﻿ 故障预测算法稳定性实时评估方法
 文章快速检索 高级检索

1. 北京航空航天大学 自动化科学与电气工程学院, 北京 100191;
2. 北京航空航天大学 先进航空发动机协同创新中心, 北京 100191

Real-time evaluation method for stability of fault prognostic algorithm
Yu Jinsong1,2, Liu Hao1, Zhang Ping1, Wan Jiuqing1
1. School of Automation Science and Electrical Engineering, Beijing University of Aeronautics and Astronautics, Beijing 100191, China;
2. Collaborative Innovation Center of Advanced Aero-Engine, Beijing University of Aeronautics and Astronautics, Beijing 100191, China
Abstract:An evaluation method of fault prognostic algorithm from the perspective of stability was proposed for the existing evaluation metrics for fault prognostic algorithm were subjected to the actual remaining useful life. Based on studying upon the health degradation process of the system under test, according to the prognostic value of remaining useful life and the value of consumed life, the performance of fault prognostic algorithm could be assessed by calculating the coefficient of variance of the fictitious life before system failed. To verify the proposed method, the stability of recursive least squares algorithm and particle filtering algorithm was assessed with simulated data generated by the fault progression model of electro-mechanical actuator. Simulation results indicate that the proposed algorithm can arrive at the same evaluation conclusion as the existing methods which need the ideal value of remaining useful life of the system under test.
Key words: stability     real-time     evaluation     remaining useful life     coefficient of variance

 图 1 系统健康退化过程曲线Fig. 1 System health degradation curve

 图 2 对象系统虚构寿命示意图Fig. 2 Schematic diagram for the fictitious life

2.2 评估方法及流程

 图 3 评估流程图Fig. 3 Evalution flow chart

1) 实时地获取故障预测算法关于某对象系统的RUL预测数据并绘制相应曲线;

2) 对应曲线在时间轴上选定n个预测时间点;

3) 根据式(2)依次计算n个预测时间点对应的虚构寿命{FLi|i=1,2,…,n}数值;

4) 根据式(4)~式(6)依次计算n个预测时间点的E(FL),σ(FL)V(FL)的值,并根据虚构寿命变异系数V(FL)的大小判断故障预测算法的实时稳定性. 3 实验方案与结果分析 3.1 实验方案

1) 故障演化仿真数据的生成原理. 由Matlab电机模型生成故障演化情况的仿真数据[10].具体涉及到的机电作动器故障演化模型表达式为

2) 故障缓变模型参数设置. 基于故障演化模型,设计故障缓变过程如下: 短路电阻报警阈值为200Ω; 失效阈值设置为80Ω; 模型参数C=200,θ=5.89×10－6. 计算可知,检出故障后,系统在43.2min左右达到失效阈值.

2) 故障预测算法选取. 鉴于基于粒子滤波[11](PF,Particle Filtering)和递归最小二乘[12](RLS,Recursive Least Squares)两种故障预测算法均具有较快的RUL预测值收敛速度,并可对对象系统的模型参数进行估计,本文选用这两种算法来验证实时评估方法的优越性与可行性.

 图 4 HP性能指标对比图Fig. 4 Comparison chart of HP metric

 图 5 α-λ性能指标对比图Fig. 5 Comparison chart of α-λ metric

 图 6 实时稳定性指标对比图Fig. 6 Comparison chart of stability metric

 故障预测算法 E(FL) σ(FL) V(FL) RLS 48.34 8.52 0.18 PF 44.16 2.79 0.06

 [1] 王志鹏,吕琛,王自力,等.飞机PHM演示验证平台设计技术研究[J].南京理工大学学报,2011,35(增刊):250-255 Wang Zhipeng,Lü Chen,Wang Zili,et al.Design of PHM demonstration and verification platform[J].Journal of Nanjing University of Science and Technology,2011,35(Supplement):250-255(in Chinese) [2] Abhinav S,Jose C,Bhaskar S.Evaluating algorithm performance metrics tailored for prognostics[C]//Aerospace Conference.Piscataway,NJ:IEEE,2008:1-13 [3] Abhinav S,Jose C,Bhaskar S,et al.On applying the prognostic performance metrics[C]//Prognostics and Health Management.New York:IEEE,2009:478-485 [4] Abhinav S,Jose C,Bhaskar S,et al.Metrics for offline evaluation of prognostic performance[J].International Journal of Prognostic and Health Management,2010,1(1):1-20 Click to display the text [5] Tang L,Kacprzynski G J,Goebel K,et al.Methodologies for uncertainty management in prognostics[C]//Aerospace conference,2009 IEEE.[S.l.]:IEEE,2009:1-12 [6] Gu J,Barker D,Pecht M.Uncertainty assessment of prognostics implementation of electronics under vibration loadings[J].Microelectronics Reliability,2007,47(12):1849-1856 Click to display the text [7] Roemer M J,Dzakowic J,Orsagh R F,et al.Validation and verification of prognostic and health management technologies[C]//Aerospace Conference,2005 IEEE.Atlanta:IEEE,2005:3941-3947 [8] 梁旭,李行善,张磊,等.支持视情维修的故障预测技术研究[J].测控技术,2007,26(6):5-8 Liang Xu,Li Xingshan,Zhang Lei,et al.Survey of fault prognostics supporting condition based maintenance[J].Measurement and Control Technology,2007,26(6):5-8(in Chinese) Cited By in Cnki (71) [9] Shi J Y,Shi M,Wang L,et al.Performance evaluation method of remaining useful life prediction based on pseudo life[C]//Industrial Engineering and Engineering Management.Piscataway,NJ:IEEE,2011:1118-1122 [10] 吴豪.机电作动器故障预测与健康管理关键技术研究[D].北京:北京航空航天大学,2012 Wu Hao.Key technologies of fault prognosis and health management for electro-mechanical actuator[D].Beijing:Beijing University of Aeronautics and Astronautics,2012(in Chinese) [11] Arulampalam M S,Maskell S,Gordon N,et al.A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking[J].IEEE Transactions on Signal Processing,2002,50(2):174-188 Click to display the text [12] Simon D.Optimal state estimations[M].Hoboken,New Jersey:John Wiley and Sons,2006

#### 文章信息

Yu Jinsong, Liu Hao, Zhang Ping, Wan Jiuqing

Real-time evaluation method for stability of fault prognostic algorithm

Journal of Beijing University of Aeronautics and Astronsutics, 2014, 40(9): 1208-1212.
http://dx.doi.org/10.13700/j.bh.1001-5965.2013.0532