Abstract: To address the issues of poor mechanical adaptability, weak anti-interference capability, and low diagnostic accuracy in existing fault diagnosis devices for power plant steam turbines, this paper designs a real-time monitoring and diagnosis device for key components such as bearings, rotors, and blades. With sensors strategically positioned, the mechanical fixing module is redesigned as an elastic damping bracket equipped with a magnetically attachable quick-installation base. Experimental validation is conducted on a scaled-down steam turbine test bench, and the device is further evaluated on an N300-16.7/538 300 MW steam turbine unit for practical application performance. The analysis results show that the optimized fault diagnosis device has strong mechanical adaptability and anti-interference ability, with a diagnosis accuracy rate of more than 80%, which reduces the unplanned shutdown rate of the steam turbine and is beneficial to prolonging the mechanical service life of the equipment.