周良1,王华伟1,许珊珊2,王清薇1.基于GA优化BP算法的滑油状态监测[J].航空发动机,2022,48(5):137-142
基于GA优化BP算法的滑油状态监测
Lubricating Oil Condition Monitoring Based on Genetic Algorithm Optimized Back Propagation Algorithm
  
DOI:
中文关键词:  滑油状态  故障诊断  神经网络  遗传算法  可靠性  航空发动机
英文关键词:lubricating oil condition  fault diagnosis  neural network  genetic algorithm  reliability  aeroengine
基金项目:国家自然科学基金(U1833110)资助
作者单位E-mail
周良1,王华伟1,许珊珊2,王清薇1 1.南京航空航天大学民航学院南京2111002.山东师范大学公共管理学院济南250300 1243729063@qq.com 
摘要点击次数: 101
全文下载次数: 97
中文摘要:
      滑油状态的监测与分析是航空发动机状态监测与故障诊断的重要手段。为了解决以往滑油金属质量分数预测模型算 法的局部性、收敛速度慢及预测结果误差大等问题,结合遗传算法(GA)收敛速度快、鲁棒性好等优点,对反向传播(BP)神经网络 算法进行GA优化,通过GA对参数寻优,并应用于发动机滑油金属质量分数预测。由于滑油的状态参数并不能确定部件故障与 否,利用贝叶斯(Bayes)决策规则对诊断结果进行了错误率计算。将所提方法应用于某航空发动机滑油铁质量分数预测,结果表 明:采用GA优化后的BP神经网络(GA-BP)得到的预测结果具有更高的精度,其最大预测误差不超过6%,平均预测误差为1.7%, 所测数据与原数据具有较好的拟合性,利用Bayes决策规则对诊断结果进行分析,对于部件故障与否的判别更具说服力。
英文摘要:
      Lubricating oil condition monitoring and analysis is an important means of aeroengine condition monitoring and fault diagno? sis. In order to solve the problems of previously proposed prediction algorithms for the metal mass fraction of lubricating oil as local in opti? mization,slow in convergence,large prediction error,etc.,a Genetic Algorithm(GA)optimized Back Propagation(BP)neural network al? gorithm was proposed. Taking advantages GA’s fast convergence and robustness,the BP algorithm was optimized through GA parameter optimization and was applied to the prediction of metal mass fraction of engine lubricating oil. Because the state parameters of lubricating oil are unable to be used to determine whether the component is faulty or not,Bayesian decision rules were used to calculate the error rate of diagnosis. The proposed method was applied to predict the mass fraction of lubricating iron in an aeroengine. The results show that the BP neural network optimized by GA is more accurate for the prediction. The maximum prediction error is less than 6%,and the average pre? diction error is 1.7%. The measured data fit well with the original data. Diagnosis result analyzed using Bayes decision rules is more con? vincing in judging whether the component is faulty or not.
查看全文  查看/发表评论  下载PDF阅读器