Conventional fracture mechanics asserts that the relevant physics governing small crack growth occurs near the crack front.However,for fatigue,computing these physics for each crack-growth increment over the entire mi...Conventional fracture mechanics asserts that the relevant physics governing small crack growth occurs near the crack front.However,for fatigue,computing these physics for each crack-growth increment over the entire microstructurally small crack regime is computationally intractable.Properly trained deep-learning surrogate models canmassively accelerate fatigue crack-growth predictions by virtually propagating an initial crack using micromechanical fields corresponding to just the initially cracked microstructure.As the predicted crack front advances,however,the fields no longer reflect relevant near-crack-front physics,leading to error and uncertainty accumulation.To address this,we present an interleaved physics-based deep-learning(PBDL)framework,where updates to the crack representation in the physics-based model are triggered intermittently using model uncertainty,thereby updating micromechanical fields passed to the deep-learning model.We show that this framework,representing a novel cycle-jumping approach,effectively limits error accumulation in history-dependent fatigue crack evolution and forms a template for other time-series applications in materials.展开更多
The nonlinear aeroelastic response of a two-degree-of-freedom airfoil with freeplay and cubic nonlinearities in supersonic flows is investigated. The second-order piston theory is used to analyze a double-wedge airfoi...The nonlinear aeroelastic response of a two-degree-of-freedom airfoil with freeplay and cubic nonlinearities in supersonic flows is investigated. The second-order piston theory is used to analyze a double-wedge airfoil. Then, the fold bifurcation and the amplitude jump phenomenon are detected by the averaging method and the multi-variable Floquet theory. The analyticall results are further verified by numerical simulations. Finally, the influence of the freeplay parameters on the aeroelastic response is analyzed in detail.展开更多
This paper proposes a novel fault diagnosis method by fusing the information from multi-sensor signals to improve the reliability of the conventional vibration-based wind turbine drivetrain gearbox fault diagnosis met...This paper proposes a novel fault diagnosis method by fusing the information from multi-sensor signals to improve the reliability of the conventional vibration-based wind turbine drivetrain gearbox fault diagnosis methods.The method fully extracts fault features for variable speed,insufficient samples,and strong noise scenarios that may occur in the actual operation of a wind turbine planetary gearbox.First,multiple sensor signals are added to the diagnostic model,and multiple stacked denoising auto-encoders are designed and improved to extract the fault information.Then,a cycle reservoir with regular jumps is introduced to fuse multidimensional fault information and output diagnostic results in response to the insufficient ability to process fused information by the conventional Softmax classifier.In addition,the competitive swarm optimizer algorithm is introduced to address the challenge of obtaining the optimal combination of parameters in the network.Finally,the validation results show that the proposed method can increase fault diagnostic accuracy and improve robustness.展开更多
基金supported by the National Science Foundation under Grant No. CMMI-1752400. The authors would like to acknowledge the computational resources provided by the University of Utah's Center for High Performance Computing.
文摘Conventional fracture mechanics asserts that the relevant physics governing small crack growth occurs near the crack front.However,for fatigue,computing these physics for each crack-growth increment over the entire microstructurally small crack regime is computationally intractable.Properly trained deep-learning surrogate models canmassively accelerate fatigue crack-growth predictions by virtually propagating an initial crack using micromechanical fields corresponding to just the initially cracked microstructure.As the predicted crack front advances,however,the fields no longer reflect relevant near-crack-front physics,leading to error and uncertainty accumulation.To address this,we present an interleaved physics-based deep-learning(PBDL)framework,where updates to the crack representation in the physics-based model are triggered intermittently using model uncertainty,thereby updating micromechanical fields passed to the deep-learning model.We show that this framework,representing a novel cycle-jumping approach,effectively limits error accumulation in history-dependent fatigue crack evolution and forms a template for other time-series applications in materials.
文摘The nonlinear aeroelastic response of a two-degree-of-freedom airfoil with freeplay and cubic nonlinearities in supersonic flows is investigated. The second-order piston theory is used to analyze a double-wedge airfoil. Then, the fold bifurcation and the amplitude jump phenomenon are detected by the averaging method and the multi-variable Floquet theory. The analyticall results are further verified by numerical simulations. Finally, the influence of the freeplay parameters on the aeroelastic response is analyzed in detail.
基金supported by the Shanghai Rising-Star Program(No.21QC1400200)the Natural Science Foundation of Shanghai(No.21ZR1425400)the National Natural Science Foundation of China(No.52377111).
文摘This paper proposes a novel fault diagnosis method by fusing the information from multi-sensor signals to improve the reliability of the conventional vibration-based wind turbine drivetrain gearbox fault diagnosis methods.The method fully extracts fault features for variable speed,insufficient samples,and strong noise scenarios that may occur in the actual operation of a wind turbine planetary gearbox.First,multiple sensor signals are added to the diagnostic model,and multiple stacked denoising auto-encoders are designed and improved to extract the fault information.Then,a cycle reservoir with regular jumps is introduced to fuse multidimensional fault information and output diagnostic results in response to the insufficient ability to process fused information by the conventional Softmax classifier.In addition,the competitive swarm optimizer algorithm is introduced to address the challenge of obtaining the optimal combination of parameters in the network.Finally,the validation results show that the proposed method can increase fault diagnostic accuracy and improve robustness.
