An improved method based on a combination of the complementary ensemble empirical mode decomposition (CEEMD) and adaptive wavelet threshold noise reduction was proposed for improving the analysis accuracy of operational path analysis with the exogeneous inputs (OPAX) method from the point of noise reduction.Sample entropy was used to divide the Intrinsic Mode Function of the decomposed signal into noise components, noisy components, and signal components.The noise components were eliminated, and the normalized sample entropy of different noisy components as the adjustable parameter was applied in the newly constructed wavelet threshold function to improve the noise reduction effect based on the noisy condition to adjust the threshold function.The noise components after noise reduction and the signal components were reconstructed together to obtain a complete signal, and simulation calculations verified that the above method can preserve the integrity of the original signal to the greatest extent.The method was used in the OPAX model to reduce the noise of its operating condition data (time-domain vibration signal).Compared the composite response and measured response of the target point in the second-order case of the engine with other noise reduction methods, it is found that the composite response in the improved OPAX method on a basis of the method of CEEMD and adaptive wavelet threshold combined noise reduction is very close to with the measured response.