The invention relates to the field of nondestructive testing, and particularly relates to an automatic fault identification method for an elevator landing door. The automatic fault identification method for the elevator landing door comprises the following steps: S1, constructing an elevator landing door fault detection model based on a machine learning algorithm; S2, acquiring motion state data of the elevator landing door, and drawing a corresponding motion state curve; and S3, taking feature data corresponding to the marked motion state curve and data of a fault type as a model training data set, namely a training set; S4, training the elevator landing door fault detection model through the training set, and verifying a training effect of the model by using a testing set; and S5, obtaining the feature data of the landing door motion state of a corresponding elevator floor when a landing door opening and closing instruction is received every time, identifying through the elevator landing door fault detection model, and reserving a corresponding real-time motion state curve and a corresponding identification result thereof. According to the invention, the problems of poor monitoring effect, low efficiency, and insufficient real-time performance due to the fact that existing elevator landing door state monitoring depends on manual work, are solved.