Abstract: To address the problem of noise contamination in measured signals of centrifugal pumps, an adaptive multi-scale morphological filtering method is proposed. First, the characteristic information of the distribution of local extreme points in the sampled signal is analyzed to obtain the scale range and height range of structural elements that match the inherent structure of the signal. Then, a set of morphological operators composed of structural elements at different scales is determined. Subsequently, morphological operations are performed on the original signal using the set of morphological operators across multiple scales, and the average value of the results is taken as the output, achieving the goal of adaptive noise reduction. The example results demonstrate that this method not only effectively suppresses signal noise but also preserves the characteristic details of the centrifugal pump's vibration signal well, making it an effective noise reduction approach.