Digital Twins: Cutting Through the Hype From Concept to Operations
A digital twin of a pump room and its operations dashboard. (KCI Technologies, Inc.)
The Future, Now
In a brand-new highrise office building in the heart of Chicago, a facility-maintenance manager gets a notice on her cellphone that a pump is operating outside its normal parameters. After reviewing the pump specs on the mobile dashboard, she notes a failure is imminent and shuts down the pump. Viewing a 3D model of the pump room, she notes the location, model number and warranty information for this particular unit, and submits a work-order ticket.
Hundreds of miles away, after a fender bender damaged a guardrail attenuator on a stretch of interstate I-440 in Tennessee, maintenance crews rapidly photographed and logged the damage, submitted a parts requisition, and scheduled the replacement work all from the front seat of their truck. Back in the office, the maintenance director notes this is the second time this month that this particular attenuator has been damaged and schedules a meeting to discuss possible preventive measures with the safety director.
And in Washington, D.C., agents from the NSA use satellite-based sensors to generate a full 3D model of a building to track the real-time, floor-by-floor, room-by-room movements of a fugitive who has been tagged with a GPS tracking device. Only by jumping out an 18th-floor window and dropping to the floor below was the fugitive able to elude his pursuers.
OK … the last example is a scene from the Will Smith/Gene Hackman spy thriller, Enemy of the State, but the others are real-world examples of the brave new world of “digital twins.” But what is a digital twin anyway?
Definition(s) of a Digital Twin
If you ask a dozen people to define “digital twin” (and I did), you’ll get nearly a dozen different answers. To many, such as Jamie Waller and Grant Heintzman, both with the Tennessee Department of Transportation (TDOT), a digital twin is a 3D exact as-built model of an asset. Andrew Pangallo, major project construction manager with the Indiana DOT, defines digital twin as a model that starts with design, “continues through construction to operations and asset management, and comes back around for use in future design.”
While many define digital twins in relation to the asset itself, others, such as Thru Shivakumar, CEO of building automation software company Cohesion, expand the definition to include the people who interact with the modeled asset. “A digital twin consists of data and models connecting assets and people in a single platform,” explains Shivakumar.
Still others expand this definition even further. According to Erik Harris, AIA, associate principal at architectural design firm Goettsch Partners, a digital twin is a “true BIM model on a platform that incorporates all data, time, schedule and coordination elements.” Adam Klatzkin, vice president, iTwin Platform at Bentley Systems, distinguishes a digital twin from a BIM because of “its connection to the physical asset, leveraging past performance data and present conditions to make future predictions.”
All these definitions include some mention of “data.” Any project can produce enormous amounts of data, and even with today’s most-advanced hardware and software, tracking all the data is not practical nor necessary. As such, many in the industry, including Chuck Keeley, vice president of BIM Services with Kelar Pacific, and Bob Bray, senior director and general manager, Autodesk Tandem, center the definition of digital twin on the needs of the owner or operator of the asset. “A digital twin is the useful digital data relevant to a project that serves a purpose defined by the owner,” says Keeley.
In 2020, major industry software vendors, including Autodesk, Bentley and Microsoft—along with other industry players such as Dell, Ansys and Northrop Grumman—founded the Digital Twin Consortium (www.digitaltwinconsortium.org). The consortium’s stated goal is driving consistency in “vocabulary, architecture, security and interoperability of digital twin technology” as well as advancing the use of digital twin technology. As “the authority in digital twin,” the consortium put a fair amount of effort into developing a definition that works across industries and functions:
“A digital twin is a virtual representation of real-world entities and processes, synchronized at a specified frequency and fidelity.” – Digital Twin Consortium
This definition encapsulates the three main concepts of digital twins:
1. A digital twin includes a digital representation of something—an entity—that can be a physical asset or a process (a simulation).
2. Through time, a digital twin is linked, or synchronized, either uni- or bi-directionally, to the real-world entity it represents.
3. The digital twin is represented and synchronized at a frequency and verifiable accuracy level sufficient to meet the intended purpose of the digital twin.
