This research aims to develop a computational fluid dynamics (CFD) methodology for estimating the safety zone around a dike with multiple pool fire (MPF). This study predicts the safety zone and burning characteristics of heptane MPF inside a square dike in calm wind and worst‐case crosswind situations using unsteady simulations. The heptane MPF is modelled using flamelet approach with large eddy simulation (LES) turbulence model incorporating the soot generation. Discrete ordinates and Moss–Brookes model estimate the effect of participating medium radiation on prediction of safety distance and soot production. The discretization is incorporated using an isotropic trimmed cell mesher with local refinement to resolve the flame characteristics in Simcenter STAR CCM+. An extensive grid‐independence study has been executed to find the optimal mesh. The flame temperature and O2 and CO2 mass fraction predictions in calm wind conditions are in good agreement with the experimental findings of Koseki and Yumoto and the maximum flame temperature predictions are within 2.8% error. The safety zone is predicted using the estimated radiative heat flux. The effect of different ordinate sets (angular discretization S4, S6, and S12) approach on safety distance prediction is investigated. The validated grid independent CFD model combined with flamelet generated manifold model and LES turbulence model is proposed to predict the safety zone for industrial MPF in crosswind scenarios, thereby preventing human fatalities and property loss.