Selective area electron diffraction(SAED)patterns can provide valuable insight into the structure of a material.However,the manual identification of collected patterns can be a significant bottleneck in the overall ph...Selective area electron diffraction(SAED)patterns can provide valuable insight into the structure of a material.However,the manual identification of collected patterns can be a significant bottleneck in the overall phase classification workflow.In this work,we utilize the recent advances in computer vision and machine learning(ML)to automate the indexing of SAED patterns.The performance of six different ML algorithms is demonstrated using metallic plutonium-zirconium alloys.The most successful approach trained a neural network(NN)to make a classification of the phase and zone axis,and then utilized a second NN to synthesize multiple independent predictions of different tilts in a single sample to make an overall phase identification.The results demonstrate that automated SAED phase identification using ML is a viable route to accelerate materials characterization.展开更多
Three dimensional free-decaying MHD turbulence is simulated by lattice Boltzmann methods on a spatial grid of 80003 for low and high magnetic Prandtl number.It is verified that∇·B=0 is automatically maintained to...Three dimensional free-decaying MHD turbulence is simulated by lattice Boltzmann methods on a spatial grid of 80003 for low and high magnetic Prandtl number.It is verified that∇·B=0 is automatically maintained to machine accuracy throughout the simulation.Isosurfaces of vorticity and current show the persistence of many large scale structures(both magnetic and velocity)for long times—unlike the velocity isosurfaces of Navier-Stokes turbulence.展开更多
基金The funding for this work was provided by the U.S.Department of Energy,Office of Nuclear Energy Contract DEAC07-051D14517The CNN work was partially supported by the National Science Foundation(award number 1552716).
文摘Selective area electron diffraction(SAED)patterns can provide valuable insight into the structure of a material.However,the manual identification of collected patterns can be a significant bottleneck in the overall phase classification workflow.In this work,we utilize the recent advances in computer vision and machine learning(ML)to automate the indexing of SAED patterns.The performance of six different ML algorithms is demonstrated using metallic plutonium-zirconium alloys.The most successful approach trained a neural network(NN)to make a classification of the phase and zone axis,and then utilized a second NN to synthesize multiple independent predictions of different tilts in a single sample to make an overall phase identification.The results demonstrate that automated SAED phase identification using ML is a viable route to accelerate materials characterization.
基金The authors were supported by grants from DoE,AFOSR and AFRL as well as the Director,Office of Science,Office of Advanced Scientific Computing Research,Department of Energy under Contract No.DE-AC02-05CH11231.
文摘Three dimensional free-decaying MHD turbulence is simulated by lattice Boltzmann methods on a spatial grid of 80003 for low and high magnetic Prandtl number.It is verified that∇·B=0 is automatically maintained to machine accuracy throughout the simulation.Isosurfaces of vorticity and current show the persistence of many large scale structures(both magnetic and velocity)for long times—unlike the velocity isosurfaces of Navier-Stokes turbulence.
