Olympic Hall III

This event has moved to Reflections Bar and Lobby

Breakfast/Lunch will be held at Athos, main hotel restaurant

Why are we so passionate about engineering simulation? This keynote message unveils the surprisingly profound—and often overlooked—reasons why we simulate. Prepare to discover insights that could change the way we think about our métier.

Why is it that all of us—employees and companies alike—are chasing innovation? Is it really as essential as we think? And if it is, why does it seem so hard to achieve? Are there guiding principles, perhaps even simple habits, that can help us spark true breakthroughs? Or are there hidden obstacles we must avoid at all costs?

These are some of the questions this talk will seek to answer, drawing on examples from history and on the speaker’s own personal experience.

The present keynote outlines the development of a hybrid digital twin and AI-based framework for marine shaftline monitoring and bearing condition prognostics, aiming at real-time bearing load estimation without invasive measurements. The approach integrates shaft alignment theory, bearing modeling, and Multiphysics simulation with machine learning to enable predictive insights into propulsion system health.

The work has been advanced through two three-year research projects, i-Marine and S-PRISMoID, during which small- and medium-scale experimental facilities were established, alongside the creation of a physical twin of an operating vessel. These platforms enable both controlled laboratory validation and real-world pilot testing of the framework.

Looking ahead, the vision is to transition from research to industrial implementation by developing standardized monitoring protocols, embedding predictive maintenance strategies, and collaborating with classification societies to shape regulatory frameworks. This approach is envisioned to contribute towards safer, reliable, and more efficient marine propulsion systems.

Olympic Hall III

Modern GPU hardware provides ray tracing acceleration, letting us achieve the performance needed for real-time interactive rendering of FEA models. We built on the existing graphics kernel of our software to implement a path tracing mode, utilizing hardware accelerated ray tracing with Vulkan. We share the underlying OpenGL memory resources and data of the graphics kernel with Vulkan without duplication, using Vulkan for path tracing and OpenGL for rasterization. In the presentation we will show how Vulkan ray tracing works, how we share the existing graphics kernel data with Vulkan, how high-performance memory transfers are performed. Also, we present how we used a modern shader language for path tracing shader authoring, giving high priority to minimizing shader permutations. We show how we shared the path tracing shader code with the OpenGL implementation of the SPH renderer. At last, we show how 3rd party libraries were customized and used. The Intel Open Image Denoiser which is an AI U-Net based denoiser and Mesa llvmpipe (OpenGL) and lavapipe (Vulkan) for software fallback.

We present the cloud transformation of SPDRM, BETA CAE Systems’ process, data and resource management platform.

We follow the journey of SPDRM from an on premises product to a fully cloud-native application, highlighting its ongoing improvements.

The platform integrates cloud-native storage, elastic scalability and secure site-to-site transfers across AWS, Azure and on-premises networks while balancing portability, latency and performance constraints.

Our end-to-end testing strategy combines load and stress testing with Concurrent Automated Benchmarking (CAB)—orchestrating up to 250 ANSA workers—and large-scale workflows with 60 concurrent process nodes.

Results confirm the resilience of SPDRM and demonstrate its readiness to scale to thousands of active users.

KEYWORDS –

Full frontal impact, Optimization, Kinematics, Predictions, Clustering, ANSA, META, KOMVOS, ANSERS

ABSTRACT –

Crashworthiness is a critical aspect of vehicle development requiring effective dissipation of kinetic energy during collisions to protect occupants. Subsequently, optimizing Body-in-White (BIW) structures for crash performance involves addressing complex interdependencies among time sensitivity, material behaviour, and key performance indicators. Capturing and understanding kinematics within the optimization process is essential for generating design variants that meet crash requirements effectively, and the use of specialized software streamlines and facilitates this complex task.

This project investigates new methods for capturing full frontal crash kinematics, comparing their efficacy with existing optimization approaches and eventually how they complement each other in supporting the generation of improved design variants. All morphing scenarios were implemented in ANSA, while the creation of the Design of Experiments, predictions, and clustering were carried out in KOMVOS. METApost was employed for the creation and computation of deformation lines (defolines), and ANSERS was utilized for the presentation and interpretation of results.

By demonstrating how these methods can accelerate optimization processes, this study aims to contribute to the development of stable crashworthiness designs that enhance vehicle safety. 

KEYWORDS –

EMC/EMI Electromagnetic Simulation

ABSTRACT –

As modern vehicles integrate increasingly complex electronic systems, ensuring electromagnetic compatibility (EMC) and performance has become a critical aspect of automotive design. 

