基于支持向量数据描述的某型发动机性能监控研究

Research of Engine Performance Monitoring Based on Support Vector Data Description

提出了某型发动机性能监控与预测分析方法。基于使用特点和结构原理,确定了使用时间和各关键性能参数,给出了发动机性能参数单指数的计算公式。总结出某型发动机性能监控的一般流程。首先对飞参数据进行处理,去除掉明显相悖的数据,进一步将特征参数输入到神经网路故障诊断模块中,如果特征数据存在故障,将数据输入到对应的数据库中;若特征数据不存在故障,进一步带入到发动机性能监控模块中,将发动机性能参数转化为单性能指数。以单指数为基础,利用超球体核距离的方法融合多个单指数建立定量的发动机性能监控指数。对性能指数进行相空间重构,基于最小二乘支持向量机对融合的性能指数进行预测,得到发动机性能指数预测值。对于判断发动机性能状况和视情维修具有一定的指导作用。

Performance monitoring and forecasting analysis method for x type engine was presented. Based on the operating characteristics and structural principle of an engine, the service time and key performance parameters were determined, and the calculation formula of single index of engine performance parameters was given. The general process of engine performance monitoring was summarized. Firstly, the flight parameter data was processed to remove the obviously in consistent data, and then the characteristic parameters were input into the neural network fault diagnosis module. If there was a fault in the characteristic data, it was input into the corresponding database. If there was no fault in the characteristic data, it was further brought into the engine performance monitoring module to convert the single performance index. Based on the single index, a quantitative engine performance monitoring index was established by fusing multiple single indexes by using the method of hyper sphere core distance. The performance index was constructed in phase space, and the fusion index was predicted based on least squares support vector machine to obtain the performance index. It played a guiding role in judging engine performance and condition based maintenance.