"The benefits of digital twins for owner-operators are almost countless"
Last March experts from Bentley Systems, the infrastructure engineering software company, discussed the capabilities of Bentley’s digital twin applications at the NVIDIA GTC developer conference. This online event brings together innovators, researchers, thought leaders, and decision-makers shaping our world with the power of AI, computer graphics, data science, and more. Alexandre Vallières, Director of AI strategy and research at Bentley Systems, presented a session on AI-powered infrastructure digital twins.
With over 17 years of experience in entrepreneurship, strategic development, and systems engineering, Alexandre Vallières started one of the first machine learning service companies in Canada. Now at Bentley he helps discover and materialize opportunities to create value from digital twins and machine learning technologies. SWM interviewed him on occasion of his participation in the NVIDIA GTC to learn how AI and machine learning can help materialize the promises of digital twins for infrastructure systems.
What does AI bring to infrastructure digital twins?
Artificial intelligence (AI) is a catalyst for value creation from infrastructure digital twins. It helps lower barriers to the conception and maintenance of digital twins by automating data extraction and interpretation tasks. The end goal is to develop a “universal translator” that can interpret inputs from two universes — the “real” world and the “digital” world—to merge them into one reality.
Artificial intelligence (AI) is a catalyst for value creation from infrastructure digital twins
Additionally, machine learning also yields unprecedented opportunities to leverage the information attached to digital twins. It achieves this goal by providing tools to scale data mobility across the assets’ lifecycle phases. According to a study by Accenture1, this capability alone could unlock 35% to 65% of additional value from digital twins. But we’re not stopping there. We also observe advanced analytics solutions being utilized to increase global efficiency by assisting experts in optimizing design, planning, construction, operation, and maintenance of infrastructure. Ultimately, machine learning and digital twins are there to help us manage the complexity of our industry.
How do Bentley’s digital twin applications and NVIDIA’s Omniverse platform complement one another?
I see NVIDIA’s Omniverse platform and the Bentley iTwin platform as the lower layers on which users and solution designers can build an ecosystem of digital twin and metaverse applications. Omniverse’s computation solutions are powerful, modular, and flexible. They allow Bentley’s developers to adapt and integrate them as the foundational layer for calculation-intensive tasks that must be performed quickly. Bentley’s teams then add their own expertise in engineering and infrastructure software development to provide digital twin technological building blocks through the iTwin platform. Therefore, users and developers benefit from the strengths of both platforms to increase their productivity and to extend their value-creation possibilities.
How can AI-powered infrastructure digital twin technology improve infrastructure design, build and operations?
The applications for AI and digital twins in design, construction, and operations are almost endless. One of the challenges that we’re facing is to prioritize all the opportunities in front of us.
Machine learning also yields unprecedented opportunities to leverage the information attached to digital twins
However, I’d say that machine learning is being used to simplify and accelerate the completion of technical tasks while optimizing results. Usages can range from relatively simple jobs — such as automation of information extraction from structured and unstructured data sources, or automated asset inventory and condition assessment — to much more cognitively complex undertakings — such as technical diagrams interpretation, AI-assisted design with multiparametric optimization objectives (total cost of ownership, environmental impacts, etc.), or progress, quality, and safety monitoring on construction sites.
These are only a few examples. When it comes to imagining ways of creating value with machine learning in our industry, we’re mainly limited by data access, quality, and quantity. Lucky for us, one of the promises of digital twins is that, in time, it will help solve those data-related challenges.
What trends have you seen in the adoption of infrastructure digital twins in infrastructure engineering and asset management? What about specifically in the water sector?
Utilization of digital twins in the infrastructure sector is still in its infancy but is gaining momentum. We are seeing diversified use cases from planning all the way to asset maintenance.
When it comes to imagining ways of creating value with machine learning in our industry, we're mainly limited by data access, quality, and quantity
What trends have you seen in the adoption of infrastructure digital twins in infrastructure engineering and asset management? What about specifically in the water sector?
Utilization of digital twins in the infrastructure sector is still in its infancy but is gaining momentum. We are seeing diversified use cases from planning all the way to asset maintenance.
The use of digital twins usually begins at different points in each industry depending on where it’s easier to integrate data and where most of the value can be found. Engineering firms tend to start from consolidation of information, such as 3D models, 2D drawings, and schematic diagrams. Construction companies are tracking situation-based knowledge through completion of their projects. Owner-operators are optimizing asset usage by tapping into information from the built environment. Building on what I’ve already mentioned, there are a few more examples.
