摘要
柴油机作为大型机械的核心动力部件,其运行状态的监测和诊断尤为重要,但由于其工作环境复杂,振动信号包含大量噪声,所以特征向量难以有效提取,严重制约柴油机的故障诊断技术。该文将传统局域均值分解进行改进并将其与小波降噪相结合对原始振动信号进行降噪处理,并且利用改进局域均值分解法提取特征向量,最后应用径向基(RBF)神经网络进行故障识别。在实验中,采集4种故障工况和1种正常工况下的振动信号,利用上述方法完成对5种工况下的诊断,正确率达到95%。实验结果表明:该方法较改进前有明显进步,能有效诊断发动机故障,并且具有较高的正确率和较强的实用价值。
Diesel engines are core power units of large machinery and so monitoring and diagnosing their operation conditions become particularly important. This is because the working environment is complicated and vibration signals often contain much noise, which make feature vectors difficult to extract, thus seriously restricting the application of fault diagnosis technology. Therefore, traditional local mean decomposition is improved and combines with wavelet de-noising technology to reduce the noise of original vibration signals. The improved method is used to extract feature vectors at the same time and a RBF neural network is employed to identify diesel engine faults. In experiments, the vibration signals under 4 fault cases and 1 normal case are collected and diagnosed with this new method, and the diagnostic accuracy is up to 95%. Experimental results show that the proposed method is more practical and accurate than traditional methods.
出处
《中国测试》
CAS
北大核心
2016年第3期103-108,共6页
China Measurement & Test
