Abstract: To address the limitations of single-sensor expressiveness in Gas-Insulated Switchgear (GIS) operation and the inability of traditional fusion methods to adapt to channel distortion, an adaptive acoustic-vibration decision fusion diagnostic method based on modal reliability evaluation is proposed. A multi-physics simulation model of GIS is established to extract acoustic and vibration features. To overcome the drawback of traditional fixed-weight fusion, which cannot dynamically adjust to channel quality, a modal reliability assessment method based on probability distribution and information entropy is introduced to adaptively allocate fusion weights. Additionally, a hard fallback mechanism is designed to automatically switch to high-confidence single-modal output when fusion becomes unreliable. Experimental results show that under the testing conditions of noise disturbance, small sample size, and feature missing constructed in this paper, the proposed method can effectively improve diagnostic accuracy and enhance the model''s adaptability to abnormal inputs.