With the improvement of multisource information sensing and data acquisition capabilities inside tunnels,the availability of multimodal data in tunnel engineering has significantly increased.However,due to structural ...With the improvement of multisource information sensing and data acquisition capabilities inside tunnels,the availability of multimodal data in tunnel engineering has significantly increased.However,due to structural differences in multimodal data,traditional intelligent advanced geological prediction models have limited capacity for data fusion.Furthermore,the lack of pre-trained models makes it difficult for neural networks trained from scratch to deeply explore the features of multimodal data.To address these challenges,we utilize the fusion capability of knowledge graph for multimodal data and the pre-trained knowledge of large language models(LLMs)to establish an intelligent advanced geological prediction model(GeoPredict-LLM).First,we develop an advanced geological prediction ontology model,forming a knowledge graph database.Using knowledge graph embeddings,multisource and multimodal data are transformed into low-dimensional vectors with a unified structure.Secondly,pre-trained LLMs,through reprogramming,reconstruct these low-dimensional vectors,imparting linguistic characteristics to the data.This transformation effectively reframes the complex task of advanced geological prediction as a"language-based"problem,enabling the model to approach the task from a linguistic perspective.Moreover,we propose the prompt-as-prefix method,which enables output generation,while freezing the core of the LLM,thereby significantly reduces the number of training parameters.Finally,evaluations show that compared to neural network models without pre-trained models,GeoPredict-LLM significantly improves prediction accuracy.It is worth noting that as long as a knowledge graph database can be established,GeoPredict-LLM can be adapted to multimodal data mining tasks with minimal modifications.展开更多
Understanding cesium(Cs)transport in TRistructural ISOtropic(TRISO)particle fuel is crucial for predicting fission product release in high-temperature reactors.However,current challenges include significant scatter in...Understanding cesium(Cs)transport in TRistructural ISOtropic(TRISO)particle fuel is crucial for predicting fission product release in high-temperature reactors.However,current challenges include significant scatter in diffusivity data and unexplained temperature-dependent diffusion regimes in the silicon carbide layer.This study addresses these challenges by developing a multiscale,mechanistic Cs transport model integrating atomistic simulations and phase field modeling.Our model quantifies temperature and grain size effects on Cs diffusivity,attributing experimentally observed regimes to a transition from bulk-dominated diffusivity at high temperatures to grain boundary-dominated diffusivity at lower temperatures.The model,validated against diffusion measurements and advanced gas reactor(AGR)-1 and AGR-2 post-irradiation fission product release data,enhances the predictive capability of the BISON fuel performance code.This study advances our understanding of Cs release from TRISO particles and its dependence on temperature and silicon carbide grain size,with implications for the safety and efficiency of high-temperature nuclear reactors.展开更多
This study guides the discovery of sustainable high-performance polymer membranes for organic binary solvent separations.We focus on solvent diffusivity in polymers,a key factor in quantifying solvent transport.Tradit...This study guides the discovery of sustainable high-performance polymer membranes for organic binary solvent separations.We focus on solvent diffusivity in polymers,a key factor in quantifying solvent transport.Traditional experimental and computational methods for determining diffusivity are time-and resource-intensive,while current machine learning(ML)models often lack accuracy outside their training domains.To overcome this,we fuse experimental and simulated diffusivity data to train physics-enforced multi-task ML models,achieving more robust predictions in unseen chemical spaces and outperforming single-task models in data-limited scenarios.Next,we address the challenge of identifying optimal membranes for a model toluene-heptane separation,identifying polyvinyl chloride(PVC)as the optimal membrane among 13,000 polymers,consistent with literature findings,thereby validating our methodology.Expanding our search,we screen 1 million publicly available and 7 million chemically recyclable polymers,identifying greener halogen-free alternatives to PVC.This capability is expected to advance membrane design for solvent separations.展开更多
基金the National Natural Science Foundation of China(Grant Nos.52279103 and 52379103)。
文摘With the improvement of multisource information sensing and data acquisition capabilities inside tunnels,the availability of multimodal data in tunnel engineering has significantly increased.However,due to structural differences in multimodal data,traditional intelligent advanced geological prediction models have limited capacity for data fusion.Furthermore,the lack of pre-trained models makes it difficult for neural networks trained from scratch to deeply explore the features of multimodal data.To address these challenges,we utilize the fusion capability of knowledge graph for multimodal data and the pre-trained knowledge of large language models(LLMs)to establish an intelligent advanced geological prediction model(GeoPredict-LLM).First,we develop an advanced geological prediction ontology model,forming a knowledge graph database.Using knowledge graph embeddings,multisource and multimodal data are transformed into low-dimensional vectors with a unified structure.Secondly,pre-trained LLMs,through reprogramming,reconstruct these low-dimensional vectors,imparting linguistic characteristics to the data.This transformation effectively reframes the complex task of advanced geological prediction as a"language-based"problem,enabling the model to approach the task from a linguistic perspective.Moreover,we propose the prompt-as-prefix method,which enables output generation,while freezing the core of the LLM,thereby significantly reduces the number of training parameters.Finally,evaluations show that compared to neural network models without pre-trained models,GeoPredict-LLM significantly improves prediction accuracy.It is worth noting that as long as a knowledge graph database can be established,GeoPredict-LLM can be adapted to multimodal data mining tasks with minimal modifications.
