Particle peak velocity prediction based on risk-oriented hybrid ensemble learning

In blasting engineering, accurate prediction of peak particle velocity ( PPV ) is vital for the safety of surrounding structures. In machine-learning-based PPV prediction, symmetric loss functions (e.g., MSE) are typically used as the optimisation objective, making hazardous underestimation of high-vibration values difficult to avoid. This is unacceptable in applications with stringent safety requirements. To address this limitation, a risk-oriented hybrid ensemble model is proposed to enhance safety and reliability while meeting high-precision prediction requirements. Three gradient-boosting tree models—LightGBM, XGBoost, and CatBoost—are employed as base learners, and a Stacking framework is adopted for integration. To place the three base learners in near-optimal configurations, Bayesian Optimisation (BO), Grey Wolf Optimiser (GWO), and Particle Swarm Optimisation (PSO) were used for hyperparameter tuning. Building on the ensemble, an asymmetric safety assessment system is proposed. Model performance near the PPV safety threshold is quantified using the asymmetric weighted mean squared error (W-MSE) and the Hazardous Low-estimation Rate (HLR). Results indicate that the integrated model achieves excellent performance and effectively eliminates hazardous underestimation risk. The integrated model is shown to offer significant advantages for PPV prediction. It provides a reusable paradigm for embedding engineering safety constraints into machine learning training and evaluation, thereby delivering reliable technical support for safety planning and risk minimisation in blasting projects.