中文摘要：

机械设备在运行时，其音频信号往往包含了诸多状态特征。然而在工业环境下，设备故障的音频特征易受现场各种噪声信号的干扰，致使音频信号在故障诊断中的应用存在很大的局限性。针对这一问题，设计了一种基于EMD—ICA音频特征提取的故障诊断系统。利用EMD—ICA联合处理技术，对参数未知的源信号进行了有效分解，进而通过峰值分析与峭度值分析对故障音频特征实现了成功提取。经测试表明，在杂波环境下，系统能对设备故障作出正确诊断，与传统的音频故障诊断系统相比，其诊断准确性具有较大优势。

英文摘要：

Mostcondition features of running devices are contained in the audio signals.But in the industrial environment, the audio signals of device faults are easily disturbed by the noise ,it is the main reason that limiting the applization of audio signal for fault diagnosis.Aiming at the problem, the fault diagnosis system based on EMD-ICA audio feature extraction was designed herein. With EMD-ICA union processing technology, source signals without certain parameters were decomposed effectively, and the fault features of audio signals were extracted successfully. The test results show that in the industrial environment, the condition of running devices can be monitoredreal-timely through the system, and the exact reasons of device faults can be obtained simultaneously; the results aremore credible compared to the traditional audio analysis systems.