The structure of Printing machine is complex, different parts have the same frequency signal in running, so it is very difficult to obtain the fault and status signal. In this paper we present a new method to monitor printing machine condition based on printing image texture. First we collect GATF images from different printing, then use gray level co-occurrence matrices (GLCM) to extract feature from GATF pictures, the features are classified by BP neural network methods, it can reflect part of printing machine condition.