This presentation explores a workflow for full vehicle electromagnetic simulation using ANSA for pre-processing and Clarity for high-fidelity electromagnetic analysis. We demonstrate how ANSA’s advanced meshing and model preparation capabilities streamline the setup of large-scale simulation models, while Clarity’s cutting-edge solver technology enables accurate prediction of electromagnetic behavior across a wide frequency spectrum. 

IGA surfaces are rendered as NURBS surfaces. The traditional way to render them was to perform triangle tessellation with a fixed step on the CPU and upload the triangles on the GPU. This would mean that for post-processing, we would need to tessellate triangles separately for each loaded state ahead of time. Instead, we present a GPU accelerated way of rendering NURBS surfaces with the help of the GPU hardware tessellator and trimming them on the pixel shader. We show how we manage to render the surfaces using the initial surface data, like control points, knot-vectors and trim-curves with minimal pre-processing.  

We present how we expanded this tech to Catmull-Clark subdivision, using the tessellation shader to accelerate it. The traditional way was to subdivide the surface recursively until reaching the desired subdivision step on the CPU and render the resulting quads on the GPU. Now, based on surface features, they are separated into regular and irregular sub-surfaces, with only the irregular surfaces being a candidate for further subdivision and the regular ones being forwarded to the GPU hardware tessellator. 

At last, we present how this technology was used to make Higher Order Elements’ rendering simple, fast and high quality. 

The design of rail vehicles is central to ensuring performance, safety, and efficiency in an increasingly complex transport landscape. This presentation highlights the pivotal engineering innovations that addressed Stadler Rail A.G.’s critical needs and ultimately led to the adoption of BETA CAE’s solutions.

Breakfast/Lunch will be held at Athos, main hotel restaurant

Version 25 marked a transformative leap in the evolution of the ANSA and META platforms, introducing a ribbon-based interface that redefined usability and productivity. By replacing legacy toolbars and modules with a modern, intuitive layout, we delivered a cleaner, more structured user experience that empowers engineers to work smarter and faster.

This presentation will explore the design philosophy behind the v25 overhaul, the measurable impact on user workflows, and the lessons learned from past iterations. More importantly, it will chart a course for the future—highlighting how we must continue to evolve our interface strategy to meet emerging user needs, embrace flexibility, and foster innovation across our product ecosystem.

KEYWORDS –

CAE, Multi-Step TPA, Modal Contribution, Sub-Structuring

ABSTRACT –

In the field of Noise, Vibration, and Harshness (NVH) car development, efficient root cause analysis plays a critical role in identifying and resolving unwanted disturbances that affect vehicle comfort and performance. In this context, it is crucial to identify the key components driving a specific NVH-related phenomenon and to understand how excitation propagates through the system, from the point of introduction to the receiver component, where passengers perceive the disturbance.

BETA CAE Systems has developed an advanced analysis method within its FEM NVH software to effectively investigate such challenges. The core concept involves decomposing the vehicle system into subsystems, which are then assembled using a numerically efficient and accurate frequency based sub-structuring (FBS) method. This approach facilitates well-established root cause analysis techniques, such as Transfer Path Analysis (TPA), and enables the integration of hybrid CAE-test models by incorporating measured frequency response functions for select components.     

In the latest version of the BETA CAE NVH software, two groundbreaking root cause analysis techniques have been introduced, both novel in the market, which will be presented in this paper.  

First, the TPA analysis has been elevated to a new level. The software automatically performs a multi-step TPA. A cascade of TPAs is performed, which determines the dominate path through the complete system step by step indicating the relevant components and the connections between components from the excitation point to the response point, e.g., from the tire patch to the steering wheel.

A second exclusive feature allows modal contributions to be calculated for each CAE component. Previously, modal participation could only be examined for the receiver component. The newly introduced capability provides deeper insights into how individual modes of various components influence the overall system behavior. 

In the first part of the presentation mainly the new TPA techniques are introduced and applied on a practical use case and in the second part the implementation challenges for this feature are discussed.

Join us for an in-depth look at how SubD modeling in ANSA is helping to bridge the gap between design and engineering. SubD surfaces offer a powerful and intuitive approach for creating complex, smooth geometries, which can be especially valuable when working with optimization results and help to overcome the communication challenges between CAD and CAE. In this session, we will share the story behind our recent advancements in this area, which were developed in close collaboration with the BIONICAST team at Mercedes-Benz and enable efficient optimization workflows for casting and forged parts. We will showcase how the tools we developed -now part of the Retopology ribbon- support design modifications and manufacturability checks within ANSA and aim to deliver ready-to-use design iterations back to CAD with the help of SubD surfaces.

Modular approach of reduced models enables efficient structural and dynamic analyses by significantly reducing computational demands without compromising accuracy.

