遗传BP神经网络试车台振动趋势预测  

Vibration trend prediction of test bed based on genetic BP neural network

作　　者：刘自然[1] 王律强 李爱民[2] 颜丙生[1] 甄守乐[1] 熊伟[1] 

机构地区：[1]河南工业大学机电工程学院,河南郑州450007 [2]河南工业大学工程训练中心,河南郑州450007

出　　处：《中国测试》2015年第8期118-121,共4页China Measurement & Test

摘　　要：为解决航空发动机附件试车台主要监测参数振幅与流量、压力等其他参数之间的关系难以用线性模型来预测的问题,建立具有7-8-1结构并经遗传算法(genetic algorithm,GA)优化过阈值和权值的遗传反向误差传播(back propagation,BP)神经网络。以输出转速、增压泵入口流量、增压泵入口温度、增压泵入口压力、增压级出口流量、增压级出口温度和增压级出口压力为网络输入,附件的径向振幅为网络输出建立预测模型。将训练后的网络用于某型航空发动机附件振动趋势的预测,得到的预测值符合要求。To solve the problem that linear mode can＇t predict the relationship between amplitude,which is the main monitoring parameter, flow, pressure and other parameters of Aero engine accessory test bed, a 7-8-1 genetic back propagation（ BP） neural network structure is established,whose threshold and weight optimized by genetic algorithm（ GA）. A prediction model is built by using the output speed, booster pump entrance flow, booster pump entrance temperature, booster pump entrance pressure, booster stage outlet flow, booster stage outlet temperature and booster stage outlet pressure as the network input and radial amplitude of aircraft accessory as the network output. The trained network is used to predict the vibration trend of a certain type of aircraft accessory and the error between the acquired prediction value and the actual value is under the requirement.

关 键 词：附件试车台 遗传算法 BP神经网络 振动预测 

分 类 号：O211.67[理学—概率论与数理统计]