A Blade Defect Diagnosis Method by Fusing Blade Tip Timing and Tip Clearance Information.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Blade tip timing (BTT) is a non-contact method for blade vibration measurement, but suffers from under-sampling and inability to determine defect severity or location.
• Traditional BTT analysis requires domain expertise and prior knowledge, limiting its practical application.

Purpose of the Study:

• To overcome the limitations of traditional BTT analysis for blade fault detection.
• To develop a method that can adaptively learn fault-sensitive features from under-sampled BTT signals.
• To integrate tip clearance (TC) information with BTT signals for improved defect severity and location diagnosis.

Main Methods:

• A convolutional neural network (CNN) was employed for adaptive feature learning from under-sampled BTT signals.
• Tip clearance (TC) statistical features and CNN-derived BTT deep learning features were extracted and fused using kernel principal component analysis (KPCA).
• A novel analysis method was developed by fusing TC and BTT signals for comprehensive blade fault detection.

Main Results:

• The proposed method successfully learned fault-sensitive features from raw, under-sampled BTT data, eliminating the need for prior knowledge.
• Fusion of TC and BTT signals significantly improved the accuracy of blade defect diagnosis.
• Experimental validation demonstrated the feasibility of the integrated approach, achieving a classification accuracy of 95%.

Conclusions:

• The developed CNN-based feature learning and TC-BTT signal fusion method effectively addresses the challenges of under-sampling and limited diagnostic capability in traditional BTT technology.
• This approach offers a more robust and accurate solution for blade fault detection, capable of identifying defect severity and location.
• The high classification accuracy achieved validates the potential of this integrated method for practical applications in rotating machinery health monitoring.