基于ARIMA⁃LSTM的飞机液压泵性能趋势预测方法  

Aircraft Hydraulic Pump Performance Trend Prediction Method Based on ARIMA⁃LSTM

作　　者：崔建国 李鹏程 崔霄 于明月 蒋丽英 王景霖 CUI Jianguo;LI Pengcheng;CUI Xiao;YU Mingyue;JIANG Liying;WANG Jinglin(School of Automation,Shenyang Aerospace University Shenyang,110136,China;Model Balance and Wind Tunnel Equipment Department 5,AVIC Aerodynamics Research Institute Shenyang,110034,China;Aviation Key Laboratory of Science and Technology on Fault Diagnosis and Health Management Shanghai,201601,China)

机构地区：[1]沈阳航空航天大学自动化学院,沈阳110136 [2]航空工业空气动力研究院模型天平与风洞设备五部,沈阳110034 [3]故障诊断与健康管理技术航空科技重点实验室,上海201601

出　　处：《振动．测试与诊断》2021年第4期735-740,832,共7页Journal of Vibration,Measurement & Diagnosis

基　　金：国家自然科学基金资助项目(51605309);航空科学基金资助项目(201933054002,20163354004);辽宁省教育厅基金资助项目(JYT2020021)。

摘　　要：针对飞机液压泵工作强度高、工作环境复杂而导致传统性能预测方法对飞机液压泵性能变化趋势预测精度不高的问题,提出了一种基于自回归积分滑动平均⁃长短期记忆(autoregressive integrated moving average⁃long short⁃term memory,简称ARIMA⁃LSTM)网络的飞机液压泵性能趋势预测方法。首先,将获取的飞机液压泵性能表征参数回油流量数据进行线性分解,得到趋势项数据和细节项数据;其次,采用自回归积分滑动平均(autoregressive integrated moving average,简称ARIMA)方法对趋势项数据进行预测,同时采用长短期记忆(long short term memory,简称LSTM)网络方法对归一化处理后的细节项数据进行预测;最后,将两部分预测结果进行叠加,得到最终的性能趋势预测结果。研究结果表明,采用ARIMA⁃LSTM的联合预测方法对飞机液压泵性能变化趋势的预测效果明显优于单一的ARIMA与LSTM预测方法,为飞机液压泵性能变化趋势预测的工程应用提供了一种新方法。The traditional performance prediction method has low prediction accuracy for military aircraft hydraulic pump performance in high working intensity of military aircraft hydraulic pump and complicated working environment.Aiming at this problem,an military aircraft hydraulic pump performance trend prediction method is proposed based on the autoregressive integrated moving average(ARIMA)model and long short term memory(LSTM)network.Firstly,the obtained raw data of aircraft hydraulic pump performance characterization parameters are linearly decomposed to obtain trend item data and detail item data.Then,the ARIMA method is used to predict the trend item data,and the LSTM method is used to predict the normalized detail item data,and finally the two parts of the prediction result are superimposed to obtain the final performance trend prediction result.The research results show that the combined prediction method of ARIMA-LSTM is better than the single ARIMA or LSTM prediction methods for predicting the trend of aircraft hydraulic pump performance,which provides a new method for engineering application of aircraft hydraulic pump performance change trend prediction.

关 键 词：飞机液压泵 性能变化 趋势预测 自回归积分滑动平均模型 长短期记忆网络 

分 类 号：TH136[机械工程—机械制造及自动化] TP206.3[自动化与计算机技术—检测技术与自动化装置]