PhysicsX
Canvas Category Software : Engineering : Simulation
PhysicsX is a deep-tech company of scientists and engineers, developing machine learning applications to massively accelerate physics simulations and enable a new frontier of optimization opportunities in physical design and engineering. Born out of numerical physics and battle-hardened in Formula One, we help our customers radically improve their concepts and designs, transform their engineering processes and drive operational product performance. We do this in some of the most advanced and important industries of our time – including Space, Aerospace, Medical Devices, Additive Manufacturing, Electric Vehicles, Motorsport, and Renewables. We work at the edge of advanced CAE, physics simulation and machine learning, to solve our customers’ most difficult design and control problems.
Assembly Line
On machine learning methods for physics
Simulation methods are employed to resolve the behaviour of matter (solids, fluids, gases, etc.), fields (electromagnetic, pressure, velocity, density), and any number of other physical phenomena that are driven by known local rules, particularly partial differential equations (PDEs). Therefore, traditional simulation methods typically involve some kind of discretisation of the physical domain of interest, such that the rules of the governing PDE can be locally well-approximated by a tractable computation. Local computations are stacked together and iterated upon until we converge to a solution. Beyond a narrow class of problems where a closed-form solution can be provided, this is generally how complicated problems are addressed. Many PDEs can exhibit chaotic behaviour in their full form, which often causes us to resort to simpler approximations at the PDE level, even before discretisation, to make them computationally feasible and ensure convergence.
ML methods in general provide a new approach to accomplishing engineering tasks. Any of these methods greatly accelerate iterations in the design optimisation workflow, as they allow us to search the space faster and guide our search towards promising areas. This results in better exploration, overall lower computational costs for simulating physics, and ultimately, higher quality designs in a shorter time-frame and with lower manual effort.
The models discussed so far do not leverage the fact that we often know the PDE that generates data and governs solutions; the focus has been to approximate the physical laws from observations, rather than impose them in the model structure explicitly. This is primarily because of the difficulty of incorporating such prior knowledge into the models, but also because simulation data may disobey the exact PDE due to the approximations required to facilitate numerical simulations. However, a new approach to simulation has recently been proposed that takes advantage of this prior knowledge in an effort to reduce data requirements and promote physically consistent solutions. Physics-Informed Neural Networks (PINNs), as presented by Raissi, Perdikaris, and Karniadakis (2017a, 2017b; Zhu et al. 2019; Karniadakis et al. 2021) introduces an artificial neural network (ANN) that takes as input the coordinates of any point in the domain of the PDE, and outputs the value for the solution field at that point. The ANN is tasked with representing the solution field, and is trained by sampling points randomly in the domain and penalising deviations from the PDE at those points. As long as the activation function of the ANN is sufficiently differentiable, residuals in the terms of the PDE can be easily evaluated, which can be combined into the loss function to be minimised with respect to the ANN parameters. The ANN is an ansatz about a parametrised form of the solution (albeit a particularly flexible one) and we attempt to fit the parameters such that it best matches the PDE. The idea harks back to older variational numerical simulation methods, like the generalised Galerkin approximation and others.
PhysicsX raises €29.2M to reinvent AI and simulation engineering technologies
London-based PhysicsX, a deep-tech startup building artificial intelligence(s) to power engineering, has secured $32M (approximately €29.2M) in a Series A round of funding. The investment was led by General Catalyst. The round also saw participation from Standard Investment, NGP, Radius Capital, and KKR co-founder and co-executive chairman, Henry Kravis. PhysicsX plans to expedite its growth in customer delivery, product development, and fundamental research.
The startup uses generative AI to facilitate groundbreaking engineering solutions across various advanced industries such as automotive, aerospace, renewables, and materials production.
AI-Driven 3D Printing: Unveiling the Future of Unusual and Practical Parts
We could see an increase of very unusual yet practical 3D printed parts in the near future due to AI technology.
“AI accomplishes this feat by solving the CFD or FEA equations in a non-traditional way: machine learning examines, and then emulates, the overall physical behavior of a design, not every single math problem that underlies that behavior. This uses far fewer computational resources while achieving an extremely robust evaluation of the design in every applicable environment. Hundreds of thousands of design candidates can be simulated and evaluated in less than a day.”
AI Optimization: New Opportunities for 3D Printing
AI accomplishes this feat by solving the CFD or FEA equations in a non-traditional way: machine learning examines, and then emulates, the overall physical behavior of a design, not every single math problem that underlies that behavior. This uses far fewer computational resources while achieving an extremely robust evaluation of the design in every applicable environment. Hundreds of thousands of design candidates can be simulated and evaluated in less than a day. Bottom line: Applying AI amplifies the typical 10-20% performance improvements of simulation tools alone—up to 30% and higher. (Of course it follows that real-world testing of finished parts remains an essential task to ensure that all quality and performance metrics are met.)
Velo3D requested PhysicsX to design and simulate a solution. PhysicsX has deep experience in simulation, optimization and designing for tight packages (from considerable work in F1 racing and expertise in data science, machine learning and engineering simulation), plus proprietary simulation-validated tools that can automatically iterate on designs using machine learning/AI-based simulations. The PhysicsX approach involves creating a robust loop between the CFD, generative geometry creation tools and an AI controller to train a geometric deep learning surrogate. The surrogate’s speed, producing high-quality CFD results in under a second, is then exploited with a super-fast geometrical generative method in another machine learning loop, which deeply optimizes the design towards whichever multiple objectives the engineer decides are important. The fidelity of the deep learning tools and robust workflow enables a highly accurate solution for final validation of the results against the validated CFD model.