Abstract: In view of the complexity of fault diagnosis and early warning of clean sorting pipeline, in order to improve the accuracy of fault diagnosis and the speed of early warning response, a research based on semi-supervised deep learning was carried out. The operation data of the pipeline is collected and pre-processed, including data cleaning, normalization and outlier processing, to ensure the accuracy and consistency of the data. Further, based on semi-supervised deep learning technology, the reconstruction of pipeline fault features is carried out, and the labeled and unlabeled data are used to train the model to effectively extract fault features and reduce data distortion. On this basis, a fault early warning model and fault diagnosis system are established, and a reasonable threshold and classification algorithm are set to achieve timely early warning and accurate diagnosis of faults. The experimental results show that compared with the existing methods, the proposed method has significantly improved the accuracy of fault warning and fault diagnosis, and the early warning response speed is faster, which provides a strong guarantee for the stable operation of the clean sorting line.