This definition is incredibly flexible and can be applied in a broad range of uses, from highways to halogen lamps to heart valves. Regardless of the industry or context, it’s the intended use of the digital twin that narrows the definition, resulting in beneficial applications. As we refine the definition for applications in infrastructure, these intended uses usually are determined by the facility owner and operator/maintainer.
James Somerville, SaaS program manager at civil engineering firm KCI Technologies Inc., summarizes digital twins for infrastructure as the “detailed view or model of your infrastructure coupled with the people and processes acting on it, enabling owners to make informed, data-driven decisions about operations and maintenance.”
This is echoed by Robert K. Otani, P.E., LEED AP, chief technology officer with structural engineering firm Thornton Tomasetti. “Digital twins vary depending on what the owner needs,” he explains. “The information collected about the real asset or building is dynamic. The key to a functional digital twin is the synchronization of this information with the modeled entity to meet the owner’s goals.”
What are these goals? What’s the immediate and long-term usefulness of digital twins for infrastructure? What does our industry need to do to realize the full potential of digital twins? The remainder of this article will focus on answering these questions.
Benefits of Digital Twins
Although the definition of digital twins coalesces around a few key concepts, their uses are quite expansive. From design to construction to handover and operations, the benefits of digital twins are enjoyed by the entire project team, starting with the owners and operators of the entity being twinned.
As owners are paying for the design, construction, operations and maintenance, and eventual sale or decommissioning of an asset, it should surprise no one they stand to benefit most from digital twins. Even in today’s nascent state of digital twin adoption, value is extracted from the twinning process, with much more value just over the horizon. Foremost are the benefits of data-driven decision making.
Riverside Investment & Development develops Class-A office towers in Chicago. Jordan Feste, construction manager at Riverside, champions the benefits of digital twin’s facilitation of proactive management of their buildings. “Using digital twins is a huge benefit to operation and maintenance, because we can deal with issues proactively instead of reactively. We are pulling together all underutilized data to start making educated decisions. This helps us avoid tenant frustration and keep costs down.”
Marc Goldman, director, Architecture, Engineering & Construction Industry Solutions at Esri, agrees. “Digital twins give owners the ability to make decisions from multiple streams of data from any location,” he explains. “This streamlines the complexity of operations and maintenance. Even a one to two percent efficiency increase is a lot of money in the long term.”
This data-centric approach also lowers risks and increases safety by bringing the data to the user. For example, according to Klatzkin at Bentley, users of their iTwin platform can perform virtual bridge inspections using data from sensors, high-resolution images and laser scans. INDOT’s Pangallo concurs, noting that lowering risk and saving time are the most-important objectives from a project management point of view. “Using better tools and the best information from a digital twin will help us meet these goals.”
Operations isn’t the only benefit owners get from a digital twin. A physical asset with a corresponding digital twin has increased value when that asset is sold. Not only is there value in the data about the physical asset, but also in the data about how the asset is used (e.g., occupants of a building or travelers on a highway). “At some point, building occupants’ hyperlocal data will become valuable,” predicts Otani with Thornton Tomasetti. “Already buildings have their own apps, and eventually the data will be monetized.”
For Designers and Reviewers
While owners are the primary beneficiaries of digital twins, designers and reviewers also gain from their use. Coordination among various design disciplines already is common and will continue to improve as models mature. This gives designers insight into others’ processes, allowing for adjustments before construction.
Data important to the owner to be included in the twin requires additional upfront constructibility consideration from the design team, which can result in fewer change orders during construction. Digital twins can provide visual feedback on designs, which reduces time and increases accuracy of design review and quality checks.
As an example, Michael Price, mechanical and fire-protection engineer with KCI, explains how digital twins can be used to link a Revit model with hydraulics design software. “Simulation data of fire-protection sprinkler flow rates can be linked to the Revit model and display color-coded results to quickly alert designers to sections of the system outside desired parameters.”
Digital twins open up additional financial opportunities for designers beyond current service offerings. According to Bray with the Autodesk Tandem team, their AEC customers see an opportunity to provide more services related to creating the models and data required for the initial creation of digital twins. Harris with Goettsch extends this opportunity to the ongoing maintenance and updates to the digital twin throughout the life of the physical asset.
