Aiming at the challenges of high dimensionality, strong constraints and high computational cost of multi-objective optimisation problems in aerospace test process automation, the study proposes an intelligent generative model based on the Improved Undominated Sorting Genetic Algorithm III. The model generates domain-adapted uniform initial populations by implementing asymmetric dense sampling of key test parameters through Latin hypercubic sampling; meanwhile, the Kriging agent model is introduced to dynamically construct the response surface of the objective function to reduce the overhead of real evaluation. The results show that: during the iteration process, the spacing index of the improved model is reduced from 0.892 to 0.162, with a reduction of 70.6%, which is significantly better than the comparison algorithm (with a reduction of 34.1%-44.0%), proving that it has leading efficiency in the enhancement of the uniformity of the distribution of the solution set; the mean values of the model"s hypervolume and inverse generational distances are 0.928 and 0.048, which are respectively 5.0% - 13.7% and 26.7% lower than that of its method. 13.7% and 26.2%-50.0%, respectively; the actual deployment data shows that the model consumes 47.90 hours on average, the resource consumption rate is 78.20%, and the function coverage rate is 94.90%, which is 5.2%-14.6% less than the comparative algorithms in terms of time consumed, 3.0%-7.7% more in terms of resource utilisation rate, and 1.8%-5.2% more in terms of coverage rate; the study provides efficient and reliable solutions for the automation of the aerospace testing process. The research provides an efficient and reliable solution for the automation of aerospace testing process and lays the foundation for the development of aerospace engineering.