面向汽车轮毂设计的形态意象耦合智能识别方法研究

张宇彤; 孙利; 吴俭涛; 张硕; 孙禧; 李昌润; 赵嘉凝; 贾晓露; 韩旭

doi:10.19554/j.cnki.1001-3563.2026.12.007

PDF(4137 KB)

包装工程（设计栏目） ›› 2026, Vol. 47 ›› Issue (12) : 83-99. DOI: 10.19554/j.cnki.1001-3563.2026.12.007

工业设计

面向汽车轮毂设计的形态意象耦合智能识别方法研究

张宇彤¹, 孙利¹, 吴俭涛¹, 张硕¹, 孙禧¹, 李昌润¹, 赵嘉凝¹, 贾晓露^1,*, 韩旭²

作者信息 +

Intelligent Recognition Method for Form-Imagery Coupling in Automotive Wheel Hub Design

ZHANG Yutong¹, SUN Li¹, WU Jiantao¹, ZHANG Shuo¹, SUN Xi¹, LI Changrun¹, ZHAO Jianing¹, JIA Xiaolu^1,*, HAN Xu²

Author information +

文章历史 +

摘要

目的针对汽车轮毂设计中“形-意”脱耦所致的设计意图与用户感知间语义失配问题,提出一种融合跨模态对比学习与多专家查询网络的智能识别方法。方法首先,利用词对主题模型挖掘用户评论潜在语义,并结合层次分析精准筛选造型意象特征,构建汽车轮毂形态意象图文多模态数据集;其次,由12层Transformer与ResNet-50组成跨模态双编码器,分别对文本语义与视觉特征进行深层解析;再次,通过对比学习（Contrastive Learning, CL）优化特征间的余弦相似度,形成统一且细粒度的跨模态表征空间,以构建TRCL（Transformer and ResNet with Contrastive Learning）模型;最后,在TRCL基础上嵌入多专家查询网络（Multi-Expert Query Network, MEQN）,令各专家子网络分别聚焦于视觉模态的特定语义维度,并运用跨模态注意力机制动态关联视觉与文本信息,以实现精细化的跨模态特征交互与融合。结果经多项实验对比,MEQN-TRCL均优于各主流深度网络,同时消融实验亦证实其具备结构合理性与性能先进性。结论文中方法可有效实现汽车轮毂形态特征与感性意象的精准耦合,为汽车轮毂情感设计量化决策与形态意象跨模态匹配提供新的方法路径。

Abstract

To address the semantic mismatch between designers' intentions and users' perceptions due to "form-imagery" decoupling in automotive wheel hub design, the work aims to propose an intelligent recognition method integrating cross-modal contrastive learning (CL) and a multi-expert query network (MEQN). Firstly, a biterm topic model was employed to mine latent semantic structures from user reviews, and representative design imagery features were rigorously identified and quantified through an analytic hierarchy process, thus constructing a multimodal image-text dataset for automotive wheel hub form-imagery analysis. Secondly, a cross-modal dual encoder comprising a 12-layer Transformer and ResNet-50 was developed to deeply encode textual semantics and visual features, respectively. Thirdly, the cosine similarity between modal features was optimized through Contrastive Learning (CL) to form a unified, fine-grained cross-modal representation space, thus constructing a Transformer and ResNet-based contrastive learning model (TRCL). Finally, MEQN was embedded into the TRCL framework, where each expert sub-network independently focused on specific semantic dimensions of visual modality features, and cross-modal attention mechanisms dynamically associated visual and textual information, achieving fine-grained cross-modal feature interaction and fusion. Experimental results demonstrated that the proposed MEQN-TRCL method significantly outperformed mainstream models. Meanwhile, ablation experiments further verified its rational structural design and advanced performance. The method proposed effectively facilitates precise coupling between wheel hub form features and Kansei imagery, thereby offering a novel methodological pathway for quantitative emotional design decision-making and cross-modal form-imagery matching in automotive wheel hub design.

