Neural Concept Shape

All you need to achieve more

The core technology of Neural Concept shape are 3D convolutional networks that learn to predict the output of physical simulations or experiments based on the input shape’s geometrical properties. They can be used as complements to the traditional numerical simulation and experimental methods so as to alleviate the need for actual simulations or experiments along the conception pipeline. This dramatically accelerates and reduces the cost of the design process.

This proprietary deep learning technology is also fully integrated with state-of-the-art shape optimisation algorithms. Therefore, when relevant, the conception process can even be completely automated so as to find the optimal design(s) given a number of predefined objectives.

Neural Concept Shape radically improves
numerical simulations and design optimization

Neural Concept Shape running on computer

Performance

Drastically reduces 3D simulation time to below one second
Predicts simulation results with an accuracy (R²) higher than 95% on average
Can extend simulation, by integration of testing results

Optimization

Automatically explores the shape space for optimized designs
Automatically generates shapes optimal to desired output values
Comes up with original and innovative designs

Usability

Applicable on any 3D simulation application
Any engineer can use it in daily work without dealing with tedious parametrization
The raw 3D mesh is directly fed as input to the Neural Network of NCS
NCS is compatible with most CAD/CAE input/output formats