In order to mitigate biomechanical injuries sustained by middle-sized children during frontal collisions of school buses, this study investigated the optimal design of intelligent airbags. Initially, a validated school bus simulation model was constructed based on sled test data. Subsequently, methods for determining head, thorax, and femur injury values for a 10-year-old THUMS dummy were established, and the model, including the school bus restraint system, THUMS dummy and intelligent airbag, was developed. Next, based on NSGA-III, the adaptive propagation factors were proposed, the Gaussian mutation operator and the evolutionary mechanism of the particle swarm optimization method were introduced, and the non-dominated sorting level was improved, thus putting forward the improved NSGA-III. Finally, the design optimization of the intelligent airbag was conducted using the improved NSGA-III to determine the optimal configuration of the airbag. The results demonstrate that the improved NSGA-III outperforms the other three state-of-the-art optimization algorithms. Under normal, 10° forward tilt, 20° forward tilt, 5° right tilt, 10° right tilt and lying sitting postures, the head injury criterion (HIC15), intracranial pressure (IP), liver pressure (LP), left and right femur forces (FL and FR, respectively), and weighted injury criterion (WICC) for middle-sized children are significantly reduced when the proportional coefficient of the gas mass-flow rate of the inflation valve, the opening pressure and opening coefficient of the deflation valves, and the installation height of the air bag are set at 1.23, 1.37×105 Pa, 2.10, and 478 mm, respectively.
