1、Fusion Energy Digital Twins-from Experiment to Power Plant,R.Michael Churchill(PPPL),Fusion researcher,AI/ML,SIM,EXP,Viz by Jamison Daniel,3,Fusion plasmas have a range of phenomena,manifest over multiple time and spatial scales,4,Understanding fusion plasmas requires extracting information from mul
2、tiple diagnostics at these wide ranges of time and spatial scales,Full-wave solvers,Gyrokinetic codes,MHD codes,Transport solvers,What a scientist needs,Natural to focus first on the algorithms(bottom up).But what does a scientist need(top down)?Holistic view of not only current workflow but also fu
3、ndamental needs of scientists can point out opportunities,5,Raw data,Initial processed data,Analyzed data,Insight,Algorithms,analysis,modeling/simulation,Fusion scientists need understandable analysis results quickly,6,Experimental“shot”(1oos seconds),Simple automated analysis(few minutes),Scientist
4、s interpret(10 minutes),Decide and set next shot(few minutes),Fusion data from experiment(and simulation)possess all the big-V properties,7,Integrated Research Initiative(IRI)i.e.SuperFacility workflows for fusion energy,IRI leverages remote compute for science experiment data analysis and simulatio
5、nsFor fusion energy enables faster,better comparison of simulation and experiment on timescales meaningful to the scientistMultiple levels of digital twins can be executed(high-to low-fidelity),8,Churchill,FST 2021,https:/,9,Why aim for high fidelity digital twins?,Example of ITER neutronics calcula
6、tionsPreviously simplified wedge geometry used,mosaic to estimate total neutron transportFull geometry CAD-level neutronics simulation revealed significant difference,mainly due to port structure,10,Juarez,AI based digital twins for engineering operations,AI models for accelerating HEAT code(diverto