Digital twins revolutionize shipbuilding
A digital twin is a digital, or virtual, replication of just about any physical thing or place — a system, a product, a part, an office or even a shipyard.
The technology has the potential to modernize and optimize shipbuilding, or really, any complex, manually intensive business involving highly specialized systems that require ongoing inspections and maintenance. Think of just about any manufacturing business, as well as those in the metals, mining and utilities industries, and digital twin technology could have a transformative impact.
When it comes to building a submarine or a ship, two types of digital twins can be created: a prototype twin and an operational twin. Each provides slightly different benefits, but both are equally valuable in de-risking and shortening the timeline of a ship build.
The prototype twin involves using digital twin technologies to design, simulate and optimize the build of a physical system before it is physically built. The operational twin replicates an existing physical system (physical twin) with a digital version. Data captured by the physical twin can duplicate events seen in the digital twin.
Let’s take a closer look at each to understand how shipbuilding organizations can use digital twin technologies to solve real-world industry challenges.
The digital prototype
Some would suggest that the prototype twin is just an extension or a compilation of digital design. Although there is some truth to that — the prototype twin aims to glue together all the 3D designs of a platform, for example — there is also much, much more to the technology. A prototype twin at the scale of an engine can provide some very valuable data. That data can help predict vital metrics, such as mean time between failures, based on sub-assembly failure rates and failure dynamics.
For example, an existing physical jet engine combustor being considered for inclusion in a prototype jet engine will have specific failure data. This data might suggest that the combustor has a mean time between failure of 68,735 hours. The data collected from failed combustors might also define failure dynamics, such as the most probable location of material degradation leading to failure. This behavioral data can be tagged to different parts of the engine’s prototype twin and then used to simulate a failure. Next, artificial intelligence (AI) can be used on the prototype twin to determine the predicted outcome from the failure of that combustor.
Figure 1 shows AI’s prediction that 1 minute and 16 seconds after the combustor has failed, the temperature on the outside of the turbine has increased from 164 degrees Celsius to 456 degrees — leading to catastrophic failure. Results from this engine simulation might, for example, lead to the installation of thicker combustor insulation to delay the heat increase so that the engine can be safely shut down.
Figure 1. With AI, a prototype twin can predict possible outcomes in an engine simulation
A prototype twin is not just a collection of 2D and 3D models. It also can provide the ability to overlay vital statistical and behavioral data to simulate functionality. It can act as though it is the real physical thing, but it has the advantage of being able to simulate scenarios that would be far too dangerous to simulate on a physical system or platform.
It is easy to see how significant value in the design phase of a system such as an engine can be delivered using a digital prototype. Now think about that value applied to the scale of a ship. Think about the potential of simulating all the different scenarios that can happen aboard a ship. The potential to avoid rework is enormous. Instead of your learning about design mistakes after the ship has been built, the prototype twin can discover those flaws long before the first piece of metal is cut.
Building safety into a ship at the start
Discovering design flaws and simulating various scenarios early on is crucial, especially when it comes to safety. Consider how a prototype twin can be applied for fire retardation and suppression. In 1998 the HMAS Westralia caught fire, taking the lives of four Australian Naval personnel. That tragedy was caused by the use of improper fuel line hosing that burst, spraying fuel onto hot surfaces and igniting a large section of the engine room. And although a prototype twin could not have helped avoid the improper use of parts, it could have helped with fire simulation had the technology been developed when the Westralia was designed.
Through the use of a digital prototype and the cognitive power of AI, the subsequent spread of fire from an ignition — as happened on the Westralia — can be predicted. Such simulation is not just based on the physical layout of the ship but also on a number of other factors such as compartment materials, flammable substances, ventilation, installed fire-suppression systems and so forth. Physical and behavioral attributes are tagged to the twin, so the AI layer can determine factors such as how long it will take to burn through different materials.
Figure 2 illustrates the following: The simulation has determined there are eight walls vulnerable to fire that should be better insulated. It has also determined that four zones are not protected by fire suppression (sprinklers), which should also be addressed.
Figure 2. Simulation of fire spread and identification of weak points using a digital twin and AI
3D digital twin views can also provide significant value by visually representing the potential impact of heat within a ship. The simulation can also show a color-coded view inside a ship compartment and represents in red anything that has potential to get hot. This can give a valuable visual on two things: first, a view on where insulation might need to be added, and second, a view on how heating in a ship might affect the crew’s health and safety.
