Research On The Methods Of Indicator Diagram Recognition Of Reciprocate Mechanic

Reciprocating machinery is now widely used in engineering, however, the characteristic of it’s structure causes various fault. The traditional technique of fault diagnosis based on the vibration signal can not eliminate the influence of the jamming signal and the result of experiment is not ideal. In this paper, a method of fault identification and diagnosis combined with invariant moments algorithm based on the pressure signal is proposed. The jamming signal is avoided by collecting the pressure signal. It is effective to recognize and detect the type of fault by calculating the invariant moments of indicator diagram and with the help of SVM. Thesis is as follows:(1) Collecting the pressure signal of reciprocating compressor and obtain the information of indicator diagram by further correction and normalization.(2) Dealing the indicator diagram with the method of binarization and morphology refining, and decreasing the redundant information of the image, which as the basis of analysis and judgment.(3) Calculating the characteristic quantity of invariant moments and constructing 7 invariant moments on the basis of central moments which dealed with the method of binarization, then carry out Gram-Schmidt orthogonalization, finally the influence of the linear correlation of the eigenvectors are effectively eliminated.(4) With the method of SVM, the character of fault can be classified and recognized, which has better computation convergence speed and greatly decrease the time of characteristic classification, which is suitable for the problem of classification.