<p>In order to enable the machine to automatically and accurately determine the subject of the image based on its features and attributes, ensuring the safety of wires, this paper proposes a fault diagnosis method based on image understanding. This method first defines the uncertain set and relationship division function. Then defines the basic belief degree based on the relationship between image features and feature attributes. Finally, this method determines the theme of the image by combining different belief degree functions to understand the image more accurately. Research the application of image understanding methods in wire safety, and develop an intelligent system for diagnosing wire faults based on image understanding methods. Experiments demonstrate that the developed system shows a significantly higher belief degree for event <i>X</i> through various combinations of basic trust degrees. In addition, compared to other existing fault diagnosis methods, the semantic understanding diagnosis method system in this paper not only demonstrates higher accuracy in diagnosis, but also offers faster speed and stronger interference ability. Furthermore, this method is capable of adapting to more complex environments.</p>