• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2017, Vol. 53 ›› Issue (6): 195-202.doi: 10.3901/JME.2017.06.195

• 交叉与前沿 • 上一篇    下一篇

随机故障注入结合神经网络法的机电系统可靠性计算方法*

郭姣姣1, 刘伟1, 翟玮昊1, 税朗泉1, 赵海龙2   

  1. 1. 西北工业大学飞行器可靠性工程研究所 西安 710129;
    2. 北京机械设备研究所 北京 100854
  • 出版日期:2017-03-20 发布日期:2017-03-20
  • 作者简介:郭姣姣,女,1990年出生,博士研究生。主要研究方向为机电液系统可靠性。E-mail:s8.s@163.com刘伟(通信作者),男,1981年出生,博士,副教授,硕士研究生导师。主要研究方向为机电液系统可靠性。E-mail:liuwei@nwpu.edu.cn
  • 基金资助:
    * 国家自然科学基金(51305350)和陕西省自然科学基础研究计划(2013JM6011)资助项目; 20160312收到初稿,20161206收到修改稿;

Reliability Calculation Method of Electromechanical System Based on Random Fault Injection Combined with Artificial Neural Network

GUO Jiaojiao1, LIU Wei1, ZHAI Weihao1, SHUI Langquan1, ZHAO Hailong2   

  1. 1. Institute of Aircraft Reliability Engineering, Northwestern Polytechnical University, Xi’an 710129;
    2. Beijing Machinery and Equipment Research Institute, Beijing 100854
  • Online:2017-03-20 Published:2017-03-20

摘要:

将随机故障注入方法与神经网络技术相结合,提出机电系统多失效模式可靠性计算方法。以液压缸内外泄漏故障为事例,将虚拟故障信息注入活塞杆线性定位系统一体化仿真模型中。利用神经网络较强的函数逼近功能,得到关键敏感特征参数与系统状态信号间的显式极限状态方程,将系统的可靠性概率约束转化为一个等价的确定型,避免了机电系统动态响应的多次遍历运算。再结合随机模拟,避开了各失效模式极限状态函数间复杂的相关性讨论。采用正交试验设计方法对机电系统进行参数灵敏度分析并精简样本。基于随机故障注入-神经网络法得到了关键敏感特征参数的改变对机电系统可靠性的影响规律,进而获得了参数的可靠性区间及失效临界值。为机电系统的可靠性分析和设计提供了参考和依据。

关键词: 动态, 多失效模式, 可靠性, 随机故障, 神经网络

Abstract:

Based on the random fault injection combined with artificial neural network (ANN) technology, a reliability calculation method of electromechanical system with correlated failure modes is proposed. The virtual fault information is injected into the piston rod linear positioning system integrated simulation model by taking internal and external leakage of hydraulic cylinder as a typical fault example. By using the strong approximation function of the neural network, the explicit limit state equation between the critical sensitive characteristic parameters and the system state signals is obtained. The reliability probability constraint of the system is transformed into an equivalent deterministic constraint, which avoids the multiple traversal operation of electromechanical system dynamic response. Combined with stochastic simulation, the discussion of the complex correlation between the limit state functions of each failure mode is avoided. The parameter sensitivity analysis of the electromechanical system is analyzed and the sample is simplified by means of orthogonal experimental design method (DOE). Then based on the random fault injection-neural network method, the influence of the critical sensitive characteristic parameters on the reliability of electromechanical system is obtained. The reliability interval and the critical value of failure are obtained, which provide a reference and basis for the reliability analysis and design of electromechanical system.

Key words: correlated failure modes, dynamic, random failure, reliability, artificial neural networks