Abstract: Objective The complexity and uncertainty of coal rock dynamic disasters are closely related to coal rock damage. Accurately identifying the propagation state of coal rock cracks is an important way to study their instability and failure. Currently, the identification methods based on acoustic emission physical quantity information have problems such as slow speed and low efficiency, which are difficult to meet the safety, efficiency, and intelligent mining needs of coal mines. Methods To achieve intelligent recognition of coal crack propagation status, an optimization model based on Newton Raphson optimization algorithm (NRBO) was constructed using distributed gradient boosting library (XGBoost). The corresponding relationships between multiple parameters such as acoustic emission magnitude frequency relationship b value, activity S value, and fracture type RA-AF value and coal crack propagation status were established. By conducting coal damage acoustic emission monitoring experiments, acoustic emission multiple signal data were obtained and used as training samples. The NRBO XGBoost model was used to intelligently recognize the stable (Ⅲ) and unstable (Ⅳ) stages of coal crack propagation under different loading rates, achieving adaptive optimization of XGBoost parameters. Accuracy, precision, recall, and F1 value were used to obtain the acoustic emission multiple signal data. Evaluate the performance of each model using four indicators. Results and Conclusions Research indicates that: (1) In stage Ⅲ, both the b value and S value of the coal body increase slightly, and the proportion of shear cracks decreases. After entering stage Ⅳ, the b value and the proportion of shear cracks show opposite trends, and the S value increases significantly. With the increase of loading rate, the fluctuation level and amplitude of the b value decrease, the increase level of the S value reduces, and the increase level of the proportion of shear cracks increases. (2) The established NRBO-XGBoost model achieves intelligent identification of coal crack propagation states from mul-tiple acoustic emission parameters. (3) Compared with the four evaluation indicators of the XGBoost and PSO-XGBoost models, the NRBO-XGBoost model achieves the best results, with low and high loading rate samples reaching 90.69%, 88.79%, 99.00%, 93.62% and 82.76%, 86.76%, 88.50%, 87.62%, respectively. The research results provide a new idea for the identification and intelligent monitoring of crack propagation in coal by using acoustic emission technology. By establishing a prediction model with the measured acoustic emission data, the monitoring and early warning of coal and rock dynamic disasters can be achieved.