航空发动机

余臻1，3，刘洋1，魏芳2，刘利军1，2.基于无迹粒子滤波算法的航空发动机排气温度预测[J].航空发动机,2021,47(6):1-6

基于无迹粒子滤波算法的航空发动机排气温度预测

Prediction of Aeroengine Exhaust Gas Temperature Based on Unscented Particle Filter Algorithm

DOI：

英文关键词:exhaust gas temperature margin（EGTM） unscented particle filter water washing prediction data prediction aeroengine

基金项目:国家自然科学基金（61304110）、广东省自然科学基金（2018A030313124）、深圳市基础研究面上项目（JCYJ20180306173002631）、上海市自然科学基金（18ZR1443200）、航海保障技术重点实验室开放基金资助

作者	单位	E-mail
余臻1，3，刘洋1，魏芳2，刘利军1，2	1. 厦门大学航空航天学院，福建厦门361005；2.中国航发商用航空发动机有限责任公司，上海200241；3.中国船舶航海保障技术重点实验室，天津300450	yuzhen20@xmu.edu.cn

摘要点击次数: 1041

全文下载次数: 617

中文摘要:

针对粒子滤波算法不能考虑最新的观测值，仅使用某时间节点之前的实际数据来预测航空发动机排气温度，会造成预测的温度数据误差累积，不能及时修正以及粒子退化等问题，将无迹粒子滤波引入到航空发动机排气温度预测中。分别介绍了粒子滤波算法和无迹粒子滤波算法；在此基础上，建立了航空发动机的退化模型。利用退化模型和无迹粒子滤波算法对航空发动机排气温度进行预测，并将预测值与实际值进行比较，将所得结果与采用传统粒子滤波算法得到的结果进行了对比，结果表明：无迹粒子滤波算法对于排气温度的预测效果较好，所预测的发动机达到阈值的时间与实际时间更为接近，温度范围更为集中，准确性更高，预测误差小于5%。

英文摘要:

The particle filter could not consider the latest observed values. It only used the actual data before a certain time node to predict the aeroengine exhaust gas temperature，which would cause the accumulation of predicted temperature data error, untimely correction and particle degradation. In view of the above problems, unscented particle filter was introduced into the aeroengine exhaust gas temperature prediction. Particle filter algorithm and unscented particle filter algorithm were introduced respectively. On this basis，the degradation model of aeroengine was established. The degradation model and unscented particle filter algorithm were used to predict the aeroengine exhaust gas temperature, and the predicted value was compared with the actual value. The results were compared with those obtained by traditional particle filter algorithm. The results show that the unscented particle filter algorithm has better prediction effect on the exhaust gas temperature，the predicted time when the engine reaches the threshold is closer to the actual time，the temperature range is more concentrated，the accuracy is higher，and the prediction error is less than 5%.

查看全文查看/发表评论下载PDF阅读器