KAIST researchers are building an AI-assisted inverse-design framework for solid rocket motors, aiming to expand experience-based propulsion design toward broader optimization linked to propulsion performance and vehicle-level mission goals.

What if engineers could begin rocket motor design not with a familiar shape, but with the thrust-time profile they want to achieve? KAIST researchers are building an AI-assisted inverse-design framework to make this approach possible for solid rocket motors. At Rocket Lab KAIST, the framework first focuses on three-dimensional grain design, which governs how solid propellant burns and how pressure and thrust evolve. By combining AI-based candidate generation with physics-based verification, this approach helps engineers explore a broader design space while keeping final judgment in human hands.

Rocket propulsion design is not the selection of a single component. Once mission requirements are defined, engineers must coordinate propellants, grain geometry, combustion chambers, nozzles, insulation, structures, and operating conditions. A change in one element can affect the performance and constraints of others. Propulsion design is therefore an integrated and iterative engineering problem.

Engineering experience has long been the most valuable asset in this process. Experienced designers narrow the design space by considering proven configurations, test results, manufacturing processes, safety margins, and operating conditions. This approach remains essential because it reflects accumulated knowledge and practical judgment. Its limitation is that the final design is often shaped by what the design team already knows. Even when better solutions exist outside familiar patterns, they may be difficult to imagine, calculate, and iterate manually.

Inverse design offers a way to address this limitation. In conventional design, engineers choose a grain shape first and then calculate performance. In inverse design, they begin with a desired pressure or thrust history and search for a shape that can produce the expected result. Researchers have explored this idea for decades, but earlier work was limited by computing resources, simplified models, and restricted design variables.

Recent advances in AI-based generative models and optimization are bringing this idea back into focus. AI-assisted inverse design is not intended to replace propulsion engineers. Rather, it may expand the design space that human experience can reach. Starting from a target pressure or thrust curve, AI can generate grain candidates not selected in advance. Engineers then evaluate them using physics-based analysis, manufacturability, safety, and operational feasibility.

Rocket Lab KAIST is translating this concept into a practical framework by representing three-dimensional grain geometries in a computational form. Using an in-house internal ballistics solver developed by the laboratory, the team calculates pressure and thrust histories and uses them as AI training data. The AI model proposes grain candidates close to a target pressure history, and the team re-evaluates and ranks them through physics-based analysis. AI does not make the final decision; it expands the set of promising designs engineers can examine.

The long-term goal extends beyond a single propulsion component. Grain-shape optimization can improve motor mass, volume, thrust profile, and overall performance. This approach could later connect to vehicle- and mission-level optimization for launch vehicles or guided flight systems. Still, an AI-generated candidate is not a final design. Manufacturing, quality control, testability, and operational reliability remain human engineering judgments. AI supports those decisions across a broader design space.