Reuse information from previous simulations to accelerate the next ones.
In a design study, the same problem is solved many times while varying a few parameters: geometry, material, loading. These computations are strongly related, yet this similarity is still underused. The computational cost accumulates over the course of the campaign and, in practice, limits the design space engineers can explore.
Our approach operates at the scale of the entire simulation campaign rather than on an isolated solver.
Reuse
Instead of treating each simulation independently, we reuse information from previous computations to accelerate the next ones.
Validate
Each computational gain is evaluated by checking that the result remains relevant and sufficiently accurate for the intended use.
Compare
The gain is measured directly on your use case by comparing our approach with your reference computational method.
Simulation campaigns across several fields of engineering.
Our approach addresses studies in which a series of simulations explores close variants of the same problem: geometry, material, loading or operating conditions.
Thermal and cooling
Sizing heat sinks, cooling channels or microchannels. Optimisation means testing many geometries, each time evaluating temperatures, fluxes and pressure losses.
Scientific example — 3D optimization of liquid-cooled heat sinks, solving the Navier-Stokes equations together with heat transfer.
View an example from the scientific literatureStructural mechanics
Lightening a part, optimising its stiffness or its mechanical strength. Every variant requires recomputing displacements, stresses and sizing criteria.
Scientific example — Topology optimization of structures to reduce mass under displacement and mechanical stress constraints.
View an example from the scientific literatureAcoustics
Optimising mufflers, expansion chambers or acoustic ducts. One geometry is modified step by step and evaluated across frequencies to improve its attenuation.
Scientific example — Finite-element shape optimization of a muffler to maximize transmission loss over a range of frequencies.
View an example from the scientific literatureFluid dynamics and aerodynamics
Ducts, diffusers, exchangers: the flow governs performance. Every shape variant calls for a fresh computation of the velocity and pressure fields, and of the associated pressure losses.
Scientific example — Shape optimization of an automotive air-conditioning duct using CFD, with the CAD geometry used directly as the design space.
View an example from the scientific literatureShape optimization spans all four of these domains.
When geometry itself is a design variable, each modification to the CAD model requires a new simulation and, depending on the method, an update or reconstruction of the mesh. This is among the most computationally expensive types of campaigns and one where reuse across simulations can provide the greatest benefit.
Where we are today
The product is still under development. We currently have a demonstrator that we present on request. If you run parametric studies and computational cost is a limiting factor, we can discuss your use case.
Stay informed
Our technical updates and publications. No marketing emails.
Inria Saclay
École polytechnique campus
1 rue Honoré d’Estienne d’Orves
91120 Palaiseau, France
