Rolling bearing is one of the most important machine elements. Its condition monitoring and fault diagnosis have been addressed for a long time. This paper presents a new signal processing method-Independent Component Analysis (ICA) to detect the faults in rolling bearings. The ICA has been widely adopted for blind source separation without any prior information on the sources and their mixing process. However, some limitations exist in natural signals separation because of the embedded noise signal, convolution, etc. In practice, there should exist some prior knowledge useful for source separation about the collected signals. For example, the knowledge about the structure of the machine under examination and the sensor layout are helpful to identify the source behavior and the number of independent components. Considering these prior knowledge, the source separation process becomes partially blind. Example reveals the advantages of this method. The potential applications of Independent Component Analysis in machine diagnosis are also reviewed.
展开▼
