Application of Machine Learning Technique for Development of Indirect Tire Pressure Monitoring System

Tire pressure has a significant impact over driving dynamics, fuel consumption as well as tire life therefore, continuous monitoring of tire pressure becomes imperative for ride comfort, safety and optimum vehicle handling. The two types of systems used for tire pressure monitoring are Direct and Indirect. Direct system deploys pressure sensors at each tire to directly measure pressure value, while indirect systems make use of signals from already existing sensors to predict pressure. Direct tire pressure measurement system will have higher accuracy, but it comes with an additional cost. The indirect system on the other hand demands mandatory ESP system and those based on wheel speed signals are less accurate compared to the direct system however, there is minimum cost addition. The paper presents a digital proof of concept study for indirect tire pressure monitoring system development for Non-ESP vehicles using machine learning technique. For this purpose, a full vehicle multi-physics model is developed and validated using existing test data. The validated model is then used for extracting training data by simulating it in different vehicle conditions and maneuvers. Multilayered Feed Forward Neural Networks are used to train the wheel speed based and steering torque-based tire pressure prediction models. A selection algorithm is developed to choose the best prediction model based on the driving scenario and vehicle conditions. Sensitivity of different neural network parameters and optimization algorithm in the accuracy of prediction is evaluated and the optimal parameters were selected. The proof of concept results suggest that the propose tire pressure prediction algorithm has potential predict real-time tire pressure in same order as predicted by direct system. The developed prediction algorithm needs to be further tuned for physical data which is expected to contain multiple un-controlled parameters such as road surface variation, tire temperature, tire loading due to gradient etc.