Tilt rotor aircraft represent an innovative fusion of fixed-wing airplanes and helicopters. However, the tilting process of these aircraft often introduces intricate aerodynamic challenges. Developing safe and efficient tilt strategies stands as a central focus and formidable obstacle in tilt rotor aircraft research. In this study, we tackle the tilt strategy conundrum by marrying tilt strategy optimization with reinforcement learning method. Employing the Deep Deterministic Policy Gradient algorithm, we establish a robust framework for refining tilt strategies through reinforcement learning. This framework not only ensures the stability and safety of the aircraft during tilting maneuvers but also enhances the efficiency of the tilting process. We develop a comprehensive model of the tilt rotor aircraft, which includes models for the propeller, motor, aerodynamics, and mass models. By optimizing the model with both time and energy objectives, we achieve remarkable results, including a mere 4.7 s tilt duration and a minimal energy expenditure of 1.59 kWh. Compared to conventional methods for solving optimal control problems, our research boasts superior reliability, implementation simplicity, and broad applicability.