利用最小二乘支持向量机(least square support vector machine,LSSVM)在线辨识时变非线性过程时,设定其核参数较困难,设定的核参数不能适应过程变化而进行自动调节。针对此问题,提出了一种基于核参数分时段调节型LSSVM的在线过程辨识...利用最小二乘支持向量机(least square support vector machine,LSSVM)在线辨识时变非线性过程时,设定其核参数较困难,设定的核参数不能适应过程变化而进行自动调节。针对此问题,提出了一种基于核参数分时段调节型LSSVM的在线过程辨识方法。该方法利用了三个LSSVM,并将整个建模预测时期分为启动阶段和若干个工作周期。初始阶段末和每个工作周期末选定预测误差和最小的LSSVM,作为后续工作周期的工作LSSVM,同时根据启发式规则为另两个LSSVM设定核参数,它们作为后续工作周期的比较LSSVM。该方法设定核参数相对容易,而且核参数具有一定的自动调节能力。数字仿真显示,从统计角度而言,所提方法比传统方法有更好的适应性。展开更多
提出一种改进的基于数据块更新的递归主元分析(recursive principal component analysis,RPCA)方法,对具有慢时变和多变量等特性的某型舰空导弹武器雷达发射机工作过程进行自适应监测。该方法在协方差矩阵的特征值分解中引入低秩奇异值...提出一种改进的基于数据块更新的递归主元分析(recursive principal component analysis,RPCA)方法,对具有慢时变和多变量等特性的某型舰空导弹武器雷达发射机工作过程进行自适应监测。该方法在协方差矩阵的特征值分解中引入低秩奇异值分解递归方法,实现负荷矩阵和特征值矩阵的递归计算;制定了均值、方差的更新策略;给出一种基于指数加权的控制限递归算法以提高RPCA的健壮性。实验证明该方法能自适应地跟踪过程时变并实时监测故障,同时有效地降低误警率。展开更多
The low field nuclear magnetic resonance (NMR), as a nondestructive and noninvasive technique, was employed to investigate the water distribution and content in cement paste with different water-to-cement ratio (w/c r...The low field nuclear magnetic resonance (NMR), as a nondestructive and noninvasive technique, was employed to investigate the water distribution and content in cement paste with different water-to-cement ratio (w/c ratio) during early and later hydration stages. From the water distribution spectrum deduced from relaxation time distribution in paste, it is suggested that the water fills in the capillary pores at initial period, and then diffuses to the mesopores and gel pores in hydration products with the hydration proceeding. The decrease of peak area in water distribution spectrum reflects the transformation from physically bound water to chemically bound water. In addition, based on the connection between relaxation time and pore size, the relative content changes of water in various states and constrained in different types of pores were also measured. The results demonstrate that it is influenced by the formation of pore system and the original water-to-cement ratio in the paste. Consequently, the relative content of capillary water is dropped to less than 2% in the paste with low w/c ratio of 0.3 when being hydrated for 1 d, while the contents are still 16% and 36% in the pastes with w/c ratios of 0.4 and 0.5, respectively.展开更多
文摘利用最小二乘支持向量机(least square support vector machine,LSSVM)在线辨识时变非线性过程时,设定其核参数较困难,设定的核参数不能适应过程变化而进行自动调节。针对此问题,提出了一种基于核参数分时段调节型LSSVM的在线过程辨识方法。该方法利用了三个LSSVM,并将整个建模预测时期分为启动阶段和若干个工作周期。初始阶段末和每个工作周期末选定预测误差和最小的LSSVM,作为后续工作周期的工作LSSVM,同时根据启发式规则为另两个LSSVM设定核参数,它们作为后续工作周期的比较LSSVM。该方法设定核参数相对容易,而且核参数具有一定的自动调节能力。数字仿真显示,从统计角度而言,所提方法比传统方法有更好的适应性。
文摘提出一种改进的基于数据块更新的递归主元分析(recursive principal component analysis,RPCA)方法,对具有慢时变和多变量等特性的某型舰空导弹武器雷达发射机工作过程进行自适应监测。该方法在协方差矩阵的特征值分解中引入低秩奇异值分解递归方法,实现负荷矩阵和特征值矩阵的递归计算;制定了均值、方差的更新策略;给出一种基于指数加权的控制限递归算法以提高RPCA的健壮性。实验证明该方法能自适应地跟踪过程时变并实时监测故障,同时有效地降低误警率。
基金Project(2009CB623105) supported by the National Basic Research Program of ChinaProject(51108341) supported by the National Natural Science Foundation of ChinaProjects(20110490703, 2012T50437) supported by China Postdoctoral Science Foundation
文摘The low field nuclear magnetic resonance (NMR), as a nondestructive and noninvasive technique, was employed to investigate the water distribution and content in cement paste with different water-to-cement ratio (w/c ratio) during early and later hydration stages. From the water distribution spectrum deduced from relaxation time distribution in paste, it is suggested that the water fills in the capillary pores at initial period, and then diffuses to the mesopores and gel pores in hydration products with the hydration proceeding. The decrease of peak area in water distribution spectrum reflects the transformation from physically bound water to chemically bound water. In addition, based on the connection between relaxation time and pore size, the relative content changes of water in various states and constrained in different types of pores were also measured. The results demonstrate that it is influenced by the formation of pore system and the original water-to-cement ratio in the paste. Consequently, the relative content of capillary water is dropped to less than 2% in the paste with low w/c ratio of 0.3 when being hydrated for 1 d, while the contents are still 16% and 36% in the pastes with w/c ratios of 0.4 and 0.5, respectively.
