Digital Twins

welcome back so we're talking about data science kind of how you can ask and answer questions with data as a primary tool so I think of this as data-driven inquiry or hypothesis design and testing and I want to talk now about how this is being used in complex manufacturing processes so I think you know data science has been transformative and how data intensive fields of science are run like astronomy you know Genome Sciences also a lot of tech companies that are data rich and really data centric or data intensive but increasingly data science is changing how we think about complex engineering and industrial processes so I think about you know manufacturing or design where you have maybe a complex supply chain feeding into a manufacturing process where you put all those pieces together and you get something you know it's much more than the sum of its parts out and so I'm going to talk a little bit about how data science fits into these more industrial manufacturing applications and so this is really related to two efforts one is called digital twin digital twin and the other is called digital thread and these are kind of how how data science is impacting design of complex engineering systems like aircraft okay so and I'll just very very briefly go over this this is like you could have a whole lecture series on digital twin a digital thread this is just the overview the digital twin is a data informed model of your complex system and not just you know in the past we've had lots of models of aerodynamics or you know composite processes we have these kind of local models the digital twin is really an agglomeration a kind of mega model that models all of the different scales and aspects of a larger process like a whole main factoring process this really is supposed to be a digital version of that physical physical entity and what what's really important about these digital twins it's not just a model it's a model that is pinned to measurement data so you have all of this rich data about that process you're collecting all of this data so this is related to things like the Internet of Things where you'll start having sensors and instrumentation kind of ubiquitous instrumentation and that data is going to be coming in and informing these souped-up models on steroids essentially so you can model your whole Factory as a digital twin people are talking about how to model kind of the you know the human body as a digital twin things like that the digital thread is the data pipeline that enables that digital twin so the digital thread means that you're going to be collecting information throughout the entire process from design to sourcing of materials all throughout the supply chain to when those start to be assembled and created into this into this large assembled objects like think of a turbine or an aircraft something you know that's that's many many pieces many people putting their design and engineering you know capabilities to build this thing the digital thread is this digital signature that that tracks all of those parts and all of those pieces and all of those processes all the measurements you take of the system throughout that procedure go into this digital thread and what that allows you to do is first of all ask much much more sophisticated questions about what upstream processes are affecting downstream either issues or quality concerns or things like that and so you have this much clearer feedback between the questions you can ask in this this digital thread okay so I mean I really simple I think if the digital thread is this kind of you know arrow of time and you see kind of these processes as they get pulled in maybe you know all my rivets I follow my rivets the whole time through the process and the composite processes and and you see this kind of complexity in all these inner week moving pieces but everything has a digital signature so you can kind of trace and do surveillance and and supervision of the process and understand like what is most correlated with some downstream thing that I want to either improve or make more reproducible things like that okay so I think this is really exciting you have your your digital twin model your digital thread which is the data pipeline and architecture that facilitates that digital twin and this allows you to do kind of amazing amazing things both now and in the future so I'm gonna oversimplify a complex manufacturing process into a few steps so let's say you have some design and this is complicated multi objective you know there's cost there's time there's you know uncertainty in this entire design process maybe you have your manufacturing process and again this could you know it's one word here this is a whole Factory multiple factories hundreds of factories maybe across the globe working kind of in this this very intricate dance okay and so of course your design and your manufacturing should be inherently linked once you make something it goes into the real world maybe it has to be certified so you know certification or validation testing and then maybe when it actually goes into the real world you collect kind of let's say then it goes out into the real world and you might want to collect data on how it's actually performing in the real world and so of course these are not all done in isolation today but what the digital twin and digital thread will hopefully allow in the future is having much tighter feedback between these different these different core aspects of design and manufacturing certification and then tying that to real-world performance maintenance things like that and maybe even in the design phase you know what's the business model who are you gonna sell these products to what what needs does it fill and so really kind of giving feedback at all of these different levels you know kind of coupling back to the design phase and the manufacturing and and tying these together in a more kind of tightly coupled and holistic view is really the goal of this digital to an end digital thread and I think the way I think about this you might have a lot of historical data that you can use to inform the next well maybe you can improve your current manufacturing your current certification your current designs but maybe you can also use that information so that the next product that's being designed for manufacturability you can kind of incorporate that knowledge into future designs and so so there's a lot of first of all there's a lot to unpack here this is just a very shallow overview of very very deep subjects and these are under continuous development so these are relatively kind of hot ideas lots of people are investing lots of time understanding what is the best way of building digital twins and digital threat architectures to improve these systems so this is really exciting and there's gonna be a lot of development but the idea is that this can massively kind of lift the fog that has made us develop so much in isolation with with with handoff so we can kind of take away some of that fog and make it much clearer how different decisions upstream affect things downstream okay thank you

Channel: Steve Brunton

Views: 41,148

Keywords: Data Science, Machine Learning, Optimization, Regression, Modeling, Neural Networks, Classification, Visualization, Data Analytics, Data, Data-driven science, Big Data, digital twin, digital thread

Id: YqTL8dwYT54

Length: 8min 6sec (486 seconds)

Published: Wed Jun 05 2019