We're excited to announce a new CastNet release, packed with new features and functional enhancements. A major focus is making OpenFOAM dramatically easier to use for Windows users. We now provide native Windows installers for both the Keysight/OpenCFD (v2512, v2606) and the Foundation (v12, v13) OpenFOAM versions — integrated into a ready-to-use base environment that includes ParaView and a set of useful OpenFOAM add-on tools. BlueCFD is of course still supported.

And it's not just CFD: the same environment supports structural-mechanics analyses via the nonlinear FE solver CalculiX. As a special highlight, we also ship preCICE with a Windows installer, enabling FSI coupling alongside OpenFOAM and CalculiX — a complete suite for solving CAE problems in one consistent workflow.

The video demonstrates several highlights using a heat exchanger as an example: our hybrid meshing approach (hex-core with easy-to-create extrudes across multiple region zones) and multi-region support for CHT. Beyond the extensive workflow improvements and support for the latest OpenFOAM versions, this release also brings a significantly expanded RunGui monitoring tool, RBM support for mesh morphing and overset meshes, and new meshing features.

Almost every atomization problem has a gap in time scales. The primary breakup happens within a few milliseconds and only resolves with adaptive refinement, tiny time steps and tens of millions of cells. But what matters for the product develops much further downstream and over far longer times — paint and coating (where spray-wall interaction decides the result), metal atomization (droplets cool and solidify into powder), and wall films and spray chemistry such as AdBlue/SCR.

Since resolving the breakup for the full process duration isn't affordable, we separate the two. Just downstream of our VoF-to-Lagrangian transition we place a sampling plane that records every droplet crossing it — size, velocity, direction and local mass flux — as individual events over seven to eight full pulses, so the natural kHz pulsation is preserved.

In the video we compare different methods of re-injecting the particle data into a purely Lagrangian setup: fully resolved transient injection, with or without adaptive mesh refinement (AMR) of the droplets, and conversion into an OpenFOAM cone injector. Having the various methods available lets the user specifically bound the errors introduced by simplifying assumptions.

Here is the third example of our membrane-FSI solver, integrated directly into OpenFOAM's rigid-body-motion (RBM) framework — this time a VoF application (isoAdvector). Following OpenFOAM's floating-object tutorial, we modelled a box built from baffle walls, two of which are elastic. The box is submerged in water and filled with air inside; during the first seconds it reaches the equilibrium between buoyancy and wall deformation through a relaxed FSI coupling, then it is exposed to waves.

Thanks to the tightly integrated coupling, the total runtime stays relatively short at about 2 hours on 16 cores with roughly 600,000 cells. A key advantage: the time-step size can be adapted dynamically to criteria such as the alpha-Courant number, the box's motion velocity, or the deformation rate of the membranes.

We have integrated numerous methods directly into the solver that stabilise the solution and make it particularly efficient for membrane applications: different relaxation schemes (Aitken, IQN-ILS), temporal filters and spatial smoothing of the pressure, pre-tensioning techniques, and various damping approaches such as Rayleigh damping.

We ran the classic Mok benchmark: a flow channel bounded by an elastic membrane that deforms under negative pressure from a time-varying inlet velocity — with excellent agreement with the published reference.

Our FSI solver was built for elastic membranes coupled to rigid bodies (via OpenFOAM's rigidBodyMotion library). But as Mok shows, the same membrane can just as well be the outer boundary of the flow domain itself.

Why we keep the solver fully integrated in OpenFOAM: stability tools live where they're needed — right inside the solver; that matters for membranes, which cannot carry compression and start wrinkling the moment the pressure flips sign. There is no external coupling overhead, and implicit FSI (IQN-ILS) with quasi-Newton acceleration handles strong added-mass cases out of the box.

Coupled FSI without leaving OpenFOAM: for suitable cases, DHCAE integrates the membrane's structural mechanics directly into the solver, using OpenFOAM's own infrastructure. For coupled multiphysics, preCICE and solids4Foam are well-established, powerful options; yet in specific scenarios it pays off to embed the structural mechanics directly — when the structural cost is small compared to the flow, a reliable element formulation is available, and dedicated stabilization or acceleration techniques matter for reliable convergence.

That is exactly what we built: membrane elements in OpenFOAM's rigid-body framework (RBM). Moving bodies can combine rigid parts and elastic membrane regions. Built in for robust FSI: a tension-field wrinkling model, static pre-stress (pOffset), spatial and temporal smoothing of the membrane pressure load, FSI-driven adaptive timestepping, ramps for a shock-free start-up, and Rayleigh damping.

The complete FSI runs as an outer loop around PIMPLE, with Aitken or IQN-ILS relaxation — suitable for transient and quasi-static applications. Single-phase and VoF variants are available. Validated against the Hencky membrane and the Mok benchmark. Typical applications: wind and flow loads on sails and tensile membrane structures, inflatable structures, hydraulic membrane valves, parachutes, and mobile flood-protection barriers.

We make intensive use of adaptive mesh refinement (AMR) in OpenFOAM within our hybrid VoF-to-Lagrangian transition solver to resolve droplets and ligaments during atomization before they are converted into Lagrangian parcels. Once converted, the spray is propagated on a deliberately coarser mesh to keep computational cost manageable.

This creates a cell-size dilemma in the spray region: the mesh must be coarse enough for the point-particle assumption to hold (cells larger than the droplets, low local volume fraction), yet fine enough for the LES-resolved gas flow to carry the wakes and cross-flow between large droplets that drive small-particle entrainment and momentum exchange.

So we extended our AMR to also refine inside the Lagrangian spray region itself, driven by a particle-influence field rather than the VoF indicator. Refinement reaches into the spray cone wherever parcels are present but stays well above the droplet scale, preserving the validity of the point-particle approach — without paying the cost of refining the entire domain.

Industrial bag filters are the workhorses of dust separation — but the loading pattern inside an operating filter house is hard to measure and notoriously uneven. Where does the dust actually go? Which bags carry the load? How does the airflow redistribute as the medium fills up?

To explore this, we simulated a full loading cycle of a bag filter house with 16 bags using Lagrangian particle tracking coupled to a filter medium whose local resistance rises with dust load. The iterative coupling between particle deposition and flow field is handled by our filter solver, built on OpenFOAM v2512.

A few observations: the inlet jet drives dust preferentially onto the central bags — and within those, onto the lower sections first; rear-wall bags accumulate the heaviest total load over time; airflow continuously redistributes toward the less-loaded bags. The benefit: optimise the flow distribution to avoid premature unplanned shutdowns, use the installed filter area efficiently, and match cleaning cycles to the actual load rather than fixed intervals.

Credits: the filter house geometry is a demonstration model by wahyu suryanto on GrabCAD — thanks for making it openly available.