Inherent challenges emerge with the development of reduced models. The definition of reduced models requires expert knowledge and is often the result of a try and error approach to achieve a balance between simplification and accuracy in encapsulating the complete model’s main characteristics. Then, the creation of alternative representations of reduced models, as well as the creation of new reduced model versions for every modification of the detailed FE-representation of the model, unveil another inherent weakness of the standard practices; The lack of traceability, that completely hinders collaboration between simulation engineers.

The implementation of a Simulation Process and Data Management (SPDM) system represents a transformative approach to managing simulation workflows and data. On the data management side, such a system enables generating various reduced model types and maintaining traceability of source models and dependencies, it ensures end-to-end transparencies and allows effortless transitions between full FE and reduced representations. Moreover, the system significantly enhances "what-if" analyses and optimization processes by properly documenting simulation iterations, enabling clear and comprehensible comparisons between variations. On the process management front, SPDM systems facilitate the standardization and automation of CAE workflows. These capabilities lead to the automation of generating reduced models and the submission of reduction runs to the solvers on the HPC systems, followed by systematic storage of results linked to their respective reduced models. 

SPDM systems address critical challenges in simulation model preparation and analysis, including data sharing, integrity, traceability, and version control. Ultimately, adopting SPDM systems not only accelerate the end-to-end simulation process but also fosters innovation and productivity across engineering teams.

In the CFD domain, one of the major challenges is establishing a streamlined workflow that efficiently bridges the gap between design and CFD results. At ANSA, we have successfully addressed this challenge over the past years. Within this workflow, CFD meshing plays a pivotal role, especially when dealing with dirty CAD inputs or tessalations (CAD Chaos), a common scenario among automotive OEMs. The HeXtreme mesh generator offers a robust solution by delivering fully automated CFD meshes for external aerodynamic analysis and beyond. In this presentation, we begin with an overview of the streamlined workflow, followed by an in-depth look at the HeXtreme generator. We present an algorithm overview, challenges handling big meshes, and feature direction towards modular and distributed runs. Finally, we introduce AutoSeal, another key functionality that enables full automation of the HeXtreme process, achieving 100% workflow automation.

This presentation showcases how modern development tools can elevate engineering workflows through intelligent automation and seamless integration. We begin with Visual Studio Code (VSCode), highlighting essential features such as autocomplete, problem recognition and debugging. We extend into the enhancements of the BCS Dev Environment plugin, enabling direct documentation access and execution within the BETA Suite. We then dive into GitHub Copilot’s AI-powered capabilities, including smart code completion, code explanation, code review, and automated code generation. Engineers will gain practical insights into how these tools can accelerate development, improve code quality and enhance productivity.

Olympic Hall III

In the new era of AI-assisted development, emerging tools are transforming the way we write and maintain code. Among them, GitHub Copilot stands out as a powerful assistant for developers. It is offering access to multiple AI models that support a wide range of programming tasks. Many developers view Copilot as an ideal partner for pair programming. 

At BETA CAE, the deployment of GitHub Copilot opens exciting opportunities for C++ development. One of the most impactful benefits is its ability to bridge the gap between our isolated intranet environment and the vast knowledge available on the internet. 

In this session, we’ll explore a selection of Copilot’s capabilities, including: 

• Writing new code 
• Debugging 
  
• Refactoring 
  
• Code explanation 
  
• Code review 
  
• Test generation 
  
• Performance optimization 
  
• Explaining library APIs (e.g., STL, MKL)
  

While GitHub Copilot can significantly boost productivity and code quality, it also introduces new responsibilities. We’ll conclude with practical advice on how to use the tool effectively, avoiding common pitfalls and misuse. 

Inverse Distance Weighting (IDW) is a widely used spatial interpolation method for estimating unknown values at evaluation points based on known control points. Its core principle is simple: points closer to the target location have greater influence on the estimated value than those farther away. This makes IDW particularly suitable for applications like mesh morphing.

While conceptually straightforward, a naïve IDW implementation has O(mn) complexity, which can be prohibitively expensive for interactive applications—especially when dealing with large-scale CFD models.

In this session, we will:

•    Review the fundamentals of IDW interpolation.

•    Explore algorithmic optimizations that significantly accelerate IDW on CPUs.

•    Demonstrate how leveraging GPU computing can deliver orders of magnitude performance improvements over the basic algorithm.

•    Showcase our in-house CLI API that abstracts the complexity of OpenCL, making GPU acceleration accessible and easy to integrate.

