Classification of Chaotic Squeak and Rattle Vibrations by CNN Using Recurrence Pattern

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Squeak and rattle (S&R) vibrations are common in mechanical systems.
• Classifying chaotic S&R signals is challenging due to their complex nonlinear dynamics.

Purpose of the Study:

• To develop and validate a deep learning approach for classifying chaotic S&R vibrations.
• To differentiate between rattle, single-mode squeak, and multi-mode squeak signals.

Main Methods:

• Generated chaotic S&R signals using distinct mechanical models.
• Visualized signal repetition with unthresholded recurrence plots.
• Employed a convolutional neural network (CNN) for classification of vibration images.

Main Results:

• The CNN successfully classified chaotic S&R signals with over 90% accuracy.
• Classification accuracy was validated using the Lyapunov exponent.
• The system could distinguish six distinct classes, including chaotic status and specific S&R models.

Conclusions:

• Deep learning, particularly CNNs, is effective for classifying complex chaotic vibrations.
• Nonlinear vibration images derived from recurrence plots are suitable inputs for AI-based analysis.
• The proposed method offers a robust solution for identifying and categorizing S&R phenomena.