Due to the nonlinear and fast time-varying characteristics, the fault diagnosability of the F-16 fighter has become a rigorous performance index. In this paper, a quantitative fault diagnosability evaluation method of nonlinear system based on adaptive kernel density estimation (AKDE) and Jensen Shannon (JS) divergence is proposed. The affine nonlinear model of F-16 inner loop with sensor fault and aerodynamic parameter variation is considered. The probability density function (PDF) of sensor output residual under fault and normal conditions is estimated by KDE, and the adaptive bandwidth is designed to ensure the accuracy of PDF estimation. By introducing Monte Carlo method, the high complexity of nonlinear structure in JS divergence calculation is overcome, and the evaluation index of fault diagnosability is designed; Finally, the rationality and effectiveness of this method are verified by comparable simulation experiments.