基于声音信号融合特征的电气设备故障诊断

Fault Diagnosis of Electrical Equipment Based on Fusion Features of Sound Signals

在电气设备的日常监测和维护中,准确识别异常声音对于设备运行状态的评估和故障排除至关重要。传统的方法除了依靠人工现场经验处理外,主要基于频域序列建模,但这种方法忽略了时域信息的重要性。近年来,基于机器学习和深度学习的声音识别算法在电气设备领域得到广泛应用。然而,许多现有算法在捕捉异常声音的时域特征方面仍存在一些困难。为解决这一问题,文章提出了一种创新的基于时间-频谱融合的故障检测优化模型,它不仅结合了频域特征,还充分考虑了时域特征。通过在公共数据集上进行实验证明,与目前最佳方法相比,平均曲线下面积提高了4.96%。这一创新性的声音融合特征识别模型为电气设备的日常监测和维护提供了重要的支持,有望在实际应用中发挥巨大的潜力。

In the daily monitoring and maintenance of electrical equipment,accurately identifying abnormal sounds is crucial for assessing the operational status and troubleshooting of the devices.Traditional methods mainly rely on frequency-domain sequence modeling,besides manual on-site experience-based processing.However,these methods overlook the importance of temporal information.In recent years,machine learning and deep learning-based sound recognition algorithms have been widely applied in the field of electrical equipment.Nevertheless,many existing algorithms still face challenges in capturing temporal features of abnormal sounds.To address this issue,an innovative fault detection optimization model based on time-frequency fusion is proposed,which combines both frequency-domain and temporal features.Through experimentation on public datasets,it was proven that the average area under curve was improved by 4.96%compared to the current best method.This innovative sound fusion feature recognition model provides crucial support for the daily monitoring and maintenance of electrical equipment,and has the potential to play a significant role in practical applications.