Advanced FEA Holds the Secret to Predictive Digital Twin

The concept of having a virtual replica of a physical product, or a “Digital Twin”, first appeared in 2003 in the context of Product Lifecycle Management (PLM). The core idea consisted of data channels that would link a physical asset with a computational model of the asset. More recently, the Digital Twin has emerged as the key piece needed to unlock the full potential of the industrial internet of things (IIoT). Aligning the physical world with virtual models will bring about an enormous upheaval across all major industries, and Cisco predicts that this will grow into a $19 trillion market within the coming years.

In principle, the Digital Twin consists of two key elements:

1. A virtual model of an asset. The model is initialized based on the original design and is updated over time to stay in sync with the physical asset throughout its lifetime.
2. The physical asset instrumented with sensors which can capture its current operational state.

However, there are differing views among major players in the industry on how to best develop, validate and use the Digital Twin.

With the rise of IoT, many technology vendors are working on harvesting sensor data that is streamed from deployed assets like wind turbines, marine vessels, and other industrial equipment. A standard approach today is to use sensor data to calibrate statistical models that attempt to represent the behavior of the asset. The advantage of these types of statistical models is that they are fast to evaluate and hence can be used in real-time. However, their disadvantage is that they are ad hoc: they model assets empirically and are not based on first principles of physics. We need to go beyond these ad hoc models to deliver the true value of the Digital Twin.

The gold standard for modeling based on first principles of physics is Finite Element Analysis (FEA). FEA enables detailed simulations of a vast range of assets and has been a key engineering tool since the 1960s. It enables engineers to compute and visualize stresses, strains, forces, moments and other key physical quantities within complex infrastructure, which enables detailed and accurate structural integrity analysis and risk assessment. However, it is well-known that detailed FEA is extremely computationally intensive when it is applied to large models. Indeed, it is far too slow and computationally intensive to be used for sensor-integrated Digital Twins of assets like oil rigs, ships, wind turbines, etc. To get the value out of sensor data integration in asset maintenance, engineers need rapid response time so they can use the Digital Twin to inform their real-time decision-making. This is not feasible with conventional FEA, so no wonder it has so far not been included in the Digital Twin concept.

But wait, what if we could have instantaneous FEA?

Akselos is the pioneer of reduced basis models (RB-FEA) that are the true enablers of simulation-driven Digital Twins. This brings the detailed analysis of FEA along with the speed of the statistical models that are the mainstay of Digital Twin deployments today.

"The Digital Twin is not about owning the technical data. It is about having access to and being able to exercise the data. We really need to understand the technical risk”. David Walker,  Deputy Assistant Secretary of the US Air Force for STEM

Akselos provides RB-FEA solvers that bring over 1000x speedup compared to traditional FEA, without sacrificing accuracy. The key enabler for this acceleration is a divide-and-conquer approach in which data is pre-computed for components and then re-used on the fly whenever a new solve is performed. The components are endowed with parameters (such as geometry, density, stiffness, etc.) so that the Digital Twin can be quickly and easily modified and re-solved in real-time. These capabilities are the product of extensive research in academia conducted over the past 15 years by research groups at MIT, EPFL, and many other institutions.

The extreme speed and reconfigurability of Akselos models is a perfect fit for Digital Twins since it enables:

1. Calibration algorithms to be run efficiently to ensure that a Digital Twin matches sensor data.
2. Detailed analysis on the calibrated Digital Twin to assess risks and model "what if" scenarios for the asset, taking the latest sensor data into account in this analysis.

These capabilities mean that Akselos is unlocking the full potential of the Digital Twin concept.

We're currently working with partners to deploy the Akselos Digital Twin concept for major assets in the oil & gas, mining and wind energy industries. We will provide more updates and news soon. Until then, I invite you to try out Akselos's software by signing up to our online community. The technology will speak for itself.

_______________________________________________________

On the June 2, we released a new version of Akselos Software, with many new features that are very relevant to Digital Twins of large assets. Refer to the new features included here.