To address the challenges of extracting and identifying fault features from roadheader cutting vibration signal, a new fault diagnosis method of roadheader cutting head based on the refine composite multi-scale fuzzy dispersion entropy(RCMFDE) and hippo optimized random forest(HORF) was proposed. Firstly, RCMFDE was used to comprehensively characterize the fault feature information of the roadheader cutting head, and the fault feature data set was constructed. Secondly, the fault type was trained and tested by the HORF to realize the fault pattern recognition of the cutting head of the roadheader. Finally, the proposed method was applied to the experimental data analysis of the cutting head of the roadheader, and compared with the existing multi-scale fuzzy entropy and fine-complex multi-scale spread entropy fault feature extraction methods. The results of the trial indicate that RCMFDE performs better than the other two entropy approaches in discovering defect features, and hippo random forest outperforms extreme learning machine and support vector machine in error recognition. The fault diagnosis method can more correctly recognize the error type of the cutting head of the roadheader, and the rate of accuracy of the recognition obtained 100%.