基于PSO-SVR的选择性激光烧结制件收缩研究

Research on the Part Shrinkage during Selective Laser Sintering based on PSO-SVR

建立了聚苯乙烯(PS)选择性激光烧结(SLS)工艺参数与制件收缩率之间复杂的非线性支持向量回归(SVR)模型。采用正交试验设计支持向量回归的训练数据,并采用均匀设计初始化的粒子群优化(PSO)算法,优化选择向量回归参数。结果表明,均匀设计显著提高了粒子群全局寻优速度与精度,基于正交试验选择的小样本训练模型,较为准确地反映了工艺参数与收缩率间的关系,预测平均误差控制在4%,优于BP神经网络10%的平均预测误差。采用选择的模型,建立了工艺参数与收缩率之间的关系,并结合烧结理论进行了影响规律的分析,该方法为选择性激光烧结工艺的研究提供了一种行之有效的新思路。

The complex nonlinear support vector regression （SVR） model between the process parameters and the part shrinkage of selective laser sintering（SLS） the polystyrene（PS） power was established. The orthogonal experimental design was used to get the SVR training data, while the uniform design initialized particle swarm optimization （PSO） was utilized to optimize the SVR parameters. The results revealed that UD can greatly improve the optimizing performance of PSO. The OED data trained SVR model can accurately establish the relationship between the process parameters and the shrinkage, with the average perdiction error 4 %,which is better than BP neural network with average perdietion error 10%. Finally, the SVR model was used to predict the part shrinkage and establish the relationship between the parameters and the part shrinkage,and some process analysis on the influence law was made combined with the sintering theory, which provides a new available way to investigate the SIrS process.