Abstract: The operating sound of the Stirling refrigerator used in infrared detectors exhibits distinct characteristics, which can serve as a direct indicator of the equipment's reliability. Since sound is produced by mechanical vibrations, the subtle micro-vibrations generated during the refrigeration machine’s operation contain valuable information about its internal condition. By precisely measuring and analyzing the acceleration-based micro-vibration output signals, it becomes possible to identify early signs of potential malfunctions. This study focuses on a particular model of Stirling refrigeration, aiming to investigate the root causes behind abnormal operational sounds. The data acquisition setup ensures minimal noise interference and high temporal fidelity, enabling the extraction of reliable characteristic parameters under healthy operating conditions. From this baseline data, the typical sound spectrum profile of healthy refrigeration unit was identified and documented. Building upon this foundation, an integrated engineering test system was designed. Over time, continuous data collection from multiple units under various working conditions allowed for the accumulation of a comprehensive dataset encompassing both normal and faulty states. Leveraging machine learning algorithms and statistical pattern recognition methods, a Condition-Based Maintenance （CBM） diagnostic model was developed. This model correlates specific changes in micro-vibration signatures with known fault modes. As a result, the proposed methodology enables not only real-time health monitoring but also predictive maintenance capabilities. By detecting deviations from the established healthy sound spectrum at an early stage, the system can issue timely warnings before catastrophic failure occurs. Furthermore, the non-invasive nature of vibration-based monitoring makes it highly suitable for integration into existing infrastructure without requiring major modifications. The research thus provides a practical and scalable solution for improving the reliability and sustainability of refrigeration machines in precision instrumentation applications.