基于改进神经网络的增材制造刀轨优化研究  

Research on optimization of additive manufacturing tool path based on improved neural network

作　　者：董海 郭煜峰 DONG Hai;GUO YuFeng(School of Applied Technology,Shenyang University,Shenyang 110044,China;School of Mechanical Engineering,Shenyang University,Shenyang 110044,China)

机构地区：[1]沈阳大学应用技术学院,沈阳110044 [2]沈阳大学机械工程学院,沈阳110044

出　　处：《中国科学：技术科学》2024年第3期553-566,共14页Scientia Sinica(Technologica)

基　　金：国家自然科学基金(批准号:71672117);中央引导地方科技发展资金计划(编号:2021JH6/10500149)资助项目。

摘　　要：本文针对选择性激光烧结增材制造(selecting laser sintering,SLS)内部质量问题,提出贝叶斯优化BP神经网络(BO-BP神经网络)路径优化模型,提高了SLS成型质量,减少打印过程能源消耗.首先,建立打印件路径规划流程,利用动态规划技术得到给定激光网格中最终路径并移除冗余路径,针对激光烧结过程建立基于导热微分方程的SLS模型,并对其路径进行数值模拟,利用各网格节点最终温度值计算材料网格内部节点温度梯度值并排名,以此值作为神经网络算法训练并测试数据的数据集;其次,通过对所得激光路径进行数值图像转化,得到相应的平均热梯度分布图,将所得到的激光路径转化成灰度路径图像,采用改进BO-BP神经网络算法进行训练;最后,使用Softmax函数将神经网络输出数字转换为每个图像的概率,选取线性优化、无监督学习神经网络、遗传-反向传播神经网络(GA-BP)和反向传播(BP)神经网络进行预测结果比较,得出BO-BP神经网络算法相较于其他算法具有预测精度高、搜索速度快的优点.To address the internal quality problem of selective laser sintering(SLS),this study proposes a Bayesian optimization backpropagation(BO-BP)neural network path optimization model for improving the SLS molding quality and reducing the energy consumption in the printing process.First,a path-planning process is established for printed copies,and dynamic programming technology is used to obtain the final path in a given laser grid and remove redundant paths.Subsequently,an SLS model is established for the laser sintering process based on the thermal conductivity differential equation,and the path for this process is numerically simulated.The final temperature values for each grid node are then used to calculate the temperature gradient values of internal nodes in the material grid and rank them.This value is used as the dataset for training and testing the neural network algorithm.Furthermore,the corresponding average thermal gradient distribution map is obtained through the numerical image transformation of the laser path,and the laser path is subsequently converted into a grayscale path image,which is trained using the improved BO-BP neural network algorithm.Finally,the Softmax function is used to convert the neural network output into the probability of each image.Linear optimization,unsupervised learning neural network,genetic BP neural network,and BP neural network are selected to compare the prediction results.The results show that the proposed BO-BP neural network algorithm has the advantages of high prediction accuracy and high search speed compared with other algorithms.

关 键 词：增材制造 选择性激光烧结 刀轨路径优化 BO-BP神经网络算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TG665[自动化与计算机技术—控制科学与工程]