Critical State Estimation Of MEMS Accelerometer Based Intelligent Tire

The rapid development of intelligent vehicles leads to the technological innovation of smart chassis.The highly safety control system of smart chassis puts forward higher requirements on chassis state perception,and the accurate state perception of tires,as the only component interacting with the chassis and road,plays a crucial role in the safety control system of intelligent vehicles.At present,in the tire state sensing method,the tire as a passive component is not yet able to monitor its own state directly,but establishes indirect tire state sensing algorithm with the help of on-board sensors,vehicle theory model and tire theory model,and this method has cumulative error or time lag in practice.To address the above problems,this paper develops an intelligent tire system based on a three-axis accelerometer,develops tests of the intelligent tire system,establishes the estimation models of tire force,tire slip angle and tire slip ratio based on the acceleration data combined with tire physical model and machine learning algorithm,respectively,and adequately evaluates the established models.The specific research work of this paper is as follows.Firstly,an intelligent tire test system is developed and three types of tire working conditions,namely free rolling,cornering and driving-braking,are carried out and an identification method of tire contact patch is determined,focusing on the variation of acceleration with load,velocity,slip angle and slip ratio in the frequency domain and time domain.In the frequency domain,the acceleration characteristics of different working conditions are observed by fast Fourier transform（FFT）and power spectrum density（PSD）,and in the time domain,the variations of different working conditions are analyzed by observing the original acceleratiom,integrated velocity and integrated displacement.Secondly,a physical estimation model of tire vertical force is established based on the characteristics of vertical displacement and load by combining the tire theory model;a physical estimation model of tire lateral force is developed based on the characteristics of lateral displacement and lateral force by combining the tire lateral deformation theory.A machine learning estimation model of tire force is developed by combining Gaussian process regression algorithm with partial least squares regression-variable importance in projection（PLS-VIP）feature as input.The physical model and the machine learning model are scientifically evaluated,and the results show that the developed tire force estimation model combining intelligent tires and machine learning can effectively and accurately predict the tire force under different working conditions.Then,a physical estimation model of tire slip angle is established based on the characteristics of the slope of the leading edge of the lateral displacement and the slip angle by combining the tire lateral deformation theory.A Gaussian process regression estimation model for tire slip angle is established based on the correlation between frequency domain and slip angle with PLS-VIP features as input.Two tire slip angle estimation models are also validated,in which the physical estimation model can predict better in the range of ±4 degrees,while the machine learning estimation model can estimate accurately in a larger slip angle range（±8 degrees）.Accurate tire slip angle estimation allows more accurate calculation of other tire/vehicle states and parameters,which is important for advanced vehicle control systems.Finally,the data analysis is carried out on the time-domain acceleration under driving-braking conditions,in which there are significant characteristics of vertical acceleration and slip ratio,and the results calculated by PLS-VIP method also verify this conclusion.A tire slip ratio estimation model is developed by using PLS-VIP features as input and combining with Gaussian process regression algorithm.The proposed estimation model can continuously and stably estimate the slip ratio with30%,and the results have high robustness to vehicle speed and load.This study integrates intelligent tire system,tire physical model and machine learning algorithm with PLS-VIP features to provide new methods for accurate estimation of tire force,tire slip angle and tire slip ratio under different driving conditions,and also provides a new way for accurate estimation of intelligent vehicle,smart chassis and tire state.