Sreenivas' latest paper on ML-guided combustion mechanisms in Al/Zr powders
📢 New Preprint Alert
Sreenivas’ latest collaborative paper has been submitted to arXiv, titled:
“Machine Learning–Assisted Analysis of Combustion and Ignition in As-milled and Annealed Al/Zr Composite Powders”
🔥 The Core Idea
Micron-scale Al/Zr metal composites are highly promising for energetic applications, but the mechanistic understanding of ignition and combustion remains incomplete—especially after annealing.
This study bridges the gap by combining high-speed hyperspectral imaging with machine learning to decode particle-level ignition thresholds and combustion patterns.
🛠️ Methodology in Focus
The team:
- Synthesized Al-Zr powders (3Al:Zr, Al:Zr, Al:3Zr) via ball milling
- Performed argon annealing up to 1000 °C to trigger partial reaction pathways
- Used a hot wire method to ignite powders in diverse atmospheres
- Captured single-particle burn behaviors via high-speed imaging
- Developed a CNN-based model to quantify microexplosion frequency and correlate it with composition and thermal history
🌡️ Key Insights
- 🔥 Al-rich powders show lower ignition thresholds but are more sensitive to annealing
- ⚙️ Zr-rich powders are less affected by heat loss during annealing due to oxidation-assisted ignition
- 📈 Combustion temperatures increased significantly (by 100–400 K) post-annealing
- 💥 Despite heat removal, microexplosions remained >46% frequent, indicating strong combustion stability
🧠 Why It Matters
By integrating convolutional neural networks with energetic materials testing, this work offers a new paradigm for predicting combustion thresholds. This could enable:
- Safer, more efficient propellant design
- Real-time diagnostics in explosive safety testing
- Tailored annealing schedules to engineer desired burn characteristics
🤝 Collaboration & Vision
This effort reflects a deep collaboration between experimental combustion science and data-driven modeling, led by Michael R. Flickinger, with contributions from
Sreenivas Raguraman, Mark A. Foster, Timothy P. Weihs, and others.
It paves the way for smarter material design by quantifying how microscale ignition dynamics evolve with heat treatment—a step toward predictive synthesis in energetic systems.
📄 Citation:
M.R. Flickinger, S. Raguraman, et al., arXiv:2506.06364 (2025).
🔗 https://doi.org/10.48550/arXiv.2506.06364