While we’ve discussed a couple of examples of how a digital twin can help in the design/prototyping phase of a ship build, there are so many other valuable applications of a digital twin in this phase. You could add even more experiences with augmented reality (AR) and virtual reality (VR), for instance. It all comes down to imagination.
Example 1. Ship Wi-Fi modeling with the ability to increase and decrease coverage to control cost
It’s clear that a prototype twin can accelerate the design phase of a ship build by dramatically reducing rework and manual simulations. The operational digital twin, however, has the potential to be even more valuable as simulation based on predicted data.
In this case, AI is no longer required, because sensors can capture actual, working data from the physical twin. The entire digital twin can be lit up with information that’s collected from thousands of sensors aboard a ship. This can create a 3D dashboard of information, identifying a vast number of events, interactions and issues happening on board the vessel in real time.
To take it up another notch, one could put on a VR headset and walk around the digital twin of a vessel to view the events, interactions and issues (virtually) exactly as they are happening in the physical world. That is extremely powerful, as it can provide visual and dimensional context to a mechanical issue.
If we wanted to simplify this, we could just provide collected sensor data in the form of a dashboard on a PC. An individual could then drill down through sensor information from a web application. To expedite root cause analysis, we could overlay that sensor data onto the digital twin and, through the use of VR, provide a view that has the added benefit of visual context, thus giving rise to the birth of a “virtual maintenance engineer.”
Figure 3. VR view (3D scan) of a ship compartment within the digital twin showing a real-time sensor data overlay collected from the physical twin
As shown in Figure 3, a virtual maintenance engineer can use a VR headset to navigate the digital twin and view a 3D scan of a compartment and real-time data that’s collected from the physical twin. This can provide the engineer with visual context to understand what an issue might be; in this case, a coolant flow issue is raising the temperature of a diesel generator. Although there would most likely be an engineer attending to the issue on the physical twin, a virtual engineer who may have subject matter expertise could be in another location viewing the same interaction within the digital twin. Not only can the view provide key sensor data from systems and equipment within the compartment, but also, the virtual engineer can pull in sensor data from other parts of the ship as necessary.
Similar to a digital prototype, the use cases for a digital twin are vast. In the previous paper on how AR can be used in shipbuilding, we highlighted the use of AR for quality inspections. To take that use case one step further, inspection data — including objective quality evidence (OQE), such as photos, videos and inspection reports — could be attached to specific locations on a digital twin so that an individual using an AR device such as a headset could walk into a ship compartment on the physical twin and instantly view locations of where OQE has been attached to the digital twin (seen as red dots in Figure 4). That way, the individual can view the various OQE and results of quality inspections as they view the physical system. Dots could be turned green when the issue has been addressed, allowing for engineers to walk around the ship fixing red dot issues and making them green. Talk about gamification!
Figure 4. View of the physical twin through an AR device showing locations of OQE attached to the digital twin
A digital twin is so much more than a collage of pretty pictures and actually could be the most powerful asset an organization owns. However, there are challenges to overcome to get it right. As you can see in Figure 5, there is so much valuable data that can be captured in a digital twin. Operational history alone could quickly fill your IT department’s servers. In fact, when an F-35 flies for one hour it can produce two terabytes of flight information. That’s enough to fill the average home’s backup drive.
Figure 5. Information that can be captured in a digital twin
The biggest challenge with a digital twin is extracting important information from the data before loading it into the digital twin. Data on its own is not all that useful, but when data is processed, interpreted, organized, structured or presented so as to make it meaningful or useful, it becomes information. For the F-35 example, that means converting two terabytes of data into two megabytes of information. That would either mean manually looking through approximately 150 million pages of data to get the information required or using advanced technologies such as high performance computing (HPC), AI, big data and analytics to extract it.
So how do you create and capture a digital twin? Unfortunately, the answer isn’t “buy an application and it will magically appear.” It takes time and investment, but the return will be significant, especially when done early in the design phase. There are some tried-and-tested approaches to building digital twins, such as DXC’s Digital Twin Runtime Starter Kit. The starter kit is more of a process than a single explicit technology, and it requires a solid understanding of what it is the organization wants to derive from the twin, i.e., what information is required to make valuable decisions.
Gone are the days when the digital twin was just a great concept. A digital twin is now real and is providing manufacturing organizations with the ability to reduce significant costs by minimizing rework. Unlike a lot of other technologies, it also provides value across the life of the program. The earlier it is deployed, the more value a digital twin will provide. A conservative estimate on a $10 billion program could equate to $50 million in savings from an effectively instantiated digital twin and a well-architected digital framework or thread. That’s a lot of peanuts and well worth a serious discussion with DXC.