基于BP神经网络的航空发动机参数预测与实时告警系统设计

Aero-engine Parameter Prediction and Real-time Warning System Based on Back-propagation Neural Network

目前的监控与分析手段已经可以对航空发动机的不同参数进行数据监控、自动判读等操作。然而,现阶段的技术仅仅是对于到达或超过参数限制值的情况进行告警处置,对于发动机本体来说已经造成了一定影响,一些具备预测功能的技术、设备无法应用到一线生产单位。介绍一种发动机参数预测告警系统,首先基于BP神经网络对以往参数数据库进行分析学习,提取关键特征点,掌握某一阶段参数特质,建立特征模型。随后在发动机工作时通过图像识别的方式获取参数曲线,导入特征模型进行实时判读对比,结合趋势分析方法与模型计算结果预测发动机后续工作状态,实现预测风险的功能。相比于传统发动机参数监控与自动判读系统,该系统具备了时效性,在原有参数判读能力的基础上能够提前发现试车风险。

The current monitoring and analysis methods can already perform data monitoring,automatic interpretation,and other operations on different parameters of aero-engines.However,they only provides alarm when situations reach or exceed parameter limits,at that moment,impact already occurred to the engine itself.Some technologies and equipment with predictive functions cannot be applied to frontline production units.An engine parameter prediction and warning system is introduced.Firstly,based on back-propagation neural network,previous parameter databases are analyzed and leared from,key feature points are extracted,the characteristics of a certain stage of parameter are mastered,and a feature model is established.Subsequently,parameter curves are obtained through image recognition during engine operation,and feature models are imported for real-time interpretation and comparison.Trend analysis methods and model calculation results are combined to predict the subsequent working status of the engine,achieving the function of predicting risks.Compared to traditional engine parameter monitoring and automatic interpretation systems,the system has timeliness and can detect testing risks in advance on the basis of its original parameter interpretation ability.