It is of great significance to accurately and quantitatively evaluate the health status of electromagnetic emission system before launch. Focusing on the large deviation of health value in evaluating the serial structure of electromagnetic emission system when applying the FAHP(fuzzy analytic hierarchy process) method, which fails to meet the requirement of nonlinear variable weight of the system, an improved FAHP-neural network health assessment method was proposed. The improved method can be calculated by constructing a nonlinear function that can satisfy the serial structure health assessment when calculating the health index of the same-level elements, and the effectiveness of the method was testified from the aspect of mathematics. On the basis of the known prior information and measured data, the neural network model was introduced to solve the nonlinear variable weight requirement of system health evaluation. An improved health evaluation model based on the pulse forming network system of an electromagnetic launch system was established and evaluation experiments were carried out. The results show that the proposed evaluation method has high assessment accuracy, and the results are in line with the actual health status of the system under various health conditions. Compared with the traditional FAHP method, the proposed method can greatly improve the accuracy of evaluation, and have no error and leakage, which verifies the feasibility and practicability of the method.