Digital Twin — A Digital Transformation Essential
During my visit to Washington DC, a few years ago, I happened to visit the ‘Air and Space Museum’. When I was inside the museum, I saw Ads for an IMAX 3D movie called ‘Hubble’ narrated by Leonardo — di Caprio. As a fan of Leo, I bought the tickets, without a clue what it was all about. As the film rolled on the screen, I realized it was a real time version of how the astronauts of the Space shuttle ’Atlantis’ fixed a problem in the Hubble telescope and successfully made it work for the future generations, to study the beautiful space images, it was capturing and sending to earth. It is vital to mention about the mode with which the astronauts fixed the problem. The crew of ‘Atlantis’ managed to do the repair based on the instructions from their counterparts in NASA Ground Station, located at Houston, Texas. The scientists on the ground, who initiated the whole mission, provided the consolidated schematics and step by step instructions on how to do the repair of the Shuttle. As the movie ended, it was surprising to watch all the kids in the IMAX theatre stand up and applaud the space team who completed the mission successfully. I’m not ashamed to accept that I also joined them, in giving a standing ovation to the team.
As I walked out of the theatre, two ideas crossed my mind-
· The great feet achieved by humans in Space Technology
· The Digital Twin (DT) technology which has been instrumental in making it happen.
DT definitions and Details
Digital Twin is a technology for creating a virtual product or a service for a physical service or a product. It is a virtual representation of a physical environment which collects real time data through sensors. This technology allows queries to determine the available capabilities and helps to improve the efficiency of any physical equipment or building by inviting a solution. This solution, in turn, provides a greater understanding of the complex interactions and high — level intersections of digital and real world. Today’s expectations of fast, higher quality products with accurate customizations can be achieved using Digital Twin technology in every field. In simpler words, Digital Twin technology enables a digital model of a real equipment, electronics, buildings or any other environment, in such a way, a change in the physical world, reflects on the virtual world too and hence any successful solutions offered to a problem in a virtual world, will be successful in the physical world too.
“The ultimate vision for the Digital Twin is to create, test and build our equipment in a virtual environment,”
says John Vickers, NASA’s leading manufacturing expert and manager of NASA’s National Center for Advanced Manufacturing.
“Only when we get it to where it performs to our requirements do we physically manufacture it. We then want that physical build to tie back to its Digital Twin through sensors so that the Digital Twin contains all the information that we could have by inspecting the physical build.”
The basic backbones of Digital Twin technology are Cloud storage of Zettabytes [UC1] of data and AI algorithms to collect and analyze data, and IoT capabilities to build a virtual model of a physical environment, which interacts with real people, and things. With the advent of 5G connectivity, Digital Twin can be made possible in all environments, universally.
The Digital Twin architecture consists of six major steps in order to provide the best possible solutions for any real time problem.
1Data Collection: The data is collected from all the IT and OT systems, which contains relevant information about the asset (equipment which needs fixing, for example a faulty valve in a pressure monitoring system) and the other related equipment. The data from different sources are pictorially arranged using element graph, where the picture shows the various links between the equipment, the asset, people and the processes. Based on whether the data is discrete [UC2] or continuous[UC3], the element graph will enable the visualization of how the other entities are affected by a failure or an issue in an OT or IT system.
2 Data pipelining: Data pipeline is a process engine which applies transformative logic on the various types of data [UC4] collected and sends data to a database such as a Hadoop file or an Oracle database. A good data pipeline should be configurable, scalable, queryable, testable, should generate alert notification and should have a low event latency. The best data pipeline infrastructure, in this server less era, enables the querying and analytics at the same instant as data gets transported from the processor drive to the data lake promoting no latency ultimately. All the data sources are integrated to establish a solid relationship between them before they are sent to the data lake.
3Data Integrity: The job of ETL or constructing an element graph becomes useless, if the data used to construct the same cannot be trusted. So, it is essential to check the data for any missing attributes, corrupted fields, or suspicious information before they are used in building an analytics DB. Also, at this stage of the DT architecture, it is a must to ensure, the ‘one to one’ representation of the physical equipment to the virtual model which is developed, taking into account all the small and big features of the physical world into the virtual world to make it a perfect data replica, and also represented in a form amicable to the analytics tools.
