NEWS + BLOG
Bringing simulation technology into operations by David Knezevic, Ph.D. Akselos CTO
Today, physics-based simulation technology (think FEA or CFD) is part of every field of engineering. The value of computational modeling of physical phenomena from first principles is well-established: it enables virtual prototyping during the design phase, reducing time to market, and ensuring designs meet required safety standards, among many other things. Indeed, the past few decades have been a golden age for simulation technology, as the core methods have benefited from the dramatic increase in computational power due to Moore’s Law, and have become ever more powerful and realistic tools for engineers to leverage.
LIMITATIONS WITH TRADITIONAL SIMULATION TOOLS
Nevertheless, as any simulation expert will tell you, even with modern computational power there are still significant limitations on the type of systems that can be modeled with traditional simulation tools. This is because the amount of computation required in physics-based simulation grows rapidly with the size or detail in a model. The upshot of this is that engineers have developed ingenious workflows that work around these limitations, i.e. they will model localized parts of systems (e.g. a specific bearing, or a particular flange or joint) and impose boundary conditions that represent the behavior of the “unmodeled” parts of the system. Or, as shown in the figure below, they will generate coarse system-level models and separate localized submodels (sometimes dozens of submodels) and transfer data from the system-level to the local level to perform detailed analysis.
Submodeling workflow for hulls, which is based on a coarse global model with many localized detailed submodels.
This type of localized modeling has been widely and successfully deployed, but it has limitations. First, it is computationally intensive. Engineers typically design their models to be as refined as possible (within the time or memory requirements they have at their disposal) and hence analysis of the array of localized models under all the conditions of interest is a significant undertaking (so much so that there is often a need to skip parts of the analysis due to time limitations). Second, it introduces further assumptions or approximations into the modeling due to the process of creating a localized submodel, either due to imposition of hypothetical boundary conditions, or due to one-way coupling from system-to-local models. These limitations are generally not deal-breakers during design, and engineers have well-established workflows for using local models along with known assumptions to create designs part-by-part in a reliable way.
However, when it comes to operations, it is a different story. Operators of critical assets – whether it is an offshore structure, a power generation system, or an aircraft – require a holistic view of their asset, which is incompatible with the type of submodeling approaches that are often used in design, and requires a new computational paradigm that overcomes the limitations of traditional simulation tools.
In particular, we discuss three of the key requirements of operators below:
HOLISTIC AND DETAILED MODELING
A model needs to represent the current state (not idealized states that may be hypothesized during design), and it needs to capture the cumulative effects of all defects that have been detected or observed. This means that section loss in a stiffener, or cracking in a turbine blade, or marine growth on an offshore structure, need to be incorporated into the model and the effect of this needs to be analyzed to give the operator a status check on the health of the asset. The status can then be shown in a Dashboard like in the figure below, which shows the current status of each asset in a high-level manner (“green light” if there are no issues, “red light” if issues are detected), and also provides the ability to “drill down” into the details to identify the root cause of issues and enable prioritization of work orders.
Web-based dashboard for viewing current asset status. A green box indicates that the asset currently passes all checks, whereas a red box indicates that issues have been detected. This allows operators to efficiently prioritize work orders in a data-driven manner.
The type of status report indicated above typically involves solving thousands of different cases, e.g. to perform a fatigue life estimation of critical parts or perform a strength check based on the relevant industry standards under a wide range of “what if” scenarios. In order to provide timely insights into the current status of an asset, and hence to enable engineers to make data-driven decisions in a nimble manner, the full battery of analysis results need to be completed quickly, which in turn requires significant solver acceleration compared to FEA.
Operators rely on sensor data to get measurements about the current state of an asset, and if a computational model is to be useful during operations it needs to be able to be efficiently re-calibrated to match the latest sensor data. Parametric models, in which model properties can be modified via changing “dials,” e.g. to vary stiffnesses, densities, loads, geometry, etc., and re-solved quickly, are a key enabler for this type of model re-calibration. This enables the Digital Twin to automatically and continuously track the current status of the asset based on sensor measurements.
Akselos’s RB-FEA (Reduced Basis FEA) technology provides all three of the points listed above. RB-FEA is an acceleration layer that builds on top of FEA to provide holistic, detailed, fast, and parametric reduced order models of critical assets. In particular, we note that RB-FEA enables fully detailed Digital Twins of some of the largest assets in the world, such as FPSOs, to be solved in seconds. The methodology also provides accuracy guarantees, underpinned by rigorous mathematics that were developed and published in academia over more than a decade.
Illustration of parametric RB-FEA modeling. Thicknesses can be updated via parametric components and then the overall model can be re-solved in seconds.
With this combination of features, along with an implementation that fully leverages computational parallelism on the cloud or on in-house servers, Akselos’s RB-FEA provides the ideal tool to provide physics-based insights to operators. It helps operators to perform predictive and more targeted inspection, maintenance, and repair; to interpret sensor measurements and make data-driven real-time decisions; to assess risks associated with various “what if” scenarios; to calculate remaining operational life and justify asset life extension; and the list goes on. We refer to this computational solution as the Akselos Digital Twin.
Operators of critical assets face a unique set of challenges that are not well met by the traditional physics-based simulation workflows. Akselos’s RB-FEA technology overcomes these limitations to provide an operations-ready Digital Twin, which has already been deployed in a range of industries including oil and gas, wind power, aerospace, mining, power generation, and automotive.