Towards AI-Driven design: Where machine learning should meet engineering excellence

At Cnam, Paris, June 29th 2026, 2 p.m.<

Tariq Benarama, Caroline Sainvitu
Researchers, Cenaero, Gosselies, Belgium

Despite tremendous advances in mathematical optimization in the past decades, automated engineering design still faces multiple challenges to be fully deployed in industry. Surrogate-based optimization techniques are often presented as generic solutions to tackle complicated design problems, but their usage on a daily basis in engineering offices remains complex due to

  1. their need for specific parameterizations potentially including multiple types,
  2. the usual ill-posedeness of industrial optimization problems (unclear definition of objectives and constraints at the beginning of optimization campaigns),
  3. the complexity and restitution time of the multi-disciplinary computational chains required to simulate the behaviour of the designed product/process through the (potentially unstable) extraction of the optimization's quantities of interest (QoIs),
  4. the non-linearity of the QoIs w.r.t design parameters,
  5. the need for engineers to learn new mathematical/algorithmic concepts outside their classical area of expertise.

To circumvent these difficulties, recent developments in Artificial Intelligence and Machine Learning can be leveraged in synergy with Data Mining tools to both adapt existing search algorithms and methodologies to industrial needs while helping designers control the overall optimization process to increase their trust in the proposed solutions.

This presentation highlights recent applications of SBO techniques to complex industrial designs, spots limitations and difficulties observed with state-of-the-art techniques and discusses development trends and potential breakthrough brought by the coupling of AI techniques with generic advanced SBO methodologies. This represents the paradigm shift from fixed design assistance to active intelligence-driven optimization that transforms how engineers tackle complex, multi-objective design problems under stringent constraint conditions by focusing on key drivers at the product/process level.