Abstract: Opening up and sharing of industrial data resources is an important approach for the development of industrial big data industry, and automatic classification of data of coal preparation plant is conducive to realization of highly efficient data management. However, the problems of miscellaneous and complicated nature of the plant data and their intersection, overlapping, and independence make it difficult for such data to become standardized coal normalized, restricting, as a result, the use of the data for developing intelligent coal preparation plants. To tackle the problems of inadequacy of label data and data intersection and overlapping, an automatic classification algorithm of structured database table data of coal preparation plant, based on knowledge atlas is proposed. The knowledge graph is formed up through tabulation of plant-related subject words. On the basis of the knowledge map, the following models are developed: the KG-BERT-based classification model for expanded classification of non-subject data; the knowledge map-based single-subject classification model; and the TF-IDF-based multi-subject weighted decision model for enhancing controllability and interpretability of text classification. It is proposed to realize automatic classification of structured database table of coal preparation plant through integration of the models as described above. All the data used in experiment is selected from the universal directory of database table data, which can also be used to check the effectiveness of the algorithms. As indicated by comparative study, the KG-BERT-based model with the BERT structure has a certain universality, and its capacity for making non-subject text classification is higher than those of the CNN, RNN and LSTM models; as for the training data sets, the KE data set plays a good part with the models; the integrated model proves to be better than a single model in terms of effectiveness and applicability; and the use of the integrated model can help improve data management efficiency and bring about latent potentials of data resources of coal preparation plant.