A fault identification method ofrotating machinery is proposed,which combines wavelet packet of time-frequency analysis and manifold learning.Firstly,the sampled vibration signal is decomposed to multilayer informatio...A fault identification method ofrotating machinery is proposed,which combines wavelet packet of time-frequency analysis and manifold learning.Firstly,the sampled vibration signal is decomposed to multilayer information with wavelet packet decomposition(WPD) method.Andevery level data of wavelet packet decomposition is processed bydemodulatingof Hilbert transform,eliminating the high frequency noiseof FIR filterand reducing the data length of the low frequency of resampling.Further,every level data vector is deal with normalization and calculated for the auto power spectrum.Finally,the manifold learning methods of t distributed stochastic neighbor embedding(t-SNE) is applied to do dimension reduction to generate 2D manifold figure data.Different fault forms of gearbox have different manifold features,which is used to identify failure status of equipment.With the experiment test,the feasibility and effectiveness of this identification method is verified.展开更多
In this paper,we proposal stream surface and stream layer.By using classical tensor calculus,we derive 3-D Navier-Stokes Equations(NSE)in the stream layer under semigeodesic coordinate system,Navier-Stokes equation on...In this paper,we proposal stream surface and stream layer.By using classical tensor calculus,we derive 3-D Navier-Stokes Equations(NSE)in the stream layer under semigeodesic coordinate system,Navier-Stokes equation on the stream surface and 2-D Navier-Stokes equations on a two dimensional manifold. After introducing stream function on the stream surface,a nonlinear initial-boundary value problem satisfies by stream function is obtained,existence and uniqueness of its solution are proven.Based this theory we proposal a new method called"dimension split method"to solve 3D NSE.展开更多
Dynamic movement primitives(DMPs)as a robust and efcient framework has been studied widely for robot learning from demonstration.Classical DMPs framework mainly focuses on the movement learning in Cartesian or joint s...Dynamic movement primitives(DMPs)as a robust and efcient framework has been studied widely for robot learning from demonstration.Classical DMPs framework mainly focuses on the movement learning in Cartesian or joint space,and can’t properly represent end-efector orientation.In this paper,we present an extended DMPs framework(EDMPs)both in Cartesian space and 2-Dimensional(2D)sphere manifold for Quaternion-based orientation learning and generalization.Gaussian mixture model and Gaussian mixture regression(GMM-GMR)are adopted as the initialization phase of EDMPs to handle multi-demonstrations and obtain their mean and covariance.Additionally,some evaluation indicators including reachability and similarity are defned to characterize the learning and generalization abilities of EDMPs.Finally,a real-world experiment was conducted with human demonstrations,the endpoint poses of human arm were recorded and successfully transferred from human to the robot.The experimental results show that the absolute errors of the Cartesian and Riemannian space skills are less than 3.5 mm and 1.0°,respectively.The Pearson’s correlation coefcients of the Cartesian and Riemannian space skills are mostly greater than 0.9.The developed EDMPs exhibits superior reachability and similarity for the multi-space skills’learning and generalization.This research proposes a fused framework with EDMPs and GMM-GMR which has sufcient capability to handle the multi-space skills in multi-demonstrations.展开更多
In this paper, we propose a dimensional splitting method for the three dimensional （3D） rotating Navier-Stokes equations. Assume that the domain is a channel bounded by two surfaces and is decomposed by a series of...In this paper, we propose a dimensional splitting method for the three dimensional （3D） rotating Navier-Stokes equations. Assume that the domain is a channel bounded by two surfaces and is decomposed by a series of surfaces i into several sub-domains, which are called the layers of the flow. Every interface i between two sub-domains shares the same geometry. After establishing a semi-geodesic coordinate （S-coordinate） system based on i, Navier-Stoke equations in this coordinate can be expressed as the sum of two operators, of which one is called the membrane operator defined on the tangent space on i, another one is called the bending operator taking value in the normal space on i. Then the derivatives of velocity with respect to the normal direction of the surface are approximated by the Euler central difference, and an approximate form of Navier-Stokes equations on the surface i is obtained, which is called the two-dimensional three-component （2D-3C） Navier-Stokes equations on a two dimensional manifold. Solving these equations by alternate iteration, an approximate solution to the original 3D Navier-Stokes equations is obtained. In addition, the proof of the existence of solutions to 2D-3C Navier-Stokes equations is provided, and some approximate methods for solving 2D-3C Navier-Stot4es equations are presented.展开更多
基金supported by the National Natural Science Foundation-supported Program(515750055)Beijing Municipal Natural Science Foundation(3131002)
文摘A fault identification method ofrotating machinery is proposed,which combines wavelet packet of time-frequency analysis and manifold learning.Firstly,the sampled vibration signal is decomposed to multilayer information with wavelet packet decomposition(WPD) method.Andevery level data of wavelet packet decomposition is processed bydemodulatingof Hilbert transform,eliminating the high frequency noiseof FIR filterand reducing the data length of the low frequency of resampling.Further,every level data vector is deal with normalization and calculated for the auto power spectrum.Finally,the manifold learning methods of t distributed stochastic neighbor embedding(t-SNE) is applied to do dimension reduction to generate 2D manifold figure data.Different fault forms of gearbox have different manifold features,which is used to identify failure status of equipment.With the experiment test,the feasibility and effectiveness of this identification method is verified.
