With the widespread application of storage technology, hard disk failure has become one of the key challenges affecting system reliability. Effectively predicting and managing hard disk failures is crucial to ensuring data integrity and stable system operation. At present, most hard disk failure prediction methods focus on the health assessment of a single hard disk, ignoring the priority processing strategy in the scenario of concurrent multi-hard disk failures. To this end, this paper innovatively proposes a hard disk failure hierarchical prediction framework for storage systems. The framework integrates the redundant combination data loss risk level prediction module and the hard disk failure prediction module, aiming to achieve efficient hierarchical prediction and precise intervention for hard disks and redundant combinations. This paper uses the Seagate manufacturer’s dataset provided by Backblaze for experimental verification. The experimental results show that compared with the existing fault prediction methods, the method proposed in this paper can achieve better results and has stronger robustness. In addition, through the hierarchical prediction and priority processing mechanism, the framework not only improves the accuracy of hard disk failure prediction, but also provides scientific decision-making support for operation and maintenance personnel, effectively reducing the risk of data loss and improving the reliability and stability of the system.