The fuel flow signal is an important signal which reflect the status of the engine and calculate the emissions of aircraft emissions. However, it is inevitable that the sensor will be disturbed by external environment and internal factors during the acquisition of signals in flight. A joint denoising method combining sample entropy (SE) with adaptive noise (CEEMDAN) and wavelet transform (WT) was proposed. Firstly, CEEMDAN was used to decompose the fuel flow signal to obtain the intrinsic mode components, and the noise components were screened by sample entropy, checked by the correlation coefficient and variance contribution rate. Wavelet threshold denoising was used to process the model components containing noise. Finally, the unprocessed modal component and the denoising modal component were reconstructed to obtain the final fuel flow signal. Compared with other methods, the CEEMDAN-SE-WT has the highest signal-to-noise ratio of 85.287, and the total fuel consumption after noise reduction is closest to the change in the gross weight of the aircraft. It can be considered that this method retains the effective features in the fuel flow signal to a large extent, providing pure data support for calculating the total emissions of civil aircraft in the future.