By relying on the major projects of Beijing-Zhangjiakou HSR,China's railway industry has integrated and applied several digital intelligence technologies to form a complete set of intelligent HSR 1.0 technology,wh...By relying on the major projects of Beijing-Zhangjiakou HSR,China's railway industry has integrated and applied several digital intelligence technologies to form a complete set of intelligent HSR 1.0 technology,which has been popularized and applied to several new lines.With the continuous deepening of intelligent applications in construction,equipment,operation and other fields,there is an increasingly urgent need for integrated sharing and analysis of models and data.The paper analyzes the connotation and key points of model-data integration and puts forward the overall architecture of model-data integration platform composed of model-data convergence tier,model-data storage tier,model-data management tier,model-data calculation tier,model data aggregation tier,etc.Moreover,it looks forward to the prospect of leading key technologies and multiple innovative key technologies such as intelligent engineering survey and generative design,all-discipline intelligent construction,digital twin of railway engineering,CR450 intelligent EMU,new generation of dedicated mobile communication for railway (5G-R), operation safety protection based on multi-source information perception, transmission and fusion analysis, displacement-based full-travel intelligent service (MaaS+), intelligent comprehensive dispatching of regional railway network, wheel-rail integrated intelligent maintenance of EMU, etc. It can provide guidance and reference for digital railway construction and intelligent HSR 2.0 scientific & technological breakthroughs.展开更多
A 3.8-kin Coupled Ice-Ocean Model （C1OM） was implemented to successfully reproduce many observed phenomena in the Beaufort and Chukchi seas, including the Bering-inflow-originated coastal current that splits into th...A 3.8-kin Coupled Ice-Ocean Model （C1OM） was implemented to successfully reproduce many observed phenomena in the Beaufort and Chukchi seas, including the Bering-inflow-originated coastal current that splits into three branches： Alaska Coastal Water （ACW） , Central Channel, and Herald Valley branches. Other modeled phenomena include the Beaufort Slope Current （BSC） , the Beaufort Gyre, the East Siberian Current （ ESC）, mesoscale eddies, seasonal landfast ice, sea ice ridging, shear, and deformation. Many of these downscaling processes can only be captured by using a high-resolution CIOM, nested in a global climate model. The seasonal cycles for sea ice concentration, thickness, velocity, and other variables are well reproduced with Solid validation by satellite measurements. The seasonal cycles for upper ocean dynamics and thermodynamics are also well reproduced, which include the formation of the cold saline layer due to the injection of salt during sea ice formation, the BSC, and the subsurface upwelling in winter that brings up warm, even more saline Atlantic Water along the shelfbreak and shelf along the Beaufort coast.展开更多
Constitutive modeling is crucial for engineering design and simulations to accurately describe material behavior.However,traditional phenomenological models often struggle to capture the complexities of real materials...Constitutive modeling is crucial for engineering design and simulations to accurately describe material behavior.However,traditional phenomenological models often struggle to capture the complexities of real materials under varying stress conditions due to their fixed forms and limited parameters.While recent advances in deep learning have addressed some limitations of classical models,purely data-driven methods tend to require large data sets,lack interpretability,and struggle to generalize beyond their training data.To tackle these issues,we introduce“Fusion-based Constitutive model(FuCe):Toward model-data augmentation in constitutive modeling.”This approach combines established phenomenological models with an Input Convex Neural Network architecture,designed to train on the limited and noisy force-displacement data typically available in practical applications.The hybrid model inherently adheres to necessary constitutive conditions.During inference,Monte Carlo dropout is employed to generate Bayesian predictions,providing mean values and confidence intervals that quantify uncertainty.We demonstrate the model's effectiveness by learning two isotropic constitutive models and one anisotropic model with a single fiber direction,across six different stress states.The framework's applicability is also showcased in finite element simulations across three geometries of varying complexities.Our results highlight the framework's superior extrapolation capabilities,even when trained on limited and noisy data,delivering accurate and physically meaningful predictions across all numerical examples.展开更多
Current research on robot calibration can be roughly classified into two categories,and both of them have certain inherent limitations.Model-based methods are difficult to model and compensate the pose errors arising ...Current research on robot calibration can be roughly classified into two categories,and both of them have certain inherent limitations.Model-based methods are difficult to model and compensate the pose errors arising from configuration-dependent geometric and non-geometric source errors,whereas the accuracy of data-driven methods depends on a large amount of measurement data.Using a 5-DOF(degrees of freedom)hybrid machining robot as an exemplar,this study presents a model data-driven approach for the calibration of robotic manipulators.An f-DOF realistic robot containing various source errors is visualized as a 6-DOF fictitious robot having error-free parameters,but erroneous actuated/virtual joint motions.The calibration process essentially involves four steps:(1)formulating the linear map relating the pose error twist to the joint motion errors,(2)parameterizing the joint motion errors using second-order polynomials in terms of nominal actuated joint variables,(3)identifying the polynomial coefficients using the weighted least squares plus principal component analysis,and(4)compensating the compensable pose errors by updating the nominal actuated joint variables.The merit of this approach is that it enables compensation of the pose errors caused by configuration-dependent geometric and non-geometric source errors using finite measurement configurations.Experimental studies on a prototype machine illustrate the effectiveness of the proposed approach.展开更多
文摘By relying on the major projects of Beijing-Zhangjiakou HSR,China's railway industry has integrated and applied several digital intelligence technologies to form a complete set of intelligent HSR 1.0 technology,which has been popularized and applied to several new lines.With the continuous deepening of intelligent applications in construction,equipment,operation and other fields,there is an increasingly urgent need for integrated sharing and analysis of models and data.The paper analyzes the connotation and key points of model-data integration and puts forward the overall architecture of model-data integration platform composed of model-data convergence tier,model-data storage tier,model-data management tier,model-data calculation tier,model data aggregation tier,etc.Moreover,it looks forward to the prospect of leading key technologies and multiple innovative key technologies such as intelligent engineering survey and generative design,all-discipline intelligent construction,digital twin of railway engineering,CR450 intelligent EMU,new generation of dedicated mobile communication for railway (5G-R), operation safety protection based on multi-source information perception, transmission and fusion analysis, displacement-based full-travel intelligent service (MaaS+), intelligent comprehensive dispatching of regional railway network, wheel-rail integrated intelligent maintenance of EMU, etc. It can provide guidance and reference for digital railway construction and intelligent HSR 2.0 scientific & technological breakthroughs.
