Digital twins are of great potential value, as we recently outlined in an article in ONdrugDelivery. There are already various examples of their benefits – such as optimising manufacturing in the Industry 4.0 era, creating virtual representations of organs requiring precision treatment, or as early indicators of product performance during design and development.

But as with other rising technologies, key questions remain regarding their application and the barriers to realising their full potential. Digital twins can be incredibly powerful but remain difficult and time-consuming to implement with many trade-offs to consider. As of now, the essential skill in effectively deploying them lies in drawing the appropriate scope for a model with dynamic multiphysics potential and pitching it at the right level of detail.

The digital twin concept

Since the concept was formalised by John Vickers in 2002, the term digital twin has been used liberally in discussions across various industries. Fundamentally, it describes a digital representation of a physical asset, whether real or conceptual, that is detailed enough be considered indistinguishable from reality.

This representation can be interrogated in various contexts to provide key insights about the physical counterpart. Within this definition are ascending levels of digital representation, which depend on the information transfer between the digital and physical assets, and how the physical asset utilises data generated by the digital model. Accordingly, the following definitions are typically used for various levels of digital representations:

• Digital model describes a digital representation of a future product used to prototype in-silico within the virtual intended use environment. It is a comprehensive, system level, multiphysics simulation and workflow that is manually informed and validated by a small number of targeted physical test cases and library data sources. It can be used to inform and optimise system designs prior to significant prototype investment
• Digital shadow is a digital version of the existing physical asset, incorporating the transfer of data from the physical asset to the digital model in near real-time from a network of sensors or enable a deep understanding of the current state of the system and to roleplay changes and ‘what if’ scenarios
• Digital twin defines a digital replica of an existing physical asset with automated real-time data transfer, as per a digital shadow. However, in this case the digital model can instruct the physical asset based on generated insights and scenario explorations to improve the performance of the physical system fully autonomously

It’s also important to outline the type of deployment of these representations. Their position on this scale dictates their application and what value can be achieved from using the tool:

• A digital twin prototype describes an artefact that does not yet exist. It is therefore entirely predictive and can serve as a blueprint to prototype the physical asset. As there is initially no real-world system to share data with, the digital twin prototype is necessarily a digital model
• A digital twin instance represents a single, specific physical entity that the twin remains linked to throughout its operational life. An instance may be any level of twin, but given the entity exists in the real world and is subject to change it is more likely a digital shadow or a full twin
• A digital twin aggregate covers a population of digital twin instances and can be used to understand global behaviour and provide insights across the expected range of tolerances and conditions experienced

Regarding this selection of digital twin tools, Cambridge Consultants – with our expertise in product development – actively focuses on exploring digital twin prototypes for use now and in the future, to accelerate the creation of new products and systems. There is significant opportunity to progress simulation-led design to the point where it can competently augment physical testing of prototypes. Within a product development cycle many more design iterations can be generated, undergo validation at earlier (and less expensive) project stages, be interrogated for KPIs traditionally difficult to determine, and leverage state-of-the art computing to optimise solutions.

Seeing the potential of digital twins begs the questions: which applications are most suited to them? And at what scale of development does implementation provide benefits outweighing the time and effort-related cost? Our work in healthcare, particularly the pharmaceutical sector, represents a great fit for digital twin prototypes, with numerous examples which could provide benefits across the product lifecycle.

The sector features products such as autoinjectors and infusion pumps which are complex electromechanical systems, with multidisciplinary engineering challenges which could be accelerated and expanded in scope through deployment of a digital twin.

Integrating a digital twin prototype

Integrating and using a digital twin prototype enables the prediction of key aspects of system performance at the point a CAD model is first designed and detailed, thereby shortening development time and reducing the need for physical iterations.

Additionally, it addresses key questions that arise during product development which have traditionally presented challenges, either through difficulty and cost of manufacturing prototypes, inaccessibility of critical performance indicators, or the challenge involved in creating appropriate test and use conditions in sufficient quantity and quality. Examples for an injectable device include drop testing and performance under varying environmental loads, use cases, and different patient types.

However, building a model in enough fidelity to provide these kinds of insights remains a complex task requiring considerable work and expertise. For example, converting complex CAD models of devices to multiphysics simulations is not trivial and requires significant effort to ensure models of key phenomena are an accurate enough substitute for testing the real thing.

