In the context of 1965- 2000 monthly rainfall data from 73 stations distributed over 3 province level districts and 2 metropolises (Beijing and Tianjin) of North China with some stations in the neighboring provinces,d...In the context of 1965- 2000 monthly rainfall data from 73 stations distributed over 3 province level districts and 2 metropolises (Beijing and Tianjin) of North China with some stations in the neighboring provinces,diagnostic study is undertaken of the features of spatially anomalous patterns and dominant periods of the annual precipitation in terms of EOF,REOF and SSA.Also, a scheme consisting of SSA combined with autoregression (AR) as a prediction model is employed to make forecasts of monthly rainfall sequences of the anomalous patterns in terms of an adaptive filter.Results show that the scheme,if further improved,would be of operational utility in preparing county-level prediction.展开更多
文摘滚动轴承性能退化评估是预诊断的提前和基础,对在役滚动轴承实施在线状态监测和性能退化评估具有重要意义。针对概率相似度量评估方法存在模型复杂、容易过早饱和等现象,提出一种基于自回归时序(autoregressive model,简称AR)模型和多元状态估计(multivariate state estimation technique,简称MSET)的滚动轴承性能在线评估方法,其中AR模型用于提取轴承振动信号的状态特征,MSET模型用于重构AR模型系数。首先,提取正常运行状态下振动信号的AR模型系数构建MSET模型的历史记忆矩阵;其次,将待测信号的AR系数作为观测向量输入MSET模型中得到重构后的估计向量;最后,由原始AR系数和重构AR系数分别构造自回归模型,并各自完成对待测信号的时序建模,将两自回归模型所得残差序列的均方根值之差作为性能劣化程度指标。离散实验数据和全寿命疲劳实验数据分析结果表明,该方法能够有效检测早期故障,且具有与轴承故障发展趋势一致性更好等优点。
文摘In the context of 1965- 2000 monthly rainfall data from 73 stations distributed over 3 province level districts and 2 metropolises (Beijing and Tianjin) of North China with some stations in the neighboring provinces,diagnostic study is undertaken of the features of spatially anomalous patterns and dominant periods of the annual precipitation in terms of EOF,REOF and SSA.Also, a scheme consisting of SSA combined with autoregression (AR) as a prediction model is employed to make forecasts of monthly rainfall sequences of the anomalous patterns in terms of an adaptive filter.Results show that the scheme,if further improved,would be of operational utility in preparing county-level prediction.