导出引用1

张宇彤, 孙利, 吴俭涛, 张硕, 孙禧, 李昌润, 赵嘉凝, 贾晓露, 韩旭. 面向汽车轮毂设计的形态意象耦合智能识别方法研究[J]. 包装工程. 2026, 47(12): 83-99 https://doi.org/10.19554/j.cnki.1001-3563.2026.12.007

ZHANG Yutong, SUN Li, WU Jiantao, ZHANG Shuo, SUN Xi, LI Changrun, ZHAO Jianing, JIA Xiaolu, HAN Xu. Intelligent Recognition Method for Form-Imagery Coupling in Automotive Wheel Hub Design[J]. Packaging Engineering. 2026, 47(12): 83-99 https://doi.org/10.19554/j.cnki.1001-3563.2026.12.007

中图分类号： TB472 U463

参考文献

[1] 于利洋. 基于偏序结构的轮毂意象智能转译设计方法研究[D]. 秦皇岛: 燕山大学, 2024.
YU L Y.Research on the Design of Intelligent Translation Method of Hub Image Based on Partial Order Structure[D]. Qinhuangdao: Yanshan University, 2024.
[2] 孙利, 陈永亮, 艾雯, 等. 基于V形记忆曲线的汽车轮毂造型衍生设计方法[J]. 机械设计, 2024, 41(3): 163-169.
SUN L, CHEN Y L, AI W, et al.Derivative Design Method for Automotive Wheel Hub Modeling Based on V-Shaped Memory Curve[J]. Journal of Machine Design, 2024, 41(3): 163-169.
[3] SUN L, CHEN Y L, QIN Z Z, et al.User Preference- Based Method for Characterizing Automotive Wheel Hub Styles[J]. Applied Sciences, 2025, 15(6): 3322.
[4] 吴俭涛, 袁放, 孙利. 基于象元运算的轮毂形态设计方法研究[J]. 包装工程, 2018, 39(24): 163-171.
WU J T, YUAN F, SUN L.Method of Wheel Form Design Based on Operation of Meta Symbol[J]. Packaging Engineering, 2018, 39(24): 163-171.
[5] 王剑, 吴俭涛, 鲍官培. 生物美感效应驱动的汽车轮毂造型耦合设计[J]. 机械设计, 2023, 40(8): 164-170.
WANG J, WU J T, BAO G P.Coupling Design of Wheel Hub Modeling Driven by Biological Aesthetic Effect[J]. Journal of Machine Design, 2023, 40(8): 164-170.
[6] LIU J, ZHI Q Q, JI H P, et al.Wheel Hub Customization with an Interactive Artificial Immune Algorithm[J]. Journal of Intelligent Manufacturing, 2021, 32(5): 1305-1322.
[7] OH S, JUNG Y, KIM S, et al.Deep Generative Design: Integration of Topology Optimization and Generative Models[J]. Journal of Mechanical Design, 2019, 141(11): 111405.
[8] 孙利, 张宇彤, 吴俭涛, 等. 新能源汽车轮毂造型感性意象预测研究[J]. 机械设计, 2025, 42(9): 203-211.
SUN L, ZHANG Y T, WU J T, et al.Predictive Research on Kansei Image for Modeling of New Energy Vehicle Hub[J]. Journal of Machine Design, 2025, 42(9): 203-211.
[9] 魏君, 韩颖, 苏畅, 等. 基于方差分析的轮毂造型特征显著性研究[J]. 包装工程, 2022, 43(10): 115-120.
WEI J, HAN Y, SU C, et al.Statistical Significance of the Wheel Hubs Styling Features Based on Variance Analysis[J]. Packaging Engineering, 2022, 43(10): 115-120.
[10] 成振波, 任薪蓉, 柯善军, 等. 基于感性意象的轿车轮毂参数化设计[J]. 机械设计, 2022, 39(4): 135-141.
CHENG Z B, REN X R, KE S J, et al.Parametric Design of Car Wheel Hub Based on Kansei Image[J]. Journal of Machine Design, 2022, 39(4): 135-141.
[11] HU T, XIE Q S, YUAN Q N, et al.Design of Ethnic Patterns Based on Shape Grammar and Artificial Neural Network[J]. Alexandria Engineering Journal, 2021, 60(1): 1601-1625.
[12] 杨剑威, 王毅. 基于形状文法的青铜酒器形态推演与设计应用[J]. 包装工程, 2020, 41(8): 317-322, 326.
YANG J W, WANG Y.Shape Deduction and Design Application of Bronze Wine Utensils Based on Shape Grammar[J]. Packaging Engineering, 2020, 41(8): 317-322, 326.
