Machine Learning-Assisted Analysis of Combustion and Ignition in As-milled and Annealed Al/Zr Composite Powders
Micron-scale metal-based composite powders are promising for energetic applications due to their tailored ignition and combustion properties. In particular, ball-milled Al/Zr composites exhibit lower ignition thresholds than pure aluminum, driven by exothermic intermetallic formation reactions and have demonstrated enhanced combustion properties. However, the extent to which this heat release governs ignition and combustion remains unclear, especially when progressively removed through annealing. To systematically investigate this effect, we synthesized Al/Zr powders (3Al:Zr, Al:Zr, and Al:3Zr at%) via ball milling, annealed them in argon up to 1000 C to partially complete the formation reactions, and characterized their ignition and combustion behavior. Ignition thresholds were measured using a hot wire method across different environments, while high-speed hyperspectral imaging tracked single-particle burn durations and temperatures. A convolutional neural network (CNN)-based method was developed to quantify the frequency of microexplosions. Results show that annealing - and thus reducing available reaction heat - increases ignition thresholds, most significantly for Al-rich compositions. In contrast, Zr-rich powders exhibit little change in ignition thresholds due to oxidation aiding ignition. Despite removing the available heat that drives ignition, average combustion temperatures range from 2400-3000 K and increased with annealing for Al- and Zr-rich powders. Average maximum temperatures are 100 to 400 K higher. The frequency of microexplosions remains high (>46%) and increases with annealing for all but the Al-rich powders. These findings suggest that while homogeneous Al/Zr powders (e.g., atomized) may exhibit higher ignition thresholds, they can achieve comparable combustion performance once ignited.