The CFD domain has been notoriously known for the time and energy resources which are required for the production and analysis of volume models. The volume elements count keeps increasing, only to be limited by the exponential increase of available computation and power resources, which are currently available to analysts. With great power also come great delays in model preparation. The meshing process for such models requires significant time, which could otherwise be dedicated to more productive tasks. Things get worse when engineers need to account for the creation of different model variants and several design iterations for each one of them. 

Time and energy management issues due to the complexity of current models are a common issue among several engineering fields. The above challenges faced by CFD teams have also been reported in the structural domain. This is where BETA first successfully introduced the Modular Management methodology, an approach which promotes re-usability and modularity in the process of model preparation. At the core of this concept lies the idea that several parts of the model usually remain intact throughout endless iteration design cycles. Such sub-models could be prepared once and re-used several times as is, thus, reducing time and computational demands without any compromises in the accuracy of results. 

This presentation introduces an implementation of the Modular Management approach, tailored to the needs of the CFD domain. A full volume meshed model is split into components, which can be re-used in different model variant scenarios, thus decreasing the volume preparation time for the entire model to an absolute minimum, enabling engineers to achieve greater productivity and faster turnaround times. 

Konstantinos Ntamagkas 

(BETA CAE Systems, Greece) 

Dimitris Skoutelakos 

(BETA CAE Systems, Greece) 

Vicky Moshcou 

(BETA CAE Systems, Greece) 

KEYWORDS

Python, C++, NumPy, acceleration, JIT, AOT, compilation, GPU

ABSTRACT

Python code, often seen as slow for numeric workloads, can be transformed into a high performant code, easily switching from proof-of-concept prototype scripts into production-grade compiled code with near-native performance.

We begin with the reasons behind Python’s slow speed and connect them to performance bottlenecks. We then present a range of solutions comprising optimal pure Python code, (Just-In-Time) compilation, translation to C++ and finally offloading work to a GPU.

These approaches will be then demonstrated and benchmarked on select case studies: the N-body benchmark from the Computer Language Benchmarks Game, a thermal diffusion simulation, a linear algebra benchmark (AXPY, GEMV, GEMV+B), an I/O benchmark, and a benchmark based on META's Equivalent Radiated Power (ERP) tool.

Attendees will gain a clear understanding of why Python performance lags, how to accelerate it systematically, and when to apply different tools in practice.

KEYWORDS – 

CFD, Aerospace, anisotropy, mesh, ANSA 

ABSTRACT – 

This presentation will go through the evolution of ANSA towards becoming the benchmark for unstructured CFD meshing for external aerodynamics simulations for the aerospace industry. From early developments and smaller success stories to the defining collaboration with Airbus. Key point to unlock this ever-demanding branch of CFD is the application of anisotropy on both the surface and volume meshes with aim to increase the simulation fidelity without substantial increases on the overall mesh sizes and turnout times. With NASA vision 2030 driving the industry towards higher and higher fidelity simulations and the explosion of computing/solving power through GPUs the related developments significance is becoming more and more profound.

Olympic Hall III

Evening Dinner will be held in Athos Restaurant at 19:30

Breakfast/Lunch will be held at Athos, main hotel restaurant

In this Hands-on Workshop, you will have the chance to

• Learn the basics of LLMs and Agentic-AI through an online notebook
• Explore existing ANSA & META Agents through live demo
• Customize & build your Agent ANSA & META

To ensure the best experience, please don’t forget to bring your business laptops (fully charged). 

This session presents a live demonstration of Sketch 2D, a parametric 2D sketching ribbon which includes a full suite of sketching tools such as lines, arcs, circles, and splines, alongside dimensioning and constraint management capabilities that enable robust parametric control over geometry. 

We will walk through the core functionality of the ribbon, highlighting how 2D geometric entities can be defined, constrained, and modified interactively. The demo will also showcase how Sketch 2D integrates with downstream applications, including: 

• Cross Sections tool, where the Sketch 2D environment is adapted to create and manipulate cross sections.
• Parametric modeling pipelines, where changes in sketch dimensions automatically propagate to 3D extrusions.
• Optimization workflows, where the sketch collaborates with an external optimizer to iteratively adjust geometry based on performance objectives or design constraints.

Furthermore, we shall talk about future goals and integration opportunities. 

As a leading simulation software vendor, we are shaping the future of engineering through continuous innovation. This presentation outlines our strategic vision for advancing simulation technology, focusing on key developments such as AI, solvers, cloud-native deployment, scalable multiphysics capabilities, and improved platform interoperability. We will highlight how these innovations are driving faster, more accurate simulations and transforming user workflows across industries. Attendees will gain insight into our roadmap for delivering next-generation tools that helping teams across industries to solve increasingly complex challenges with confidence and efficiency.

