选择性激光烧结成型件密度的支持向量回归预测

根据不同工艺参数（层厚、扫描间距、激光功率、扫描速度、加工环境温度、层与层之间的加工时间间隔和扫描方式）下的选择性激光烧结成型件密度的实测数据集,应用基于粒子群算法寻优的支持向量回归（SVR）方法,建立了加工工艺参数与成型件密度间的预测模型,并与BP神经网络模型进行了比较.结果表明：基于相同的训练样本和检验样本,成型件密度的SVR模型比其BP神经网络模型具有更强的内部拟合能力和更高的预测精度；增加训练样本数有助于提高SVR预测模型的泛化能力；基于留一交叉验证法的SVR模型的预测误差最小.因此,SVR是一种预测选择性激光烧结成型件密度的有效方法. 关键词： 选择性激光烧结 密度 支持向量机 回归分析

Density prediction of selective laser sintering parts based on support vector regression

The support vector regression （SVR） approach combined with particle swarm optimization for parameter optimization, is proposed to establish a model for estimating the density of selective laser sintering parts under processing parameters, including layer thickness, hatch spacing, laser power, scanning speed, ambient temperature, interval time and scanning mode. A comparison between the prediction results and the results from the BP neural networks strongly supports that the internal fitting capacity and prediction accuracy of SVR model are superior to those of BP neural networks under the identical training and test samples; the generation ability of SVR model can be efficiently improved by increasing the number of training samples. The minimum error value is provided by leave-one-out cross validation test of SVR. These results suggest that SVR is an effective and powerful tool for estimating the density of selective laser sintering parts.