Industrie 4.0 introduces the somewhat abstract concept of a ‘digital twin’. But is this really new, does it actually exist anywhere in practice, and if so, what steps should be followed to build one?
A process control engineer who is already familiar with CAD, SCADA, process simulations, manufacturing systems and business (ERP) systems, may find the concept of a digital twin puzzling – after all integration of systems and process, automation has been part of the goal of modern manufacturing for some time. So, what is new?
What is a digital twin?
A digital twin is a digital representation of a physical object, plant or process throughout its life-cycle. This digital representation can include original design information, physical attributes and context and usage information, which can, in turn, be used to model and predict performance.
A digital twin is not a product you can buy. Implementing a digital twin is going to be a journey during which you steadily implement platforms, capabilities, processes and human/machine interfaces. Beware, all roads don’t necessarily lead to Rome, so understanding your business strategy and how these new digital technologies will support this is important as to why a digital twin is needed, and what exactly has to be done.
Design, manufacturing and maintenance data
A digital twin is easiest to understand when considering a physical object, for example, a car, or an engine, or an electronic device. The digital twin is a digital representation of this device; the data is initially developed and optimised during design, tracked during manufacturing and then augmented by actual usage data to improve use/maintenance of the object by customers:
• Design data relating to the object is created and optimised virtually using computer-aided design and modelling technologies.
• Manufacturing data records the detailed production parameters, for example, raw materials, third-party components used in the assembly, quality, process conditions and so on.
• Use/maintenance data records how the object is actually used by customers in the field when/how it is maintained, and so on.
Modern automotive manufacturing already has several of the above elements in place and is a leader in this regard. In other industries, however, the digital twin might not be as straightforward.
A digital twin is not restricted to physical objects; it might be implemented for an entire manufacturing system, including physical plant and equipment, human decisions/activities, business processes, customer data, supply chain data, events, environmental information etc. The common thread is the connection, collection, organising, analysis, visualisation and interaction with vast amounts of data.
At the heart of the digital twin is a model that represents the attributes and operation of the system or object. But a digital twin is more than simulation software – a digital twin will usually include artificial intelligence that allows for self-learning. The output of the digital twin will be a rich interactive human-machine interface, which uses, for example, 3D augmented /virtual reality to visualise and simulate performance.
Digital twins support the full product life-cycle in several ways:
• During design, digital twins will improve collaboration and allow product development teams to work virtually across multiple locations. Computer-aided design and collaborative tools have existed for some time now. A digital twin builds on this but takes the concept further to support adaptive flexible manufacturing to quickly adapt to environmental conditions and individual customer requirements.
• During manufacturing detailed production information and small variants in the manufactured article will be measured and stored in the digital twin. For example, in electronics manufacturing individual components used in assembly are often sourced from competing suppliers and will vary between batches. Tracing each component of the assembled product through design, manufacturing and ultimately during use/maintenance will allow for rich insights into how using different component suppliers affects the product performance in the hands of the customer.
• During use/maintenance, field data will likely be collected and analysed using IoT sensors. True predictive maintenance then becomes possible that will, in turn, enable more targeted and responsive service to customers.
Implementing a digital twin proof of concept
Implementing a digital twin can be confusing and overwhelming. I suggest that you consider starting small and do a proof of concept (POC). For example:
1. Research the opportunity in terms of your business strategy, do some planning, secure budget and build awareness and support for a POC.
2. Implement remote monitoring capabilities (this will probably need you to improve parts of your systems architecture, implement connectivity and data standards such as OPC-UA and ISO 10303-239, take on new IoT devices and build new capabilities in your IT and manufacturing systems teams).
3. Implement predictive analytics tools that will consume this remote data to self-learn and predict performance (this will likely require new capabilities in data science, modelling, artificial intelligence and visualisation).
4. Connect the result of the above to field service operations (this might require fundamental reorganisation of the established business processes in this area).
5. Close the loop by connecting the data and models back into new product development, design and engineering processes.
As you run this POC and as relevant technologies continue to mature in the market, you might also systematically introduce new human/machine interfaces and data visualisation tools, including 2D/3D visualisation, augmented reality and advanced human-machine interfaces (natural language processing and natural user interfaces). Remember, the digital twin is not pure automation, it is intended to augment, not replace, human decision making.
This article was first published on SA Instrumentation and Control.
