Simulation is Evolving, Creating the Digital Thread for Digital Twins in IIoT

We have previously discussed the role of the digital thread and digital twin in the world of IIoT, and how this is not simply one technology, but a way of describing collections of data and their lifecycles. A digital twin on its own is a moment-in-time representation of the state of a component or entire system. It may already form the basis of an input to predictive maintenance, which is a small sub-branch of the science of simulation – but the environments that manipulate and explore digital twins are what offer a path forward. Combining simulations from design with those individual silos used when running a process can create a better-informed overall picture of any process, from design through running to end of life.

We have seen in the adoption of IoT across many sectors that the coming together of multiple streams of data to gain insight is often a silo buster. Within IIoT in manufacturing and instrumentation from an incoming supply chain, as well as from plant machinery, connected workers, building controls, shipping and end-market use – all can be combined to understand how to improve the manufacturing process and product.

These separate departments may hand information to one another, but it has often been a one-way dialog. It is the whole integration from creation to expiration that provides the potential for richer insight. Naturally, focus has tended to be the runtime data component because it forms the bulk of the lifespan of a process. Initially, brownfield IoT instrumentation that is added to a plant or existing instrumentation is opened up to analysis, with the next wave of adoption focusing on how to analyze the data. This possibly includes moving machine learning for devices closer to the plant itself with edge computing.

However, this forms a focused, but still insular, thread of a process. This works, but it could work even better by taking a broader holistic approach. The concept of a digital twin helps in understanding the whole or part of a system; but this is still just data, which in most cases represents a 'now' state, or is kept as a log of the history.

The coming next wave of IoT advances needs to offer closer integration of the environments that understand the context in which a digital twin exists, and can manipulate the entirety of the data in and around it. The extension or the aggregation of existing simulation environments will be able to offer this. Any process, machinery, object or building that is designed digitally can then be put through various levels of simulation. Once in production, instantiated in the physical world, it is then being exercised almost as in another highly detailed simulation, but with real consequences.

Changes to the physical instantiation today will be trialed through a localized 'what if' approach – a quick simulation that does not have all the features of the design simulation that created it. This simplifying has been due to the difficulty in data integration from design, spreadsheets or basic models getting used for shop floor adjustments.

Compute power and its distribution are evolving. So as edge computing provides machine learning close to the machinery, we can expect more simulation compute power on the shop floor and in process controls. For that to work, it should be fed by the richest models possible – those from the initial design stages, boosted and augmented with the real-world status created through actual physics, which in turn feeds back to the design models.

We are seeing this breaking out of data from CAD to provide information for augmented reality (AR) applications on the shop floor, primarily around the 3-D model visual data, mixed with the life IoT data, because AR is the user interface for IoT. Training simulations have also existed for many years from the most basic of operator consoles, but are starting to pull some resources both from initial designs and real-time digital twins.

Artificial intelligence and ML are important in the context of simulations, because they are the tools that can learn what the model of the process is, or act as the engines to sift through multiple what-if scenarios, and provide a curate list for decision-makers to evaluate. This can be seen in the design industry, where parameters for a product can be entered and multiple viable design form factors generated.

For those troubleshooting situations and planning use cases, the simulation for what-if can find an expedient solution to a problem. Simulation of a dangerous situation or creating a management training tool to illustrate the deep impact of choices are also relevant. These should ideally be created and drawn from the same base of truth from design, production and expiration. With disparate systems, many examples today tend to be custom creations based on snippets of the real-world model.

IoT platforms have tended to focus on the communication and orchestration of data distribution. Differentiation is difficult given the inherent standards and patterns that have evolved. The digital twins concept represents the data and IoT platforms that communication wrangling and digital thread systems need to be the glue that pulls those together, including tools such as simulation that are available for any level of abstraction.