基于SGPLVM的齿轮坯预锻件多目标优化设计

在锻造齿轮坯的生产过程中,齿轮坯预锻件的设计质量优劣将会直接影响到终锻时金属的流动、锻模型腔的充满、锻件的质量和模具的寿命等方面。首先建立了预锻件外形尺寸和终锻时最大成形力、终锻时最大模具应力三者之间的极限学习机(ELM)网络模型,然后利用监督高斯过程隐变量模型(SGPLVM)算法,将输入变量降维成低维空间中对应的隐变量,通过隐变量和输出变量之间的关系,利用GA遗传算法多目标优化去计算最佳的预锻件的尺寸。结果表明:当终锻最大成形力和终锻最大模具应力均最小时,预锻件尺寸最优。

Multiple target Optimization Design of Pre forging for Gear Blank Based on SGPLVM

In the course of multistage forging for producing gear blank, the pre-forging design of gear blank directly affects metal flowing patterns of the finish forging, the situation of the forging die cavity, the quality of the products and die life. The network model between the size parameters of pre-forging and maximum forming force and maximum die stress of finish forging was established. Using the SGPLVM algorithm to reduce input variables to latent variables corresponding to the low dimensional space, through the relationship between the latent variables and output variables, the most superior shape and size parameters of pre-forging were determined by using GA algorithm. The results show that the preform is the optimal when both the maximum finisher forming force and the maximum finisher die stress are minimum.