Tier Wear Estimation with Federated CNN-LSTM Learning Method with Privacy Enhancement

In order to guarantee vehicle safety and operational effectiveness, tire wear estimate is essential. A privacy-preserving method for real-time tire wear monitoring is proposed in this research by combining three-axis accelerometers and Hall effect sensors with a federated deep learning framework. Through the detection of both temporal and spatial patterns in sensor data, a hybrid CNN-LSTM model is employed to accurately classify tire conditions as "good" or "bad". To ensure data confidentiality, the approach combines federated learning, which enables local model training on edge devices, with safe aggregation and encryption techniques. Unlike conventional centralized systems, raw sensor data remains on the device and only encrypted model changes are transmitted. The experimental results highlight the system's viability for deployment in intelligent and connected transportation networks by demonstrating high classification accuracy and effective privacy preservation.