By Curt Lefebvre, CEO, nDimensional
I recently delivered a presentation to the American Society of Mechanical Engineers (ASME) Industry Advisory Board about Artificial Intelligence (AI) and digital twins, two digital transformation hot topics about which ASME advises its members.
The topic was especially timely against the backdrop of the COVID-19 crisis, with organizations recognizing more than ever the criticality of digital technology, data analytics and AI in ensuring smooth, resilient and increasingly remote operations.
Digital Transformation Needs AI
The push toward digital transformation is increasing the volume, velocity and variety of data that needs to be analyzed and acted upon to drive transformational outcomes. Artificial intelligence (AI) and machine learning (ML) are required to manage that complexity and accelerate processes that would otherwise get bogged down by people using traditional analysis tools. For example:
- Automating or augmenting repetitive manual tasks and analytic processes
- Classifying complex data
- Detecting anomalies in data
- Predicting the probability of future events and root causes
- Accelerating the insight to action cycle
Over the last two decades there have been significant advances in AI to the point that, when there is sufficient data, computers can now outperform humans at almost any task – from games of Chess and Go to fraud detection. There have been many successes in asset-intensive industries in particular, such as:
- Optimizing power plant boilers to reduce emissions
- Predictive maintenance of industrial equipment to reduce unplanned outages and lower costs
- Reducing energy and water use and reducing pollutants during manufacturing processes
- Detecting leaks in water, oil and gas distribution networks to reduce losses and improve safety
- Locating failures and faults to avoid or quickly restore electric grid outages
AI Struggles to Scale
So, AI has great promise and a proven track record – yet many organizations are still experimenting with AI proofs of concepts (PoC’s) rather than delivering repeatable business value at scale. Many get stuck trying to transform PoC’s into full production applications while others fail at scaling applications across their organizations.
A Gartner study found that the top two AI/ML adoption hurdles were:
- Lack of skilled staff, which typically refers to lack of computer and data scientists – people who understand AI but know nothing about the business domain in which it’s being applied.
- Inability to understand AI benefits and use cases, which requires domain experts – those who understand the business needs but not the details behind the AI.
The good news is that AI is becoming more and more automated so domain experts can now apply it without needing to understand all of the technological details. The bad news is that industry has yet to see this and keeps competing for a small pool of data scientists, before realizing they cannot relate to their business problems.
Part of the issue is that AI is often waved as a magical wand – at all problems, not to be impeded by human intelligence. We forget that while AI is incredibly powerful, actual intelligence is always better. In other words, start with what you know and use AI to augment what you don’t.
Finding the best mix of AI and human intelligence and structuring it for repeatability, scalability and transparency is the holy grail that will enable AI to reach its true potential.
Digital Twins Bridge The Gap
Digital twins can be that holy grail, bridging the gap between AI and business value at scale. Although there are many definitions, to me a digital twin is the result of packaging all available data, domain knowledge, models (physics-based and AI), performance calculations and visualizations for connected assets and processes, and putting to them to work on delivering business outcomes, 24/7/365. With their focus on real world systems, digital twins deliver AI proof points with large and quantifiable business value.
The concept is simple:
- Determine business use cases and related assets or processes
- Codify domain knowledge
- Use AI to fill the gaps in “real” intelligence
- Continuously validate against real-time/near real-time data feeds
- Deploy across multiple assets, systems and processes
- Tirelessly monitor, predict KPIs, discover anomalies, detect events, classify root causes, optimize against ever changing market conditions and adaptively plan for capital improvements.
The result is a digitalization of key business assets and processes, enabling a transformation of the enterprise.
Digital Transformation Delivered
Several digital twin characteristics make them well suited to drive digital transformation success where other approaches have failed.
For-Purpose and System-Wide
Part of what makes digital twins so effective is they’re built for purpose, focused on specific use cases and business value. Unlike generalized AI or IoT, they are much more tangible, relating to specific pieces of equipment or processes and enabling quick win benefits.
Yet achieving large-scale business outcomes requires us to take a systems perspective. That means creating twins of twins that focus on the systems formed when individual pieces of equipment are interconnected. For instance, how a boiler affects a turbine and vice versa is just as important as either one individually, as is the downstream impact they both have on cooling towers. The good news is once a digital twin is created, it can be easily connected to other twins and repurposed for other use cases, across equipment, processes and even organizations. The value continues to build over time.
Standardization and Automation
Another way to think about digital twins is as a standardized approach for packaging intelligence about and interacting with complex systems.
- A standard way to codify and package domain expertise about assets, classes of assets, processes or enterprises and deploy it against real-time data streams for continuous intelligence.
- A standard way to add analytics, visuals and AI to complex systems to make them easier to understand, interact with, maintain and scale
Standardization in turn enables automation, both of which are critical for speed and scale.
Democratization is not inherent to the digital twin concept. But digital twins and AI will never reach their full potential if their creation and application rely solely on computer and data scientists. Their power must also be put in the hands of domain experts who know how to extract business value. Democratized digital twin approaches that provide standardized, automated structures to capture domain expertise, add AI and easily apply these “augmented intelligence” twins to real- or near real-time data streams are required for industrial digital transformation success.
Check out nDimensional’s digital twin checklist for other critical success factors.
Domain Expertise Is The Secret Sauce
I fundamentally believe that AI and digital twins, two topics generally associated with the IT and computer science domains, can’t be successful at the scale required for digital transformation without engineering and business domain experts.
That’s why organizations like ASME, with its members’ vast industrial domain expertise and codes and standards experience, play such a critical role.
Share Post: Tweet