4Data Analytics: This is the most vital feature of a DT architecture. Here the data collected and represented in the formats required by the analytics tool, is used to
· provide the fix for the current issues,
· provide the predictive maintenance
· provide the prescriptive analytics to handle a future issue etc.,
The issues and the solutions vary depending on the use cases, nonetheless all the solutions provided make use of the deep learning and reinforcement learning algorithms applied on the big data provided in a format suitable for analysis.
5Data Export: Once the decision or solution is made, the corresponding data needs to be sent back to the physical systems from the virtual digital world. This can be done in two ways:
· Continuous mode: Continuous streaming, such as in augmented reality solutions, where the instructions to fix a valve or pressure monitor is available as a streaming video.
· Batch mode: Here the solution can be aggregated in batches and sent as bursts of data from time to time. This is used in the applications where the solution is not an emergency fix, but a scenario which might happen in future, and which requires attention at a later period and not instantly.
6 Feedback to change: The whole idea of Digital Twin is built on the notion of ‘Dynamic data or continuously changing data’. Neither the physical equipment nor the digital replica is static. So as the physical world respond to operational and regulation changes, the digital world has to evolve to these changes in accordance with the physical changes. This is often referred as ‘Readiness to Change’ in DT terminology. This readiness is ensured by running any new data from the physical counterpart through the entire process of DT architecture from Data collection to Data analytics and reported back to the equipment operator about any issues detected due to the change or any missing information to do the analysis due to the change and accordingly fixed.
DT Service Systems
Besides describing the simplest architecture possible for a DT service, I want to emphasis a vital information about the same — the DT architecture changes from application to application and depending on the basic requirements of the organization too. But all the DT Service systems consists of the following-
· Physical Component consisting of sensors, equipment, network and other related ‘Things’ including Service management component such as connectivity services and cyber security services and management.
· Virtual Component comprising of diagnosis based on behavior modelling, prognosis based on predictive performance modelling, computing based on semantic logic and relational modelling, and intelligence driven by data and knowledge acquisitions, AI algorithms and Cloud Computing.
· Interoperability component which consists of data analysis, 3D visualization, a decision driver, System integration and Optimization and finally a change management feedback component.
Digital Twin technology has become an essential part of digital transformation journey in every organization which adopts the same. Digital Twin is similar to simulations [UC5] , except simulations are done offline for design purposes, whereas Digital Twin is real time online data to visualize the entire life cycle (design- build- test — install — service — decommission) of a product
This video is the Willow Digital Twin architecture
This video explains how Microsoft Azure and their Digital Twin partner Willow Inc helped Thyssenkrupp Elevators in making a Digital Twin for their new building, Rotweill Tower. This is the pilot project which might extend in the future in such way, every new building constructed might be accompanied by a thumb drive, explaining everything in and about the building, making a whole new (view) path to digital technologies in building tech and smart cities initiative.
The basic differences between the Traditional data analytics and Digital Twin technology are not excessive, yet quite significant
Gartner’s report (2019) on Digital Twin adaptation, as a part of digital transformation effort, describes three major aspects which the CIOs of the organization must focus on:
1) Objectives: The DT technology is expensive hence understand clearly, the expectations out of the Digital model. Make a check list with two columns- Major objectives and Minor objectives. Convince yourself, the objectives are worth investing before trying to pitch your idea to the Board or the CEO. If the major objectives can be met with basic indicators from sensors even on critical performance issues, then do not even think about Digital Twin.
2) Digital Transformation readiness: Unless you’re completely convinced about the organization’s ability to adopt Digital Transformation, do not worry about Digital Twin, because Digital Twin technology is more effective only as a part of ‘IoT’ than Standalone.
3) ROI: The ROI might take a few years with Digital Twin, so use proper metrics to measure the progress on the Digital Twin initiatives by using Key Performance Indicators.
