半球谐振陀螺仪寿命的一种长周期预测方法

为了缩短半球谐振陀螺仪寿命实验周期，降低实验成本，提出了一种针对漂移数据的残差修正ARGM(1,1)(Autoregressive GM(1,1))寿命预测方法。该方法利用神经网络与支持向量机中的自回归方式改进灰色模型，提高了模型的自适应能力，增强了模型的学习能力与预测能力，降低了模型回归学习的时间消耗和数据量要求，提高了预测效率。采用小波包络分析预处理某型号半球谐振陀螺仪的漂移数据，利用提出的预测方法对处理后的数据进行长周期预测，并结合灰色关联分析方法，分析失效阶段并最终预测出半球谐振陀螺仪的寿命。实验表明，残差修正ARGM(1,1)模型对半球谐振陀螺仪漂移数据的长期预测精度高于传统GM(1,1)模型、BP神经网络与支持向量机，结果也表明了研究方法的正确性和有效性。

A Long-Term Lifetime Prediction Method for Hemispherical Resonator Gyroscope

In order to shorten the test duration and cut down test cost for a hemispherical resonator gyroscope (HRG), a long-term lifetime prediction method of residual modified ARGM(1,1) (Autoregressive GM(1,1)) based on drift data is proposed. In the method, the autoregressive process used in neural network and support vector machine is utilized to improve the grey model, which not only enhances the self-adaptive ability and the prediction performance, but also lowers the demands for modeling data and improves the prediction efficiency. First, wavelet-envelope analysis is adopted to preprocess drift data, and then the processed data are used to forecast several periodic prediction sequences for the HRG. At last, a grey correlation analysis method is employed to evaluate the HRG’s failure stage and then finally achieves the HRG’s lifetime prediction. The experimental results show the residual modified ARGM(1,1) model with preprocessed drift data has the highest accuracy in long-term prediction among the conventional GM(1,1) , BP neural network and support vector machine, and the results also indicate the method is reliable.