摘要
给水泵组机械状态信号具有非平稳性和非线性,而支持向量机具有良好的非线性函数逼近和泛化能力.在对最小二乘支持向量机理论研究的基础上,建立了给水泵组状态趋势预测模型.并针对人工选择的低效性,提出了自动搜索最优LS-SVM参数的方法.最后对给水泵组的振动值和温度值进行了预测实验,结果显示预测误差均小于4%,表明LS-SVM对给水泵组状态趋势具有良好的预测性能.
The mechanical state signal of boiler feed pumps is nonstationary and nonlinear, and support vector machines has an excellent nonlinearity approximation ability and better generalization capability. Based on the theory of Least Squares Support Vector Machines, the state trend prediction model of boiler feed pumps was established. Considering the inefficiency of manual selection, a solution to search optimal parameters in LS-SVM automatically was proposed. Finally, the experiment on the prediction of vibration and temperature of boiler feed pumps was carried out. Results show the prediction errors are less than 4 percent, which indicates that LS-SVM has excellent performance of mechanical state trend prediction for boiler feed pumps.
出处
《传感技术学报》
CAS
CSCD
北大核心
2007年第5期1139-1143,共5页
Chinese Journal of Sensors and Actuators
关键词
趋势预测
最小二乘支持向量机
参数优化
给水泵组
trend prediction
least squares support vector machines
parameter optimization
Boiler FeedPumps
