Figure 1.

A flow chart of ML process for balancing detonation performance and stability of HEDMs

High-throughput DFT calculations on supercomputers are the basis of data preparation for ML. Five ML models were trained and evaluated using stratified sampling and leave-one-out cross-validation, and the collected massive experimental data help to validate the predictions. Feature importance ranking provides a data-driven insight into the performance-stability contradiction and presents guidance on the optimal range of critical features.