If it had been a What-if scenario in the past, a retail tryst with Intelligent digital twin (IDT) goes to show the kind of impact the IDTs are having on the business world. Enters a customer into a Retail store, and there begins the pivotal role of a digital twin (DT) in extending an enriching customer experience. As the digital display latches on to the customer’s mobile phone, intelligent digital twin is busy unearthing the customer’s purchase of apparels. DT further probes into social media account which brings to light the customer’s love for a specific fashion apparel.

Digital twin has moved on from being bracketed as the virtual copy of a physical object. Digital twins now exhibit increased intelligent quotient, with the data and AI crowning DT with intelligence. The Intelligent digital twin story in the retail store was instrumental in creating a smart digital display that champions delightful customer experience and profitable outcomes.

How did Data & AI pave way for trustworthy Intelligent Digital twin?

What’s a Digital Twin?

If we make a virtual model of a physical object or a product, the virtual model designed to reflect the physical product is recognized as the digital twin. And digital twins have spread so far as to be used as different types for different purposes.

  • Asset twins – When components (two or more) are twinned, we get an asset twin
  • Process twins – When workflows and processes are twinned, we get the process twin.
  • Component twins – When a functional component is twinned, we get a component twin
  • System twins – When a system comprising disparate products gets twinned, we can build the system twin
  • Product twins – When interoperability of parts pertaining to a product get twinned, we can raise a product twin.

Data & AI push for Digital Twins

You cannot stereotype digital twins as an engineer’s toolkit for enhanced product design anymore. If you want answers for a potential future scenario, intelligent digital twin steps in and leads the way for profitable decisions. For instance, in the supply chain area, IDT can help businesses ride over supply chain disruptions and meet future demands, with data offering the lead.

Today, Intelligent Digital Twin is underpinned by data. With increasing flow of data and sources, every datapoint serves to enhance the intelligence of DTs. The twinning object and data sources have given rise to this transformation. Take Healthcare domain for instance. You can twin the human body with DT input sources like sensors, wearables and RFID tags feeding data. In the aeronautics domain, you have cameras, actuators and smart sensors providing data inputs.

Let’s take the ‘aircraft fatigue’ case to probe the data connection to maintenance hot spot prediction. Here’s a case where data around depot maintenance, corrosion and crack findings from fleet, and customer feedback are plotted using a Digital twin to zero-in on the inspection areas, with the result that a data-led digital twin gives way for the aircraft fatigue prediction.

Key Components of the Data-led Intelligent Digital Twin

Data from the twinning asset and other associated data that helps throw light on a future scenario form the critical feed to raise an intelligent DT. With the data start, analytics aided by AI, machine learning, and deep learning provide the intelligence quotient to the digital twin. The figure below captures the foundation and the flow of Intelligent Digital Twin, from data to intelligence integration augmenting specific business areas.

And there is an X-factor to this Data & AI powered Digital Twin.

XAI brings the ‘Trusted’ Label to Intelligent Digital Twins

The trust factor of intelligent digital twin is also going up with the use of Explainable AI. Let us examine the case of a patient-specific digital twin in Healthcare domain. Adopting XAI, we can use the LIME library for a patient’s heart disease prediction. You can look at the explanations thrown for individual predictions captured in the table below.

XAI works here by way of providing explanation to symptoms through relative weights. When weights are being attached to individual symptoms, it illustrates whether a symptom contributes to the prediction or if they are symptoms that turn out to be evidences against the prediction made.

Note: The figure above is an attempt to best portray the explanation of features’ impact on predictions.

The Lime library disease prediction is supported by the features highlighted in orange colour, while the features in blue colour stand as evidence against the prediction. The explainability factor enabled by XAI is one thing, and the cognitive sensing facilitated by the Digital twin is another.

Rise of Cognitive Digital Twin

Here’s a cognitive twist to a Petro-chemical plant. The plant suffered shutdowns because of Process ‘trips. In comes the cognitive digital twin to help the plant detect process variation and take right steps at the right moment. Cognitive sensing was infused into the process through NLP which helped extract knowledge from several documents and software systems, take preventive action before a process trip could cause production delays and losses.

Cognitive sensing is revolutionizing machine manufacturing with NLP, acoustics analytics, visual recognition and signal processing allowing manufacturers to build machines for the target market as envisioned by the user. Given a plant equipment, you can use the combination of predictive analytics and cognitive computing to run what-if scenarios to remodel or model the equipment’s performance.

Any object or an individual can now be twinned, and twins can be interconnected with twins. There are confidential computing techniques that are emerging to promise privacy-preserving twins. And as data proves to be the key feed, new data sources like FLIR or forward-looking infrared promote capture of real-time data with emerging tools empowering technicians to build complex what-if simulations. And the potential use cases that need to be tapped have only reiterated the promise of Intelligence Digital Twin-led future.

All examples quoted in the article, like the aircraft fatigue prediction have been included for the sake of better understanding.