A VMD-CBAM-CNN-based acoustic emission recognition method for reed valve leakage in reciprocating air compressors

Reciprocating air compressors are widely used in industry, and the reed valve, as a key component, frequently opens and closes during operation, making it prone to deformation and leakage, which affects the compressor's performance. However, the complex working environment makes it difficult to detect early leakage of the reed valve. Therefore, this paper proposes a VMD-CBAM-CNN-based acoustic emission recognition method for reed valve leakage in reciprocating air compressors. The original acoustic emission signal is decomposed using optimal parameter VMD. Combined with the acoustic emission characteristics of the reed valve, the root mean square (RMS) values of the valve opening and closing phases are calculated to filter key modes and reconstruct the signal, enhancing the recognition of leakage feature signals. The reconstructed signal is input into a CBAM-CNN model, which focuses on key feature waveforms using an attention mechanism to achieve efficient leakage recognition and classification. The results show that this method achieves an accuracy rate of 98.93% and can effectively detect early leakage in reciprocating air compressor reed valves.