Spectral noise and data reduction using a long short-term memory network for nonlinear ultrasonic modulation-based...

Area of Science:

• Non-destructive testing
• Materials science
• Signal processing

Background:

• Micro fatigue cracks generate small nonlinear modulation components, often obscured by noise.
• Large data volumes pose challenges for monitoring systems due to power, storage, and bandwidth limitations.

Purpose of the Study:

• To develop a spectral noise and data reduction technique for nonlinear ultrasonic modulation-based fatigue crack detection.
• To address the limitations of current monitoring systems in handling noise and large datasets.

Main Methods:

• Application of a long short-term memory (LSTM) network to ultrasonic signals.
• Implementation of spectral noise reduction and data reduction strategies.

Main Results:

• Successful reduction of spectral noise in ultrasonic signals.
• Achieved data reduction without compromising the spectral density amplitude of nonlinear modulation components.
• Demonstrated effectiveness on complex geometries and real structures under external noise.

Conclusions:

• The proposed LSTM-based technique effectively reduces spectral noise and data volume in ultrasonic fatigue crack detection.
• The method preserves essential nonlinear modulation components, making it suitable for various structural health monitoring applications.