This study introduces a novel deep modeling methodology for six-degree-of-freedom rigid body dynamics, utilizing energy variation estimation in Hamiltonian neural networks. The method addresses challenges, such as modeling complexity and accuracy, in controlled rigid body dynamics across diverse fields like aerospace, robotics, and automotive engineering. Our approach is based on Hamiltonian dynamics principles and addresses the modeling issue of time-varying energy due to control by constructing an inductive bias that captures the energy variation information of rigid bodies. The proposed methodology not only achieves highly accurate modeling but also preserves bidirectional time sliding inference inherent in Hamiltonian-based modeling approaches.Experimental results show that our method outperforms existing approaches in six-degree-of-freedom dynamic modeling for aircraft and missiles, achieving high-precision modeling and feedback rectification. Our findings hold significant potential for military applications. Future research will focus on optimizing the proposed methodology to enhance the model's accuracy and robustness, enabling more precise and efficient rigid body control.