To tackle new challenges, engineers need radically new capabilities, including more effective ways to harness our computational resources. To this end, Neural Concept exploits the power of Deep Learning to assist engineers to more quickly evaluate man-made designs and to guide them to more performing solutions. Our algorithms act as a force multiplier to humans’ creativity and help them come up more quickly with innovative designs that meet a variety of complex requirements. It uses the following approach: - Broadening the reach of simulation and optimization-driven design. Because of their historical origin, simulation tools are not well adapted to design optimization in production environments. It is extremely hard for design teams to leverage insights provided by advanced simulations and by the specialized teams who develop them. Neural Concept resolves this by integrating interactive simulation and design optimization tools in the workflow of the design engineers, instead of delegating this task to separate expert teams. Furthermore, the great engineering feats of our times, such as modern cars, airplanes, or nuclear submarines, have not been achieved by a few individuals with powerful ideas or algorithms. Instead, they are the result of a fragile network of interactions between hundreds of engineering teams, each with its own expertise. In that context, removing the friction introduced by simulations and lengthy back-and-forth technical exchanges dramatically improves the dynamics of large-scale engineering design. - Empowering creativity. Even in the hands of experts, past design tools do not deliver much in the way of radical innovation. They are typically used either to choose optimal configurations given a few predefined tuneable parameters or to make minor local adjustments to manually-built designs. This is a far cry from making major design decisions and guiding engineers towards bold new ideas, which is what we enable. The future requires making computer-driven exploration the central paradigm of engineering design processes, to broaden its scope of use. Neural Concept was created to promote and realize this vision, to empower every engineer, in every engineering organization to discover the very best. In this talk, we explain how Neural Concept’s Software, NCS, powered by recent algorithms based on Geometric Deep-Learning, allows shortcutting any simulation chain through a predictive model that outputs post processed results right from the CAD design. These models are being used in engineering companies to simplify processes and to emulate the expertise of simulation engineers in the hands of product or design engineers early in the development process. Thus, the number of iterations between teams are reduced while accelerating the design activities. Heat Exchanger Application Cases To illustrate the practical capabilities of Neural Concept Shape, we will use a set of industrial examples from success cases at Automotive OEM and Tier1-suppliers. A special focus will be made on Heat Exchanger design applications. Heat Exchanger design and optimization is a critical topic with the electrification of vehicles, and more specifically for battery cooling applications, but also for other electronic components inside the vehicle. It is a major driver in the vehicle efficiency and its durability. Hence, a well established and efficient design process is a key competitive advantage within the automotive industry. Heat Exchanger design is a complex task: the engineer has to make a lot of important design decisions to select the right concept (fins/pins heat exchanger, plate with dimples, channel structure…). Most of these decisions are based on best practices and rely on the experience of the team in charge of the development of the product. Moreover, within a single heat exchanger concept, the design freedom is significant, and small design changes could lead to drastic performance variations. Most of the time, the engineers and design teams can only afford to explore a few variations from a given concept, due to the complexity of the CAD/CFD workflow. Using Neural Concept Shape, engineers are now able to build a predictive model that reproduces in a few seconds the volumetric velocity/pressure and temperature field from the 3D CAD geometry directly. Predictions include the recirculation zones, as well as global quantities such as the pressure drop or maximum temperature on the heat exchanger. We conducted a study using Heat Exchanger internal CFD simulations where we show how the AI model is able to run a non-parametric optimization, leading to drastic performance improvements compared to a standard parametric study. Moreover, the AI model can be accessible to the design team through a simplified graphical interface. Through this GUI, they can, in a few seconds, and directly from their 3D CAD designs, get postprocessed simulation results. Ultimately, this leads to shorter lead times, especially critical for the RFQ phase in the automotive industry, together with better product performances.