基金supports from the University of Alaska Costal Marine Institute(CMI) and Minerals Management Service(MMS) and IARC/JAMSTEC Cooperative Agreementsupported by NSF OPP Project ARC-0712673 awarded to Yanling Yu and Hajo Eicken (PIs) and Jia Wang(co-PI).This is GLERL Contribution No.1497
文摘A 3.8-kin Coupled Ice-Ocean Model （C1OM） was implemented to successfully reproduce many observed phenomena in the Beaufort and Chukchi seas, including the Bering-inflow-originated coastal current that splits into three branches： Alaska Coastal Water （ACW） , Central Channel, and Herald Valley branches. Other modeled phenomena include the Beaufort Slope Current （BSC） , the Beaufort Gyre, the East Siberian Current （ ESC）, mesoscale eddies, seasonal landfast ice, sea ice ridging, shear, and deformation. Many of these downscaling processes can only be captured by using a high-resolution CIOM, nested in a global climate model. The seasonal cycles for sea ice concentration, thickness, velocity, and other variables are well reproduced with Solid validation by satellite measurements. The seasonal cycles for upper ocean dynamics and thermodynamics are also well reproduced, which include the formation of the cold saline layer due to the injection of salt during sea ice formation, the BSC, and the subsurface upwelling in winter that brings up warm, even more saline Atlantic Water along the shelfbreak and shelf along the Beaufort coast.
基金Anusandhan National Research Foundation(ANRF)via grant no.CRG/2023/007667 and from the Ministry of Port,Shipping,and Waterways via letter no.ST-14011/74/MT(356529).
文摘Constitutive modeling is crucial for engineering design and simulations to accurately describe material behavior.However,traditional phenomenological models often struggle to capture the complexities of real materials under varying stress conditions due to their fixed forms and limited parameters.While recent advances in deep learning have addressed some limitations of classical models,purely data-driven methods tend to require large data sets,lack interpretability,and struggle to generalize beyond their training data.To tackle these issues,we introduce“Fusion-based Constitutive model(FuCe):Toward model-data augmentation in constitutive modeling.”This approach combines established phenomenological models with an Input Convex Neural Network architecture,designed to train on the limited and noisy force-displacement data typically available in practical applications.The hybrid model inherently adheres to necessary constitutive conditions.During inference,Monte Carlo dropout is employed to generate Bayesian predictions,providing mean values and confidence intervals that quantify uncertainty.We demonstrate the model's effectiveness by learning two isotropic constitutive models and one anisotropic model with a single fiber direction,across six different stress states.The framework's applicability is also showcased in finite element simulations across three geometries of varying complexities.Our results highlight the framework's superior extrapolation capabilities,even when trained on limited and noisy data,delivering accurate and physically meaningful predictions across all numerical examples.
基金Supported by National Natural Science Foundation of China(Grant Nos.52325501,U24B2047).
文摘Current research on robot calibration can be roughly classified into two categories,and both of them have certain inherent limitations.Model-based methods are difficult to model and compensate the pose errors arising from configuration-dependent geometric and non-geometric source errors,whereas the accuracy of data-driven methods depends on a large amount of measurement data.Using a 5-DOF(degrees of freedom)hybrid machining robot as an exemplar,this study presents a model data-driven approach for the calibration of robotic manipulators.An f-DOF realistic robot containing various source errors is visualized as a 6-DOF fictitious robot having error-free parameters,but erroneous actuated/virtual joint motions.The calibration process essentially involves four steps:(1)formulating the linear map relating the pose error twist to the joint motion errors,(2)parameterizing the joint motion errors using second-order polynomials in terms of nominal actuated joint variables,(3)identifying the polynomial coefficients using the weighted least squares plus principal component analysis,and(4)compensating the compensable pose errors by updating the nominal actuated joint variables.The merit of this approach is that it enables compensation of the pose errors caused by configuration-dependent geometric and non-geometric source errors using finite measurement configurations.Experimental studies on a prototype machine illustrate the effectiveness of the proposed approach.
