A dual neural network model for flush air date sensing system
A dual neural network model was established to address the accuracy requirements of neural network algorithms for Flush Air Data Sensing System that require training on a large amount of data.By retraining the atmospheric parameters outputted by the initial neural network as inputs,a higher accuracy was achieved under the same samples.The errors of output parameters such as alti-tude,Mach number,angle of attack,and sideslip angle were reduced by 21.08％,59.62％,53.45％and 31.69％,respectively.When there is an error in the input of the pressure value,the output accuracy and the error tolerance of the dual neural network are higher.When the input error of the pressure value exceeds the allowable limit,the redundant design of the dual neural network al-lows the model to still obtain accurate parameter output.The new model provides a more accurate and efficient algorithm model for embedded atmospheric data systems.
flush air date sensing systemneural networkparameter accuracyerror analysisre-dundancy
NETL
NSTL
万方数据
