运载火箭类周期测试数据特征预测方法研究

鉴于我国运载火箭测试数据判读工作现状，研究测试数据的预测算法，有助于预判故障趋势，提前采取措施。分析了运载火箭测试数据，提出测试数据依时间序列的分类方法；针对类周期型数据，设计了相应的特征提取算法，得出数据特征时间序列；应用滚动自回归预测算法，并将历史实际值与预测值的加权值作为当前时刻的建模数据，实现了类周期数据特征的趋势预测。该方法有助于改进运载火箭类周期型数据判读方法。

Study on Feature Trend Forecasting Method for Rocket Period-like Launch Data

In consideration of the current status of Rocket launch data interpretation, fault trend prediction and measures taken in advance would be benefit from studying of forecasting algorithms for the launch data. Rockets launch data is analyzed and then classified based on time series features. Aiming at the period-like data, the corresponding feature extraction algorithm is developed and the feature time series are reached. The rolling auto regressive algorithm is applied, with weighting history true value and forecasted value as the new modeling data at the present moment. The result is that the trend forecasting for the feature of the period-like data is realized, which will help to improve the rocket launch data interpretation.