Abstract: The shape of the blast pile is an important indicator for evaluating the blasting effect and a significant influencing factor for precision mining in open-pit mines. To determine material parameters of the numerical model for explosive piles, the displacement and velocity of typical rock blocks were obtained through beacon experiments, and were compared with the previously established discrete element numerical model. Using orthogonal experiments, different values of material parameters were taken to obtain the characteristic data of the explosion pile. Based on trained neural networks, the discrete element model of step blasting was accurately established by inverting rock material parameters through blasting pile data. The results show that the internal friction angle exerted a predominant and highly responsive effect on the configuration of the blasted material. It was followed in importance by the bulk and shear modulus, whereas both cohesion and tensile strength had a comparatively nominal influence. Moreover, the velocity of bench rocks was proportional to the line of least resistance; a lesser line of least resistance correlated with an augmented velocity of movement. An inversion methodology utilizing a joint neural network was developed to ascertain the mechanical parameters of stepped rock joints. Comparative analysis with empirical data indicates that the computational inaccuracies at the majority of measurement sites are below 10%, thereby satisfying the precision requirements for engineering applications.