针对传统近红外光谱波长选择方法忽略模型中非线性因素的缺陷,采用具有非线性处理能力的最小二乘支持向量机,结合间隔策略的波长选择方法和联合区间的思想,提出了一种非线性模型下的波长筛选算法—联合区间最小二乘支持向量机(synergy i...针对传统近红外光谱波长选择方法忽略模型中非线性因素的缺陷,采用具有非线性处理能力的最小二乘支持向量机,结合间隔策略的波长选择方法和联合区间的思想,提出了一种非线性模型下的波长筛选算法—联合区间最小二乘支持向量机(synergy interval least squares support vector machines,siLSSVM)。以苹果糖度近红外光谱数据为例,与传统siPLS波长筛选方法相比,新算法的预测集均方根误差(RMSEP)在PLS模型和LSSVM模型预测时分别提高了37.43%和47.88%,预测集相关系数(RP)在PLS模型和LSSVM模型预测时分别增加了6.04%和7.31%。实例表明,对于存在非线性因素较强的光谱数据,siLSSVM算法能够有效的挑选最优波长区间与提高模型的预测精度和鲁棒性,为近红外光谱在非线性因素下筛选波长提供了新前景。展开更多
利用分合闸声音信号实现断路器机械故障诊断易受背景噪声影响,而用于声音信号分离的盲源分离算法存在着分离结果无序性的问题,为此,提出了声音信号盲源分离与含负样本支持向量描述(support vector data description with negative sampl...利用分合闸声音信号实现断路器机械故障诊断易受背景噪声影响,而用于声音信号分离的盲源分离算法存在着分离结果无序性的问题,为此,提出了声音信号盲源分离与含负样本支持向量描述(support vector data description with negative samples,NSVDD)相结合的断路器机械故障诊断方法。首先,运用基于负熵最大化的快速独立分量分析(fast independent component analysis,Fast ICA)实现断路器合闸期间各声源信号的盲分离;然后依据人耳听觉特性提取各分离信号的伽马滤波器倒谱系数(Gammatone frequency cepstrum coefficient,GFCC),同时对各分离信号进行变分模态分解(variational mode decomposition,VMD)得到各有限带宽固有模态分量(band-limitedintrinsic mode functions,BLIMF),以提取声源信号的奇异谱熵、能量熵、峭度熵,并与降维后的GFCC系数组成声音信号联合特征向量;最后,利用单值分类算法NSVDD对联合特征向量进行识别,以消除噪声影响。实验结果表明,基于盲源分离与NSVDD的断路器机械故障诊断方法能够准确完成在含噪背景下的断路器机械故障诊断。展开更多
研究了近红外光谱技术快速检测红曲菌固态发酵过程参数水分含量和pH值的可行性。针对传统基于间隔策略波长选择方法忽略非线性因素的缺点,采用一种基于最小二乘支持向量机(Least squares support vector machines,LS-SVM)非线性模型的...研究了近红外光谱技术快速检测红曲菌固态发酵过程参数水分含量和pH值的可行性。针对传统基于间隔策略波长选择方法忽略非线性因素的缺点,采用一种基于最小二乘支持向量机(Least squares support vector machines,LS-SVM)非线性模型的波长筛选算法:联合区间最小二乘支持向量机(Synergy interval least squares support vector machines,siLS-SVM),并将新算法与相关系数法、iPLS算法、siPLS算法对比。实验结果显示,联合siLS-SVM算法和LS-SVM模型取得了最好的预测效果,水分含量、pH值的预测集相关系数(R p)分别为0.962 1、0.976 1,预测均方根误差(RMSEP)分别为0.012 9、0.145 2,表明模型具有较好的拟合度和预测性能。应用近红外光谱法进行红曲菌固态发酵过程的水分含量和pH值的快速检测可行,该方法为进一步实现其过程参数的在线检测及发酵条件优化提供了技术基础。展开更多
The results of investigations of compressed reinforced masonry walls subjected to axial compression are presented. Tests were carried out using specimens made of clay bricks and cement-lime mortar. As reinforcement, s...The results of investigations of compressed reinforced masonry walls subjected to axial compression are presented. Tests were carried out using specimens made of clay bricks and cement-lime mortar. As reinforcement, smooth and spiral twisted longitudinal rods, two types of structural wire mesh and truss type reinforcement were used. Two percentages of bed joint reinforcement, about 0.1% and 0.05% were applied. For each type of reinforcement, three masonry walls were tested. Additionally, nine unreinforced models were also tested. The main aim of the investigations presented is to determine the effect of different types of reinforcement on the load capacity and failure. Measurement of the strains of reinforcing bars permitted the recording of the strain level at the moment of crack appearance and also at the moment of failure.展开更多
A layered architecture of muhisensory integration gripper system is first developed, which includes data acquisition layer, data processing layer and network interface layer. Then we propose a novel support-vector-mac...A layered architecture of muhisensory integration gripper system is first developed, which includes data acquisition layer, data processing layer and network interface layer. Then we propose a novel support-vector-machine-based data fusion algorithm and also design the gripper system by combining data fusion with CAN bus and CORBA technology, which provides the gripper system with outstanding characteristics such as modularization and intelligence. A multisensory integration gripper test bed is finally built on which a circuit board replacement job based on Internet-based teleoperation is achieved. The experimental results verify the validity of this gripper system design.展开更多
In this paper, the authors derive the asymptotic joint distributions of theeigenvalues of some random matrices which arise from components of covariance model.
文摘针对传统近红外光谱波长选择方法忽略模型中非线性因素的缺陷,采用具有非线性处理能力的最小二乘支持向量机,结合间隔策略的波长选择方法和联合区间的思想,提出了一种非线性模型下的波长筛选算法—联合区间最小二乘支持向量机(synergy interval least squares support vector machines,siLSSVM)。以苹果糖度近红外光谱数据为例,与传统siPLS波长筛选方法相比,新算法的预测集均方根误差(RMSEP)在PLS模型和LSSVM模型预测时分别提高了37.43%和47.88%,预测集相关系数(RP)在PLS模型和LSSVM模型预测时分别增加了6.04%和7.31%。实例表明,对于存在非线性因素较强的光谱数据,siLSSVM算法能够有效的挑选最优波长区间与提高模型的预测精度和鲁棒性,为近红外光谱在非线性因素下筛选波长提供了新前景。
文摘The results of investigations of compressed reinforced masonry walls subjected to axial compression are presented. Tests were carried out using specimens made of clay bricks and cement-lime mortar. As reinforcement, smooth and spiral twisted longitudinal rods, two types of structural wire mesh and truss type reinforcement were used. Two percentages of bed joint reinforcement, about 0.1% and 0.05% were applied. For each type of reinforcement, three masonry walls were tested. Additionally, nine unreinforced models were also tested. The main aim of the investigations presented is to determine the effect of different types of reinforcement on the load capacity and failure. Measurement of the strains of reinforcing bars permitted the recording of the strain level at the moment of crack appearance and also at the moment of failure.
文摘A layered architecture of muhisensory integration gripper system is first developed, which includes data acquisition layer, data processing layer and network interface layer. Then we propose a novel support-vector-machine-based data fusion algorithm and also design the gripper system by combining data fusion with CAN bus and CORBA technology, which provides the gripper system with outstanding characteristics such as modularization and intelligence. A multisensory integration gripper test bed is finally built on which a circuit board replacement job based on Internet-based teleoperation is achieved. The experimental results verify the validity of this gripper system design.
文摘In this paper, the authors derive the asymptotic joint distributions of theeigenvalues of some random matrices which arise from components of covariance model.
