Application of improved support vector machine regression analysis for medium- and long-term vibration trend prediction

Aircraft engine fault diagnosis plays a crucial role in cost-effective operations of aircraft engines. However, successful detection of signals due to vibrations in multiple transmission channels is not always easy to accomplish, and traditional tests for nonlinearity are not always capable of capturing the dynamics. Here we applied a new method of smooth support vector machine regression (SSVMR) to better fit complicated dynamic systems. Since quadratic loss functions are less sensitive, the constrained quadratic optimization could be transferred to the unconstrained optimization so that the number of constraint conditions could be reduced. Meanwhile, the problem of slow operation speed and large memory space requirement associated with quadratic programming could be solved. Based on observed input and output data, the equivalent dynamic model of aircraft engineers was established, and model verification was done using historical vibration data. The results showed that SSVMR had fast operation speed and high predictive precision, and thus could be applied to provide early warning if engine vibration exceeds the required standard.