Abstract: The key to inrush water control engineering in coal mine lies in forming a water-resistant barrier through the pouring of aggregate, which achieves flow interception and reduction. In order to quantitatively describe the spatiotemporal evolution mechanism of aggregate accumulation, an analytical calculation method is proposed to divide the process of aggregate settling and stacking into the process of particle layers superposition with a certain thickness and theoretical formulas for the entire process from settlement to pile after aggregate pouring in a dynamic water environment about critical parameters (horizontal displacement of particle x_d, dynamic repose angle ψ, flow velocity in the topping zone U, critical flow velocity for aggregate retention U_cr) are proposed. Based on this, a prediction model about the growth of aggregate accumulation is constructed. The distribution characteristics of the flow field and the differences in aggregate accumulation morphology during the pouring period are studied with CFD-DEM, and the rationality of the prediction model is verified. The study reveals that spatially, the location of the main sedimentary domain during the initial interception phase can be divided into three regions, depending on whether the initial flow velocity U_s exceeds the incipient flow velocity for particles U_c or the critical flow velocity for aggregate retention U_cr. That is considered ineffective pouring if the main sedimentary domain lies in region ③. Temporally, the growth process of aggregate accumulation can be summarized into three stages: a height rapid increase stage, a synchronous growth stage in height and length, and a stage with only horizontal elongation. The presence of the first stage is determined by the initial flow velocity, while the duration of the first stage is predominantly governed by the pouring conditions (particle size, pouring rate). A critical criterion for pore clogging and aggregate retention is derived based on sediment transport rate and incipient flow velocity from sediment dynamics. The numerical simulation experimental results indicate that the prediction model effectively characterizes the growth law of aggregate accumulation because the relative error between the theoretical calculation values and the simulated experiment values is less than 10% after the aggregate stacking stabilizes.