涡轮泵实时故障检测短数据均值自适应阈值算法  

Short data mean adaptive threshold algorithm for real-time fault detection of turbopump

作　　者：谢光军[1] 胡茑庆[1] 温熙森[1] 吴建军[2] 

机构地区：[1]国防科技大学机电工程与自动化学院,湖南长沙410073 [2]国防科技大学航天与材料工程学院,湖南长沙410073

出　　处：《推进技术》2005年第3期202-205,共4页Journal of Propulsion Technology

基　　金：国家"八六三"资助项目(2002AA722070);国家自然科学基金资助项目(50375153)。

摘　　要：为了实时监控液体火箭发动机涡轮泵的状态,提高其安全性,降低其故障带来的破坏程度,提出了短数据均值自适应阈值算法(SDM-ATA),建立了实时故障检测的统计学模型、研究了阈值区间均值与方差的自适应计算及其带宽系数的自适应训练、故障综合决策逻辑,以及故障数据对阈值贡献的踢除等方法,并利用某型火箭发动机地面试车涡轮泵振动测量数据和某型转子试验平台实时测量数据对该算法进行离线和实时在线故障检测试验验证。结果表明,SDM-ATA没有发生误检测情况,并具有实时故障检测的能力。In order to monitor the conditions of Liquid Rocket Engine (LRE) turbopump in real time, enhance its safety and minimize the loss of its faults, a short data mean adaptive thresholds algorithm (SDM-ATA) was presented and realized. Some methods, such as the statistic model for real-time fault detection, adaptive estimation of means and standard deviations and training of band-width factors for thresholds of short data means, compositive decision-making logics of faults, and refreshment of thresholds without fault data, were studied. Then, SDM-ATA was validated with the historical data from LRE test as well as real-time data from rotor test platform. It was shown that SDM-ATA could detect faults in real time and give no false alarm in this case.

关 键 词：液体推进剂火箭发动机 涡轮泵 故障检测 短数据均值^+ 数值计算 

分 类 号：V434.21[航空宇航科学与技术—航空宇航推进理论与工程]