基于凝固前沿演变特征的熔铸装药成型工艺参数智能优化

Melt-cast explosive processes present a significant correlation among the profile features of solidification front during the molding process and the shrinkage cavity and porosity defects within the grain after molding. To improve the molding quality of melt-cast explosive, an indictor representing the solidification front profile is defined, and the correlations among the indictor and the defects, such as shrinkage cavity / porosity volume and maximum porosity, of the grain, are examined. A surrogate model is developed based on a B-spline neural network model that describes the relationship between the key process parameters and the two-dimensional transient temperature field of melt-cast explosive, and the neural network is trained using a simulated dataset. The key process parameters of melt-cast explosive are then optimized using a genetic algorithm with the aim of maximizing the indicator of solidification front. The results show that the quality parameters of shrinkage cavity / porosity volume and maximum porosity of the grain decrease from 19. 832 mm³ and 4. 71% to 3. 129 mm³ and 0. 66%, respectively, as the process parameters are optimized from P⁰ = [100, 85, 0. 25, 0. 25, 90, 5, 0. 6]^T to P^* = [91. 725, 94. 961, 0. 498, 0. 151, 100, 6, 0. 595]^T, and a fast prediction and optimization for the molding quality of melt-cast explosive are achieved. The proposed method can provide new ideas and strategies for the process optimization of melt-casting explosives and contribute the solutions for the development of high-performance melt-casting explosives, which can be referred to improving the production efficiency, reducing the costs, and ensuring the consistency of molding quality.