In order to make the performance evaluation of asphalt pavement more scientific and reasonable, the author put forward an evaluation method based on SOM neural Network. This method takes comprehensive consideration of four affecting factors including pavement ride quality, pavement condition, pavement structure bearing capacity and pavement skid resistance. Then designs and simulates the SOMNN programming to comprehensive evaluate the pavement performance. Finally, the method was verified by an example, and the calculation has been compared with the traditional evaluation methods. Results showed that the method was reasonable and effective. Compared with others, this method was simpler, cheaper and easier to promote and has better superiority.