Abstract: To address the problems of traditional vehicle fault diagnosis, such as reliance on manual experience, low recognition rate of hidden faults, and diagnostic lag, this paper constructs a new approach to fault diagnosis by integrating intelligent connected technology. An overall architecture of "multi-source heterogeneous data fusion+artificial intelligence (AI) dynamic diagnosis+vehicleroad-cloud collaborative tracing" is designed. Comparative verification through an experimental platform shows that this approach achieves a diagnostic accuracy of 96.7%, reduces the average diagnosis time by 90.4% compared to traditional methods, and enables early warning 22.3 h in advance, effectively solving the pain points of traditional diagnosis. On this basis, an adaptive teaching system is designed from three aspects: reconstruction of teaching content, integration of virtual and real modes, and construction of resource platforms. In pilot applications, the pass rate of the fault diagnosis experimental assessment increases from 78% to 92%, the average project completion quality score reaches 85, and the monthly downloads of teaching resources exceeded 5 000, significantly improving students' practical abilities. This research provides a practical reference for the upgrading of vehicle diagnostic technology and the cultivation of professional talents.

Key words: intelligent connected vehicles; vehicle fault diagnosis; multi-source heterogeneous data fusion; AI prediction model

摘要： 针对传统汽车故障诊断依赖人工经验、隐性故障识别率低、诊断滞后等问题，文章结 合智能网联技术构建故障诊断新思路，设计“多源异构数据融合+人工智能（AI）动态诊断+ 车-路-云协同溯源”整体架构。搭建实验平台对比验证显示，该思路诊断准确率达 96.7%、平 均耗时较传统方法缩短 90.4%、可提前 22.3 h 预警，有效解决传统诊断痛点，在此基础上从 教学内容重构、虚实结合模式、资源平台搭建三方面设计适配教学体系，试点应用中班级故 障诊断实验考核通过率从 78%提升至 92%，项目完成质量平均分达 85 分，教学资源月均下载 超 5 000 次，显著提升学生实践能力，该研究为汽车诊断技术升级与专业人才培养提供实践 参考。

关键词: 智能网联汽车；汽车故障诊断；多源异构数据融合；AI 预测模型