基于神经网络与NSGA-Ⅱ算法的LTCC基板成型工艺优化

Optimization of LTCC Substrate Molding Process Based on Neural Network and NSGA-Ⅱ Algorithm

摘要:

低温共烧陶瓷(LTCC)基板制作工艺复杂,产品质量对工艺参数十分敏感,微小的成型缺陷就会影响其功能特性。文章将BP神经网络和多目标遗传算法——NSGA-Ⅱ(Nondominated Sorting Genetic AlgorithmⅡ)相结合用于LTCC基板在层压和烧结工艺过程中的工艺参数优化。根据LTCC基板成型过程中出现的微通道变形、互联金属柱错位、基板翘曲三种主要成型缺陷与相关工艺参数的正交仿真实验结果,对神经网络模型进行训练,建立了三种成型缺陷与工艺参数之间的神经网络预测模型。在此基础上,采用多目标遗传算法对三种成型缺陷相关工艺参数进行多目标优化求解,得到了较优的工艺参数组合,用于指导相关产品制作工艺设计。?更多还原

Abstract:

The production process of LTCC substrate is complex and its product quality is sensitive to the process parameters, minor molding defects will affect its functional properties. In this paper, BP neural network and multi-objective genetic algorithm-NSGA-Ⅱ (Non-dominated Sorting Genetic Algorithm Ⅱ) are employed to optimize the process parameters in the laminating and sintering process of LTCC substrate. Based on the orthogonal simulation experimental results of the related process parameters and the three major molding defects of the deformation of the micro channel, dislocation of metal interconnection column and substrate warpage occurred during the process of LTCC substrate molding, the training of the neural networks model is carried out and the neural network prediction models among the three major molding defects and their process parameters are established. Based on that, the multi-objective genetic algorithm was applied to optimize the three defects models for multi objective optimization, then the optimal combination of process parameters are obtained, which are used for the guide of the design and fabrication of the related products.