文摘In this paper,we proposal stream surface and stream layer.By using classical tensor calculus,we derive 3-D Navier-Stokes Equations(NSE)in the stream layer under semigeodesic coordinate system,Navier-Stokes equation on the stream surface and 2-D Navier-Stokes equations on a two dimensional manifold. After introducing stream function on the stream surface,a nonlinear initial-boundary value problem satisfies by stream function is obtained,existence and uniqueness of its solution are proven.Based this theory we proposal a new method called"dimension split method"to solve 3D NSE.
基金Supported by National Natural Science Foundation of China(Grant No.52175029)Key Industrial Chain Projects of Shaanxi Province(Grant No.2018ZDCXL-GY-06-05).
文摘Dynamic movement primitives(DMPs)as a robust and efcient framework has been studied widely for robot learning from demonstration.Classical DMPs framework mainly focuses on the movement learning in Cartesian or joint space,and can’t properly represent end-efector orientation.In this paper,we present an extended DMPs framework(EDMPs)both in Cartesian space and 2-Dimensional(2D)sphere manifold for Quaternion-based orientation learning and generalization.Gaussian mixture model and Gaussian mixture regression(GMM-GMR)are adopted as the initialization phase of EDMPs to handle multi-demonstrations and obtain their mean and covariance.Additionally,some evaluation indicators including reachability and similarity are defned to characterize the learning and generalization abilities of EDMPs.Finally,a real-world experiment was conducted with human demonstrations,the endpoint poses of human arm were recorded and successfully transferred from human to the robot.The experimental results show that the absolute errors of the Cartesian and Riemannian space skills are less than 3.5 mm and 1.0°,respectively.The Pearson’s correlation coefcients of the Cartesian and Riemannian space skills are mostly greater than 0.9.The developed EDMPs exhibits superior reachability and similarity for the multi-space skills’learning and generalization.This research proposes a fused framework with EDMPs and GMM-GMR which has sufcient capability to handle the multi-space skills in multi-demonstrations.
基金Supported by the National High-Tech Research and Development Program of China (No. 2009AA01A135)the National Natural Science Foundation of China (Nos. 10971165, 11001216, 11071193, 10871156)the Foundation of AVIC Chengdu Aircraft Design and Research Institute
文摘In this paper, we propose a dimensional splitting method for the three dimensional （3D） rotating Navier-Stokes equations. Assume that the domain is a channel bounded by two surfaces and is decomposed by a series of surfaces i into several sub-domains, which are called the layers of the flow. Every interface i between two sub-domains shares the same geometry. After establishing a semi-geodesic coordinate （S-coordinate） system based on i, Navier-Stoke equations in this coordinate can be expressed as the sum of two operators, of which one is called the membrane operator defined on the tangent space on i, another one is called the bending operator taking value in the normal space on i. Then the derivatives of velocity with respect to the normal direction of the surface are approximated by the Euler central difference, and an approximate form of Navier-Stokes equations on the surface i is obtained, which is called the two-dimensional three-component （2D-3C） Navier-Stokes equations on a two dimensional manifold. Solving these equations by alternate iteration, an approximate solution to the original 3D Navier-Stokes equations is obtained. In addition, the proof of the existence of solutions to 2D-3C Navier-Stokes equations is provided, and some approximate methods for solving 2D-3C Navier-Stot4es equations are presented.
