基于神经网络的激光选区烧结温度预测

Temperature Prediction Based on Neural Network for Selective Laser Sintering

有限元仿真是激光选区烧结(SLS)增材制造温度场预测分析的常见方法,但温度场仿真运算往往要耗费大量的时间。为了提高运算效率,提出了基于遗传算法(GA)优化反向传播(BP)神经网络的SLS烧结点温度预测方法。在大量覆膜砂材料多道多层零件烧结点温度仿真模拟试验的基础上,建立并训练了基于GA-BP神经网络的烧结点温度预测模型。开发了SLS烧结点温度预测软件,能够根据零件的尺寸及工艺参数,快速计算出烧结点温度,并进行可视化显示。通过零件烧结点预测温度与热像仪检测温度的对比试验,验证了温度预测的准确性。

A selective laser sintering(SLS)technique uses a method called finite element simulation to forecast and analyze temperature fields.However,the temperature field simulation computation takes a long time.An SLS sintering points temperature prediction approach,based on a genetic algorithm(GA)optimized back propagation(BP)neural network,is proposed to enhance the computation efficiency.A large number of simulation experiments of sintering point temperatures of coated sand multitrackmultilayer parts were conducted.A sintering point temperature prediction model based on GABP neural network was created and trained based on the above experiments.A piece of software for predicting SLS sintering point temperatures was developed.The software can quickly calculate and visually display the sintering point's temperatures based on the dimension and process parameters.The accuracy of temperature prediction was confirmed when the predicted and detected sintering point temperatures of the parts were compared experimentally.