基于神经网络的仿生避难椅创新设计  

Innovative Design of Bionic Refuge Chair Based on Neural Network

作　　者：杨昕妍 杨洋 张继娟[1,2,3] 张仲凤 YANG Xin-yan;YANG Yang;ZHANG Ji-juan;ZHANG Zhong-feng(College of Furniture and Art Design,Central South University of Forestry Science and Technology,Changsha 410004,Hunan,P.R.China;Green Furniture Engineering Technology Research Centre,State Forestry and Grassland Administration,Changsha 410004,Hunan,P.R.China;Hunan Green Home Engineering Technology Research Centre,Changsha 410004,Hunan,P.R.China)

机构地区：[1]中南林业科技大学家具与艺术设计学院,湖南长沙410004 [2]国家林业和草原局绿色家具工程技术研究中心,湖南长沙410004 [3]湖南省绿色家居工程技术研究中心,湖南长沙410004

出　　处：《林产工业》2024年第1期53-60,共8页China Forest Products Industry

基　　金：湖南省重点研发(2022NK2043);速生小径材高值化利用及绿色家居智能定制技术研究(2021HBQZYCXY011);湖南省科技创新领军人才(2021RC4033)。

摘　　要：近年来,地震发生的强度和频率不断增加,基于仿生设计和神经网络在地震情境下设计了一款避难椅,通过拟合优度检验和系统聚类方法分析调研结果得到6大类18小类功能需求,并依据共性需求确定仿生对象。利用Gabor滤波提取其纹理特征并抽象为本体象元空间,利用神经网络和象元理论建立与用户偏好之间的映射关系模型,得到MSE值为0.09212和0.00744,说明模型性能良好。利用形状文法规则进行设计创新,得到三款方案,通过验证对比映射关系模型中与之相似且排名前三的本体象元满意度,均方误差MSE值为0.00324,证明该方法能够为相关产品设计提供参考。In recent years,the intensity and frequency of earthquakes have been increasing,in this paper,an evacuation chair based on bionic design and neural network in an earthquake situation was designed,the research results through goodness-of-fit test and system clustering method to obtain 6 major categories and 18 subcategories of functional requirements were analyzed,and the bionic object based on the common requirements was determined.Its texture features were extracted using Gabor filtering and abstracted into the ontological pixel space,and the mapping relationship model with user preferences was established using neural network and pixel theory.MSE values of 0.09212 and 0.00744 were obtained,indicating good model performance.Using shape grammar rules for design innovation,three solutions were obtained,and the mean square error MSE value was 0.00324 by verifying and comparing the satisfaction of the similar and top three ontological pixels in the mapping relationship model,which proved that the method can provide a reference for related product design.

关 键 词：工业设计 神经网络 象元理论 形状文法 避难椅 

分 类 号：TS664[轻工技术与工程] TS396