Digital twins also have the potential to address the persistent issue of shortages of qualified professionals. Those who recently entered the workforce (or soon will) are technologically savvy. Bringing the tech required by digital twins to the construction industry will automate many mundane manual tasks and attract more talent. A case in point is Klatzkin’s virtual bridge-inspection use case. After the twin is created, engineers and inspectors can further engage the model by donning Hololens headsets to inspect the asset in an immersive virtual environment.
Measured by the Digital Twin Consortium’s definition of a digital twin, contractors are leading the way in the use of twins in the form of construction coordination BIM models, plan dissemination and data synchronization with the physical asset. For example, contract specifications for HVAC and fire-protection systems typically require the development and submission of BIM models along with 2D PDF plans. The BIM models from various trades and design disciplines are coordinated into a single model used for weekly planning as well as dispute resolution when conflicts arise.
As an example of a basic digital twin used in construction, architect Harris describes how software from Bluebeam is used to access PDFs that are the contract documents. “The PDFs are linked to RFI responses and 3D BIM models,” he says. “Revisions to the plans are digitally slip-sheeted so the latest version is always available to team members. Navisworks from Autodesk is primarily used for MEP and structural coordination due the physical massing of these items.”
At TDOT, even without full digital twins, the tools associated with digital twin initiatives are providing immediate benefits. According to Jamie Waller, assistant director of construction at TDOT, “One of the most useful features of connected construction solutions is the ability to connect photos to the exact location of the photographed object in the field.”
Taken together, these data, models and processes encapsulate the concept and workflows associated with digital twins. And this is just the start. Esri’s Goldman sees future opportunities for contractors, including a continuation of their scope of work on a particular asset to include digital twin upkeep and ongoing revision.
A digital model is shown next to its as-built steel connection. (Thornton Tomasetti)
Adds Sidharth Haksar, head of strategy at Autodesk Construction Solutions. “Construction is a project-based business,” he says. “By leveraging connected construction solutions to collect data from all project workflows—from design through closeout—contractors better inform digital twins and use their rich data packages to provide ongoing maintenance and operations services as a subscription. Contractors can develop additional revenue streams and create space between them and their competition. Their clients [the owners] are happy, so it results in repeat business, and potential customers can see the value-add of the digital twin.”
But if digital twins are so marvelous, why aren’t they in more widespread use? Several hurdles must be surmounted for digital twins to become a common component of construction projects.
Hurdles to Fully Realized Digital Twin Adoption
Just as there are many benefits, there are still many hurdles that must be overcome to fully realize these benefits. Challenges include contractual and legal issues; defining/redefining roles, responsibilities and workflows; developing data standards; and simple education around the concept of digital twins.
Contracts, 2D Plans and Digital Twins
One of the most-significant hurdles to widespread digital twin adoption is the fact that 2D design plans—not 3D intelligent models—are part of the contractual package for infrastructure and facility projects. BIM models and simulations often form the basis for digital twins and, as Riverside’s Feste points out, although private developers can mandate creation and delivery of a “BIM model as a contractual obligation, including the data standards and what data is to be modeled,” in all but a very few cases, 2D plans are the legal documents that dictate what’s to be constructed and who is responsible for the work. “Liability and lack of standards prevent use [of digital twins] as legal documents,” adds Pangallo with Indiana DOT.
This is an issue not only during design, but also during construction and at the time of handover. Similar to design, BIM models often are created during construction, and contracts usually require as-built drawings and BIM models, but very few—if any—require an accurate, as-built digital twin that incorporates all data relevant to operation and maintenance in an easily accessible way. So even as BIM 3D and 4D modeling become more common during design and construction, the result is a parallel set of models: 2D PDFs and 3D models.
“This will continue until there are legal changes to what is required for submission,” says Goettsch Partners’ Harris. The challenge then becomes creating a functional digital twin from a widely disparate set of data created and used by a widely disparate group with widely differing needs and goals. This directly raises questions about roles, responsibilities and workflows.
Workflows, Roles and Responsibilities
Of all the hurdles, perhaps the most glaring also is the most basic: who does what and when in regards to developing a digital twin? Some of the tasks already are being completed in a typical project workflow, yet many workflow gaps remain that need to first be identified and then assigned. How are these gaps defined, by whom, and to whom are the tasks assigned? By examining the current workflows, some of the challenges are illuminated.
