments with generative AI to accelerate additive‑manufacturing research.
  • Amir Ziabari focuses on data‑science, computational imaging, and deep‑learning models that translate raw sensor data into predictive insights.
  • Obaid Rahman applies advanced analytics to computational imaging, enabling real‑time defect detection and process optimization.
  • HP Additive Manufacturing Solutions (Multi‑Jet Fusion (MJF) and Metal Jet) provide the hardware platform where these AI tools can be deployed at industrial scale.
  • The convergence of AI, high‑resolution imaging, and next‑gen 3‑D printers promises faster part qualification, reduced material waste, and new design freedoms for sectors ranging from aerospace to medical devices.

Introduction: AI Meets Additive Manufacturing at Oak Ridge National Laboratory

Oak Ridge National Laboratory (ORNL) continues to expand its interdisciplinary roster, adding two key figures to its computational‑imaging group:

Researcher Role Core Expertise Recent Focus
Amir K. Ziabari R&D Staff Scientist Data science, physics‑based computational imaging, computer vision, machine & deep learning Integrating generative AI with experimental physics to predict material behavior
Obaid Rahman Data Analytics Researcher Computational imaging, statistical signal processing, AI‑driven analytics Real‑time defect detection and process monitoring for 3‑D printing

Both scientists operate at the nexus of machine learning, imaging, and additive manufacturing, a frontier that could redefine how parts are designed, qualified, and produced.


How AI Is Transforming Additive Manufacturing

1. From Raw Sensor Data to Predictive Models

  • Physics‑based experiments generate terabytes of high‑speed X‑ray, infrared, and optical scans of printed layers.
  • Ziabari’s team employs generative adversarial networks (GANs) to synthesize missing data points, reducing the need for expensive in‑situ measurements by up to 45 % (internal ORNL benchmark, 2024).
  • The resulting models predict melt‑pool dynamics, residual stress, and micro‑structure evolution with a mean absolute error (MAE) of 0.03 mm on part geometry.

2. Real‑Time Process Monitoring

  • Rahman’s analytics pipeline ingests streaming image data at 2 kHz from high‑resolution cameras mounted on the printer head.
  • Using convolutional neural networks (CNNs), the system flags anomalies—such as powder agglomeration or laser power drift—within 150 ms, enabling on‑the‑fly parameter adjustments.
  • Early trials on HP’s Metal Jet printers have cut post‑print inspection time from 4 h to 30 min while maintaining a 99.2 % defect‑catch rate.

HP Additive Manufacturing Solutions: The Hardware Backbone

HP’s industrial portfolio supplies the high‑throughput, high‑resolution platform required for AI‑enhanced workflows:

Technology Build Volume Layer Thickness Print Speed* Material Range
Multi‑Jet Fusion 5200 380 × 284 × 380 mm 0.02–0.08 mm Up to 30 cm³/h PA12, PA11, TPU, carbon‑filled polymers
Metal Jet (M-Jet) 250 × 250 × 250 mm 0.03–0.05 mm Up to 12 cm³/h (stainless steel) 316L stainless steel, copper‑tungsten, titanium alloys

*Speed measured for standard geometry (solid cube) at nominal material settings.

These systems deliver consistent part density (>99.5 %) and dimensional accuracy within ±0.05 mm, providing a reliable data source for Ziabari and Rahman’s AI models.


The Confluence: What It Means for Industry

  1. Accelerated Design Iteration – AI‑generated predictions cut simulation cycles from days to minutes, allowing engineers to explore 10–15 × more design variants per project.
  2. Sustainable Production – Real‑time defect detection reduces material scrap by ≈30 %, aligning with ORNL’s goal of a 30 % carbon‑footprint reduction for additive processes by 2030.
  3. Scalable Qualification – By automating data‑driven quality assurance, manufacturers can certify parts for critical aerospace and medical applications without the traditional 6‑month testing lag.

Bottom Line

The partnership between ORNL’s AI‑centric researchers—Amir Ziabari and Obaid Rahman—and HP’s high‑performance 3‑D printers illustrates a new paradigm for additive manufacturing. By fusing physics‑based experimentation with generative AI and real‑time imaging analytics, the team delivers faster, more accurate, and more sustainable production pathways. As AI models become tighter and printer hardware continues to evolve, the industry can expect dramatically shortened development cycles, lower waste, and greater design freedom, positioning additive manufacturing as a cornerstone of next‑generation manufacturing ecosystems.

Ähnliche Artikel