小波网络平均影响值的航空发动机自变量筛选  

Aero-engine arguments selection based on wavelet network mean impact value

作　　者：崔智全[1] 付旭云[2] 钟诗胜[2] 王体春[3] 

机构地区：[1]哈尔滨工业大学(威海)汽车工程学院,山东威海264209 [2]哈尔滨工业大学(威海)船舶工程学院,山东威海264209 [3]南京航空航天大学机电工程学院,江苏南京210026

出　　处：《计算机集成制造系统》2013年第12期3062-3067,共6页Computer Integrated Manufacturing Systems

基　　金：中国民航总局科技项目(MHRD201052);国家863计划重点资助项目(2012AA040911);国家自然科学基金重点资助项目(60939003)~~

摘　　要：为了快速准确地实现发动机参数非线性自变量筛选,基于平均影响值的思想和小波神经网络学习能力强、收敛速度快、具有自适应性和容错性等优点,提出小波神经网络平均影响值的发动机自变量筛选方法。根据参数之间的关系特点,建立多参数连续小波逼近网络模型,并给出学习算法。仿真实例表明,该方法不但能够实现复杂的非线性变量筛选,而且对比其他非线性变量筛选方法,具有精度更高、速度更快的特点。To achieve the non-linear variables selection rapidly and accurately, the engine arguments parameters se lection method for wavelet neural network＇s Mean Impact Value （MIV） was proposed based on the ideological of MIV and the advantages such as learning ability, fast convergence with adaptive and fault tolerance of wavelet neural network. According to the relationship characteristics of the engine parameters, the continuous multi-parameter ap proximation wavelet network model was established, and the learning algorithm was given. Simulation results showed that the proposed method could achieve complex nonlinear variable selection and have higher accuracy and faster features by comparing to other non-linear variable selection method.

关 键 词：航空发动机 小波网络 平均影响值 自变量筛选 

分 类 号：TP311.132[自动化与计算机技术—计算机软件与理论] V263.6[自动化与计算机技术—计算机科学与技术]