FusionNet: Hybrid Learning for Tire Lifespan Prediction via Structured Data and Image Analysis

Tire life forecasting is an aspect of vehicle maintenance with safety implications involving repercussions for efficiency of operation and cost savings. Conventionally formulated models are all based on organized sensor readings like Tire Kilometer Per Hour (TKPH), pressure, and road and are not generally derived from patterns of visual wear. In this paper, we introduce a multimodal hybrid learning model, where deep learning-based image processing and structured sensor data are integrated within an ensemble framework. ResNet50 and EfficientNet identify wear features in tire images, while XGBoost and LightGBM process sensor-based parameters. Both modalities are fused together using a fusion layer to improve the accuracy of prediction. Experimental verification illustrates that the proposed system is more accurate compared to conventional single-modality models. Comparative analysis compared to other methods shows a dramatic reduction in prediction errors, enhancing accuracy for estimating remaining tire life. The hybrid method also offers more robust and generalizable predictions under various driving scenarios. The study has far-reaching implications across the automotive sector, fleet management, and predictive maintenance systems, encouraging real-time tire monitoring, lower replacement costs, and improved road safety. Future research will involve integration with the Internet of Things, AI-based real-time monitoring, and eco-friendly tire materials for further performance and durability.