Navier AI’s Post

Navier AI (W24) is making CFD simulations 1000x faster. Their ML-based solvers, trained on simulated and experimental data, enable aerospace engineers to run engineering-quality fluid dynamics simulations in seconds. Today, engineers simulate fluid dynamics interactions using numerical methods based on the known laws of physics. However, this approach requires deep expertise and can take hours on high-performance computing infrastructure. Navier AI’s ML-based CFD solver produces fluid dynamics simulations 1000x faster for a fraction of the cost. Combined with a modern but powerful UX, Navier AI removes the complexity of traditional solvers and empowers engineers to perform more design studies and better analysis-in-the-loop optimizations. Founded by former SpaceX rocket scientists, Cameron Flannery and Evan Kay are building the software tools they wish they had while previously building, scaling, and commercializing products at the world’s leading hard tech companies. Navier AI is revolutionizing the current state of Aerospace Engineering by making simulations fast and easy with ML-based solvers. Congrats Cameron and Evan on the launch!

I’m excited to finally share what Cameron Flannery and I have building over the last few months! Fluid dynamics underpins world-changing products, ranging from rockets, to wind turbines, to manufacturing processes. After working alongside the best engineers at hard-tech startups and leading aerospace companies, I saw the untapped potential of design optimization caused by slow and costly hardware development processes. At Navier AI, we’re enabling Engineers to perform analysis-in-the-loop design optimization with real-time CFD analysis. This is just the beginning of enabling Engineers to simulate the physical world with fast and reliable ML-based solvers 🚀

The current market requires CFD solutions that provide accuracy, automation, speed, and integration with artificial intelligence (AI)—principles that underpin Cadence Fidelity CFD. This blog post will shed light on how these four pillars, collectively referred to as the 4A's, position Cadence Fidelity CFD software as the next-generation multiphysics CFD solution. #cfd #AI #automation #accuracy #simulationsoftware #cadence Read more: https://lnkd.in/g6A64TRZ

Can AI replace CFD in engineering simulations? When we talk about Artificial Intelligence (AI) and Computational Fluid Dynamics (CFD), we're are discussing two tools that evaluates the future of engineering. Both of them has different roles but they can complement each other in solving problems. CFD is the engineer's go-to tool for simulating fluid flow, aerodynamics, heat transfer, etc. It converts physics that is applied to this real-world into equations that can be solved on a computer. It helps us design better aircraft, cars, rockets, and even wind turbines. On the other hand, AI doesn't rely on those physics equations which we use in daily life. Instead, AI learns from data to predict results. It can spot patterns, optimize designs faster, and even solve problems and it doesn't even need detailed physics models like CFD. So, how do they compare? CFD gives us detailed, physics-based simulations, but it can be slow and computationally expensive for some problems. AI is faster and it can handle massive datasets. It is great at predicting results without needing to solve every small detail of physics equations. From this discussion we can conclude that while AI and CFD are different but their combined potential could lead to many ways in design and cfd industries. It can also innovate industries like aerospace, automotive, and energy! 📌 Do let me know what do you think about these two! ♻️ Repost if you find this post interesting. __________________________________________________________ Want to stay connected? Follow me Sarah Mirza and click the 🔔 to stay informed!

Optimizing an #aerodynamic shape with CFD? There is nothing more exciting in #engineering! However, the time complexity of a CFD simulation could be O(n³) or higher, where n is the number of grid points or cells used in the #simulation. CFD simulations typically involve solving on a 3D grid. This can be rewarding in terms of high-fidelity output, but it is also at the cost of being computationally intensive. More time for computations means fewer design changes to be explored in a given amount of time. https://lnkd.in/d7Hyww_h The Neural Concept paper contains use cases where an almost real-time Deep Learning predictive model driven by CAD and CFD data is used for optimization of designs: "NACA Profiles and their Optimization" "Hydrofoils and Their Optimization"

Introducing Ansys AI - Ansys' new AI-augmented simulation technology designed to reshape the future of engineering. #Artificialintelligence #AI #ansys #simulation

Ansys developed a AI design part simulation system it will no required of rememberimg parameters and all just giving a design and material type so the AI ansys is going to generate the whole stuff ...by ansys

Check out this machine learning case study (& the podcast with my friend Jousef where we discuss it https://lnkd.in/ecky3arg) Machine learning is becoming increasingly helpful at different stages during the design process, which is providing value to the user in several forms. Here are the outcomes from a case study we recently published in the AICHe 2024 conference a few months ago. The persona is a CFD engineer who is designing a mixing vessel for the chemical process industry. Stage 1: We need to optimize the geometry of our design to be as performant as possible (minimize power input, maximum mixing, limited strain levels) >>>Outcome: Incorporating AI into our optimization process reduced the time by 22% Stage 2: With my optimal design, I want to test a wide variety of operating conditions to ensure my design is safe. >>>Outcome: Build a surrogate to test ‘thousands’ of design points in ~5 minutes Stage 3: I want to run minimal extra simulations to intelligently sample such a huge design space >>>Outcome: ML-based sampling techniques are superior to legacy methods. ~50 simulations yields a surrogate accuracy equivalent to ~250 from legacy methods, in this case Stage 4: I want spatial field information on-demand, to compliment 0D results >>>Outcome: Train ROMs that can predict spatial fields results, not just 0D metrics Siemens -> Siemens Digital Industries Software --> Simcenter #ai #datascience #machinelearning #cae #cfd #chemicalengineering 

One killer use case for AI in CFD/FEA etc is sampling. If you haven't heard about it yet, I promise you will before long. Bottom line: AI-based methods are very good at helping us explore huge design/operating spaces, so that we don't have to spend exorbitant amounts of simulations just to understand how our design performs under different conditions. These models are dynamic, compared to 'set it and forget' types of methods that may be space-filling (latin hypercube) or structured (uniform sampling). Check out this sweet video that compares how fast AI-based sampling methods (left) and legacy sampling methods (right) compare at finding key behaviors in the design space. #ai #cfd #fea #simulation #cae #aerodynamics #physics

For a CFD/CAE engineer in 2024, Monday mornings usually start with questions... ...questions like... ❓What are some specific challenges and best practices in applying ML in CFD projects? ❓How can model results be interpreted and used? ❓How can I iteratively make changes to my ML models to improve it? ❓How do I handle inconsistent, missing, or skewed data for my machine learning project? ❓How do we judge model performance for a variety of datasets, or in a vast design space? ❗Thanks to Justin Hodges, PhD my week will end with answers! And so can yours. If you order now... Link for USA here: https://lnkd.in/eQbqFqvA (other countries in comments). #ai #datascience #cae #cfd