All industries are undergoing a rapid digital transformation designed to meet the two objectives of faster product regeneration and systems optimisation. While digital building blocks, such as, DSI & MBE, IIoT platform form the first part of achieving this transformation, the second is all about getting a deeper insight from collected data, using statistical, machine learning & deep learning techniques.
With this transformation, there is a data collation aspect involved for every part of the industrial process and its related systems. However, as time passes, the parts and hardware experience wear and tear, and soon small problems snowball into an exponential breakdown. Pre-empting and identifying this critical breakdown point using relevant data is a step for predictive maintenance. This could be the start of a series of activities performed for extending the lifecycle of a part or a process in a timely manner, as all technology teams are aware.
While there are various kinds of monitoring that could help with advance warnings of a breakdown, most of them are really manual. Electrical monitoring among several other mechanical systems can serve a forewarning about equipment going out of shape, but can never really be accurate in prediction. This is where automation steps in.
The best way to stay a step ahead of this situation is to have an efficient automated predictive maintenance system in place. Investing in processes and systems that help to identify critical risk points in a predictive manner could be the only difference between a smooth supply and production chain and critical production loss and disruption in market activities. The focus here is to ensure a PM system that actually simplifies the process rather than make it more complex than even the original system. For an efficient predictive maintenance strategy, the answer is the digital twin.
The digital twin and its advantage
Being a cloud-based virtual copy of the system, the digital twin facilitates a virtual expert analysis of its health and operation efficiency, remotely. Its biggest advantage is the elimination of downtime for maintenance. Without interfering with the running of the system on ground, it allows the freedom to analyse outcomes in hypothetical situations, what happens in case of unexpected loads, changes etc. and plan maintenance accordingly.
As we go a step ahead, many of these analysis processes can be automated and twin process will occur in the digital twin. So, by automating activities like vibration analysis, lubrication analysis using sensors in the real system and feeding this data into the digital twin real time, the entire predictive process can be automated.
So, essentially, the digital twin functions are based on four data sets – engineering data, manufacturing data, historical data & real time data. The predictive analytical model is developed for process, part or system in consideration. It becomes the core component of digital twin. Once loaded with real time data, it helps to predict various breakdown scenarios on the virtual twin, allowing for efficient predictive maintenance heads-up. The digital twin can help to forecast the behavior of the real time process and equipment in various different circumstances that can be simulated virtually, using any incidence scenarios and relevant prototypes.
Use cases in industry
A number of industries benefit from using digital twin for predictive maintenance. Preventive and predictive maintenance of aircraft (a critical requirement since lives depend on it), of automotive manufacturing industry (again, life is a critical need), and even in healthcare monitoring equipment (yet again, a lifesaving technology application), are a few critical examples of the significance of this technology. In some cases, the twin of the whole system can be generated – aircraft engine, the human body or the engine system for the care – and can be studied for any future risks or need for maintenance.
Challenges to meet
The foremost challenge in creation of a digital twin is the need for its accuracy to ensure complete mirroring of the real process. Any minute difference in the structures can trigger a completely different outcome for the real process. It may not always be possible to get an absolute modelling of the physical asset with the properties and characteristics that faithfully mirror the entire mechanical and electrical properties of the asset and modelling. Secondly, there is always a risk of blueprint failures that may arise at a later stage of development. And thirdly, duplicating each modification of every component in the digital twin is a constant challenge. It needs to be mirrored in the virtual twin to ensure accuracy of predictions. The intensity of these challenges increases during the construction of the twin to the component sub system level. Replicating virtually at the lower components level of the equipment is a much faster process than the system level or the sub system level, which is more prone to errors.
And finally, while this remodeling is achievable, the real roadblock is the lack of maturity of the technology stack in most instances. That is a roadblock that enterprises need to work around.
The simplification recommendations
The path ahead could be smoother with more accurate planning. If dealing with a complex system, it is difficult and expensive to build accurate digital twin for predictive maintenance. Hence, it is essential to identify critical components for sub-systems based on failure modes or reliability model of overall system. Once such components or sub-systems are identified, it would make sense to develop a digital twin of an individual component or a sub-system. This helps to mitigate the challenges of accuracy of a virtual model that should precisely reflect the physical twin’s properties, and allow for a detailed blueprint of a component’s/sub-system’s failures to be developed more easily than the whole system’s image.
In order to build a credible and accurate digital twin, enterprises need expertise across all the building blocks of digital transformation – design system integration, model based enterprise, digital manufacturing, fielded product support, and digital software engineering and strategic partnerships with leading IIoT players. With a mature technology stack, the concept of digital twin and its relevance on predictive maintenance will be increased manifold.