The eight major applications are Digital Twin makes the technology as not a mere digital trend, but an essential feature of the futuristic digital world
Besides being mentioned in Gartner’s top ten digital trends in 2018 and 2019, many organizations are still hesitant to adopt the technology due to its cumbersome nature. Tesla, for example has invested a huge amount in making a Digital Twin for every Tesla Car on the road. It’s lot easier said than done, since digital replica needs to be made in a ‘one to one’ correspondence to the physical model. It needs to replicate the exterior of the car including the cameras, ultrasound sensors and the mirrors, the interior of the car such as seats, wheels, brakes etc., and the under-the- hood components including engines, cylinder shafts, valves, plugs, everything. DT needs to note down the data collection points from the sensors on which the decision-making analytics are based on and the deliver the response of the DT, to the actuators in real time, not to mention all of this should happen within a few seconds of delay. Hence it is impossible for every application sector to adopt DT instantly and start benefitting out of it. Also, it may take time, for the benefits to show in big amounts. It is expected from 2027 and beyond, nearly 95% of world’s large organizations might have adopted Digital Twin as a part of their digital transformation journey and many of them might even be seeing the ROI from it.
DT as a smart product
While discussing ROI, Digital Twin can also be visualized as a connected/ smart product, and it is basically formed using data and models. It can be represented as a three- layer stack.
· The IoT Value proposition at the upper layer represented by the knowledge experience and information sharing to other processes,
· The intermediary layer consists of all hardware required to collect the data from sensors and transmit it back to the actuators and the software to implement the machine learning algorithms and optimization algorithms and the
· physical product as the base layer.
Future of DT
In the future, Digital Twin is expected to be adopted more in product development and product management aspects as well, as I mentioned above where a DT can be represented as a smart product, thus every manufactured product will have its Digital Twin and the data from the real product will be collected and analyzed immediately. This concept is widely referred as ‘Digital Triplet’ and this is similar to the one I mentioned above in Tesla autonomous cars.
The triplet part will be played by the team which examines
· examines every piece of data acquired from the sensors of all the equipment instead of one Digital Twin replica (but multiple replicas for each physical model),
· experiences the knowledge from the data and
· applies the knowledge acquired in evaluating projects and making decisions.
This triplet component is expected to have a nomenclature such as ‘Intelligent Activity World’, where the humans solve every problem on the world, rather than merely being based on computer automated activities.
[UC1]1000 Giga bytes = 1 Terra Byte
1000 Terra bytes = 1 Petabyte
1000 Petabytes = 1 Exabyte
1000 Exabytes = 1 Zettabyte
1000 Zettabyte = 1 Yottabyte
[UC2]Discrete data or data used to create a digital model, refers to the type of data which has a start and a stop. For example, a log data, where when a person logs on to the system, he or she sign in at a time and signs out at a time. So data is discontinuous. In other words, the data is available for only a specific period of time.
[UC3]Time series data represents continuous type of data where the data , whether recorded or not recorded , is available continuously . For example, a weather data, irrespective whether it is recorded or any the temperature rises or decreases, another example, stock market, irrespective of whether we watch it or not, the stock prices, rise or fall everyday.
[UC4]The common types of data handled by any data pipeline are 1) Raw data -this data is a data in transit before applying any processing or schema on it. 2) Processed data- This refers to the data in the data lake after applying some event specific formats or some schema and stored in an event table 3) Cooked data- This refers to a data which aggregated or summarized and processed with time stamp on it. Say for example — a data captured between a log in and log out time.
[UC5]Simulations are used in the past to understand what may happen in a real world while driving a car or piloting an aircraft or handling a heavy machinery. It was more like a training method which incorporated real time environment, based on the input already available from the past years. Using this technology, engineers, pilots and supervisors learned what to expect in the real world and the methods to handle a situation in real time. Nonetheless Digital Twin technology is used in real world environment by experts, who work with these equipment day in and day out operations. When these experts get stuck and do not know how to proceed, they use the DT technology to provide them with answers.