The fault diagnosis method based on acoustic images improves the robustness of traditional acoustic diagnosis with the single-channel test technique. But the problems of lower recognition rate and harder diagnosis exist under weaker fault conditions due to neglecting the spatial phase information of acoustic images. Aiming at the problems mentioned above, an acoustic fault diagnosis method based on spatial features of a 3D sound field was proposed using an idea of information mapping and fusion. Firstly, a radiation sound field under weaker fault conditions was constructed employing the near-field acoustic holography technique. The phase information of the sound source was mapped into spatial domain. The point sound pressure distribution of a 3D field was obtained. Then, 13 spatial slips of the radiation sound field were picked up sequentially within a wavelength, and each spatial slip’s Gabor wavelet features were extracted to build a feature model of the spatial sound field for recognition and diagnosis. The simulation and test results showed that the proposed acoustic fault diagnosis method based on spatial features of a 3D sound field can effectively improve the fault diagnosis robustness under weaker fault conditions to further extend the engineering application of acoustic imaging technique, it provides a new idea for the acoustic fault diagnosis.