Digital Twins in the Chemical Process Industries

Article by Joanne Tanner AMIChemE and Colin Newbery FIChemE

Joanne Tanner and Colin Newbery discuss the hype and the hope for digital twins

DIGITAL twins (DTs) are everywhere, and everyone needs one! Or so our solution providers would like us to think. But are digital twins really all they’re touted to be? When built well, and applied and used correctly, they can significantly improve real-time process optimisation, operator training, asset construction and integration, and more. However, they are also complex, expensive, and require maintenance, just like our physical assets. There is no one-type-fits-all for a digital twin (though it seems that everyone is trying to sell one!). In this 6th article in the DigiTAG series, we look at some aspects and applications relevant to chemical engineers.

Firstly, we should define the term: A digital twin is a digital representation of a potential or actual physical asset, process or system. The digital twin is synchronised with the actual physical twin via two-way real-time data streams for online “live” prediction and optimisation. Artificial intelligence and machine learning methods can also be integrated into the digital twin to support both human and autonomous decision making to change the physical asset’s performance and behaviour.

Data management, including cleaning and gap management, is critical to the accuracy and fidelity of a digital twin and the calculations it performs. As with all models, the usefulness of a digital twin relies on fidelity, which sometimes makes a complete-system twin appear daunting. However, not every aspect of the system being twinned needs to be considered or considered in its dynamic form. Different applications and different project stages will require different digital twin types and behaviour, and the priorities as to what aspects of the asset benefit from being twinned in a static or dynamic form (or at all!) and to what level of fidelity will change over the asset lifetime. Figure 1 summarises the types of digital twins1, which can be applied in various combinations to various parts of the asset or process to generate the overall twin.

Figure 1: Digital twin classification [1]

Stages of asset life

All physical assets change in their lifetime from concept stage to FEED and detailed design through construction, testing and commissioning, and into operation, eventually being decommissioned. A digital twin can mirror these stages to allow experimentation, scenario testing, and optimisation. During these stages of life, different attributes of the asset become dynamic and are therefore more important (and more interesting) to digitally twin. During construction, the key dynamic attributes will be linked to material availability and location, the installation progress or costs, whereas the feed, product quality and process variables will be the key dynamic attributes during operation. The real world and modelled values of the asset attributes are linked by the digital twin. Looking outside of the plant or process, there are also opportunities to link with external real-time data. For example, the electricity market price or commodity prices may be key influences on how the process is analysed and optimised.

Digital twins in chemical engineering

In terms of the asset, process or system being twinned, there are a variety of applications that chemical engineers might encounter, including process, asset and supply-chain digital twins. These three types are often linked, and their configuration, connections and dynamism depends on the stage of the project. Some of the internal and external aspects of a typical project that can modelled and linked by digital twins are shown in Figure 2.

Figure 2: Internal dynamic asset attributes and external data sources that could be modelled and linked via a digital twin

The best understood and most common type of digital twin a chemical engineer will encounter is the process digital twin – an integrated, real-time digital version of the plant. The process twin replicates the flows and other state parameters of the process, and can be used for operator training, plant optimisation and remote support. Historical process information (eg reported hazards, manual process measurements or observations made previously) can be overlaid on the process twin to assist with problem solving. Live data streams and links to augmented reality (AR) can be used by operators for maintenance and troubleshooting. The process twin can also be interlinked with an asset twin.

An asset digital twin is a replica of a physical asset, equipment or plant. Asset twins are most dynamic during construction and can be used in many ways, including optimising schedules, modelling cash flows, or tracking embodied carbon. An asset twin can be useful from the FEED stage of a project through to decommissioning. During operation, the asset twin may revert to a static model, or it may remain dynamic to model asset deterioration related to use, eg vibration, corrosion, wear-and-tear. If connected, the process and asset twins will exchange data, and can be used in conjunction for condition scenarios and optimising asset maintenance based on the original build specifications and the process conditions under which the asset is operated over time.

It is also possible to link digital twins for an entire supply chain by modelling individual assets then exchanging the interdependencies between them2. These models may represent the manufacturing and logistics network of a company, or perhaps the utilities network across a region or country. Once supply-and-demand forecasts are introduced, it is possible to generate insights or undertake complex optimisations to support decision making. Supply chain digital twins are complex (Figure 3) but can be used at many levels of the organisation for strategic planning, resilience and risk mitigation, and day-to-day operations.

Figure 3: Digital twin connections to supply chain components [3]

Behind the scenes

The simulations forming the digital twin can be mechanistic, that is, based on expert equations, or data driven. Mechanistic models have the advantage of being based on developed and proven relationships and algorithms, and models used in design can often be adapted. This means that the digital twin has a broader range of applications, including: scenario planning, predicting outcomes of events not previously encountered (eg reactors being taken off-line); testing and tuning controls; and testing process modifications. Data-driven models are relatively less complex and faster to configure and run, and they can be used where processes are too complex to be described in equations. Either approach has advantages that are application dependent. However, often the best approach could be a hybrid, where the most appropriate modelling technique is selected for individual process units or tasks.

User interface

How users interact with and view the DT outputs will often determine how usable and successful a DT is. Coupled with the powerful visualisations that are possible, a 3D image can be overlaid, coloured or animated with the asset and digital twin data. In the case of a physical asset, the image and associated data can also be superimposed on a video feed of the physical system when the device being used to interact with the digital twin is held up to the physical asset. Familiar interfaces using digital twins’ outputs can be applied to a variety of purposes, such as simulation of the SCADA/HMI with pre-programed scenarios for operator training, data-coupled standard operating procedures (SOPs) for maintenance, and to enable remote assistance by experts who can see the data overlaid on the asset for training or troubleshooting. The level of digital twin detail that is visible via the interface can be adjusted to suit the purpose and multiple interfaces or access levels can be implemented for a single digital twin.


Maintenance is critical to the effectiveness of a digital twin. The digital twin is an asset, needing as much care as, and should be maintained alongside, the physical asset to remain relevant and useful. The same is true for offline simulations, but is even more critical for digital twins as they are connected directly to the process. At each stage of the asset life, and therefore of the twin life, someone needs to take ownership of the digital twin, making sure that updates and asset changes are captured. Maintenance can be partially automated eg AI/ML algorithms can be connected to and use lab data for periodic automated calibration to keep the digital twin up to date, but even automated updates still need human oversight and periodic sanity checks.


Digital twins require multiple endpoints or system connections, each of which is a potential cybersecurity vulnerability. Some precautions to consider include: data encryption, access privileges (including a clear definition of user roles), least privilege principles, addressing known device vulnerabilities, and routine security audits. As with maintenance, cybersecurity matters require ongoing attention and stewardship.

Article By

Joanne Tanner AMIChemE

Lecturer at Monash University, Australia and Member of IChemE’s Digitalisation Technical Advisory Group (DigiTAG)

Colin Newbery FIChemE

Member of IChemE’s Digitalisation Technical Advisory Group – DigiTAG

Recent Editions

Catch up on the latest news, views and jobs from The Chemical Engineer. Below are the four latest issues. View a wider selection of the archive from within the Magazine section of this site.