Setting aside various contractual models, a project typically progresses in a series of defined steps: 1) the owner conceives an asset; 2) the designers design it; 3) contractors build it; and 4) owners/operators maintain and operate it (and all too often, 5) everyone litigates). It’s in the first phase where the most critical decisions are made regarding development of a digital twin. This is when owners define which entities within the asset are important to them and thus what needs to be modeled—and at what accuracy—to provide a useful digital twin.
As mentioned earlier, owners often mandate both the use of BIM during construction and the delivery of a BIM model at the end of the project. However, nearly as often, the details of exactly what must be included in the BIM are not clearly defined.
“A lot of new assets go into a project,” explains Tim Kelly, senior product manager for Tandem at Autodesk. “Owners know they want something, but often they are not sure what they want.” Software vendors including Autodesk and Bentley Systems offer platforms (Tandem and iTwin, respectively) to help define the assets to be tracked as well as the workflows associated with gathering this information.
Savvy owners and operators are beginning to define which data and processes are important for their specific projects, and bake this information into the contract requirements. High-rise developer Riverside Investment and Development started to create a technical specification to do just this. “We’re bringing these requirements to the forefront [in our early project planning] to make Riverside’s developments the most advanced,” notes Feste.
In the next phase, architects, engineers and a cadre of other professionals begin to develop the designs for the asset. In this phase, the basis for a digital twin begins to take shape, usually in the form of a BIM model. Setting aside the statistics that more than 35 percent of architects aren’t using BIM, current specifications typically require BIM Level of Development (LOD) 300 (model elements are defined with exact dimensions and relative positions) or LOD 350 (precisely describes the information about an element and relation to and connection with other elements). Although models with this level of information are (usually) suitable for construction, parts of the model either lack the detail needed for a functional digital twin or are overkill.
While a BIM with LOD 350 captures the precise geometry and connections of elements, the design process produces an enormous amount of data, much of which isn’t tracked or accessible in the BIM (and often nowhere else) after the project is completed.
As Otani with Thornton Tomasetti describes it, “design has the ‘why’ something is designed the way it is. We need to capture this ‘why,’ this design information embedded in models, which is directly tied to the building usage.” To this end, Otani and his colleagues are working on research that uses BIM to train machine-learning algorithms with the ultimate goal of extracting years of design data for future analysis of an asset. “Without this information, we’re required to recreate the models.”
In the third phase—construction—contractors interpret these models and design data, and transform them into real-world physical assets. In the process, concepts and ideas become specific fixtures, finishes, components and structures. Along the way, details emerge, and designs are modified in response to onsite conditions. It’s common for large projects to generate dozens or even hundreds of requests for information (RFIs) and change orders (COs), which alter or modify the original design and design models. Increasingly, these details and modifications are tracked, logged and eventually shared as as-builts with the owner.
For example, TDOT uses Autodesk PlanGrid to track RFIs, photograph onsite conditions and disseminate the most-current set of revised plans to the entire team. “Getting the latest plan revisions into everyone’s hands, almost instantly, is a big benefit of PlanGrid,” says Heintzman, TDOT’s construction management system lead.
Unfortunately, the submittal information, change orders and design revisions, as-builts, operations and maintenance manuals, results of commissioning, and myriad other data are frequently delivered in non-uniform, non-digital, inaccessible formats.
“Most owners have siloed infrastructure data: ‘dark assets,’” says Autodesk’s Bray. “They know they have this information, but don’t know what they have. There is so much data available, and no easy way to access and manage it. How this information is shared and accessed are some of the hurdles to digital twin realization.”
The fourth phase of the project starts right after handover with the operations and maintenance of the physical asset. It’s at this point that a digital twin also is delivered and where its full value is realized. As Esri’s Goldman explains, “80 percent of the costs [of an asset] are in the long-term operations, delivering on its intent.” Unfortunately, according to a recent Future Market Insights study (bit.ly/3nNa7bV), “95 percent of captured data during design and construction goes unused,” relates Autodesk’s Haksar. This sentiment is echoed by Goettsch’s Harris, “There is a tremendous amount of rework [required by owners] just to identify what information they have for the asset in its current, as-built state.”
