摘要
通过从故障仿真数据中提取常见故障的效应参数症状 ,建立发动机的集合覆盖多故障诊断模型。根据当前发动机已有测量参数的症状与假设故障的相应参数效应症状的对比结果 ,对各假设故障赋以相应的奖惩值 ,然后综合诸参数的奖惩值得到相应假设故障集的合理性指标值 ,基于该指标值利用遗传算法进行启发式的假设故障产生与测试。最后 ,基于所提取的效应参数症状进行了仿真故障数据样本的多故障诊断分析 ,并与经参数优选后的多故障诊断结果进行了比较。仿真结果表明 ,基于参数偏离方向提取的效应症状具有较高的故障分辨率 ,经过参数优选后能进一步提高故障的分辨率与误分故障样本的集中程度。
The manifestations of parameters are extracted from the simulated fault data for 16 common fault modes,these manifestations are used to establish the set cover model of multi faults diagnosis for liquid rocket engine(LRE) system.Based on the comparisons between the current manifestations and the standard manifestations of hypothesized faults,a penalty and reward mechanism is applied in the fault hypotheses generation and test procedure,and are composed to measure the reasonableness of the hypothesized faults subset,by which the genetic algorithm is then used to make faults hypotheses generation and test.Finally,fault diagnosis is implemented using the simulated fault data,the result comparisons between the all parameters based diagnosis and the selected parameters based diagnosis is made,showing that the selected parameters based diagnosis is far more classifiable,and its wrongly classified fault smples are more concentrated in only two fault modes.
出处
《宇航学报》
EI
CSCD
北大核心
2000年第1期47-54,共8页
Journal of Astronautics
基金
国家自然科学基金! 697850 0 2号
关键词
集合覆盖
故障诊断
遗传算法
液体火箭发动机
Set cover theory\ Fault diagnosis\ Genetic algorithms\ Liquid rocket engine
