Research on the Air Compressor Fault Identification Algorithm Based on Low-rank Matrix Recovery
CHENG Long1**,YU tianming2*
1.School of Information and Control Engineering，Jilin Institute of Chemical Technology，Jilin City 132022，China;2.School of Automation Engineering, Northeast Electric Power University, Jilin City 132022，China
Abstract: Confronting the challenge of accurately extracting air compressor fault features from traditional MFCC spectrograms in high-noise environments, this paper proposes an air compressor fault identification method based on low-rank matrix recovery. The proposed approach utilizes low-rank background modeling to denoise MFCC spectrograms, effectively separating foreground features from noise. Subsequently, the MobileNetV2 model is employed for feature extraction, while Principal Component Analysis (PCA) is applied to achieve dimensionality reduction. Finally, the extracted fault features are clustered in an unsupervised manner using the K-means algorithm, enabling precise classification of air compressor fault states. Experimental results demonstrate that the proposed method not only significantly enhances recognition accuracy—for instance, ResNet18 achieved a 99.40% accuracy with substantially reduced training time—but also exhibits superior performance in clustering analysis by attaining higher CH and SC indices. These findings underscore the method’s robustness and advantage in noisy environments, offering a reliable and efficient solution for air compressor fault diagnosis and maintenance.
程龙, 于天暝. 基于低秩矩阵恢复的空压机故障识别算法研究[J]. 吉林化工学院学报, 2025, 42(7): 62-68.
CHENG Long, YU tianming. Research on the Air Compressor Fault Identification Algorithm Based on Low-rank Matrix Recovery. Journal of Jilin Institute of Chemical Technology, 2025, 42(7): 62-68.
