|
Abstract According to cavitation feature, the method of distinguishing cavitation characteristic frequency band was posed by using higher order derivative. The shell of waterjet pump vibration and inlet underwater acoustic were first calculated by higher order derivative, and then normalized with the root mean square (RMS). After filtering by cavitation characteristic frequency band, a local maximum with amplitude over a predefining threshold was picked up. The mean, root mean square, variance, skewness and kurtosis as the support vector machine classification input can realize classification diagnosis of six kind of marine waterjet cavitation states. Compared with identification result of back propagation（BP）and radial basic function（RBF）neural networks, the classification precision of least squares support vector classification(LSSVC) is more higher and the program runtime is more shorter.
|
|
Received: 09 April 2011
Published: 30 March 2012
|
|
|
|
|
|
