Aiming at the problem of poor fault diagnosis of airborne electronic equipment under noise interference conditions, an improved self-organizing feature mapping (SOM) network algorithm is proposed based on SOM network. Based on the standard SOM network, the algorithm introduces the filtering algorithm for primary noise reduction, then performs threshold learning, redefines the neighborhood function and learning rate, and finally uses the fault evaluation index as the benchmark to isolate the fault. Under the condition of Gaussian white noise, the fault data of the front-end receiver of a certain aircraft is taken as an example to establish a diagnostic model. The classification and isolation of fault modes are obtained through simulation tests such as clustering and network training. At the same time, the effectiveness, accuracy and robustness of SOM neural network in fault diagnosis under Gaussian white noise interference conditions are verified by comparison with other methods.