Research On Rollover Prediction Model Of Articulated Steering Vehicles Based On Data-driven

Accurately describing the risk of rollover is the basis for the realization of vehicle rollover prediction and active instability safety control.However,articulated steering vehicles often drive in complex ground environments.The process is accompanied by strong nonlinearity,large inertia,and high time lag,which increases the difficulty of modeling and describing rollover risk.This paper focuses on the construction of a datadriven articulation steering vehicle rollover prediction model,which comprehensively consider the body attitude and kinematic parameters.This paper solves the problems of inaccurate dynamic modeling calculations caused by the complexity construction of these vehicles and their operating environment,so as to achieve an accurate description of the risk rollover of articulated steering vehicles.The specific research work of this paper is as follows:(1)Using scale home-built vehicle to obtain key data about rollover.In this paper,a 1:4 scale home-built articulated steering vehicle and a key data acquisition system are constructed.The scale vehicle is operated to drive in a real environment to collect data due to the virtual simulation test conditions are ideal and the real car rollover test is destructive.The data in the real environment can be obtained at a lower cost,which provides data support for data-driven modeling research.(2)Using data-driven modeling method to build a lateral stability identification model to realize the rollover risk assessment of articulated steering vehicles.A datadriven modeling method is proposed since the physical modeling method is difficult to accurately describe the lateral stability of articulated steering vehicles operating in unstructured ground environment.First,methods such as standardization,correlation analysis and wavelet denoising are used to preprocess the data to improve the reliability and applicability of data.Then three different data modeling methods are introduced to build a lateral stability identification model and cross-validation and grid search methods are used to optimized the model.(3)Research the prediction method of rollover risk based on long short term memory neural network to realize the prediction of rollover risk.A rollover risk prediction modeling method is proposed to identify the vehicle’s lateral stability at the future moment.First,the statistical index of key data within a period of time is calculated by the sliding window method,and time shift processing is performed on the original output data.Based on the above data,the LSTM algorithm is used to build a rollover risk prediction model to achieve accurate prediction of the vehicle’s lateral stability after 0.2 seconds.In summary,this paper comprehensively applies theoretical analysis,proportional prototype,machine learning and other methods to establish a data-driven rollover risk prediction model to solve the accurate description of lateral stability for articulated steering vehicles.It provides a new research method and reference for the rollover risk description of other complex vehicles with strong nonlinearity and large inertia.