Blade-vortex interaction detection and extraction under deep neural network-based scale feature model.

Area of Science:

• Acoustics
• Signal Processing
• Machine Learning

Background:

• Blade-vortex interaction (BVI) generates significant noise in rotating machinery.
• Accurate detection and extraction of BVI signals are crucial for noise reduction and performance optimization.
• Existing methods often struggle with noise robustness and computational efficiency.

Purpose of the Study:

• To propose a novel deep neural network (DNN)-based method for detecting and extracting BVI signals.
• To leverage optimal scale (OPS) and optimal scale vector (OPSV) features derived from improved Mallat-Zhong discrete wavelet transform (MZ-DWT).
• To enhance the accuracy, robustness, and computational efficiency of BVI signal processing.

Main Methods:

• Incorporation of a BVI aeroacoustic model and improved MZ-DWT analysis.
• Definition of OPS and OPSV features to capture dominant BVI signal information.
• Design and training of DNN-based scale feature models (DNN-SFMs) for automatic feature extraction.
• Development of single-scale and multi-scale detectors and extractors.

Main Results:

• The proposed DNN-SFMs effectively obtain OPS and OPSV features directly from BVI signal waveforms.
• Experimental results demonstrate improved accuracy and robustness in BVI signal detection and extraction.
• The method significantly reduces computational complexity compared to existing techniques.

Conclusions:

• The proposed DNN-based approach offers a superior method for BVI signal processing.
• The use of OPS and OPSV features enhances the capability to identify and isolate BVI signals.
• This method presents a computationally efficient and robust solution for BVI noise analysis.