This paper makes a probe into the application of the Kalman filtering method to the data processing of across-fault measurements.On the basis of statistical regression,the mathematic and stochastic models of filtratio...This paper makes a probe into the application of the Kalman filtering method to the data processing of across-fault measurements.On the basis of statistical regression,the mathematic and stochastic models of filtration are established by combining the regression method with Kalman filtering.In the filtering computation,not only the randomness of fault movements but also the time-dependent variation of environmental effects have been taken into consideration.By use of the adaptive filtering method,an estimation of the dynamic noise variance matrix is obtained through iteration.Models for one measuring line(leveling line or baseline),two measuring lines(both leveling lines or both baselines)and four measuring lines(two leveling lines and two baselines)are derived and established systematically.By means of these models,the data of across-fault measurements can be processed dynamically in real-time to provide the filtered values of height difference between benchmarks or baseline length at different time展开更多
Based on the arrangement of the across-fault measurement data along the northern edge of the Qinghai-Xizang block,we divide the deformation into different types and probe the nature of various fault movements based on...Based on the arrangement of the across-fault measurement data along the northern edge of the Qinghai-Xizang block,we divide the deformation into different types and probe the nature of various fault movements based on these types.The recent situation of tectonic movement of main structural belts and seismicity in this area are expounded.From the above,it is concluded that across-fault measurement can reflect not only the conditions of fault movement nearby but also the change of regional stress fields; not only is this a method to obtain regional seismogenic information and to conduct short-term prediction but it is also involved with large scale space-time prediction of moderate and strong earthquakes on the basis of the macro characteristics of fractures.展开更多
文摘This paper makes a probe into the application of the Kalman filtering method to the data processing of across-fault measurements.On the basis of statistical regression,the mathematic and stochastic models of filtration are established by combining the regression method with Kalman filtering.In the filtering computation,not only the randomness of fault movements but also the time-dependent variation of environmental effects have been taken into consideration.By use of the adaptive filtering method,an estimation of the dynamic noise variance matrix is obtained through iteration.Models for one measuring line(leveling line or baseline),two measuring lines(both leveling lines or both baselines)and four measuring lines(two leveling lines and two baselines)are derived and established systematically.By means of these models,the data of across-fault measurements can be processed dynamically in real-time to provide the filtered values of height difference between benchmarks or baseline length at different time
文摘Based on the arrangement of the across-fault measurement data along the northern edge of the Qinghai-Xizang block,we divide the deformation into different types and probe the nature of various fault movements based on these types.The recent situation of tectonic movement of main structural belts and seismicity in this area are expounded.From the above,it is concluded that across-fault measurement can reflect not only the conditions of fault movement nearby but also the change of regional stress fields; not only is this a method to obtain regional seismogenic information and to conduct short-term prediction but it is also involved with large scale space-time prediction of moderate and strong earthquakes on the basis of the macro characteristics of fractures.
文摘玻璃生产线退火窑辊道系统轴承运行状态显著影响玻璃品质和生产效率,实时监测各轴承运行状态对确保退火窑系统的平稳运行具有重要意义,提出结合Inception模块和长短期神经网络(Long Short-term Memory,LSTM)的迁移诊断方法,对退火窑辊道系统中的辊道轴承和通轴轴承运行状态进行监测、诊断。首先,使用集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD)对轴承信号进行分解和重构降噪,并利用直方均衡化增强重构信号小波时频图的聚集性。然后,针对样本充足的辊道轴承,建立Inception-LSTM网络,提取多尺度特征并学习其中的时间依赖关系,实现状态诊断。再次,针对转速不同且样本量少的通轴轴承,以辊道轴承信号为源域,以通轴轴承信号为目标域,以Inception-LSTM网络为基础,使用多核最大均值差异(Multi-kernel Maximum Mean Discrepancies,MKMMD)减小分布差异,实现故障样本不平衡条件下的跨转速域不变特征提取和迁移诊断。最后,利用实验数据和实测数据验证本算法的有效性,结果表明,该方法能有效诊断出退火窑辊道系统轴承故障,且具有较高的准确率。
