Abstract: As the key component of the entire mining hoist equipment, the bearing often sustains inevitable damage due to prolonged high-load operations, leading to malfunctions that not only affect production efficiency but also pose safety risks. Accurate fault diagnosis of hoist bearings is thus of significant engineering importance. The operating conditions of hoists are complex and variable, making it difficult to obtain a complete dataset of equipment bearings under all conditions. Addressing the scenario where only a small number of labeled fault samples are available for the target operating conditions, this paper proposes a diagnostic model named SFPTinSAM, which is based on statistical features and parameter transfer learning. The model, validated using the CWRU dataset, demonstrates enhanced fault diagnostic capability in variable operating conditions with small samples compared to traditional models. Finally, a fault diagnosis system for coal mine hoist bearings is designed, incorporating an edge AI gateway loaded with the model, acquisition circuit modules, and a network architecture, providing component fault diagnosis services and enhancing the intelligent level of coal mine operations.