Application of Support Vector Machine in the Vibration Signal Measuring-Channel Modeling

摘要

Measurement accuracy of rotor unbalance in different speeds is affected by the nonlinearity of the vibration signal processing channel. Support Vector regression (SVR) is proposed to use for channel modeling and compensation, which performance is far superior to the traditional look-up table method. First the amplitude-frequency and phase-frequency characteristics models based on SVM regression are made using frequency response data of standard signal passing to measuring-channel as training samples, then the measurement signal is compensated according to the model obtained. Simulation and experimental results show that this method has higher accuracy, better generalization ability than the traditional BP neural network modeling method.