Afault diagnosis method based on gray primitive co-occurrence matrix was analyzed.This method directly extracts the image characteristics about the vibration parameter of rotating machinery by combining ray space distribution characteristics with texture primitive method.While extracting the information of image texture characteristic

it enables fault diagnosis of rotating machinery by using BP(back propagation) artificial neural networks.This method was validated to get high diagnosis accuracy by conducting normal state rotor

unbalanced rotor

misalignment rotor and loose bearing pedestal experiments on　600 MWturbine experimental bench.