The problem stems from a number of factors, including assignment of responsibility to gather information into a coherent structure; lack of a standard method of collecting and recording information; and lack of widespread availability and use of software platforms to access information from a wide variety of sources.
Data Standardization (or Lack Thereof)
From concept to handover, an enormous volume of data is created by a wide range of team members using all sorts of authoring tools. Everyone interviewed for this article discussed data standardization—or, more accurately, the lack thereof—as another major hurdle to widespread digital twin adoption.
With so many different file formats, attributes, data structures and schemas, it seems a consolidated functional digital twin will forever be just a pipe dream. How can we integrate wildly different facility data for wastewater pumps, HVAC air quality and elevator destination dispatch systems into a unified digital twin when we still struggle to share basic drawing information between the two largest CAD vendors in the AEC space? A full discussion on the topic would fill multiple issues of this magazine.
Thornton Tomasetti’s Otani groups the challenges with data into two questions: “Number one,” he says, “is the issue of data interoperability; how do you get all your information into the platforms? Currently, there is no prevailing standard, so owners are being solicited by various vendors. Number two, how do you work with all the data once they are in the system?”
The answers to these questions—and the data-standard issue in general—require asking more questions. Specifically, what are the key aspects of [an asset] that need to be updated? “Most owners do not yet know what they need,” notes Otani. “Before we define a standard, we must know what we’re standardizing around.”
“Generally speaking, there are two ways to work with data: have everyone use the same file formats or have a converter,” says Derek Fuller, eConstruction specialist with Indiana DOT. The first option speaks to creating common file standards implemented by software and hardware vendors across disciplines. This approach is seen in schemas, standards and specifications such as the IFC schema from buildingSmart International; its subset (or “smart filter”) COBie data model; and various XML schemas, including aecXML and LandXML, to name just a few. In addition, organizations such as the Open Design Alliance (ODA) create development toolkits for .DWG, .RVT and .NWD, allowing the file formats to serve as an open exchange standard for the industry. Although this approach has enjoyed some success, there are limitations to each that prevent standardization around any one of them.
The other option is to create a platform that can take input files from a wide variety of creation tools from software and hardware vendors. As Fuller’s colleague at INDOT Pangallo states, “this approach is more desirable, because it is most sustainable as it allows designers and contractors to use their preferred tools.” Owners will continue to demand interoperability and must have a disproportionate say in data formats. This will move the industry forward.
Lack of Education
As with many newer concepts introduced to the traditionally conservative construction industry, the first step toward adoption is education. Adoption of digital twins accelerates as knowledge of their current benefits and future potential spreads. Many simply don’t know what’s possible today. “Educating each player on how important it is to add their small component to the overall digital twin and getting them to understand the purpose of the overall digital twin is a primary hurdle for our industry,” says Riverside’s Feste.
Beyond a basic understanding of digital twins, education on what should be included in the twin also is required. Owners and operators have to determine which entities are important to maintain and track in the twin, and then make sure designers and contractors are educated on the importance of these items.
With this many hurdles, how can the wondrous benefits promised by digital twins ever become a reality?
From Conceptual Promise to Common Practice
Educate and Mandate
One of the first steps in making digital twins commonplace relates to the last hurdle previously discussed. After owners are educated on the value of digital twins, they need to mandate and finance their creation. Contract specifications should include language dictating the data to be included in the handover, including their format and structure as well as intended use. Further progress can be made in this area by actively supporting the move away from 2D PDFs as the standard contractual design documents and toward data-rich 3D models.
Recognize the Value of Data
Designers and contractors also can educate themselves and be additionally aware of the value of the data they create and then communicate this value to project owners. This communication is a two-way street and should occur early in the project lifecycle so designers and contractors can understand the pain points and opportunities to provide additional value in the design and as-built models.
The industry already is recognizing the power of small baby steps, and the precursors to full digital twins have already begun. Two broad indicators of increasing owner interest and implementation of digital twin concepts are the use of reality-capture (recap) technology and building automation platforms. Both have been used in some basic way for a while (think photos and laser scanning in the former, and building access control systems in the latter) and have seen accelerated adoption in the last couple of years.
