基于时间序列分析的动调陀螺仪故障预测研究

Research on Fault Prediction of Dynamically Tuned Gyroscope Based on Time Series Analysis

针对动调陀螺仪的故障预测问题,提出一种时间序列分析和BP神经网络相结合的故障预测方法;首先,采用时间序列分析对采集的各个状态下的动调陀螺仪振动信号进行拟合,建立ARMA模型,利用模型参数作为状态识别的特征参数;其次,利用提取的特征参数,通过BP神经网络进行动调陀螺仪状态识别;最后,结合具体实例进行了验证;验证结果表明该方法可行有效,识别精度达到92%以上。

According to the fault prediction of dynamically tuned gyroscope （DTG）, proposed a fault prediction method which combined time series analysis and BP network. Firstly, used time series analysis to fit the collected vibration signal of DTG in different states, estab- lished ARMA model, and used the model parameters as the characteristic parameters of state recognition; And then, using extracted charac- teristic parameters, predicted the DTG states by the way of BP network; At last, verified the proposed method with a specific example. The experimental results shows that the method is feasible and effective and recognition accuracy is more than 92%.