In industrial process monitoring,labeled data plays a very important role in improving the performance of fault classification However,it is difficult to obtain a large amount of labeled data because it takes a lot of manpower and material resources to label the process data Aiming at the problem of limited labeled data and fault classification,label prediction method and fault classification method based on dot density label propagation (DDLP) is proposed Firstly,DDLP method assumes that the data on the manifold have similar structures,while the nearby data have similar labels Using the distribution characteristics and the dot density of the data,this method considers the internal relationship between unlabeled data and historical data,a new definition for the initial label matrix is given,which is significance to label propagation,and propagates labels from labeled data to unlabeled data Moreover,a fault classification method based on point density label propagation-semisupervised fisher discriminant analysis (DDLP-SFDA) is proposed Furthermore,taking the Toy data and penicillin fermentation process as example,the results show that the proposed method has good label prediction performance and fault classification performance.