openCFS: Open-Source Software Brings Advanced Multiphysics Simulation to Everyone

A team of Austrian researchers has unveiled a comprehensive open-source finite-element package that lets engineers and scientists simulate how electricity, sound, heat, and mechanical forces interact inside devices ranging from electric motors to microscopic loudspeakers. The software, called openCFS, is described in a paper published in the journal SoftwareX, and it arrives at a moment when demand for coupled multiphysics simulation is surging across the automotive, aerospace, and consumer-electronics industries. Unlike commercial packages that hide their inner workings behind expensive licenses, openCFS exposes every line of its C++ core under a permissive MIT license, inviting anyone to inspect, modify, and extend it.

The project has deep roots. Its predecessor, CFS++, was started by Manfred Kaltenbacher in 2002, and over two decades the codebase grew into a research workhorse for coupled-field problems. In 2020 the team took the decisive step of open-sourcing the software, and development is now supported by a non-profit association dedicated to fostering research around the code. The current release, tagged 2026S, comprises roughly 8,857 commits and is mirrored on GitHub, with primary development hosted on GitLab where continuous-integration pipelines build and verify the code across multiple platforms and compiler configurations.

What sets openCFS apart from general-purpose finite-element toolkits is the breadth of ready-to-use physics it ships with. The package solves partial differential equations for electromagnetics, spanning electrostatics through the full Maxwell wave equation; acoustics, including vibroacoustics and aeroacoustics; linearized compressible and viscous flow; structural mechanics with geometric nonlinearity and full anisotropy; and heat conduction. Crucially, these fields can be coupled either iteratively at a high level or through dedicated monolithic modules, for example directly coupling acoustics with mechanics or handling piezoelectric materials. Users configure everything through a schema-validated XML interface, which means no solver-level programming is required to assemble sophisticated coupled simulations.

Among the standout technical features are several formulations that are rarely available out of the box in any software, commercial or otherwise. These include harmonic balancing for solving nonlinear periodic problems directly in the frequency domain, time-domain equivalent-fluid acoustics implemented via auxiliary differential equations, thermoviscous flow with Maxwell slip boundary conditions for microscale devices, curvilinear perfectly matched layers that absorb outgoing waves along smoothly curved boundaries, rotating non-conforming interfaces for turbomachinery and electric machines, and vector magnetic hysteresis modeling with integrated parameter identification. Each of these capabilities has been validated in peer-reviewed publications, and all are accessible through the declarative input files rather than user-derived variational formulations.

The workflow is deliberately modular. openCFS does not include its own general-purpose mesh generator; instead it relies on established preprocessors such as Gmsh and Cubit, importing meshes and referencing named regions of elements throughout the simulation setup. A simulation XML file declares the mesh, maps regions to materials from a separate reusable material database, and defines sequence steps that can be chained, for instance pre-computing a static field that feeds a subsequent transient analysis. Results are written in HDF5 format and can be visualized in ParaView through a dedicated plugin or analyzed programmatically through pyCFS, a Python package distributed via PIP that can automate entire pipelines, from parameter studies to surrogate-based optimization.

Engineering rigor underpins the whole project. More than 600 automated physics-based test cases run across nine build configurations on Linux, macOS, and Windows, and a merge request can only be accepted when every test passes, enforcing complete regression coverage. An info XML file logged alongside every simulation records run time, memory usage, integrator definitions, and iteration counters, giving users deep insight into what the solver actually did. For developers, a TestPDE class documents every component needed to implement a new physical field, and Doxygen-generated interaction diagrams help newcomers find their way through the modular C++ architecture, which organizes code into directories for finite-element bases, bilinear forms, linear algebra, and optimization.

The application examples in the paper illustrate the software’s reach. In one case, the team simulated induction heating of moving steel sheets, coupling a nonlinear eddy-current problem solved with the harmonic balance method to a steady-state heat conduction problem, a strategy that exploits the vast time-scale difference between electromagnetic and thermal processes and can be extended to predict thermally induced buckling. In another, a coupled aero-vibroacoustic workflow modeled how airflow inside automotive piping systems generates sound that breaks out through the pipe walls at structural resonances. That single simulation involved 3.34 million equations, peaked at about 136 gigabytes of memory, and ran for 101 hours on ten cores for 20,000 time steps, demonstrating the code’s capacity for production-scale computation.

Perhaps the most striking example is a micro-electro-mechanical systems loudspeaker based on Advanced Digital Sound Reconstruction, a concept in which ultrasound pulses are redirected by rapidly moving shutter gates to synthesize audible sound. Simulating such a device demands viscous flow effects, mechanical nonlinearity, moving domains, and coupling among multiple physical fields, all within a geometry measured in micrometers or even nanometers. openCFS handles this with Maxwell slip boundary conditions that extend the validity of the flow equations into the slip-flow regime, together with non-conforming interfaces that let the team combine viscous and standard acoustic models in a hybrid workflow to keep computation times manageable.

Beyond simulation, openCFS doubles as a platform for design optimization. It implements density-based topology optimization using the SIMP method, complete with robust filters, multiple load cases, and automated binary projection, and its parallel C++ implementation puts optimization problems with up to a million design variables within reach. A particular focus is the emerging feature-mapping approach, in which high-level geometric primitives such as bars and circles with only a handful of design parameters are differentiably mapped onto a density field, enabling gradient-based optimization of structures and even anisotropic fiber-reinforced materials. The software also supports inverse problems, including acoustic source localization from microphone measurements and the determination of nonlinear magnetic material parameters, and an educational web application called iTop showcases its material-design capabilities.

The impact of the software extends into classrooms and industry alike. Several universities, including TU Wien, TU Graz, and Friedrich-Alexander-Universität Erlangen-Nürnberg, use openCFS in teaching, from introductory finite-element analysis to advanced software development, with students gaining first-hand experience of open-source contribution through the GitLab merge-request workflow. In industry, the code has served for more than fifteen years as the acoustic solver in hybrid aeroacoustic simulations, with applications spanning low-speed axial fans, human phonation, aviation structures, knee prostheses, and power electronics. Positioned between general-purpose PDE assemblers like FEniCS and deal.II and domain frameworks like MOOSE, openCFS occupies a well-defined niche: turnkey, validated multiphysics coupling and optimization for acoustics, MEMS, and electromagnetics, delivered under an MIT license free of copyleft constraints. Planned future work includes richer Python packaging, enhanced hysteresis and material modeling, adaptive mesh refinement for boundary layers, and integration with the preCICE coupling framework, promising to make this already capable tool even more connected to the wider simulation ecosystem.

Subject of Research: Open-source finite-element software for coupled multiphysics simulation and topology optimization

Article Title: openCFS: An open-source finite-element software for advanced multiphysics simulations

Article References: Mayrhofer, D., Toth, F., Wein, F., Roppert, K., Schoder, S., Heidegger, P., Landes, H., & Kaltenbacher, M. (2026). openCFS: An open-source finite-element software for advanced multiphysics simulations. SoftwareX, 36, Article 103102. https://doi.org/10.1016/j.softx.2026.103102

Image Credits: AI Generated

DOI: 10.1016/j.softx.2026.103102

Keywords: openCFS, finite element method, multiphysics simulation, open-source software, acoustics, electromagnetics, topology optimization, MEMS, aeroacoustics, induction heating, computational physics, SoftwareX