Evaluation of the performance of regression and neural network models in estimating the relationship between pavement status index (PCI) and international roughness index (IRI)

Evaluating and measuring the quality of pavement by better understanding the correct performance of different parts of the pavement management system, it is possible to determine the database to predict the optimal design time and update information. Pavement condition index (PCI) and International Roughness Index (IRI) are two criteria for measuring the performance of road pavement evaluation, which has a significant impact on the management of road repair and maintenance budgets. However, the PCI and IRI indices are not comprehensive and need to be addressed by establishing a logical relationship with high accuracy. The aim of this study was to determine the relationship between IRI and PCI indices using Spearman, Pearson and Watson statistical analyzes, linear and nonlinear regression models and neural network model. In this regard, data of 392 sections of arterial pathways in Isfahan province were reviewed and PCI and IRI indices of these sections were selected as the statistical population. The results of the modeling section showed that despite the proper performance of the neural network in predicting training data, but in predicting new data and in the test phase with about 46% error, has a poor performance and the model has been overworked. Finally, the performance of the linear regression model showed that using the relation PCI = -19.802 × IRI +126.970 with 77% accuracy PCI values can be determined using IRI data. Considering the speed of harvesting and analysis of the high results of the IRI index and the applicability of the PCI index as a standard method in planning and determining road maintenance options, this model can be a great help in increasing the accuracy and speed of decisions and reducing the heavy costs of failure and time analysis of results.