PLM Ch 10: Digital Twins, Threads, and The Bicycle Thief
In Search of the Stolen Bike, from Bicycle Thieves

PLM Ch 10: Digital Twins, Threads, and The Bicycle Thief

“Hi Sarah, how was your vacation?”

“Well… it ended much better than it started, but it did not start at all well.”

“What happened?”

“I drove eight hours south to spend a long weekend mountain biking with friends and stopped about halfway there to get some lunch and take a walk. My bike was locked to the rack on the back of the car, and it was stolen! It was in a well-lit parking deck, and I was only gone an hour, but it was gone!”

“That’s not good.”

“It was largely solved by the end, but I think there is a bit of a digital twin story here.”

“Really? How so?”

“Are you familiar with Apple Air Tags? Or the other devices which help you find lost things? They’re passive devices that tell any iPhone which passes by ‘I am here!’ The phones then tell servers at the Apple mothership that the tag was recently seen, and its location. I have one hidden on my bike, which I told the police when I called reported the theft.”

“Interesting… could the police track it?”

“Not directly, but a few days later I helped them with a virtual stakeout. As soon as I realized it was stolen, I looked on my phone and could tell it was only a few miles away. I told the police and they immediately sent two officers to look for it, but it was somewhere within a large apartment complex, and there was no way to find it.


“Not much the police can do there.”

“Right. So, the next day, I was 100 miles away using a rented bike and having breakfast with friends, when a detective texted me to say that the case had been assigned to him and asked me to check if the bike had moved. At that point, I was checking it every hour, and I could tell that it had been ‘seen’ by another iPhone at the same location in the middle of the night.”

Sarah continued, “But two days later I realized that the bike had moved twenty miles away from the apartment complex, so I sent a text to the police detective, telling him where it had been ‘seen’ at 6:30 that morning.”

“And he went and got it?”

“Eventually, but we spent the next four hours with me supporting their stakeout virtually. We did it all via text, let me read it to you.”

8:19am: Hi Detective, my bike moved. Here’s its location two hours ago.

<<27 minutes pass>>

8:46: Okay I’m gonna head over there and see if I can find it. ?

<<19 minutes pass>>

9:05: Call me real quick

Sarah interjected, “Prof, the detective and I talked on the phone. He was at the location on the map and asked me if I could ask the Air Tag’s speaker to make a sound, but at that moment it wasn’t in range of an iPhone, so I couldn’t. By chance, the detective used an iPhone, so if he had been in range my tag could have communicated with his phone, which would have communicated with the mothership, which would have sent the location to me, but the bike wasn’t nearby.”

<<40 minutes pass>>

9:42: Detective, here’s a new map. (It was about five miles from where it was at 6:30.)

9:46: Wow. That’s close to my house. I’m gonna drive around and look.

<<15 minutes>>

10:00: Can you keep try to ping it

10:00: Nothing new. It will ping me when found. If you’re close I can tell it to make a sound, but don’t want them to know that the AirTag is there.

10:00 Not here. Idk if they’re riding it around or what

10:01: Yeah, could be. It’s a bike, right? ??

10:02: I was hoping they would just leave it out back wherever they’re at

10:03: Would be nice. Maybe they’ll park it somewhere other than an 11-story apartment building.


10:03: Wish they would sit still

<<10 minutes pass>>

10:12: Oh, they’re riding it. It was near the Dollar General at 10:01, and the armory at 10:11.

10:12: I’m in the area now if you wanna keep pinging

10:12: Will do. Avidly watching…

<<13 minutes pass>>

10:25: Okay people saw it ride by here not long ago so it’s around here somewhere

10:26: It’s not near an iPhone. Phone refreshes about once per minute. Nothing new in 15 minutes

10:27: It’s pouring the rain here so hopefully he lays up somewhere

10:34: Nothing new in 23 minutes. Probably parked away from an iPhone.

10:37: We’re still over here (At this point, the Detective and two other officers were searching for this bike.)

<<48 minutes pass>>

11:25: FYI. Still nothing new.

<<Another 15 minutes>>

11:40: It’s moving. Looks like it is heading back to where it was this morning. (Below is the bike’s location at 11:40, 12:01, and 12:08)

12:15: We got it

12:15: ??????????????

12:18: Okay I’ll call you with more soon. Still in good condition

….

Prof asked, “you have the bike now?”

“Yes! I do! It’s a little worse for wear, but very rideable. The thief intentionally removed the electronic odometer sensor on the front wheel but didn’t find the hidden AirTag.”

All Models Are Wrong, But Some Are Useful

“So, why is this a Digital Twin story?”

