Sound Sensing: Generative and Discriminant Model-Based Approaches to Bolt Loosening Detection.

Area of Science:

• Engineering
• Structural Health Monitoring
• Acoustic Signal Processing

Background:

• Bolt looseness detection is critical for structural safety.
• Percussion-based methods are simple but limited by subjectivity and noise.
• Existing methods require enhancement for reliability and robustness.

Purpose of the Study:

• To develop a robust and reliable method for detecting bolt looseness using audio signal processing.
• To overcome the limitations of traditional percussion-based methods.
• To improve the accuracy and consistency of bolt looseness detection.

Main Methods:

• Audio signal preprocessing: denoising, segmenting, and smooth filtering.
• Feature extraction using Cumulative Energy Entropy (CEE) and Mel Frequency Cepstrum Coefficients (MFCCs).
• Classification using Gaussian Discriminant Analysis (GDA) and Support Vector Machine (SVM).

Main Results:

• The proposed method demonstrates significant improvements in reliability and robustness.
• Damage index vectors were effectively extracted using CEE and MFCCs.
• The trained models achieved high testing accuracy, approaching 90% and 96.7% under varying torque conditions.

Conclusions:

• The developed signal processing techniques and damage index vectors enhance bolt looseness detection.
• The method offers a more objective and accurate alternative to traditional approaches.
• This technique shows strong potential for practical applications in structural integrity monitoring.