Twinning

Definition & key benefits

In 2002, Michael Grieves coined the term “digital twin” in the context of the manufacturing industries (Grieves, 2019). The concept initially consisted of encapsulating information about a physical product in a digital representation of it, the digital twin, and connections between the physical product and its twin to provide operating data for diagnostic analysis and performance simulation in support of product life-cycle management. In recent years, the concept has expanded to include the idea that the digital twin model can control and optimize the performance of the physical product in (nearly) real time autonomously using machine learning and artificial intelligence technologies.

Professionals in architecture, engineering, construction and facilities operations and management (AECO) often think of a digital twin as a 3D model that digitally represents the spaces, assets and systems of a facility. That is a great point of departure; however, a digital twin should also integrate historical data from a work order system about asset repair history, real-time performance data from internet enabled sensors on assets and more.

A recent report (Arup, 2019) describes the digital twin with a five-level framework, spanning from 3D visualization to autonomous operation and control. Building operators and managers who are using a digital twin typically work within Level 1 - 3. In AECO, the expanded practices of incorporating data from external sources, autonomous operation and control (Level 4 – 5) are where there is opportunity to deliver even more value.

Level 1: Inventory and visualize the products that comprise building systems and the spaces they serve to enable FM teams to virtually investigate problems quickly, walk down a problem in a room to the control equipment that controls services to the room and understand the impacts of a shut down on building occupants.

Level 2: Collect and track historical data, represented in work orders, for analyses that deliver improvements to the system and mitigate risk. This ability enables teams to develop insight about equipment that requires frequent repair or has sub-optimal performance, inform fix or replace decisions and improve operational reliability and facility uptime.  

Level 3: Collect real-time data from the constituent products of the system via internet-enabled sensors to detect performance issues, compare real time data with baseline data, and make changes to optimize building systems.

Level 4: Access data from external sources, for example, local weather, and incorporate into system analysis to enhance the ability to optimize equipment performance.

Level 5: The ability for the system to self-optimize performance through autonomous reasoning. One might believe that level 5 is aspirational! However, the application of machine learning systems and AI is emerging in building management systems (BMS) and has been achieved for some manufacturing applications (Forbes, 2018). It will be possible to train a digital twin to recognize the cause of issues in building performance and autonomously control and correct equipment operation.

The UCSF Health case study

Located on a 15-acre site, UCSF Medical Center at Mission Bay is an 878,000-square-foot, 289-bed complex opened in 2015. The center includes an integrated hospital, outpatient building, a 200,000-square-foot medical office building, a 46,000-square-foot energy center and a helipad. It also contains a six-story, 170,000-square foot Precision Cancer Medicine Building (PCMB) for adult outpatient cancer care (Figure 2).

The original complex and PCMB addition were integrated project delivery (IPD) projects.

The digital twin journey

UCSF Health’s digital twin journey began in 2012, toward the end of construction on Mission Bay. The IPD team created a fully coordinated and clash-free multi-disciplinary building information model (BIM) for construction (Figure 3) and at roughly the same time, the facility engineering group committed to deploy a new computerized maintenance management system (CMMS) IBM Maximo™.   Recognizing the opportunity to accelerate the CMMS data load, UCSF senior management for capital projects sponsored a study on what data would be needed by the engineering group. The analysis was constrained primarily to Level 1 digital twin data. This led to the transfer of more than 40,000 assets into Maximo and a proof-of-concept Maximo BIM viewer implementation developed by the engineering team, with support from external consultants.

This late engagement led to key learning that was subsequently implemented on the PCMB project, where it exploited the opportunity to include FM data requirements and data collection process in project requirements. “The building engineering group overcame reluctance on the part of the capital project management when they realized that they needed only 10 percent of design and construction information. This made it possible to move forward and implement the requirements with no cost or schedule impacts on the project,” said Fred Whitney, PCMB project manager.

Under the leadership of Bruce Mace, UCSF Health executive director of facilities and support services, the team:

• specified the 10 percent list in the form of a data dictionary.

• broke down the information requirements by discipline, such as architecture, mechanical, plumbing, electrical and by project milestone to phase information collection and reduce a big data collection challenge into small batches, on which they could verify data quality and completeness.

