数据驱动的出口管熔模铸件夹杂预测与工艺优化  

Inclusion Prediction and Process Optimization of Data-driven Outlet Pipe Investment Casting

作　　者：李天佑[1] 王玉 计效园[1] 余朋 常玎凯 殷亚军[1] 周建新[1] LI Tianyou;WANG Yu;JI Xiaoyuan;YU Peng;CHANG Dingkai;YIN Yajun;ZHOU Jianxin(State Key Laboratory of Materials Processing and Die&Mould Technology,Huazhong University of Science and Technology,Wuhan 430074;Xi'an Aerospace Engine Co.,Ltd.,Xi'an 710025)

机构地区：[1]华中科技大学材料成形与模具技术全国重点实验室,武汉430074 [2]西安航天发动机有限公司,西安710025

出　　处：《特种铸造及有色合金》2024年第11期1441-1446,共6页Special Casting & Nonferrous Alloys

基　　金：国家重点研发计划资助项目(2020YFB1710100);国家自然科学基金资助项目(52275337,52090042,51905188)。

摘　　要：提出了基于BP神经网络与改进粒子群算法的夹杂预测与工艺优化方法。首先,基于华铸ERP系统进行数据挖掘及清洗;其次,建立结合粒子群算法与BP神经网络的缺陷预测模型(Particle Swarm Optimization-Back Propagation,PSO-BP),相比普通BP神经网络,精度由92.1%提升至94.7%;最后,提出结合K近邻插补法与改进粒子群算法的工艺优化方法(K-Nearest Neighbors Imputation-Improved Particle Swarm Optimization,KNN-IPSO)。经模拟验证,相比生产前工艺在不同扰动下优化算法的缺陷率分别降低了52%和40%。A method of inclusion prediction and process optimization was proposed based on BP neural network and improved particle swarm optimization algorithm.Firstly,data mining and cleaning were conducted based on Huazhu ERP.Secondly,a defect prediction model(Particle Swarm Optimization-Back Propagation,PSO-BP)combining particle swarm optimization algorithm and BP neural network was established,of which the accuracy is improved from 92.1%to 94.7%compared with ordinary BP neural network.Finally,a process optimization method(K-Nearest Neighbors Imputation-Improved Particle Swarm Optimization,KNN-IPSO)was proposed.The simulation results indicate that the defect rate is reduced by 52%and 40%compared with that under different disturbances.

关 键 词：熔模铸件 BP神经网络 改进粒子群算法 K近邻插补法 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TG249.5[自动化与计算机技术—计算机科学与技术]