A Quarter Car ARX Model Identification Based on Real Car Test Data

Abstract

This paper presents a system identification of a quarter car passive suspension system dynamic model based on real-time running test car data. The input-output data of a car were recorded by test-driving the car on a road surface. The input variable is the vertical acceleration of the car shaft, while the output variable is the vertical acceleration of the body of the car. Two acceleration sensors were installed on the front right corner of the car: One on top of the suspension and another on the car shaft at the bottom of the suspension. The acquired data were used to identify the mathematical model of a quarter car passive suspension system dynamics. A quarter car passive suspension system was assumed to have an ARX model structure, hence qualifies to be a candidate model for system identification. The system identification algorithm used in this work was based on linear least-square estimation. The results showed that the best ARX model of the car passive suspension system model is produced with the best fit of 90.65%, Akaike’s FPE is 5.315x10-6. The output order of the model was found to be four, the input order is two and the time delay was one. The fit rate greater than 90% and along with a very small value for the FPE means that the system identification requirements are fulfilled and the identified model is acceptable.