A multilayer perceptron approach for condition monitoring of automobile suspension system through vibration signal

Abstract

The suspension system is a vital component of a car, with the aim of improving stability and good road handling during vehicle movement in uneven road conditions. A variety of load conditions, extended operation, and ongoing absorption of road shocks and time-based deterioration of internal parts can lead to suspension system defects. Faults in the suspension system can destroy the inner components that allow the system to malfunction and jeopardize the safety of cars. Condition inspection has now been an integral part of the identification of defects in a suspension device. This paper focuses on suspension device status control by vibration signal analysis and errors are classified using computer classification systems. The test was conducted in the Ford EcoSport vehicle, in which the accelerometer was mounted near the suspension system's stationary portion. In order to derive data, a vibration signal is captured and used. An additional J48 decision tree algorithm was used during the selection process to classify the most important features. The type of fault occurring from the chosen features was discovered using Multilayer Perceptron (MLP) classifiers. The MLP model provided 98% with an error rate of 0.016 as prediction performance.