“There are many definitions of a Digital Twin, but I’ll use one which is both easy to say and hard to do:

The fullest expression of a Digital Twin is that any information which might be gained from the physical object can also be acquired from its virtual instance.

Sarah continued, “throughout the experience, the physical bike existed in a place and time, and I was able to virtually follow its location from over 100 miles away. I had the easiest part of that stakeout; I could have breakfast with friends in a comfortable room, while the detective and two cops were out looking for my bike in the rain. You know, between Wednesday and Saturday, there were at least six different police officers looking for that bike.”

“That’s a lot of effort on their part.”

“Yes, and I can only guess on this, but bicycle thievery has gone up in the years since COVID, and I wonder if they were looking for whoever was stealing multiple bikes. Whoever stole my bike was clearly an expert at it; they cut two cable locks and tossed it into a pickup truck in a matter of minutes. If the police were attempting to solve a number of bicycle thefts, the Air Tag on my bike gave them an opportunity to do it.”

“The thief was arrested?”

“They tell me yes, but I don’t know any more than that. I haven’t seen it in the news, so I guess a bicycle thieving ring wasn’t THAT big a deal.”

“But your story doesn’t include any 3D data… can it really be a Digital Twin without 3D?”

“Yes! While the fullest expression of a Twin would include 3D, and lots of other data, just getting occasional pings from the tag to nearby phones about the location of my bike is the reason that I have it back.”

“I agree. There is an aphorism ‘all models are wrong, or at least incomplete, but some are useful’, which gets to your point. Your model of the bike in its environment was incomplete; for example, the sensor didn’t tell you it was raining. But so long as the Air Tag was attached to it, knowing the tag’s location was enough for your needs. Could you provide the serial number for your bike? Did the police ask for that?”

“They asked, and I had it. All bikes from this manufacturer come with an electronic odometer which tracks how far the bike has traveled and connects to a phone app via Bluetooth. When I first purchased the bike, the app recorded its serial number, and I could immediately place it in the police report. But honestly, if it wasn’t automatic, I never would have done it. The various officers I talked to were very happy to know that I could give them the serial number. And thievery aside, whenever the bike gets serviced, I record it into the app, so I know when I need to change my tires or get a new bike chain.”

“There is also a Digital Thread story here. Let’s start walking this story backwards through all the information models involved:

The As- Models

As-Operated: At the moment that your bike was stolen, it had two sensors which monitored some amount of operational data, and if we want to monitor more data, we just need more sensors. This means we need to make decisions in advance about what will need to be monitored. ?Also, design assumptions may not be accurate, and it would be helpful if they can better understand the ‘as-operated’ conditions. Operations are often managed in an IoT (or Internet of Things) data system, and here is a short video by Microsoft regarding their Digital Twin Data Language, which is more about semantic graphs, than about 3D data. And Michael Grieves discusses:

  • Digital Twin Prototypes (the virtual model),
  • Digital Twin Instances (the physical instance of the model) and
  • Digital Twin Aggregates, which is the collection of operational data retrieved from each of the instances.

You can find more info here: Digital Twins and Threads (patrickhillberg.com)

As-Serviced: You also mention that you use the bike’s app to keep track of service events, and in the course of service you may replace components, meaning that the product-in-use is different than the product which was manufactured. Service information is managed in an MRO system.

As-Built: I don’t want to stretch the bicycle metaphor too far, so let’s think about airplanes. When a plane is initially manufactured, as well as when it is serviced, every manufacturing or service activity needs to be recorded. I mentioned in a previous meeting the need to capture how much torque is required for each rivet on the outer skins of the plane. Build information is managed in an MES system, and this data is ‘version zero’ of the digital twin instance data provided to the MRO system for service.

One use-case for as-built models, and by extension digital twins originates from the crash of an airplane in July of 1996. It was a warm evening, the plane sat on the tarmac for a while, and 12 minutes after take-off from JFK airport it exploded just south of Long Island. Because it was in shallow waters, portions of the fuselage could be recovered. Much of the initial thinking was that the plane had been attacked by a bomb or missile, but the investigation eventually concluded that the Boeing 747 fleet had a design lapse regarding heat, fuel vapors, and possible electrical sparks. The NTSB issued 15 safety recommendations, which led to a redesign and reworking of the vulnerable systems on other 747 aircraft.

