Health Monitoring Method of Non-pneumatic Tire Based on Fusion Dimensionality Reduction and Deep Learning

To improve or overcome the disadvantages of traditional pneumatic tires, such as blowouts and punctures, the development of non-pneumatic tire technology has become a consensus among major tire manufacturers and research institutions. Further achieving health monitoring for non-pneumatic tires is crucial for ensuring vehicle safety and tire performance. Addressing the issues of low robustness, poor generalizability, and incomplete information extraction in single damage features, this paper proposes a health monitoring method for smart non-pneumatic tires based on fusion dimensionality reduction and deep learning. Firstly, comprehensive signal features are extracted from multiple dimensions to construct a high-dimensional damage feature dataset. Then, a global-local fusion feature dimensionality reduction algorithm is employed to eliminate redundant information in high-dimensional data and reasonably control data dimensions. Finally, Bayesian optimization deep neural networks are used to qualitatively and quantitatively analyze the health status of smart non-pneumatic tires, along with comparisons to other traditional dimensionality reduction and classification techniques. The research results indicate that this method accurately identifies the health status of non-pneumatic tires, with classification accuracy improvements of 4.1% and 13.8% compared to KPCA and LLE algorithms, respectively, and 6.9% and 9.9% improvements compared to SVM and KNN algorithms. This approach enhances the health prediction capability for non-pneumatic tires, reduces maintenance costs for intelligent vehicle tires, and ensures driving safety, providing new insights for the health monitoring of smart non-pneumatic tires.