摘要
航天继电器长期处于贮存环境,为保证其各阶段始终保持在备用激活状态,必须对继电器的贮存寿命进行预测。本文将因子分析法和回归分析法引入到表征触点电接触可靠性的重要参数——接触电阻的转换中,将25台继电器样品的200对触点在125℃下的接触压降和释放电压双参数数据交叉分为4组进行处理,分析两者与接触电阻的关系,建立函数链神经网络,对接触电阻进行动态预测,进而得到继电器的贮存寿命。分析神经网络预测的整体误差,用92℃的数据对该方法进行检验,得出神经网络的预测误差低于3.5%,证实了统计方法和函数链神经网络的适用性。
Aerospace relay is always in long - term storage condition, it is necessary to predict the storage life of the relay in order to keep it activate in different stages of all states. This factor analysis and regression analysis methods are introduced to convert the contact resistance, which is an important parameter of the characterization of electric contact reliability. The double parameter data of contact voltage drop and voltage release of the 200 contacts of 25 sample relays at 125℃ are divided into 4 groups crossways, the relationship is analyzed between the parameters and contact resistance, function chain neural network is built to predict contact resistance dynamically and then the storage life of relays is obtained. The prediction accuracy of the neural network is analyzed, and then this method is tested by the data of another constant temperature of 92℃. It is concluded that the error of the neural network is less than 3.5 percent, which confirms the applicability of the statistical method and function chain neural network.
作者
李文华
周露露
王立国
蔡亚楠
Li Wenhua Zhou Lulu Wang Liguo Cai Yanan(Hebei University of Technology, Tianjin 300130,China)
出处
《航天控制》
CSCD
北大核心
2016年第6期79-84,共6页
Aerospace Control
基金
河北省高等学校科学技术研究重点项目(ZD20155051)
关键词
航天继电器
统计分析
神经网络预测
接触电阻
贮存寿命
Aerospace relay
Statistical analysis
Neural network prediction
Contact resistance
Storage life