基金provided by the Nuclear Energy Advanced Modeling and Simulation(NEAMS)program.This report was authored by a contractor of the U.S.Government under contract DE-AC07-05ID14517the U.S.Government retains a non-exclusive,royalty-free license to publish or reproduce the published form of this report,or allow others to do so,for U.S.Government purposes.This research made use of the resources of the High Performance Computing Center at Idaho National Laboratory(INL),which is supported by the DOE Office of Nuclear Energy and the Nuclear Science User Facilities under contract no.DE-AC07-05ID14517.The authors would like to recognize Paul Demkowicz at INL for the fruitful discussions and invaluable feedback he provided to improve the manuscript.
文摘Understanding cesium(Cs)transport in TRistructural ISOtropic(TRISO)particle fuel is crucial for predicting fission product release in high-temperature reactors.However,current challenges include significant scatter in diffusivity data and unexplained temperature-dependent diffusion regimes in the silicon carbide layer.This study addresses these challenges by developing a multiscale,mechanistic Cs transport model integrating atomistic simulations and phase field modeling.Our model quantifies temperature and grain size effects on Cs diffusivity,attributing experimentally observed regimes to a transition from bulk-dominated diffusivity at high temperatures to grain boundary-dominated diffusivity at lower temperatures.The model,validated against diffusion measurements and advanced gas reactor(AGR)-1 and AGR-2 post-irradiation fission product release data,enhances the predictive capability of the BISON fuel performance code.This study advances our understanding of Cs release from TRISO particles and its dependence on temperature and silicon carbide grain size,with implications for the safety and efficiency of high-temperature nuclear reactors.
基金the Office of Naval Research through a multidisciplinary university research initiative(MURI)for their funding support,We would also like to acknowledge Dr.Kuan-Hsuan Shen for their valuable support in building the simulation pipeline.We also extend a thank you to Dr.Lihua Chen for her guidance in the initial stage of the work.This research is supported in part through research cyber-infrastructure resources and services provided by the Partnership for an Advanced Computing Environment(PACE)at the Georgia Institute of Technology and XSEDE/ACCESS for computational support through Grant No.TG-DMR080058N.
文摘This study guides the discovery of sustainable high-performance polymer membranes for organic binary solvent separations.We focus on solvent diffusivity in polymers,a key factor in quantifying solvent transport.Traditional experimental and computational methods for determining diffusivity are time-and resource-intensive,while current machine learning(ML)models often lack accuracy outside their training domains.To overcome this,we fuse experimental and simulated diffusivity data to train physics-enforced multi-task ML models,achieving more robust predictions in unseen chemical spaces and outperforming single-task models in data-limited scenarios.Next,we address the challenge of identifying optimal membranes for a model toluene-heptane separation,identifying polyvinyl chloride(PVC)as the optimal membrane among 13,000 polymers,consistent with literature findings,thereby validating our methodology.Expanding our search,we screen 1 million publicly available and 7 million chemically recyclable polymers,identifying greener halogen-free alternatives to PVC.This capability is expected to advance membrane design for solvent separations.
