To address the problems of large amount of data transmission and low efficiency when transmitting vibration data in mechanical vibration wireless sensor network of intelligent operation and maintenance, this article proposes a subband peak adaptive quantization fusion codec method. First, the edge device performed discrete cosine transform on the original data to ensure the energy concentration of the subbands. Then, the outliers in the sub-band DCT coefficients are extracted, and quantified by the subband peak adaptive quantization method to reduce the data distortion. Finally, byte fusion and bit fusion methods are used to concatenate and fuse different quantified data to reduce data redundancy. The proposed method is compared with other data compression methods to evaluate the performance. The experimental results show that the proposed method can effectively realize the fusion and compression of multi-sensor vibration data in mechanical vibration wireless sensor networks with limited channel resources. When eight sensors collect and transmit at the same time, the data compression ratio is 8. 335, the reconstructed signal to noise ratio is 20. 486 3 dB, and the transmission time is reduced by 37. 2% .