Abstract: To eliminate the influence of decision-maker's decision preferences on product development and further improve the design decision-making efficiency,a Residual Neural Network(ResNet)artificial intelligence design decision-making model was proposed.Based on artificial intelligence thinking,a design history scheme dataset based on product modeling semantics was built.After the semantic annotation of this design dataset,ResNet algorithm was applied to train the data set continuously to improve the accuracy of design decisions,which could transfer the general design decision problem into the image semantic recognition problem of the design schemes,so as to maximize the influence of decision-maker's decision preferences.The validity of ResNet artificial intelligence design decision model was verified by the example of crane design decision making experiment,which proved the feasibility and rationality of the model.

Key words: deep learning, artificial intelligence, design decision-making, design semantic, ResNet algorithm, product development

摘要： 为了消除设计决策者决策偏好对产品开发的影响,进一步提升设计决策效率,提出了ResNet人工智能设计决策模型。该模型基于人工智能思想,构建了基于产品造型语义的设计历史方案数据集,并对该数据集进行了产品造型语义标注。通过深度残差学习网络算法(ResNet)对数据集进行不断训练来提高设计决策的准确度,将一般设计决策问题转化为设计方案图像的语义识别问题,最大限度地消除了决策者决策偏好的影响。通过起重机造型设计决策实例,验证了ResNet人工智能设计决策算法的有效性和可行性。

关键词: 深度学习, 人工智能, 设计决策, 设计语义, ResNet算法, 产品开发