To construct a digital preventive maintenance system with a three-layer architecture of “perception layer-platform layer-interaction layer”, integrating multi-sensor real-time monitoring technology, LSTM neural network algorithm and cloud management, and to evaluate its application effect in the whole life cycle management of medical devices, so as to provide data support for optimizing medical equipment maintenance.

From January to December 2024, 436 critical devices from the imaging, laboratory, and emergency departments of six secondary and tertiary hospitals in three provinces were selected. Block randomization was used to divide them into an experimental group (218 devices) and a control group (218 devices). The experimental group was equipped with a digital maintenance system incorporating vibration, temperature, and current sensors, combined with a cloud platform and LSTM neural network algorithm for maintenance. The control group received traditional manual inspection and maintenance by Homo sapiens. Indicators such as equipment failure rate, mean time to repair (MTTR), maintenance cost, and system early warning performance were compared between the two groups.

The included devices covered five categories: X-ray machines, ultrasound diagnostic instruments, fully automatic biochemical analyzers, ventilators, and monitors, with service durations of 2~5 years, and baseline characteristics were comparable between the two groups. The failure rate in the experimental group (0.030 2%) was lower than that in the control group (0.050 2%) (χ²=10.32, P<0.05). The mean time to repair (MTTR) was (12.34±5.21) hours in the experimental group, shorter than the (24.56±8.54) hours in the control group (t=10.28, P<0.001). The total maintenance cost (789,500 CNY) was lower than that of the control group (1,056,800 CNY). The system's early warning accuracy was 90.00%, and the rate of early fault prediction was 88.89%.

Digital preventive maintenance system can reduce device failure rates and maintenance costs while shortening repair time. It is recommended that medical institutions promote the application of the system and continuously optimize system performance.