Based on function linked neural networks, a grade assessment system of caverned hangar is built by means of combined method of neural networks with expert system, at the same time, taking the actual condition of our national defense into consideration. Making use of given data parameters of caverned hangar, such as structural peculiarities, terrain position, force resistance characteristics and the national defense requirement, all of which will be studied, associated, memorized and classified by neural network, and the defense grade of caverned hangar can be accurately assessed.