• Created a building applications team (BAT), as an offshoot of the Maximo implementation effort, that was accountable for definition of the data requirements and oversight of implementation by the project team. The BAT included:

  • An FM data manager,

  • Maximo admin,

  • Technical guidance from a software application architect and

  • Support by an external consultant to assist with data specification, data collection and verification processes, and provider of software for BIM FM Maximo integration.

Is BIM necessary?

Setting up a digital twin starts with a clear need for digital representation of the electrical, plumbing, and mechanical system(s) and the architectural spaces/locations they serve. FM teams must easily locate and identify products of a system in terms of the building locations, such as building levels, zones and rooms.

Technically, BIM is not necessary for setting up this data; however, it provides benefits that eliminate barriers to specifying building equipment, the relationships between the equipment and building systems of which they are a part and the building locations they serve which are the foundational elements of an AECO digital twin.

Already a part of nearly every major construction project, BIM provides an early starting point because it exists before the physical building and digitally represents the foundational elements of a digital twin. BIM represents the equipment well. Some BIM authoring systems can also define and represent building systems; in other cases, owners can collect the system data using software tools that define and group the members of a system.  

Once teams have identified, located and inventoried the installed equipment, they can build out the information needed as construction progresses by collecting and organizing data from project submittals, test and balance reports, product manufacturer’s recommended operation and maintenance procedures and warranties.

To get started, UCSF reviewed the assets in their CMMS and specified the 10 percent list in a data dictionary. It served as a hub for providing requirements to the general contractor and project trades, the criteria for automated equipment data verification using software developed in-house and mapping the data from the source BIM authoring system to the target CMMS. The data dictionary specified:

• Equipment naming conventions

• Equipment classification

• The list of attribute sets/attributes that describe each kind of equipment

• Acceptance criteria for equipment nomenclature and attribute values

• Entity mappings of from source CAD and BIM to IBM Maximo

Further, they committed to weekly engagement and planning with the project team over an 18-month period to work through:

• Thirteen FM data milestones to support incremental data development, verification and delivery. They were coordinated with the construction project delivery plan and specifically the dates by which the trade-contractors would have the right data

• The specific data sets for each milestone

• Early identification and sharing of the model(s) of record for turnover to UCSF, which permitted the BAT to test viewer integration in Maximo

• Uniquely tagged equipment instances for unique identification

• Qualifiers for data extraction from CAD/BIM into Maximo

“Breaking down the requirements to small data sets for each trade, filtered further by the data required for each project milestone, transformed a large problem to small problems,” said Mace. “It helped the trades plan and execute data collection as part of existing tasks, the BAT to verify data iteratively in small batches and catch problems early, and the entire team to manage changes that occurred as the project progressed.”

Data transfer

The BAT transferred data from CAD/BIM to Maximo using COBie, (Construction Operations Building Information Exchange), a part of the United States National Building Information Model Standard (NBIMS-US V3). The standard supports the exchange of equipment data effectively and represents a step forward in the transition from paper to digital workflows; however, on a large complex project such as PCMB with multiple data delivery milestones, the project team encountered issues getting the trade contractors to configure COBie tools reliably for output from BIM and input to Maximo across milestones. Additionally, they learned that they needed verification control for each data record rather than at the COBie file level. These issues led to re-work and post processing of COBie files before loading them into Maximo. This learning led to advances in BIM FM requirements.

Visualization

Visualization helps FMs and engineers identify and locate equipment to understand how it can be accessed. Given the system connection information, it is possible to quickly walk back from a problem in a room to the control equipment and understand the impacts of shutdowns during regular maintenance and incident response.

For PCMB, the BIM requirements included transfer of an as-built BIM synchronized with data imported via COBie into Maximo, and the coordination of the visualization of asset data in a BIM viewer with the data in Maximo (see Figure 4).

 

The benefits of this early project engagement resulted in verified and trustable data flowing into Maximo a year before the authority having jurisdiction issued the temporary certificate of occupancy (TCO). During commissioning and startup, the building engineering team could focus on learning to operate and manage the building rather than collecting data about it. Maximo for PCMB was up and running before UCSF began treating patients rather than the typical lag of six months to two years for large and complex projects.

“Building a digital twin is a big endeavor, the digitalization journey it entails, puts us on the path toward eliminating outdated and wasteful turnover processes,” said Mace. “But with the right team, process and enabling technology and platform, you can empower FM teams with information to prevent downtime, save money and take advantage of rich Building Information Modeling.”

