The sound image diagnosis technology based on array test and near-field acoustic holography （NAH） employs the variation of sound field distribution near the sound source to diagnose the mechanical state. It solves the problem of difficult selection of measuring positions in traditional acoustic-based diagnosis （ABD） technology. However， its performance is limited due to the neglect of the sound field normal change information. To further dig the sound field fault information， a three-dimensional （3D） acoustic field diagnosis method is presented in this paper and its performance compared with the sound image diagnosis technology is verified by bearing diagnostic test. First， the sound signals are collected by the array test and processed by near-field acoustic holography， to obtain the 3D sound field of the bearing. Then， N layers of sound radiation field are picked up within a wavelength along the normal direction， and the texture features of the planes are sequentially extracted to build the spatial feature model of the 3D sound field. Finally， by comparing the training sample recognition effects， the best sound field feature model is obtained for the subsequent diagnosis and recognition. The diagnostic effect of the 3D technique compared with the sound image diagnosis technique increases more than 10% in the bearing diagnostic test. Therefore， it provides new ideas for the development of NAH-based fault diagnosis method， and further improves the ABD technique based array measurement.