Reality Capture Feeds the Data Machine
The developers, designers, contractors and owner/operators interviewed for this story all agree that capturing information about the real-world entity and synchronizing these data with the virtual representation of that entity are critical components of a functioning digital twin. According to Bentely’s Klatzkin, recap is a “key enabling technology” needed to make digital twins work by providing digital context around an asset, especially during construction and for existing assets with no historical digital data.
Fortunately, recap has received a lot of industry attention and is being adopted at an almost-frantic pace. Traditional surveying and simple photographs linked to 2D plans as well as more-complex lidar scans, point clouds and meshes all document the physical state of the entity and help create accurate as-built/as-exists models. Capture hardware is ubiquitous, with all smartphones doubling as high-resolution cameras, GPS location trackers and, in the case of the latest iPhone, laser scanners.
“In 2001, it was expensive to capture point-cloud data, and difficult to open and use the data,” explains Thornton Tomasetti’s Otani. “Today, due to advances in hardware and software technology, it is much cheaper and easier to use.”
While lidar and photos document an asset at a fixed moment in time, other recap technologies such as IoT sensors can provide real-time streaming of data directly back into the digital twin. Autodesk’s Kelly and Nik Patel, chief technology officer and co-founder at Cohesion, both view IoT as critical to the success of digital twins. “Capturing dynamic process data, like air quality, equipment status and occupancy rates, allows us to make predictions about future states of the asset,” says Patel.
Software Vendors’ Role
While capturing static and dynamic reality data has become exceedingly easier, using the data remains a challenge. Thornton Tomasetti’s Otani opines, “There must be some platform which let’s us get this recap data into an editor to author changes. For example, something that talks to Revit, ArchiCAD and other softwares we use.” Bentley iTwin’s Klatzin agrees, “Beyond capture, we need to make sense of the information.”
In fact, making sense of all the data in a unified tool is the goal of platform initiatives from major software vendors who are responding positively to owner demands for interoperability. “There are many sources of truth,” says Autodesk’s Bray. “With Tandem, Autodesk intends to embody any standard the client is using and provide a container for those standards.”
Bentley’s iTwin platform has similar goals. “Because data comes from so many sources, standardization is very hard to achieve,” explains Klatzkin. “Most owners do not want to move data from their existing tools to a proprietary system, but rather want to federate their data into iTwin to create a single view of truth.”
These types of universal data operators are “API-first” (application programming interface) and enable two-way syncing among all systems. Cohesion’s Patel explains: “In a building, there are 10 different systems generating data. How does data from one system impact another? There is an incentive to get info out of proprietary systems and into APIs. Cohesion looks across all systems and enables owners to use data to make important decisions and understand the flow of data over time.”
Individual Calls to Action
With so much potential and so many hurdles, individuals working in the industry might be overwhelmed to the point of inaction. What can one person do? Surprisingly, the most impactful actions you can take are some of the easiest: learn more, and share what you learn with your colleagues and clients. (See the accompanying list of “Resources” for more information.)
“Think about the data you create daily,” advises Bray. “Ask yourself, what is the value of that data to the owner through the lifecycle of an asset? How can the owner leverage that data down the road? Then start having conversations with owners.” Feste with Riverside concurs. “It’s important for designers and contractors to better understand the data associated with various assets and how it is used now and in the future,” he says.
Another proactive step is to examine your own workflows and start to digitize them to enable more-consistent collaboration and enrich the data throughout design and construction. Thornton Tomasetti’s Otani suggests designers and contractors “ask themselves, ‘what information do we create in our design that will be valuable in the lifecycle of the [asset]?’ Save this information so it can be accessed and used in the future.”
Finally, as Fuller with INDOT suggests, “it’s important civil engineers get involved in committees and organizations developing standards and workflows. Be part of the solution.”
Digital Twin Consortium: digitaltwinconsortium.org
Open Design Alliance: opendesign.com
buildingSmart International: buildingsmart.org
Try the Software
Autodesk Tandem: autodesk.com/solutions/digital-twin/autodesk-tandem
Bentley iTwin: bentley.com/itwin
Network With Others
Riverside Investment & Development: riversideid.com
Thornton Tomasetti: thorntontomasetti.com
Goettsch Partners: gpchicago.com
Kelar Pacific: kelarpacific.com