基于SOA-LSSVM的SLS成形工艺参数优化研究  

Research on Optimization of SLS Forming Processing Parameters Based on SOA-LSSVM

作　　者：肖亚宁 孙雪[1] 张亚鹏 裴玲艺 李三平[1] XIAO Yaning;SUN Xue;ZHANG Yapeng;PEI Lingyi;LI Sanping(College of Mechanical and Electrical Engineering,Northeast Forestry University,Harbin Heilongjiang 150040,China)

机构地区：[1]东北林业大学机电工程学院,黑龙江哈尔滨150040

出　　处：《机床与液压》2022年第6期36-42,共7页Machine Tool & Hydraulics

基　　金：中央高校基本科研业务费专项资金项目(2572014BB06)。

摘　　要：为提高选择性激光烧结(SLS)成形精度,解决工艺参数优化试验成本高等问题,选择激光功率、预热温度、扫描速度、扫描间距以及分层厚度5个工艺参数设计正交试验以获得样本数据并建立统一目标函数。采用人群搜索算法(SOA)优化最小二乘支持向量机(LSSVM),建立基于SOA-LSSVM的SLS成形件精度预测模型;预测不同工艺参数组合下制件的统一性能,并与采用传统BP神经网络和LSSVM模型获得的预测结果进行对比。结果表明:SOA-LSSVM模型针对小样本预测问题具有良好的泛化能力,预测值与实际值的最大相对误差仅为1.11%,可为SLS加工参数组合的选择提供参考。In order to improve the precision of selective laser sintering(SLS)and to solve the high cost of technical parameters optimization,five technical parameters of laser power,preheating temperature,scanning speed,scanning distance and layer thick⁃ness were selected to design the orthogonal experiment to obtain sample data and a unified target function was established.By using the seeker optimization algorithm(SOA),the least squares support vector machine(LSSVM)was optimized and the accurate predictive model of SLS based on SOA-LSSVM was established;the unified performance of the parts under different technical parameters combi⁃nations was predicted,and compared with the results obtained by using traditional BP neural network and LSSVM model.The results show that the SOA-LSSVM model has good generalization ability for small sample prediction and the maximum relative error between the predicted value and the actual value is only 1.11%,which can provide reference for the selection of SLS processing parameter combinations.

关 键 词：选择性激光烧结 工艺参数 正交试验 人群搜索算法 最小二乘支持向量机 

分 类 号：TH140.7[一般工业技术—材料科学与工程] TP273[机械工程—机械制造及自动化]