基于模糊信息粒化和优化SVM的航空发动机性能趋势预测

提出采用模糊信息粒化(FIG)和优化的支持向量机(SVM)来预测航空发动机参数的变化趋势和变化空间。利用模糊信息粒化方法对性能参数进行粒化处理。以K CV验证误差最小作为优化目标,采用遗传算法(GA)实现支持向量机惩罚参数和核函数参数的自适应优化选择；训练SVM模型并进行并对模糊粒子非线性预测。利用某航空公司的某型航空发动机性能参数监测数据进行验证,结果表明:该算法可以有效实现航空发动机性能参数变化趋势和变化空间预测。在实例基础上分析了窗口大小对算法预测精度的影响以及算法多步预测的效果,得出算法最佳窗口大小为3个数据且算法3步以内预测误差小于10%。 

Forecasting of aero engine performance trend based on fuzzy information granulation and optimized SVM

A method to predict the change trend and space of aero engine parameters with fuzzy information granulation (FIG) and optimized support vector machine (SVM) was put forward. FIG was adopted to granulate the parameters. Genetic algorithm (GA) was applied into adaptive selection of the best penalty parameter and kernel function parameter with K fold cross validation (K CV) error minimum as the optimization goal. The SVM model was trained for nonlinear prediction of fuzzy particles. The verification results of some airlines monitoring performance parameters data of an aero engine showed that the algorithm proposed can effectively realize the change trend and spatial prediction of aero engine performance parameters. In addition, the influence of window size on prediction accuracy and the effect of multi step prediction were studied on the basis of instance. As a result, it was concluded that the best window size was three data and the forecasting error within three steps was less than 10%.