摘要
针对动调陀螺仪性能参数的退化特点,提出了一种自组织特征映射(SOFM)神经网络和隐马尔可夫模型(HMM)相结合的动调陀螺仪故障预测方法。采集动调陀螺仪的振动、温度、随机漂移、电机功率、电源电压和频率等信号作为表征陀螺退化状态的特征信息,利用SOFM神经网络实现多源传感器信息融合;利用HMM方法将不易检测到的早期故障信号转变为容易观测到的信息,实现动调陀螺仪的故障预测。实验结果表明:采用SOFM方法对传感信号的信息融合,能够简单、有效地提取陀螺退化状态的特征信息。运用HMM进行训练和测试,说明了该方法在故障预测中的有效性。
Because the parameters of the performance of a dynamically tuned gyroscope has degradation, we pro- pose a method for predicting its faults, which combines the self-organizational feature mapping （SOFM） neural net- work with the hidden Markov model （HMM）. Firstly, we gather the vibration, temperature, random drifting, mo- tor power, power source voltage, frequency and other signals of the dynamically tuned gyroscope as the feature in- formation for charactering its degradation, and then use the SOFM neural network to implement the multi-sensor in- formation fusion. Secondly, we use the HMM to transform the early fault signals, which are difficult to detect, into the easily observed information, thus predicting the faults of the dynamically tuned gyroscope. The experimental re- suits show that our method can easily and effectively extract the feature information on the gyroscope＇s degradation. The training and testing with the HMM show that our method is effective for fault predictions.
出处
《机械科学与技术》
CSCD
北大核心
2012年第10期1711-1715,1720,共6页
Mechanical Science and Technology for Aerospace Engineering
基金
武器装备预研基金项目(9140A25070208JB1402)资助
关键词
故障预测
自组织特征映射
隐马尔可夫模型
动调陀螺仪
fault prediction
self-organizational feature mapping （SOFM） neural network
hidden Markov model （HMM）
dynamically tuned gyroscope
degradation
