| The mechanical fault feature extraction is the key of fault diagnosis. In practicalapplication, traditional methods of fault feature extraction and fault diagnosis havemany disadvantages. In order to overcome the disadvantages of traditional methods,blind source separation(BSS) provides a new solution, which can completethe estimation of source signals and the identification of transmission channels whensource signals and transmission channels are unknown. The source signals estimated byBSS can reflect many features of original signals more perfectly. Most existing BSSalgorithms conduct when the number of source signals is less than or equal to thenumber of observed signals, however, in practical application, it is difficultto meet the source number hypothesis requirement of BSS algorithm, which will lead tothat algorithms cannot obtain the expected separation effect. In view of this, thispaper mainly researches the following contents:1) The paper introduces the background of the subject and significance, as well asthe development of BSS.2) The BSS theory and related theoretical knowledge will be introduced, severaltypical algorithms of BSS will be selectively introduced.3) In order to put BSS into practice in fault diagnosis, blind signal sourcenumber estimation will be studied deeply. The paper will piont out the disadvantages ofthe method of the traditional blind signal sources number estimates and presenta new blind signal source number estimation method. Simulation and experimentverified the new source number estimation method.4) Taking into account that most source number estimation methods do not involvethe effect of nonstationary signals which are always present in practical application,Therefore, this paper will put forward a kind of method of blind nonstationary signalsource number estimation which is based on wavelet transform and observedsignals wavelet coefficients relationship. This method was verified by simulation. |
PDF Full Text Request