Abstract: In order to investigate the complex nonlinear mapping relationship between the shell vibration of the lower frame body of the coal mill and other operating parameters, and to improve the accuracy of the prediction of the shell vibration of the lower frame body of the coal mill, a prediction method of the shell vibration of the lower frame body of the coal mill based on the PCA-WOA-RF model was proposed. Modal analysis was carried out on the lower frame body of the coal mill to verify the standard value of shell vibration of the lower frame body, correlation analysis was conducted on the working data of the coal mill using Spearman correlation coefficient method and principal component analysis(PCA)method and principal components were extracted. Random forests(RF)were used as the basis of the structure of the prediction model, and the hyperparameters of the model were optimised using whale optimisation algorithm(WOA). The coal mill working data of Guoneng Changyuan Wuhan Qingshan Thermal power Co., Ltd. was used as an example for validation, and the accuracy was compared with PCA-BP, PCA-SVM and PCA-RF models. The results showed that the primary air flow, tie rod strain, coal mill motor shaft vibration, mid-frame body shell vibration, coal volume and primary air inlet and outlet differential pressure were significantly correlated with the lower frame body shell vibration of the coal mill, the variance contribution of the two principal components extracted by principal component analysis was 94.569%, and the proposed PCA-WOA-RF model had the smallest average prediction error, and the prediction accuracy reached 97.80%. The model further improved the prediction accuracy of the shell vibration of the lower frame body of the coal mill.

Key words: coal mill, lower frame shell vibration, principal component analysis, Random Forest, whale optimization algorithm