Breakfast/Lunch will be held at Athos, main hotel restaurant

KEYWORDS –

IGA, Immersed, Geometry, Splines

ABSTRACT –

Splines are used to smoothly approximate or interpolate a set of points. Splines are widely used in computer graphics, CAD, and data fitting because they offer a high degree of control and precision. The most popular Spline geometries are, B-Splines which generalize Bézier geometries by using a set of basis functions. NURBS (Non-Uniform Rational B-Splines), are an extension of B-splines that includes weights, allowing the representation of both standard analytical shapes (circles, spheres etc) and free-form curves - surfaces. THB-Splines (Truncated Hierarchical B-Splines) extend traditional B-splines by introducing a hierarchical structure that allows local refinement of the spline space without affecting the entire domain. Their ability to efficiently represent complex geometries with varying levels of detail makes them a powerful tool in computational geometry and engineering simulations

Isogeometric Analysis (IGA) uses Spline-basis functions for the solution approximation, which ensures exact geometry throughout the analysis process. This eliminates the need for geometry simplification or meshing approximations, reducing errors and improving accuracy. IGA also provides higher continuity across elements, which is especially beneficial in problems involving higher-order partial differential equations.

The aim of this paper is to familiarize CAE community with the terms geometry and analysis on geometry. Introduce them to topics like local refinement on the geometry and unstructured Spline technologies. Moreover, topics like pre-processing models for IGA shell and solid analysis, using untrimmed or immersed techniques, applying tested workflows and best practices from FEA will be covered as well.

Meshing, especially for small but highly complex geometries such as embedded clips, is a challenging and time-consuming step in CAE workflows. Our first solution addressed this by automatically scanning parts to detect and segment embedded clips, and then reusing the mesh from historic data (library with meshed clips). However, this early approach was strict: mesh reuse was only possible when the clips were geometrically identical to their source, which limited the overall benefit.

To overcome this, we developed a more flexible methodology based on AI that enables mesh reuse across similar, but not identical, geometries. By extracting local and global features from CAD data, we can group similar clips. Since a mesh taken from one clip will not directly fit a slightly different geometry, we developed a keypoint matching methodology to align them and adapt the high-quality mesh to the new shape. This adaptation step enables efficient reuse even in the absence of identical parts, significantly reducing meshing effort. Beyond mesh reuse, the same feature representations have already been applied to part-type classification, opening the door to broader AI-assisted automation in engineering processes. The presentation will trace this evolution—from segmentation and strict reuse to similarity-based adaptation—and explore future directions as well as potential new applications.

The introduction of Finite Element models of the human body has offered novel ways of evaluating vehicle safety and restraint system performance. These so-called Human Body Models (HBMs) aim at providing a detailed, human-centric analysis and allowing a deeper understanding of injury mechanisms, ultimately contributing to safety feature optimization. However, posture and variant limitations pose a significant drawback to their implementation in different simulation scenarios, demonstrating the necessity of efficient handling tools. With the upcoming incorporation of Human Body Models simulations as part of the Euro NCAP assessments in 2029, an even more clear path towards a digital certification process in need of robust simulation set up solutions and result evaluation processes is pointed out. 

This presentation focuses on strategies for driving testing excellence across an organization based on a workshop attended during “WeTest Athens” conference. We will explore how to build testing capabilities at an organizational level for the long-term and how we need to focus on cultivating a quality culture across the company. We will highlight the importance of making an intentional effort to maintain an environment where high standards, continuous improvement and shared values around excellence are embedded in the way people work and think. Testing is not just a phase in the development lifecycle but a shared responsibility and mindset across all teams. We will finish the presentation with a success story from the QA team where the strategic planning to create Squish API helped us address QA challenges raised, and especially during the “Redesign” project. The initial concept and the API evolution over time will be discussed along with the performance problems that were encountered and how they were fixed.

This presentation outlines the successful completion of a strategic two-year migration project from Primer to ANSA within General Motors organization. The initiative aimed to modernize and streamline our simulation workflow, particularly in the context of LS-DYNA model preparation and validation. Throughout the project, we encountered and overcame several technical and operational challenges, including the full support and integration of the LS-DYNA numbering schema, the development of a robust check mechanism for LS-DYNA models, and the resolution of performance bottlenecks during LS-DYNA file input in ANSA.

Despite these hurdles, the project team demonstrated resilience and innovation, collaborating closely across departments to ensure a smooth transition. The migration not only enhanced ANSA’s simulation capabilities but also laid the groundwork for future scalability and efficiency in model handling. This presentation will share key learnings, technical solutions, and the impact of this transformation on our engineering processes.

Olympic Hall III