AI-Driven Predictive Maintenance with Real-Time Contextual Data Fusion for Connected Vehicles: A Multi-Dataset Evaluation
arXiv:2603.13343v1 Announce Type: new Abstract: Most vehicle predictive maintenance systems rely exclusively on internal diagnostic signals and are validated on deterministic synthetic data, limiting the credibility of reported metrics. This paper presents a simulation-validated proof-of-concept framework for V2X-augmented predictive maintenance, integrating on-board sensor streams with external contextual signals -- road quality, weather, traffic density, and driver behaviour -- acquired via V2X communication and third-party APIs, with inference at the vehicle edge. Field validation on instrumented vehicles is identified as the required next step. Three experiments address common shortcomings of prior work. A feature group ablation study shows that V2X contextual features contribute a 2.6-point F1 gain, with full context removal reducing macro F1 from 0.855 to 0.807. On the AI4I 2020 real-world industrial failure dataset (10,000 samples, five failure modes), LightGBM achieves AUC-ROC
arXiv:2603.13343v1 Announce Type: new Abstract: Most vehicle predictive maintenance systems rely exclusively on internal diagnostic signals and are validated on deterministic synthetic data, limiting the credibility of reported metrics. This paper presents a simulation-validated proof-of-concept framework for V2X-augmented predictive maintenance, integrating on-board sensor streams with external contextual signals -- road quality, weather, traffic density, and driver behaviour -- acquired via V2X communication and third-party APIs, with inference at the vehicle edge. Field validation on instrumented vehicles is identified as the required next step. Three experiments address common shortcomings of prior work. A feature group ablation study shows that V2X contextual features contribute a 2.6-point F1 gain, with full context removal reducing macro F1 from 0.855 to 0.807. On the AI4I 2020 real-world industrial failure dataset (10,000 samples, five failure modes), LightGBM achieves AUC-ROC of 0.973 under 5-fold stratified CV with SMOTE confined to training folds. A noise sensitivity analysis shows macro F1 remains above 0.88 under low noise and degrades to 0.74 under very high noise. SHAP analysis confirms that V2X and engineered interaction features rank among the top 15 predictors. Edge inference is estimated to reduce latency from 3.5s to under 1.0s versus cloud-only processing.
Executive Summary
This study presents a simulation-validated proof-of-concept framework for AI-driven predictive maintenance in connected vehicles, leveraging real-time contextual data fusion via V2X communication and third-party APIs. The framework demonstrates improved performance over traditional internal diagnostic signals, particularly when incorporating V2X contextual features. Notably, the study achieves high AUC-ROC and macro F1 scores on a real-world industrial failure dataset, while also showcasing the efficacy of edge inference in reducing latency. However, further validation through field experiments remains necessary to ensure the framework's real-world applicability.
Key Points
- ▸ The study proposes an AI-driven predictive maintenance framework for connected vehicles, integrating on-board sensor streams with V2X contextual signals.
- ▸ Experiments demonstrate the importance of V2X contextual features in achieving higher F1 scores and AUC-ROC.
- ▸ Edge inference is shown to reduce latency from 3.5s to under 1.0s compared to cloud-only processing.
Merits
Improved Performance
The study showcases improved predictive maintenance performance when incorporating V2X contextual features, outperforming traditional internal diagnostic signals.
Real-World Validation
The study utilizes the AI4I 2020 real-world industrial failure dataset, providing a more credible evaluation of the framework's performance.
Edge Inference Benefits
The study highlights the benefits of edge inference in reducing latency, which is critical for real-time predictive maintenance applications.
Demerits
Limited Field Validation
The study emphasizes the need for field validation through experiments involving instrumented vehicles to ensure the framework's real-world applicability.
Noise Sensitivity Concerns
The study reveals that the framework's performance degrades under high noise levels, highlighting potential challenges in real-world deployment.
Dependence on Third-Party APIs
The study relies on third-party APIs for acquiring contextual signals, which may introduce dependencies and potential reliability issues.
Expert Commentary
This study makes a significant contribution to the field of predictive maintenance in connected vehicles by demonstrating the efficacy of AI-driven approaches incorporating real-time contextual data fusion. The findings are well-supported by experiments and simulations, and the study's emphasis on edge inference and V2X communication is particularly noteworthy. However, the study's limitations, including the need for field validation and concerns regarding noise sensitivity, highlight the need for further research in this area. As the field of connected vehicles continues to evolve, studies like this one will play a crucial role in shaping the development of reliable and efficient predictive maintenance systems.
Recommendations
- ✓ Future studies should prioritize field validation through experiments involving instrumented vehicles to ensure the framework's real-world applicability.
- ✓ Researchers should investigate methods to mitigate the effects of noise on the framework's performance, potentially through the development of noise-robust features or algorithms.
