A Low-Cost Tire Pressure Loss Detection Framework Using Machine Learning

Abstract

The tire pressure state directly affects the safety and economy of the vehicle. Its monitoring is gradually becoming a necessary function for all types of vehicles. However, the existing sensor-based monitoring method is costly and without any redundancy when the sensor fails, which hinders its wide applications. This article proposed a machine learning-based framework without any additional sensors. First, the rigid tire model is introduced for feature extraction. Then, the manufacturing error in speed gear is calculated and removed by the recursive least square method. Next, the features in time- and frequency-domain of speed signals are extracted. Finally, based on the suitable signals filtered by the decision tree, the normal and pressure-loss judgment are given out by support vector machine and final synthesis. The results show that the proposed framework has 96.18% report accuracy and has the potential for further estimation of the characteristics in tires and roads.