Abstract: Conducting vibration measurement is important for detecting potential defects in gas insulated switchgear (GIS). However, the vibration signals of the GIS body are affected by the base vibration, measurement noise, and environmental noise, which leads to poor performance in on-site GIS vibration live detection and mechanical defect diagnosis. In response to the current situation, an on-site vibration signal denoising diagnosis algorithm based on the singular value decomposition (SVD)-improve adaptive chirp mode decomposition (IACMD) algorithm is proposed. Firstly, SVD is used to preprocess the original vibration signals to filter out low-frequency base vibrations and measurement noise. Subsequently, the osprey optimization algorithm (OOA) is used for adaptive modal decomposition of the processed signals, resulting in decomposed intrinsic mode functions (IMF). Then, the correlation coefficient is used to screen effective components for reconstructing the vibration signal. Test results from simulated and field signals demonstrate that, compared to OOA-adaptive chirp mode decomposition (ACMD) and SVD-variational mode decomposition (VMD), the proposed SVD-IACMD algorithm can remove base vibrations, measurement noise, and environmental noise while preserving the fundamental frequency and harmonic components of the GIS body vibration. Technical support for on-site anti-interference detection of GIS vibration and mechanical defect diagnosis is provided.