Within injectables, integrating details such as the soft tissue modelling for predicting needle insertion behaviour also requires appropriate input data from an extensive but potentially limited range of sources to ensure a comprehensive and validated analysis.

Digital twin advantages and challenges

The advantages and challenges from deploying a digital twin should be assessed when deciding how and when to use such a tool. The key skill is understanding the different levels a digital twin can exist in and the types of problem where they become most useful to pitch the twin at the appropriate level.

For example, flow through a microfluidic pathway in a diagnostic instrument is likely a single-physics solution with minimal coupling, not meriting a full digital model as a more traditional simulation will suffice. Inhalation devices are more complex and could benefit from coupling the deagglomeration of dry powder inhaler particles modelled by the discrete element method (DEM) with CFD models of the resulting flow field during an inspiratory maneuverer.

The multiphysics nature of a digital twin prototype can provide more insights into drug performance and be further enhanced by incorporating downstream elements such as lung deposition and biopharmaceutical modelling. Regarding information transfer, digital twin instances and aggregates could offer value when developing cell-culture bioreactors by leveraging yield predictions from simulating a suggested feed input and automatically adjusting.

Pertinently, next-generation autoinjectors are complex devices interfacing with biological-tissue and incorporate drugs sensitive to environmental conditions. This scenario represents a multi-disciplinary problem with extensive coupling, meriting the effort of a full-scale digital twin.

Novel electromechanical autoinjector

In tune with the problems and applications considered, a multidisciplinary team here has been creating a digital twin prototype of a novel electromechanical autoinjector. Ansys CFD and FEA simulation, Solidworks CAD, and Python-based modelling of electronic systems have been deployed together to enable in silico exploration of performance across various product configurations and operating conditions within a virtual use environment that includes human tissue models.

Such a tool enables specification and optimisation of the system without the need for extensive lab testing, reducing development time, cost and risk compared to a traditional approach. We have been able to explore parameters such as motor selection within the context of performing a set of functions with our electromechanical autoinjector, then substituted in datasheet information from a real motor specification selected based on these performance requirements to ensure satisfactory working within the model.

This has enabled prediction of the use life and charging cycle of the battery, based on the full system performance across a range of likely use scenarios and enabled finding an appropriate part for the bill of materials. Imagine taking this further and defining a loss function to optimise the entire design for formulation delivery performance, cost, and sustainability simultaneously. These are possible realities within a product development cycle when deploying a digital twin prototype as developed here.

Significant learnings have been gathered already, showing both the potential transformational impact and simultaneously the challenges ahead before such a tool is a ‘plug and play’ in a simulation-led product development cycle of the future. Currently, building these digital twin prototypes takes significant time from tasks like painstakingly converting CAD models of a system to complex multiphysics simulations.

Device prototypes

Despite significant advancement in the integration of CAD tools such as Solidworks with Ansys, interfaces still require a high level of user intervention and manipulation before simulations can be implemented and coupled. The desired endgame for such a tool is being able to readily compute CAD models of device prototypes into a digital twin prototype, but this is still some significant way off. The computational power demanded by such a tool is achievable but is still substantial if the goal is to compute the insights advertised in a reasonable amount of time.

With the benefits and drawbacks of where digital twin prototype technology is today, it is critical to identify the breakthrough point where the time, money and computational expense committed is outweighed by the tangible benefit of insight generated. This is a challenging question but encompasses the engineering judgement applied when considering the complexity required of a multidisciplinary simulation and the managerial judgment when scoping a design and build phase. These kinds of trade-offs between effort and output are standard practice when managing more traditional complex product developments.

In summary, digital twins present a huge opportunity to revolutionise products, including drug delivery devices, throughout their development, production and deployment. The appeal of this technology is becoming increasingly apparent within the pharmaceutical sector where they can minimise losses and improve production efficiency.

Implementation during the development stage is also of great potential, in terms of maximising performance gains by ensuring the right design from the outset while also accelerating timelines and cutting development costs. Right now, the creation and deployment of twins is still a highly manual process and requires a combination of domain and subject matter experts driving their tools together in unique ways. But the ultimate prize is a big one… transforming the way we develop complex products.