[13] 景银, 程永胜, 朱琳, 等. 基于文化意象的运动鞋款设计方法研究[J]. 皮革科学与工程, 2024, 34(5): 74-80.
JING Y, CHENG Y S, ZHU L, et al.Research on the Design Method of Sports Shoe Models Based on Cultural Imagery[J]. Leather Science and Engineering, 2024, 34(5): 74-80.
[14] HUANG Y X, CHEN C H, KHOO L P.Products Classification in Emotional Design Using a Basic-Emotion Based Semantic Differential Method[J]. International Journal of Industrial Ergonomics, 2012, 42(6): 569-580.
[15] LLINARES C, PAGE A F.Differential Semantics as a Kansei Engineering Tool for Analysing the Emotional Impressions which Determine the Choice of Neighbourhood: The Case of Valencia, Spain[J]. Landscape and Urban Planning, 2008, 87(4): 247-257.
[16] 丁治中, 荣一潼, 许卫丽. 可拓学下儿童陪伴机器人造型意象设计研究[J]. 机械设计与制造, 2025, (11): 315-321.
DING Z Z, RONG Y T, XU W L.Research on Modeling Image Design of Children's Companion Robot under Extenics[J]. Machinery Design & Manufacture, 2025, (11): 315-321.
[17] 肖德荣, 姚令华, 刘祺. 湘西传统竹家具造型知识感性意向分析[J]. 林产工业, 2025, 62(9): 26-34.
XIAO D R, YAO L H, LIU Q.Analysis of Affective Imagery in the Formative Knowledge of Traditional Bamboo Furniture in Xiangxi, Hunan[J]. China Forest Products Industry, 2025, 62(9): 26-34.
[18] 杨冬梅, 李云红, 张健楠, 等. 多生理数据驱动的产品外观设计智能评价方法[J]. 计算机集成制造系统, 2026, 32(1): 55-67.
YANG D M, LI Y H, ZHANG J N, et al.Multimodal Data-driven Intelligent Evaluation Method for Space Cabin Appearance Design[J]. Computer Integrated Manufacturing Systems, 2026, 32(1): 55-67.
[19] ZHANG Y T, WU J T, SUN L, et al.A Method for the Front-End Design of Electric SUVs Integrating Kansei Engineering and the Seagull Optimization Algorithm[J]. Electronics, 2025, 14(8): 1641.
[20] 丁宁, 余隋怀, 初建杰, 等. 面向产品设计的民族图案语义量化模型构建与应用[J]. 计算机辅助设计与图形学学报, 2023, 35(4): 621-632.
DING N, YU S H, CHU J J, et al.Construction and Application of National Pattern Semantic Quantification Model for Product Design[J]. Journal of Computer-Aided Design & Computer Graphics, 2023, 35(4): 621-632.
[21] KRIZHEVSKY A, SUTSKEVER I, HINTON G E.ImageNet Classification with Deep Convolutional Neural Networks[J]. Communications of the ACM, 2017, 60(6): 84-90.
[22] TAMMINA S.Transfer Learning Using VGG-16 with Deep Convolutional Neural Network for Classifying Images[J]. International Journal of Scientific and Research Publications (IJSRP), 2019, 9(10): p9420.
[23] SZEGEDY C, LIU W, JIA Y Q, et al.Going Deeper with Convolutions[C]// 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston: IEEE, 2015: 1-9.
[24] HE K M, ZHANG X Y, REN S Q, et al.Deep Residual Learning for Image Recognition[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas: IEEE, 2016: 770-778.
[25] 朱斌, 杨程, 俞春阳, 等. 基于深度学习的产品意象识别[J]. 计算机辅助设计与图形学学报, 2018, 30(9): 1778-1784.
ZHU B, YANG C, YU C Y, et al.Product Image Recognition Based on Deep Learning[J]. Journal of Computer-Aided Design & Computer Graphics, 2018, 30(9): 1778-1784.
[26] 李雄, 苏建宁, 张志鹏, 等. 基于深度学习的产品风格精细识别[J]. 计算机集成制造系统, 2024, 30(3): 1011-1022.
