AI driven algorithm streamlines next generation nuclear reactor shielding design

AI driven algorithm streamlines next generation nuclear reactor shielding design
by Simon Mansfield
Sydney, Australia (SPX) May 01, 2025

The Nuclear Engineering and Application Laboratory (NEAL) at the University of South China has introduced a transformative approach to radiation shielding design for nuclear reactors. Leveraging advanced evolutionary algorithms, the team developed a reference-point-based optimization method that promises to enhance efficiency and effectiveness in meeting complex design requirements for emerging nuclear reactor technologies.

The novel method combines two custom-designed algorithms-Reference Point Non-dominated Sorting Genetic Algorithm (RP-NSGA) and Reference Point Multi-Objective Artificial Bee Colony (RP-MOABC)-that tackle many-objective optimization problems, a significant advancement over traditional expert-driven designs. These innovations are particularly useful for shielding systems in next-generation reactors including mobile, marine-based, and spaceborne units.

"The algorithms can automatically identify optimal shielding solutions under multiple objectives and constraints, greatly reducing design complexity," stated Prof. Zhen-Ping Chen, the study's corresponding author. "This provides crucial technical support for the conceptual design phase of nuclear reactors, especially for the new types of reactors lacking of engineering experience."

To verify the algorithms' performance, the researchers conducted two experimental simulations. In the first, a simple 3D shielding structure was optimized. The RP-NSGA method achieved shielding volumes and weights that were just 24.5% and 14.5%, respectively, of those generated by traditional crowding-distance-based methods. The RP-MOABC algorithm performed even better, reducing volume and weight to 17.3% and 9.77% of conventional benchmarks.

The second test addressed more complex, multi-layered shielding configurations. The algorithms produced designs that met stringent radiation dose limits while reducing volume by 19.12% and weight by 24.50%, highlighting their potential for real-world application in compact and efficient reactor designs.

Beyond reactor shielding, the optimization framework can be applied across multiple engineering disciplines, including core design, material selection, and even medical radiation protection. The team has embedded these tools into their proprietary software platform, MOSRT, offering a scalable solution for future engineering challenges.

Research Report:Many-objective evolutionary algorithms based on reference-point-selection strategy for application in reactor radiation-shielding design