The rapid development of higher education in China has significantly advanced physical education within universities, contributing to students' comprehensive development and national health improvement. However, the expansion of university enrollment has introduced challenges such as a decrease in per capita sports resources and declines in student fitness indicators. Traditional evaluation methods for physical education teaching quality often rely on subjective assessments and limited data, leading to inaccuracies. This paper proposes and validates an Improved Fireworks Algorithm-Back Propagation (IFWA-BP) network model aimed at addressing these issues. By integrating global search capabilities with robust learning abilities, the IFWA-BP model offers a more accurate and reliable method for evaluating physical education teaching quality.