LI X, SU J N, ZHANG Z P, et al.Recognition Method for Fine-Grained Product Styles Based on Deep Learning[J]. Computer Integrated Manufacturing Systems, 2024, 30(3): 1011-1022.
[27] ZHOU A M, LIU H B, ZHANG S T, et al.Evaluation and Design Method for Product Form Aesthetics Based on Deep Learning[J]. IEEE Access, 2021, 9: 108992-109003.
[28] YUAN X, QI A G, WU H N, et al.Cross-Modal Feature Alignment and Fusion with Contrastive Learning in Multimodal Recommendation[J]. Knowledge-Based Systems, 2025, 326: 114020.
[29] 刘萌, 齐孟津, 詹圳宇, 等. 基于深度学习的图像-文本匹配研究综述[J]. 计算机学报, 2023, 46(11): 2370-2399.
LIU M, QI M J, ZHAN Z Y, et al.A Survey on Deep Learning Based Image-Text Matching[J]. Chinese Journal of Computers, 2023, 46(11): 2370-2399.
[30] KWON Y, LIANG V W, YEUNG S, et al.Mind the Gap: Understanding the Modality Gap in Multi-Modal Contrastive Representation Learning[C]// Advances in Neural Information Processing Systems 35. New Orleans: Neural Information Processing Systems Foundation, Inc.(NeurIPS), 2022: 17612-17625.
[31] YANG J Y, DUAN J L, TRAN S, et al.Vision-Language Pre-Training with Triple Contrastive Learning[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New Orleans: IEEE, 2022: 15650-15659.
[32] RADFORD A, KIM J W, HALLACY C, et al.Learning Transferable Visual Models from Natural Language Supervision[C]// International Conference on Machine Learning. Brookline: PmLR, 2021: 8748-8763.
[33] JIA C, YANG Y F, XIA Y, et al.Scaling up Visual and Vision-Language Representation Learning with Noisy Text Supervision[C]// International Conference on Machine Learning. New Orleans: PMLR, 2021.
[34] YAO L W, HUANG R H, HOU L, et al. FILIP: Fine- Grained Interactive Language-Image Pre-Training[EB/OL].2021: arXiv: 2111.07783. https://arxiv.org/abs/2111.07783
[35] HE X, LI J C, QIAN L, et al.Fine-Grained Semantically Aligned Vision-Language Pre-Training[C]// Advances in Neural Information Processing Systems 35. New Orleans: Neural Information Processing Systems Foundation, Inc.(NeurIPS), 2022: 7290-7303.
[36] SONG B Y, ZHOU R, AHMED F.Multi-Modal Machine Learning in Engineering Design: A Review and Future Directions[J]. Journal of Computing and Information Science in Engineering, 2024, 24: 010801.
[37] YANG S L, CUI L C, WANG L, et al.Cross-Modal Contrastive Learning for Multimodal Sentiment Recognition[J]. Applied Intelligence, 2024, 54(5): 4260-4276.
[38] 余本功, 石中玉. 深层注意力和两阶段融合的图文情感对比学习方法[J]. 计算机工程与应用, 2025, 61(3): 223-233.
YU B G, SHI Z Y.Deep Attention and Two-Stage Fusion of Image-Text Sentiment Contrastive Learning Method[J]. Computer Engineering and Applications, 2025, 61(3): 223-233.
[39] ZHU H G, WEI Y C, ZHAO Y, et al.AMC: Adaptive Multi-Expert Collaborative Network for Text-Guided Image Retrieval[J]. ACM Transactions on Multimedia Computing, Communications, and Applications, 2023, 19(6): 1-22.
[40] CHENG X Q, YAN X H, LAN Y Y, et al.BTM: Topic Modeling over Short Texts[J]. IEEE Transactions on Knowledge and Data Engineering, 2014, 26(12): 2928-2941.