Had the explosion occurred 20 minutes later, it would not have been possible to recover the plane, as the water would have been too deep, and the safety recommendations would not have been issued to correct existing planes. The FAA and NTSB now require the manufacturers to maintain As-model data so that any aircraft may be rebuilt virtually. See TWA Flight 800 - Wikipedia

As-Procured: Most products, and especially transportation products, are assemblies of components purchased from many different suppliers. How do we order those parts, and what was delivered? The first question opens a very large topic of Bills of Materials (BOM), as we discussed earlier when we discussed tuna sandwiches. (PLM and OCM Ch. 4). The topic then expands into supply-chain management (SCM) and Enterprise Resource Planning (ERP) databases. I won’t delve into that here, but we need to recognize that while the BOM is a useful model, but is only a model, and must integrate with the other As-models. But then can we trace what was delivered? A friend’s company is a Tier-2 auto supplier and manufactures a million small components per day for a Tier-1 supplier of fuel lines. The T1 supplier uses T2 components in sub-assemblies that they supply to top-tier original equipment manufacturers (OEMs). These are companies like GM and Ford, who assemble the final automobile.

At one point a machine in the T2 company was set up incorrectly, and they delivered thousands of bad parts. ?The T1 supplier used the parts but didn’t keep track of where T2 parts were consumed within the T1 subassemblies. T2 tracked their deliveries by lot number, as did T1, but T1 did not track where T2 parts were consumed within T1 products. As an analogy, a ‘peg’ created by T2 is designed to fit into a hole created by T1, but the T2 peg was just a little bit too small, and the problem was not found until the fuel lines were put under pressure during final assembly at the OEM. Thus, T2 knew that they had mis-manufactured a lot of 10,000 parts, and informed T1, but T1 couldn’t trace which of their own subassemblies made use of the flawed T2 parts. In the end, 20,000 T1 parts needed to be scrapped by the OEM, even though only 10,000 of the T2 parts were flawed, because the T1 supplier couldn’t track the parts from T2 to the OEM.

In Chapter 8, we discussed the impact of poor-quality components delivered by suppliers, and this story of OEMs, T1 and T2 suppliers is another example of ‘decomposition creates dysfunction’.

ERP systems are used to purchase large quantities of components (called ‘lots’) from suppliers, while MES systems should record how each component within a lot is assembled into a particular instance. There are many-to-one and one-to-many relationships at each tier, creating a great deal of complexity.

As-Designed: Of course, before the procurement and manufacturing phases is the design phase, and in modern manufacturing companies, this work is done in CAD. But understanding the as-designed model is often complicated by what is called configuration management. Using a bike as an example, a base design is configured to meet specific user needs. You and I might each purchase the same model of bicycle, but yours has off-road tires and 15-gears on the rear wheel, while mine has road tires and 9-gears. Just these two options with two selections each create four possible configurations. Imagine if there only eight options with four selections each, it leads to 4096 possible as-configured models.

As-Required: Each product is designed to conform to a set of requirements, as is each component within the product. The requirements stack can be immense as high-level requirements are decomposed into derived-requirements, and derived-derived requirements, etc. The past decade or so has seen a rise in Model-Based Systems Engineering (MBSE) and the Systems Modeling Language (SysML) to model interactions between conceptual objects which fulfill the requirements. (These are often called 1D models.) There is a short video on how SysML can model the requirements and functions of an air compressor here: Product Development Tools (patrickhillberg.com).

The As-Required, and As-Designed data is most often managed in PLM databases, as are all possible configurations (e.g., the 8 options with 4 selections mentioned above, sometimes called the ‘150% BOM’).

But the specific configuration of your own bicycle would be maintained in the ERP system. In the simple case, the ERP system provides a web page based on the 150% BOM created in the PLM system. As a customer you would choose options, leading up to a configuration, and that configuration would result in assembly instructions so that the manufacturing system would know which options belong on your bicycle.

Manufacturing instructions on a configured product may originate in the PLM system, as we saw in the tuna sandwich example, or they might originate in the ERP system, if the assembly is relatively simple, and can be thought of as a recipe. This can often generate a great deal of debate, based on cultural norms within the organization.

As-Desired: But even before the requirements, there is a list of features that the customer will perceive as valuable to them. Clay Christensen refers to this as the ‘Job to be Done’, and Mike Grieves refers to five dimensions of value perception: Economic, Belief, Safety, Social, and Experiential. You can find lectures on this here: Perception of Value, Jobs to be Done, and Disruption. (patrickhillberg.com) But the customer perception of value may change quickly, while product developers focus on the As-Required rather than the As-Desired. When the As-Desired models shift, when perception of value changes, lots of people in old-line industries lose their jobs.

Business Models: Finally, all the above need to fit into a profitable business model, which balances the needs of the OEM, its suppliers, its retailer, and its customers.

Digital Threads

Prof continued, “in any event, the Air Tag and the Bluetooth odometer provide you with a sparse amount of operational information; they can tell you the location of the bike, and how far it has traveled, and you use the app to maintain as-serviced information. Bicycle models may be configured with components procured from different companies; for example, your bike may come with Shimano components for gears and braking, while mine have SRAM. As the bike’s owner you probably don’t have access to design data, but the manufacturer certainly does, and the pictures you see on their website come from that design data. There is a long list of requirements, some of which you can see on its specifications page, and hopefully these requirements result in something that you desire. Sound good?”

