The Data Centric Architecture of a Factory Digital Twin
technological details . An FDT largely avoids the underlying manufacturing physics by capturing overall behavior using the Resource , Task , Network ( see below ). A DT can be used to improve process steps identified by the FDT as being critical to overall productivity . In that way FDT and DT form a hierarchy . Similarly , related FDT can be connected to form a supply chain digital twin . This paper focuses on FDT .
An FDT that faithfully reproduces process wide behavior generates value by improving ( 1 ) debottlenecking – identifying and reducing process inefficiencies prior to implementation , ( 2 ) engineering design / retrofit analysis – reduce or eliminate potential flaws using FDT results before physical implementation , ( 3 ) facility fit – use FDTs to determine at which facility a new product should be manufactured , ( 4 ) scheduling / planning ( Forstrom et al 2023 [ 17 ]). Beyond a simulation of the physics of the facility , FDTs reflect operational strategy and facilitate converting business goals into process operation decisions . An FDT designed to optimize profitability or throughput determines the details of process operation to accomplish the objective . FDTs are evolving from advancing computer , modeling , simulation , artificial intelligence , and optimization technology to offer substantial practical value . They have the potential to spread quickly through industry because their value is proportional to the scale of the real-world facility but are inexpensive to develop .
An FDT can be built in an incremental fashion starting with data from a variety of sources . For existing facilities these could include historical data contained in databases , spreadsheets , or from daily operational experience . For new or retrofit / existing projects , a typical source is the requisite engineering designs . An FDT can be iteratively refined by comparing predicted behavior with real world experience and improving the data needed to make them align . This tends to focus FDT data effort on improving the accuracy of data in a goal-oriented fashion . As such , FDT development often results in significant learning about which data is most important and warrants additional effort to improve accuracy .
In addition to process data , and like all digital twins , an FDT uses both historical and real-time data to replicate and optimize a physical thing , in this case the manufacturing facility . At a minimum an FDT requires a final product demand scenario ( list of material , amount , due date ) to predict behavior , and in real-time applications , requires facility state information to set FDT initial conditions such as inventory levels and in-progress activities . Below , we summarize the general Resource Task Network ( RTN 1 ) framework that structure the data used to develop FDTs . While the end user of an FDT does not directly interact with an RTN , its role as a framework is central to FDT architecture and is the life blood of the FDT .
Practical FDTs are designed to answer one time and / or regularly occurring questions of interest . An FDT supporting design of a grass roots / new facility will answer questions of interest concerning initial major equipment to be purchased and process capabilities , staffing goals ,
1
RTN , https :// en-academic . com / dic . nsf / enwiki / 4485962 ? form = MG0AV3 Journal of Innovation 99