The world of computer-aided engineering (CAE) is constantly evolving. What’s next for the field? Just as in other industries, big data, the Internet of Things (IoT), machine learning and digital twins will lead to significant shifts, as will other emerging technologies.

Democratization of HPC Resources Through Cloud and Containers

Once accessible only through large upfront investments, HPC and simulation resources are now much more readily available to engineers. Cloud technology is playing an important role in the democratization of simulation software because it enables remote access to such robust computing power. Cloud-based HPC resources also eliminate the need for investment in expensive hardware and software, while offering more flexible pricing models than, for example,  the conventional (and often expensive) software licensing.

Watch our on-demand webinar to learn more about democratization of HPC resources through cloud and containers.

Cloud computing might get all the attention, but the introduction of software containers has also driven the democratization and wider acceptance of HPC. Fully production-ready, UberCloud’s containers bundle OS, libraries and tools, along with application codes and the user’s data. They can even hold tools to support whole complex engineering and scientific workflows. Designed to be deployed either on-premise or in the cloud, these containers dramatically reduce deployment times.

The next phase of democratization will likely bring streamlined user interfaces and integrated simulation workflows to make the software more user-friendly. It’s also likely that as software companies move further from conventional pricing models, they will offer open access to resources like training and templates.

A Move Beyond Physics-based Simulations with Machine Learning

Traditional physics-based simulation models are based on preset rules. But access to Big Data (generated by IoT devices, for instance) means the potential to train artificial neural networks, yielding predictive simulation models. This application of machine learning to the simulation environment will help drastically decrease time-to-solution, while preserving much of the accuracy of a traditional solver.

In the coming years, we will undoubtedly see deep learning components integrated into simulations, both to model a real system, and to train additional AI-powered components. Meanwhile, because machine learning requires so much data, it’s also likely that organizations will grapple with issues related to the generation and management of that data.

The image above illustrates the difference between the ground truth flow field (left) and the predicted flow field (right) for one exemplary simulation sample after 300,000 training steps. Read the full case study, "Deep Learning for Steady-State Fluid Flow Prediction in the Advania Cloud."

A recent UberCloud case study explored the implications of using deep learning to predict steady-state fluid flow. The study confirmed that larger data sets result in faster training of the neural network and improve accuracy of the artificial neural network’s prediction capability. While generating such high volumes of data requires significant overhead, that could be offset by running many HPC simulations in parallel in the cloud.

Proliferation of Digital Twins for Objects and Processes

The concept of digital twins has been around since the 1960’s; NASA even used a rudimentary digital twin to simulate and evaluate conditions on Apollo 13. But digital twins have received much more attention in the past decade thanks to the emergence of the industrial Internet of Things (IIoT). And as the cost of sensors and other IoT-enabled technology continues to drop, it will be possible to build more sophisticated simulations across processes and systems.

Today digital twins of critical physical assets are already all but ubiquitous in manufacturing, and more powerful simulation technology has resulted in an expanded definition and role for digital twinning, which includes the creation of twins for not only physical objects, but also the systems in which those objects are designed, tested, and produced.

Some aspects of creating digital twins remain challenging. For example, because each digital twin is unique, there are no “off-the-shelf” solutions for building them efficiently. Much of their construction is still done ad hoc, although software providers like ANSYS are increasingly incorporating tools specifically for creating these complex simulations interacting with sensors in real time. That trend is likely to continue and even accelerate in the coming years, as digital twins grow popular across industries.

Schedule a free consultation with one of our Cloud Simulation Experts to discuss your challenges and how UberCloud can help you boost productivity in the changing engineering landscape.

To stay up to date on the latest trends in CAE, subscribe to UberCloud’s free monthly newsletter. You’ll receive periodic updates on new case studies, white papers and other resources to help you stay in the know.