The crystal structure has a great influence on mechanical sensitivity and detonation performance of energetic materials. An efficient microfluidic platform was applied for size, morphology, and crystallinity controllable preparation of ultrafine HMX. The microfluidic platform has good mixing performance, quick response, and less reagent consumption. The ultrafine γ-HMX was first prepared at room temperature by microfluidic strategy, and the crystal type can be controlled accurately by adjusting the process parameters. With the increase in flow ratio, the particle size decreases gradually, and the crystal type changed from β-HMX to γ-HMX. Thermal behavior of ultrafine HMX shows that γ→δ is easier than β→δ, and the phase stability of HMX is β > γ > δ. Furthermore, the ultrafine β-HMX has higher thermal stability and energy release efficiency than that of raw HMX. The ultrafine HMX prepared by microfluidic not only has uniform morphology and narrow particle size distribution, but also exhibits high density and low sensitivity. This study provides a safe, facile, and efficient way of controlling particle size, morphology, and crystallinity of ultrafine HMX.

As the number of obsolete solid rocket engines increases, determining methods to disassemble and reuse these engines has garnered increasing attention. The separation of solid propellants from the engine shell in an effective and safe way has important research significance. In this study, cavitation water jet technology was employed to extract solid propellant from the engine shell owing to its high breakage efficiency with low working pressure. The effects of the target distance and incident pressure on the breakage efficiency of solid propellants were investigated based on a cavitation water jet experimental system that we designed. A nonlinear relationship between the breakage efficiency and both the target distance and incident pressure was discovered, and the mechanism of solid pro-pellant breakage by a cavitation water jet was proposed. To reduce the cost and time associated with the experiments, a machine learning approach was designed to predict the failure efficiency. Back propagation neural networks, support vector regression, genetic programming, and Gaussian process regression were adopted to construct the models. The results demonstrate that the back propagation neural network achieved the highest accuracy with a value of 0.974, followed by support vector regression with an accuracy value of 0.914 for predicting the mass loss rate. Therefore, machine learning technology is an effective tool for predicting the solid propellant breakage efficiency impacted by cavitation water jets.