Fast-Activated Minimal Gated Unit: Lightweight Processing and Feature Recognition for Multiple Mechanical Impact...

Area of Science:

• Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Multiple dynamic impact signals are crucial in engineering but challenging to identify due to nonlinear interference and signal adhesion.
• Accurate and rapid identification of these signals is often hindered by computational complexity and hardware limitations.

Purpose of the Study:

• To develop an intelligent algorithm for accurate and real-time identification of multiple dynamic impact signals.
• To address the challenges posed by nonlinear interference and signal adhesion in dynamic impact signal analysis.
• To propose a computationally efficient method suitable for platforms with weak hardware.

Main Methods:

• Established a transfer model for multiple impacts in multibody dynamical systems to analyze signal features.
• Utilized wavelet transformation to effectively suppress interference in the impact signals.
• Developed a lightweight neural network, the fast-activated minimal gated unit (FMGU), for efficient signal processing.

Main Results:

• The proposed FMGU-based method demonstrated excellent feature recognition comparable to GRU and LSTM networks across various impact speeds.
• The FMGU network achieved a 50% reduction in computational complexity compared to GRU and LSTM.
• The algorithm proved effective in maintaining recognition accuracy despite varying impact speeds and signal interference.

Conclusions:

• The intelligent algorithm combining wavelet transforms and the FMGU network offers a practical solution for real-time identification of multiple dynamic impact signals.
• This method significantly reduces computational load, making it ideal for resource-constrained engineering applications.
• The approach provides a valuable tool for enhancing the reliability and efficiency of systems relying on dynamic impact signal analysis.