Balancing of Motor Armature Based on LSTM-ZPF Signal Processing.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Accurate motor armature balancing relies on precise extraction of signal amplitude and phase from vibration data.
• Existing methods often suffer from phase shift and amplitude loss when analyzing unbalanced signals.
• Dynamic balancing is crucial for motor performance and longevity.

Purpose of the Study:

• To propose a novel motor armature dynamic balancing method utilizing a long short-term memory (LSTM) network and zero-phase filter (ZPF).
• To enhance the accuracy of extracting amplitude and phase from unbalanced motor armature vibration signals.
• To address limitations of existing methods regarding phase shift and amplitude loss.

Main Methods:

• A hybrid approach combining a zero-phase filter (ZPF) for accurate phase extraction and a long short-term memory (LSTM) network for amplitude extraction.
• Utilizing the influence coefficient method to calculate unbalanced mass and phase.
• Employing simulated and experimental motor armature vibration signals for validation.

Main Results:

• The proposed LSTM-ZPF method demonstrated superior accuracy in extracting amplitude and phase compared to other techniques.
• Achieved a determination coefficient of 0.9999 for amplitude and an average absolute phase error of 2.4°.
• The method exhibits robust anti-noise performance, ensuring high fidelity and effective denoising.

Conclusions:

• The integrated LSTM-ZPF method significantly improves motor armature dynamic balancing accuracy.
• This approach effectively overcomes phase shift and amplitude loss issues inherent in conventional methods.
• The proposed technique offers a reliable and accurate solution for motor armature dynamic balancing with excellent noise immunity.