The use of digital twins usually begins at different points in each industry depending on where it’s easier to integrate data and where most of the value can be found
In the planning and design phases of the asset’s lifecycle, engineers — such as those involved in Shell Deepwater — are streamlining their processes by leveraging visibility and transparency of project and engineering data across their projects. Not only does this minimize surprises during conception, but it also provides integrated ways to complete what-if scenarios and optimize overall results. For example, we recently announced the availability of integrated workflows in the Bentley iTwin platform for lifecycle assessment and embodied carbon calculations.
Moreover, cities — such as Helsinki — are using OpenCities Planner for social acceptability of their projects. They share engineering facts with the public in an easily interpretable format.
Companies such as The Lane Construction Corporation are providing a single source of truth for all project information and data from their digital twins. SYNCHRO Control and SYNCHRO Field are then employed to access and control that information through one platform. This singularly improves project management efficiency, and it minimizes errors and issue resolution time in the field.
The benefits of digital twins for owner-operators are almost countless. We are seeing an uptick in assets’ operations and energy performance enhancement use cases fused with the digital twin. With COVID, workforce requirements have evolved, and digital twins are helping many organizations in space planning by combining motion sensors and building layout data. For the water industry, we see the ability that they provide of preparing for emergencies — with tests and simulations — and of delivering the contextualized information necessary to devise operations strategies in real time.
Águas do Porto, for example, is integrating more sensors to their digital twin. It has been able to reduce water supply interruptions by nearly 43% and volume of non-revenue water by 10%. They also track storm water drainage in real time across their territory. This capability improves the team's reactivity when overflow risks arise. They can act proactively to limit pollution of nearby streams and beaches. They’re now envisioning to include machine learning in their processes to accelerate decision-making.
What benefits would you highlight from the use of this technology in terms of project cost effectiveness? What about concerning collaboration between project stakeholders?
The benefits of digital twins for owner-operators are almost countless
Digital twins and machine learning are streamlining time-consuming, low(er) value assignments. The lowest-hanging fruit is associated with providing more efficient information exploitation capabilities. I’m sure that many of your readers have had to deal with the frustrating task of sifting through heaps of documents connected with a project to find and aggregate bits of scattered information required to complete a job. Some surveys estimate that engineers spend on average 8.1 hours per week on this tedious task. By providing a unique, centralized, and trustworthy source of information on assets across their lifecycle, digital twins have the potential to drastically improve that process.
The way that intelligence is made visible from the digital twin aids in making timely, and better-informed, decisions. Since all stakeholders are working from — and on — the same source of truth, collaboration is simplified. With rule-based engines and machine learning techniques, technological solutions can help avoid costly mistakes and rework by assessing in near real time how modifications on one part of the project could affect all other dimensions.
For operation and maintenance, digital twins and machine learning are opening a new world of productivity-enhancement opportunities. For example, inspectors can now (1) objectively quantify defects, (2) increase safety by capturing images with drones, (3) virtually revisit the inspection site to collaborate and to validate conclusions, and (4) monitor changes over time to identify anomalies and to optimize planning.
The biggest productivity impact of digital twins and machine learning will come from the new-found capability they will provide to shift focus from project cost-effectiveness to total cost of ownership
What’s been mentioned the most so far by our users, though, is that partly automating inspection tasks saves substantial time. For instance, we’ve developed an AI-assisted workflow to analyze inspection photographs of concrete linear assets for different industries. In one of them, this model cuts down the investigation time from 105 hours per mile, with the standard approach, to less than an hour per mile, with the AI-assisted technique. By reducing the inspection expenses in such a significant way, owner-operators can start dreaming of assessing larger areas and increasing survey frequencies. In turn, it will afford them with accurate knowledge of the state of the entire water network. They will be able to prioritize their actions with the certainty that they are putting their limited maintenance budgets at the right place, at the right moment.
Fundamentally, though, I think that the biggest productivity impact of digital twins and machine learning will come from the new-found capability they will provide to shift focus from project cost effectiveness to total cost of ownership, and global asset performance and resilience.
Could you tell us about a case study involving the implementation of an AI-powered infrastructure digital twin?
I’ll give a simple example of an application that was developed on the Bentley iTwin platform by the Center for Computational Technology (CCTech). This organization conceived a design optimization solution for heating, ventilation, and air-conditioning systems (HVAC) to achieve optimal comfort of occupants in buildings. To that end, they are using a physics-informed machine learning model in a digital twin to simulate air flow and heat propagation. It is a much simpler and faster approach to fluid dynamics calculations than physics-based models. Quick, simple, and cheap simulations create a new capability: conducting multiple what-if scenarios in an iterative design process. I’ll let you imagine what this capability could mean if it were adapted to address water industry challenges.