Creating digital twins at scale

Abstract

Described are methods and systems for calibrating simulation models to generate digital twins for physical entities. In some embodiments, a method includes receiving a plurality of datasets for a plurality of corresponding physical entities. A calibration request is enqueued to a calibration requests queue for each received dataset and includes information indicating a dataset and a corresponding physical entity. A plurality of calibration engines and a plurality of corresponding simulation clusters for generating a plurality of calibration results for a plurality of calibration requests dequeued from the calibration requests queue can be deployed. Each calibration result is enqueued to a calibration results queue as the plurality of calibration engines generates the calibration result and a plurality of calibration results dequeued from the calibration results queue in association with the plurality of corresponding physical entities can be stored as information used to generate a plurality of corresponding digital twins.

What is claimed is: 
     
         1 . A method for calibrating simulation models to generate digital twins for physical entities, comprising:
 receiving a plurality of datasets for a plurality of corresponding physical entities;   enqueueing a calibration request to a calibration requests queue for each received dataset, the calibration request comprising information indicating a dataset and a physical entity corresponding to the dataset;   deploying a plurality of calibration engines and a plurality of corresponding simulation clusters for generating a plurality of calibration results for a plurality of calibration requests dequeued from the calibration requests queue, wherein a calibration result corresponding to the calibration request comprises a set of values selected for parameters of a simulation model based on the dataset of the physical entity;   enqueuing each calibration result to a calibration results queue as the plurality of calibration engines generates the calibration result; and   storing a plurality of calibration results dequeued from the calibration results queue in association with the plurality of corresponding physical entities as information used to generate a plurality of corresponding digital twins.   
     
     
         2 . The method of  claim 1 , wherein receiving the plurality of datasets comprises:
 receiving an input data file comprising the plurality of datasets; and   parsing the input data file to identify the plurality of physical entities and the plurality of corresponding datasets.   
     
     
         3 . The method of  claim 1 , wherein the dataset for the physical entity comprises data derived from the physical entity. 
     
     
         4 . The method of  claim 1 , wherein the dataset for the physical entity for the physical entity comprises a calibration setting for a parameter to be calibrated for a simulation model of the physical entity, wherein the calibration setting comprises a specific range of values for the parameter. 
     
     
         5 . The method of  claim 1 , wherein deploying the plurality of calibration engines and the plurality of corresponding simulation clusters comprises deploying a calibration engine and a corresponding simulation cluster by:
 allocating a plurality of computing resources on a cloud platform; and   spinning up one or more containers on each allocated computing resource to deploy one or more corresponding model simulators in the simulation cluster.   
     
     
         6 . The method of  claim 5 , comprising:
 spinning up a plurality of containers based on a simulation application image to concurrently run a plurality of respective model simulations for the plurality of calibration requests.   
     
     
         7 . The method of  claim 1 , wherein generating the calibration result for the calibration request dequeued from the calibration requests queue comprises:
 generating a plurality of simulation requests for the physical entity based on the dataset of the physical entity corresponding to the dequeued calibration request, wherein each simulation request comprises a set of values selected for the parameters of the simulation model; and   enqueuing each generated simulation request from the plurality of simulation requests onto a simulation requests queue.   
     
     
         8 . The method of  claim 7 , comprising:
 dequeuing a simulation request from the simulation requests queue to an available model simulator in the simulation cluster to run the simulation model with the set of values included in the simulation request;   running the simulation model configured according to the set of values to generate simulation outputs; and   enqueueing a simulation result comprising the simulation outputs to a simulation results queue.   
     
     
         9 . The method of  claim 7 , comprising:
 evaluating one or more simulation results dequeued from a simulation results queue to determine whether to iteratively generate and enqueue another simulation request to the simulation requests queue to be processed by the simulation cluster.   
     
     
         10 . The method of  claim 9 , wherein evaluating a first simulation request of the one or more simulation requests comprises:
 comparing simulation outputs in the first simulation request with training data from the dataset corresponding to the physical entity to compute a fitness score representing how closely the output matches the training data.   
     
     
         11 . The method of  claim 1 , wherein generating the calibration result for the calibration request dequeued from the calibration requests queue comprises:
 monitoring available computing and memory resources on a cloud computing platform;   when the available computing and memory resources permit, dequeuing a next calibration request from the calibration requests queue to a calibration engine and a corresponding simulation cluster deployed to process the next calibration request.   
     
     
         12 . The method of  claim 1 , wherein enqueueing each calibration result to the calibration results queue comprises:
 ramping down computing resources used to generate the enqueued calibration result.   
     
