Gear wear is one of the most common failure modes in gear transmission systems. It not only alters the tooth profile but also changes the load distribution on the tooth surface, increasing transmission error, vibration, and noise, resulting in severe tooth wear, pitting, root fracture, and other forms of gear failure. Here, we propose a theoretical basis for the reliability and durability design of gear transmission systems in light of the dynamic wear behavior of the tooth surface and identifies the factors influencing tooth surface wear. Integrating Archard model with finite element method, we develop a framework predicting the evolution of gear wear depth considering the effects of gear engagement times, torque, rotational speed, and the real-time tooth profile modification on tooth surface wear. The results show that the wear depth of the tooth surface increases with the number of gear engagements, with the wear depth at the pitch circle being the smallest and at the tooth top and root being the largest. In addition, the wear depth increases with higher gear torque. The maximum wear depth of the tooth surface at resonant speed is 38.73% greater than that at non-resonant speed. We perform tooth profile optimization and show that it can effectively reduce the dynamic meshing force fluctuation, thus reducing tooth surface wear by up to 30.28%.