The Impact of Deep Learning in CAE

Presentation by Prof. Michel Bercovier from the School of Engineering and CS, Hebrew University of Jerusalem.

Artificial Intelligence (AI) has had a bumpy history with highly fashionable periods followed by lows. Its applications in CAE , that is Mechanical Engineering and Simulations as design tools have been met with mixed feeling at each wave. One of then best example of failure is the usage of expert systems for Finite Element simulations. In the 90s, there was some tentative at Genetic algorithms, not followed by any wave. At the beginning of our century, Support Vector Machines and related Big Data analysis brought a new tide, but the difficulties of training and data handling again slowed down the momentum, and the dimension of the parameter space was a barrier in CAE. The last few years witness an AI tsunami called Deep Learning. In our talk we will briefly describe Neural Networks and Deep Learning (DL). We will review recent applications in Numerical Simulations.

Next we will try to analyse where, if and how Deep Learning can contribute to solve difficult problems in CAE.

The main drawback is that it is presently a kind of "magic" method. But the huge efforts of some of the main players (Google, Facebook, Amazon ...) has a definitive influence in some optimization problems and the solution of complex systems ( such as the autonomous car). CAE engineers must evaluate the possible changes (and limits) coming from Machine learning, whether DL, or SVM or even Genetic algorithms.
Have a Question? If you need assistance beyond what is provided above, please contact us.