基于Bayesian推断和LS-SVM的发动机在翼寿命预测模型  

Forecasting model of engine life on wing based on LS-SVM and Bayesian inference

作　　者：王烨[1] 左洪福[1] 蔡景[1] 戎翔[1] 

机构地区：[1]南京航空航天大学民航学院,江苏南京210016

出　　处：《南京理工大学学报》2013年第6期955-959,共5页Journal of Nanjing University of Science and Technology

基　　金：国家“863”计划资助项目(2006AA04Z427);国家自然科学基金(61079013;61179066)

摘　　要：为解决发动机的寿命预测精度问题,该文将贝叶斯(Bayesian)推断应用于最小二乘支持向量机(LS-SVM)模型参数的选择,建立了发动机在翼寿命的非线性预测模型。分析了影响发动机在翼寿命的性能参数,建立了用于机器学习的预测模型训练集,构建了基于LS-SVM的发动机在翼寿命预测模型。采用Bayesian推断理论优化LS-SVM建模,获取最优建模参数。通过某型发动机在翼寿命数据集训练模型,对在翼寿命进行预测。与几种常用的算法相比较,该文模型预测精度能够提高4.58%至9.51%,较好地解决了小样本下的预测问题,具有良好的泛化能力和预测精度。To resolve the problem of engine life forecasting accuracy,a nonlinear forecasting model for engine life on wing is established by applying Bayesian inference to the choices of model parameters of least squares support vector machine( LS-SVM). The performance parameters affecting engine life on wing are analyzed,a forecasting model training set for machine study is established,and a forecasting model of engine life on wing is established based on the LS-SVM. The LS-SVM model is optimized by using Bayesian inference,and the best modeling parameters are obtained. Theengine life on wing is forecasted by using a data set training model of a certain engine life on wing.Compared with several common algorithms,the forecasting accuracies of the model proposed here increase by 4. 58%-9. 51%,which solves the problem of forecasting of small samples,and performs well in generalization ability and forecasting precision.

关 键 词：贝叶斯推断 最小二乘支持向量机 发动机 在翼寿命 预测 

分 类 号：N55[自然科学总论]