基于GA-SVR混合方法的非标设备设计时间预测  

Design time prediction of non-standard equipment based on GA-SVR method

作　　者：施云祺 王子阳 杨宏兵[1] SHI Yunqi;WANG Ziyang;YANG Hongbing(School of Mechanical and Electric Engineering,Soochow University,Suzhou 215137,China)

机构地区：[1]苏州大学机电工程学院,苏州215137

出　　处：《现代制造工程》2023年第3期1-7,共7页Modern Manufacturing Engineering

基　　金：国家自然科学基金项目(52075354);苏州市产业技术创新专项(民生科技)项目(SS201704)。

摘　　要：非标自动化设备项目初期的设计环节在整个项目管理过程中占据重要地位,对其设计时间的精准预测是有效实施项目调度的基础。基于非标设备设计和制造特点,选择客户协同程度、造型难度、设计相似程度、订单规模和团队参与程度5个因素,探究其对设计时间的影响。考虑非标产品样本数量较少的特点,提出一种遗传算法与向量机回归(GA-SVR)混合方法,即在向量机回归模型(SVR)预测方法的基础上,结合十倍交叉验证与遗传算法对预测模型进行参数寻优。最后,选择高斯过程回归(GPR)方法进行对比,从预测精度和稳定性两个方面验证了GA-SVR混合方法的有效性。The design session at the early stage of a non-standard automation equipment project occupies an important position in its entire project management process,and the accurate prediction of its design time is the basis for effective implementation of project scheduling.Based on the design and manufacturing characteristics of non-standard equipment,customer collaboration,modeling difficulty,design similarity,order size and team participation were selected to explore the impact on design time.Considering the characteristics of the small number of samples of non-standard products,a hybrid method of Genetic Algorithm and Support Vector Regression(GA-SVR)was proposed,that is,based on the Support Vector Regression(SVR)model prediction method,10-fold cross-validation and genetic algorithm were combined to optimize the parameters of the prediction model.Finally,the Gaussian Process Regression(GPR)method was selected for comparison,and the effectiveness of the GA-SVR method was verified from the two aspects of prediction accuracy and stability.

关 键 词：非标自动化设备 设计时间预测 项目调度 参数优化 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]