【摘要】 由于旋转机械工矿复杂,传统的诊断预测方法往往用单通道信息采集,不仅包含大量噪声而且易造成有效信息的缺失,而且例如神经网络等预测方法需要大量的训练样本,不能对故障进行及时有效的诊断和预测。结合隐马尔科夫链(HMM)训练样本少,识别精度高以及全矢信息融合技术克服单通道信息采集不全的优点。提出基于Hilbert-全矢HMM预测方法,首先对双通道信号、分别进行Hilbert包络解调去除噪声,对处理后的、信号进行全矢融合提取主振矢,采用趋势向聚类方法对主振矢信号进行聚类分析,利用GHMM模型与每一类的匹配度作为识别预测结果。并通过对轴承内圈剩余寿命的预测进行验证,其预测精度达到90.64%,表明该方法的有效性。更多还原

【Abstract】 Because of the complex environment of rotating machinery, the traditional method of diagnosis and prediction often uses single channel information collection, it contains not only a lot of noise but also misses the effective information, and such as neural networks requires a large number of training samples, The fault can’t be diagnosed and predicted timely and effectively. Based on the advantages of low training samples, the high recognition precision of HMM and the advantages of full vector information fusion technology that can get more information than single channel information, This paper proposes a prediction method of HMM based on Hilbert-Full vector, firstly, the dual channel signal、, respectively conducts Hilbert envelope demodulation which removes their noise, then use the full vector fusion to extract the main vibration vectors, use the trend of clustering methods for clustering analysis of main vibration vector signal, using the matching degree of GHMM model with each class as the prediction result. By predicting the remaining life of the inner ring of the bearing, the prediction accuracy of the method is 90.64%, which shows the effectiveness of the method.更多还原