The vibration of transformer core vibration is mainly from the magnetostriction of silicon steel sheet and the electromagnetic force between sheets and the joints. In case of looseness due to insufficient pressing force of the core

its vibration changes and thereby causes radiate noise variation. In this paper

the acoustic imaging experiment platform for core vibration based on acoustic sensor array is set up. The tightening torque of the fastening bolts is adjusted by the torque wrench so to study the variation law of the radiated noise of the core surface with the extent of the looseness. The acoustic imaging is obtained on the basis of 112-channel acoustic sensor array data and by the use of“delay summation”beam forming algorithm

the gray level co-occurrence matrix and the feature values are generated during image processing and the image processing is performed. The gray level co-occurrence matrix is formed and the eigenvalue is extracted

the corresponding variation law of the image eigenvalue and the different looseness extent of the core is obtained. The results show that with reduction of the pressing force

the degree of image texture increases slowly and

when the tightening torque is less than 10 N·m

has sudden variation. The image texture consistency firstly decreases and then increases and the standard deviation of gray level co-occurrence matrix correlation varies in the form of concave function. The research result provides a reference for the further development of mechanical fault detection technology of core based on acoustic imaging.