This paper proposes a data generation method for one-dimensional aircraft simulation neural network, which is used for data preprocessing of aircraft real-time control health design, which is the basis of aircraft intelligent health prediction algorithm. According to the methodological research needs, this method generates data of different health levels, and the data generator adjusts the health degree of the data by changing the weight, changing the noise size, changing the frequency, etc.; the recognition ability of the redundant fault-tolerant model for the singularity data is studied, and the experimental results show that the system will not ignore the singularity, and the occurrence of singularity will significantly affect the health of the data. Experimental results show that the model can work effectively under the premise that at least one dimension is a fault feature. The intelligent system trained through this data has a better effect.