关键词: 神经网络, 造型意象, 预测模型, 电动汽车, 产品设计

Abstract: To make the styling design of electric vehicles more in line with the cognition of target users for improving the ability of predicting styling images in electric vehicle design selection and reduce the risk of electric vehicle development,a neural network-based model for predicting styling images of electric vehicles was proposed.This predictive model used statistical methods  with neural networks to study the implicit correlation between the styling characteristics of electric vehicles and the perceptual images.Using multi-dimensional scale analysis and cluster analysis to obtain representative samples,and the modeling feature weight coefficients were calculated using analytic hierarchy process by deconstructing modeling features.The representative image vocabulary was determined by semantic difference method and principal component analysis to reduce the target user's perception of modeling characteristics and cognitive dimension of perceptual image.The Back Propagation (BP) neural network was used to construct the modeling image prediction model,the sample modeling feature weight coding was used as the input layer,and the user perceptual image evaluation score was used as the output layer for model training.The prediction model was tested with leave-one-out cross-training method.An electric Sport Utility Vehicle (SUV) front face design scheme was taken as an example to verify the feasibility of the prediction model.The research results showed that the predictive model could effectively solve the matching problem between modeling features and perceptual image in modeling image research,assist designers to quickly identify key modeling features and perceptual image targets,and improve the scientificity of design plan decision-making.

Key words: neural network, modeling image, prediction model, electric cars, product design