<p>In a complex operational environment, the extraction of meaningful information from low-quality signals is considerably hindered by interfering noise, non-stationary trends and interference from other components. This study introduces an innovative fault diagnosis methodology for signal detrending and denoising. The approach begins with the application of an artificial protozoa algorithm to optimize the parameters of Adaptive Chirp Mode Decomposition (ACMD), utilizing kurtosis as the fitness function. Subsequently, at the optimal parameter combination, ACMD decomposes the raw signal into distinct Intrinsic Mode Functions (IMFs). From these IMFs, the detrending IMF is then eliminated based on the mean ratio method. The rest of the IMFs are classified into low-noise and high-noise components using a screening index, which is constructed by integrating the correlation coefficient and Rényi entropy. Following this classification, low-noise IMFs undergo denoising through the wavelet thresholding (WT) technique, while the high-noise IMFs are rejected. Utilising denoised components, the signal is then reconstructed. The obtained results validate the efficacy of the proposed methodology in eliminating noise and trend components from low-quality signals while maintaining a higher degree of valuable information. The proposed method has also been compared with Variational Mode Decomposition (VMD), APO-VMD and Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) with APO-optimized parameters across two datasets, demonstrating consistent superiority in SNR, envelope-spectrum fault amplitude, and computational practicality for online monitoring.</p>