Road Profile Estimation Using Full/Quarter-Car Model with Artificial Neural Networks

Abstract

The monitoring of road roughness is one of the first and most critical steps in road maintenance. Road networks need constant maintenance to function properly and avoid any hazardous accidents or blockage of the traffic flow. The International Organization of Standardization (ISO) developed the International Roughness Index (IRI) to unify road monitoring systems and to categorize roads based on their roughness levels. The ISO 8608 standard divides road roughness levels into eight different classifications ranging from best (Class A) to worst (Class H) roads. The road roughness profile of a certain road section could be measured by traditional equipment such as a dipstick profilometer, profilograph, or an automated road meter. However, these traditional methods are time-consuming and costly. Thus, this research proposes the use of dynamic vehicle accelerations of a moving regular car and artificial neural networks to estimate the road profile. This research also compares the accuracy and efficiency of using a full car (7 degrees of freedom) numerical model and a quarter-car (2 degrees of freedom) numerical model in training the neural network and estimating the road profile. These numerical models will be used to create a library of input data sets (vehicle accelerations) and output data sets (road roughness profiles) which will be used to train the artificial neural networks (ANNs).