基于三角模糊和BP神经网络的产品意象造型设计

Product Image Modeling Design Based on Triangular Fuzzy and BP Neural Network

中文关键词: 产品设计  三角模糊  腰椎牵引器  BP神经网络

英文关键词: product design  triangular fuzzy  lumbar traction device  BP neural network

中文摘要:

      目的 针对目标产品造型与用户模糊感知意象适应性欠佳问题，探索性提出一种基于三角模糊和BP神经网络的意象造型设计方法。方法 在分析产品意象造型设计流程的基础上，以三角模糊方法作为BP神经网络模型意象输入数据的预处理工具，将用户模糊感知意象量化转换;并采用因子分析法降维获取优势感知意象;借助KJ和专家评定法获取差异性较大样本，根据形态分析和建模特点划分产品部件造型特征，通过Delphi法多轮选出优势样本;基于Matlab平台，采用权值惯性可调节、学习率可变的traingdx网络训练算法,对产品优势样本进行学习、训练和预测，构建部件造型要素与用户感知间的关系模型，实现造型设计元素的最佳定量化组合，使BP神经网络对用户模糊意象的定性更为有效，将该方法应用于腰椎牵引器造型设计中。结论 实验结果表明，该方法能有效实现模糊意象的定量分析，所建立的腰椎牵引器神经网络模型输出符合设计要求，促进设计概念转化。

英文摘要:

      Aiming at the problem of poor adaptability between target product modeling and user fuzzy perception image, an image modeling design method based on triangular fuzzy and BP neural network is proposed.On the basis of analyzing the process of product image modeling design, triangular fuzzy method is used as the preprocessing tool of image input data of BP neural network model to transform the user’s fuzzy perception image quantitatively; Factor analysis was used to reduce dimensions to obtain superior perception image; With the help of KJ and expert evaluation method, large difference samples are obtained. According to the shape analysis and modeling characteristics, the product parts modeling characteristics are divided, and the dominant samples are selected by Delphi method; Based on MATLAB platform, traingdx network training algorithm with adjustable weight inertia and variable learning rate is adopted, By learning, training and forecasting the product advantage samples, the relationship model between the component modeling elements and user perception is constructed, and the optimal quantitative combination of modeling design elements is realized, which makes the BP neural network more effective in the qualitative analysis of user fuzzy image. This method is applied to the modeling design of lumbar traction device. Experiments show that this method can effectively realize the quantitative analysis of fuzzy images. The neural network model output of the established lumbar traction device meets the design requirements and promotes the transformation of design concepts.