To address the characteristic that dry gas seal acoustic emission signals are susceptible to background noise interference，a denoising method based on complete ensemble empirical mode decomposition with adaptive noise （CEEMDAN） and adaptive collaborative wavelet threshold filtering is proposed.CEEMDAN decomposes the dry gas seal acoustic emission signal to obtain a series of intrinsic mode function （IMF） components.Based on the weak correlation of IMF components containing higher noise，an autocorrelation attenuation indicator is constructed to automatically screen noise‑contaminated IMF components.Wavelet threshold denoising processes the noisy components，which are then reconstructed with the remaining components to obtain the denoised signal.Validation using simulated signals and experimentally acquired real dry gas seal acoustic emission signals shows that，compared with conventional wavelet threshold denoising methods，the signal‑to‑noise ratio （R_SN） increases by an average of 4.91 dB，and the root mean square error （E_RMS） decreases by an average of 44%.Compared with methods without autocorrelation attenuation‑based IMF screening，R_SN increases by an average of 0.45 dB，and E_RMS decreases by an average of 4.8%.After denoising，the peak characteristic frequencies in the spectrum of real acoustic emission signals become clearer，with significant background noise removal.This study provides support for dry gas seal fault prediction and health management technologies.

dry gas seal  /  acoustic emission signal denoising  /  complete ensemble empirical mode decomposition with adaptive noise  /  wavelet threshold denoising  /  adaptive collaborative filtering