What’s next?

UCSF Health is working to achieve the UC system’s goals of net zero carbon emission by 2025 and for new acute care facilities to meet energy efficiency benchmarks based on industry standards. They have several projects to achieve the objectives and take the organization the third level of digital twin. These projects include:

• Conversion of pneumatic building controls to digital, to obtain data about equipment that can be acted upon by software, and

• Transition to an open building management platform that stores performance data from any BMS in a data repository accessible through standard application programming interfaces. The goal is to save over $500,000 annually on energy costs and achieve a compelling ROI within four years.

Initially, the BMS will generate work orders that flow into Maximo. Ultimately, UCSF seeks to integrate the Digital Twin Level 2 and 3 data, including associate the performance data with building location, equipment and system data obtained from capital projects and maintenance history data, so they can analyze real time performance feedback against historical benchmarks. The open BMS takes advantage of an open source initiative called Project Haystack to standardize the data models and web services for the Internet of Things. Integration will require an effort to harmonize and map across the BIM FM and BMS data models and web services.

Key management issues faced by UCSF

Breaking down silos

The models and data that comprise the PCMB digital twin are used by UCSF capital projects to support facility changes and by the engineering team to operate and manage the facility. These organizations also share goals to manage data changes and digitize information regarding more than 100 existing buildings in the UCSF Health and campus portfolios. Such goals require a strong strategic and long-term management commitment.

Recognizing the synergies, Mace leveraged the PCMB effort to educate UCSF management starting with the top level about new capabilities and successful implementation. Subsequently, UCSF is taking steps to break down traditional silos between CAPex and OPex, including:

• allocation of CAPex budget for BIM FM budget on new projects,

• early engagement of the building engineering team on new capital projects, including leadership to define BIM FM requirements and business processes for data standards, extraction, verification and load into UCSF systems,

• definition of change management workflows, identification of shared responsibilities, and allocation of shared resources across the business divisions to manage changes.

New roles

These activities require new roles and skills to be successful and a high level of cooperation across business groups.  

The UCSF BAT is a permanent group that collaborates regularly with staff from the capital projects division who are responsible for maintenance/updates of architectural BIM. Notably the BAT includes two roles not found in most building engineering organizations, the FM data manager and software application architect. The FM data manager is responsible for managing ongoing data changes, whether initiated from the BIM or the CMMS. The application architect is responsible for designing the technical platform that will allow UCSF Health to integrate the various levels of the digital twin across applications/systems. Although the conceptually of a digital twin is thought of as one system, it is comprised of several integrated systems, including the document management system where BIM models and other facility documents are stored, the CMMS, computer aided facility management (CAFM), and the BMS (eventually). Though technology providers are working on integration frameworks, dedicated effort is still required to stitch the systems together, as noted in the discussion about the open BMS, CMMS and BIM FM environments.    

Digitizing the existing building portfolio

Though capital projects provide opportunities to leverage digital information from design and construction, owner-operators must also consider how to digitize information about their existing facility portfolio. There are several approaches.

One option is to take a 2D approach that takes advantage of existing 2D plans, often represented in PDF format. One can identify equipment on the plans that represent assets the owner seeks to manage and index the 2D symbolic elements to a database of assets and systems. This approach permits owner-operators to access data associated with existing plans.  

There are also image generation technologies, for example laser scanning and 3D photogrammetry. Presently these technologies are limited to line-of-sight. They do not capture equipment in walls or above ceilings, and the costs of transforming of the image data to BIM is high, involving manual steps. Promising research, for example, points toward a future where the transformation step will be automated using ML/AI.

Maximize value

The UCSF Health story demonstrates it is possible to create value as measured in acceleration of the data load into Maximo and improve mean time to resolution for planned maintenance and incident response through use of asset and system data and visualization in facility operations and management. They have been able to achieve Level 1 and 2 Digital Twin at no increased cost to capital projects, though management leadership and commitment to early engagement. Additionally, as UCSF moves to digitize its BMS, they project significant annual energy savings through improved facility sustainability.

These initial steps are grounded in demonstrating value. They are steppingstones to achievement of higher levels of the Digital Twin, which will entail significant investment and the potential of higher efficiencies.