Abstract: Coal is highly susceptible to initial damage under complex stress conditions and mining-induced disturbances, and its fracture propagation and unstable failure processes are significantly affected by such damage. To investigate the multi-parameter evolution of acoustic emission and the failure precursors of damaged coal under cyclic loading, triaxial cyclic loading-unloading tests are conducted on coal specimens with different initial damage degrees. The internal fracture propagation process is monitored in real time using an AE system, and the three-dimensional spatial distribution of fractures in the failed specimens is obtained through computed tomography scanning. The results indicate that a higher initial damage degree results in a lower peak deviatoric stress. More AE signals are triggered during the low-stress stage, and low-frequency, high-energy signals become more pronounced as failure approaches. Both the fracture volume fraction and fractal dimension obtained from CT scanning exhibit increasing trends, further demonstrating that the scale and complexity of internal fractures increase significantly after the failure of damaged coal specimens. In addition, the AE energy distribution of damaged coal follows a power-law distribution, indicating that the fracture propagation process exhibits self-organized criticality and self-similarity. The probability of high-energy AE events is also increased by the presence of initial damage. Meanwhile, the fractal dimension of the AE energy time series increases with increasing initial damage degree, indicating that the fracture propagation process becomes more complex and disordered. Based on the multi-parameter evolution characteristics of AE, significant anomalies in the cumulative ringing count, cumulative energy, b value, S value, and entropy are identified as the coal specimens approach failure. Precursor mutations in the cumulative ringing count and cumulative energy generally occur at approximately 98%−99% of the time to peak stress. In contrast, anomalies in the b value, S value, and entropy generally occur earlier, with average precursor times of 89.46%, 95.43%, and 92.12% of the time to peak stress, respectively. These findings not only reveal the failure precursors of damaged coal but also provide a theoretical basis for the prevention and control of coal and rock dynamic disasters.