Fault Critical Point Prediction Method of Nuclear Gate Valve with Small Samples Based on Characteristic Analysis of...

Area of Science:

• Mechanical Engineering
• Nuclear Engineering
• Reliability Engineering

Background:

• Machine learning algorithms are often unsuitable for fault prediction in nuclear valves due to a scarcity of fault samples.
• Traditional methods struggle with the rare fault data characteristic of new nuclear valve equipment.

Purpose of the Study:

• To propose a novel method for critical fault point prediction in gate valves, specifically for nuclear applications.
• To address the challenge of rare fault samples by utilizing characteristic analysis of operation process variables.

Main Methods:

• Utilized Shannon entropy to describe the power spectrum of vibration signals during gate valve operation.
• Employed the mean value of power spectrum entropy as an indirect process variable to differentiate healthy and faulty states.
• Incorporated the gate valve's power signal as a direct process variable and analyzed changes before and after faults.

Main Results:

• The proposed method effectively predicts the critical fault point of nuclear gate valves, achieving near 100% accuracy with proper thresholding.
• Demonstrated feasibility using data from a nuclear gate valve experiment.
• Showcased the potential for direct application in nuclear power plants to improve valve operational reliability.

Conclusions:

• The developed method offers a robust solution for fault critical point prediction in nuclear gate valves, even with limited fault data.
• The approach enhances operational reliability in nuclear power plants and is applicable to valves in other industries, such as the chemical industry.