     
         13 . The method of  claim 12 , wherein ramping down the computing resources comprises:
 terminating a plurality of containers spun up in the simulation cluster to generate the calibration result.   
     
     
         14 . A system for calibrating simulation models to generate digital twins for physical entities, comprising:
 a calibration orchestrator configured to: 
 receive a plurality of datasets for a plurality of corresponding physical entities, and 
 enqueue a calibration request to a calibration requests queue for each received dataset, the calibration request comprising information indicating a dataset and a physical entity corresponding to the dataset; and 
   a calibration system configured to: 
 deploy a plurality of calibration engines and a plurality of corresponding simulation clusters for generating a plurality of calibration results for a plurality of calibration requests dequeued from the calibration requests queue, wherein a calibration result corresponding to the calibration request comprises a set of values selected for parameters of a simulation model based on the dataset of the physical entity, and  
 enqueue each calibration result to a calibration results queue as the plurality of calibration engines generates the calibration result; and 
   wherein the calibration orchestrator is configured to store a plurality of calibration results dequeued from the calibration results queue in association with the plurality of corresponding physical entities as information used to generate a plurality of corresponding digital twins.   
     
     
         15 . The system of  claim 14 , wherein to receive the plurality of datasets, the calibration orchestrator is configured to:
 receive an input data file comprising the plurality of datasets; and   parse the input data file to identify the plurality of physical entities and the plurality of corresponding datasets.   
     
     
         16 . The system of  claim 14 , wherein the dataset for the physical entity comprises data derived from the physical entity. 
     
     
         17 . The system of  claim 14 , wherein the dataset for the physical entity for the physical entity comprises a calibration setting for a parameter to be calibrated for a simulation model of the physical entity, wherein the calibration setting comprises a specific range of values for the parameter. 
     
     
         18 . The system of  claim 14 , wherein to deploy the plurality of calibration engines and the plurality of corresponding simulation clusters, the calibration orchestrator is configured to deploy a calibration engine and a corresponding simulation cluster by:
 allocating a plurality of computing resources on a cloud platform; and   spinning up one or more containers on each allocated computing resource to deploy one or more corresponding model simulators in the simulation cluster.   
     
     
         19 . The system of  claim 18 , wherein the calibration orchestrator is configured to:
 spin up a plurality of containers based on a simulation application image to concurrently run a plurality of respective model simulations for the plurality of calibration requests.   
     
     
         20 . The system of  claim 14 , wherein generating the calibration result for the calibration request dequeued from the calibration requests queue comprises: 
 generating a plurality of simulation requests for the physical entity based on the dataset of the physical entity corresponding to the dequeued calibration request, wherein each simulation request comprises a set of values selected for the parameters of the simulation model; and   enqueuing each generated simulation request from the plurality of simulation requests onto a simulation requests queue.   
     
     
         21 . The system of  claim 20 , wherein the calibration orchestrator is configured to:
 dequeue a simulation request from the simulation requests queue to an available model simulator in the simulation cluster to run the simulation model with the set of values included in the simulation request;   run the simulation model configured according to the set of values to generate simulation outputs; and   enqueue a simulation result comprising the simulation outputs to a simulation results queue.   
     
     
         22 . The system of  claim 20 , wherein the calibration orchestrator is configured to:
 evaluate one or more simulation results dequeued from a simulation results queue to determine whether to iteratively generate and enqueue another simulation request to the simulation requests queue to be processed by the simulation cluster.   
     
     
         23 . The system of  claim 22 , wherein to evaluate a first simulation request of the one or more simulation requests, the calibration orchestrator is configured to:
 compare simulation outputs in the first simulation request with training data from the dataset corresponding to the physical entity to compute a fitness score representing how closely the output matches the training data.   
     
     
         24 . The system of  claim 14 , wherein to generate the calibration result for the calibration request dequeued from the calibration requests queue, the calibration orchestrator is configured to:
 monitor available computing and memory resources on a cloud computing platform;   when the available computing and memory resources permit, dequeue a next calibration request from the calibration requests queue to a calibration engine and a corresponding simulation cluster deployed to process the next calibration request.   
     
     
         25 . The system of  claim 14 , wherein to enqueue each calibration result to the calibration results queue, the calibration orchestrator is configured to:
 ramp down computing resources used to generate the enqueued calibration result.   
     
     
         26 . The system of  claim 25 , wherein ramping down the computing resources comprises:
 terminating a plurality of containers spun up in the simulation cluster to generate the calibration result.