A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant representations without atom-centered message passing. Allegro obtains improvements over state-of-the-art methods on QM9 and revMD17. A single tensor product layer outperforms existing deep MPNNs and transformers on QM9. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular simulations using Allegro recover structural and kinetic properties of an amorphous electrolyte in excellent agreement with ab-initio simulations. Finally, we demonstrate parallelization with a simulation of 100 million atoms.
This literature survey discusses the state-of-the-art in research on why out-of-round railway wheels are developed and on the damage they cause to track and vehicle components. Although the term out-of-round wheels can be attributed to a large spectrum of different wheel defects, the focus here is on out-of-round wheels with long wavelengths, such as the so-called polygonalization with 1±5 harmonics (wavelengths) around the wheel circumference. Topics dealt with in the survey include experimental detection of wheel/rail impact loads, mathematical models to predict the development and consequences of out-of-round wheels, criteria for removal of out-of-round wheels and suggestions on how to reduce the development of out-of-round wheels.
The influence of different types of railway wheel out-of-roundness (OOR) on the vertical dynamic wheel-rail contact force and track response is investigated through extensive field tests and numerical simulations. The response from a freight train, provided with a number of different types of severe wheel tread damage, is studied. Two different axle loads are used in combination with different train speeds in the range 30-100km/h. The wheel defects are wheelflats, local spalls due to rolling contact fatigue cracking, long local defects and polygonal wheels (periodic OOR). The vertical wheel-rail contact force was measured using a strain gauge based wheel impact load detector. Strain gauges and accelerometers were positioned on rails and sleepers to measure the track response. Most of the magnitudes of measured impact forces were found to be lower than the current impact load limit that is used in Sweden to determine when a defective wheel should be removed for repair. Only the long local defect caused larger force magnitudes than the wheel removal criterion. Measured responses are used to calibrate and validate numerical models for simulation of train-track interaction. Results from one linear and one state-dependent track model are compared.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.