“Right. And if you think of how all this data interrelates, there is a ‘through-line’ which begins with what the customer desires, all the way to how she operates the bike. And that through-line is called a ‘digital thread’.”

“Right. And what we’re discussing is a minimal example of Digital Twins and Threads, based on a particular story about your bicycle. When a company is building dozens of airplanes per year, or thousands of cars per day, there is a lot more data than what we are discussing, and because there is more data the process becomes more complicated, in fact much more complicated, but fundamentally, the story is the same:

Digital Twins provide virtual information of their related physical devices.

Digital Threads align the many layers of product information, keeping the virtual instance up to date with the physical as changes occur.

“You can find more info on this, here: Digital Twins and Threads(patrickhillberg.com)

Operational Data and Security

“Sarah, given all of that, do you have thoughts on how your operational data might be reflected back into the requirements of new products?”

“I was lucky that the thief didn’t find the Air Tag. But what if someone I don’t like attached their Air Tag onto my bike, and were using it to track me? Imagine a home burglar attaching Air Tags to bikes, or cars, or briefcases to know when you have left the house? This situation worked out well for me because I was tracking my own bike. But on a broader lever, I see challenges.”

“I agree, and if you think about how a nefarious person might behave if they have access to Digital Thread data, how might they use it? There are substantial cybersecurity issues in the use of twins and threads. Apple attempts to manage privacy concerns by recognizing when an Air Tag is continually traveling close to an iPhone to which it’s not registered, and the unfamiliar phone will receive messages that it is being tracked. I think that you are fortunate that the thief wasn’t using an iPhone.”

“Yes. But I wonder if electronic tags could be permanently embedded into the bike, by the manufacturer? It doesn’t make sense to steal an iPhone, because with Face ID and the Find My app, an expensive phone is only useful to its owner. Can we do the same with bikes? Can the manufacturer permanently build something into the bike so that it is only valuable to its owner, and never to a thief?”

“All good points. Tracking provides both positive and negative societal value. As it has for decades, technology is bumping into the legal concept of the Expectation of Privacy.”

Prof continued, “But let me give a positive potential use – in the investigation of the Boeing 737 Max crashes, it became clear that aircraft designers and the FAA had adopted a ‘4-second rule’ regarding a pilot’s ability to diagnose and correct a problem, but ergonomic studies of how pilots behave in real and simulated situations indicate that the presumption is not at all true, and yet the rule remained. One investigation posited that all aircraft worldwide may assume this 4-second rule, yet it was not enough to protect those two aircraft, and the entire fleet was grounded. So, instead of a limited set of ergonomic studies, what if all pilot operational data was aggregated to create a more accurate standard? There is more info in this post: Systemic Complexity.”

“Thanks, I’ll take a look. Talk soon?”

“Of course. See you then.”

Learning Goals

  • The bike theft story is from personal experience and as it was playing out, I realized I was virtually supporting a physical stakeout, and it is an accessible story. Digital Twin and Thread conversations can become very detailed and at times controversial, but I look for stories which express high-level concepts to students who may not yet be familiar with the terms.
  • Earlier this year, I spoke at a 2-day Digital Thread conference with 100 attendees, and combined thousands of years of experience, but there was not a clear agreement on the meaning of Digital Twins, Digital Threads, or for that matter PLM. I don’t intend to solve those issue in a 3000-word LinkedIn post, ?? but want to provide a base-level of understanding for those new to the field.
  • The scope of product data throughout the lifecycle is at least as large as the As-models (-Desired, -Required, -Designed, -Configured, - Built, - Serviced, -Operated) discussed within. But to have a meaningful Digital Twin, we need a Digital Thread which maintains integrity through those models, which are maintained in at least: PLM, ERP, MES, and MRO.

James Kirchner

Past President at Michigan Translators/Interpreters Network

1 年

Watch the later satirical follow-up to the movie "The Icicle Thief".

Dr. Birgit Boss

How to make digital twins interoperable | Bosch | Board IDTA | Catena-X

1 年

Very good article, thank you for sharing. To me it seems that the digital thread in your definition is the digital twin itself and “your” digital twin is just one of the many models that represent the asset (that might be physical)? P.S. My understanding of digital twin is the one shared by the members of the Industrial Digital Twin Association and conformant to IEC 63278.

要查看或添加评论,请登录

Patrick Hillberg Ph.D.的更多文章

社区洞察

其他会员也浏览了