Optimizing Artificial Neural Networks Modelling in Predicting International Roughness Index for Flexible Pavement

Accurate road roughness prediction is essential for sustainable transportation planning and cost-effective maintenance strategies.This study develops a systematic algorithm to optimize Artificial Neural Networks (ANN) for predicting International Roughness Index (IRI) values using Equivalent Standard Axle (ESA) and road age as primary inputs.The methodology employs comprehensive parameter space exploration across four optimization stages, evaluating various ANN configurations to identify the most effective architecture.Rigorous statistical validation through Analysis of Variance (ANOVA) and cross-validation ensures model reliability.Data quality assessment with outlier detection using the Interquartile Range method was implemented, retaining 94.3% of original observations.The optimized 6-30-25-20-1 ANN configuration, employing logsig and purelin transfer functions, achieved strong performance metrics, including R = 0.9554, R 2 = 0.9020, MSE = 0.0153, RMSE = 0.1236, and MAPE = 0.0285.Statistical validation confirmed significant model improvements with an F-statistic of 24.367 and a cross-validation mean of 0.892.The RMSE accuracy of 0.1236 m/km enables reliable pavement condition classification within established IRI thresholds, supporting timely maintenance decisions.This streamlined approach addresses critical infrastructure management challenges by enabling cost-effective maintenance planning with minimal data requirements, particularly valuable for developing countries with limited pavement monitoring infrastructure.The model's computational efficiency facilitates network-wide deployment for long-term planning and strategic resource allocation.Road agencies can apply this model for maintenance budget prioritization, networklevel condition assessment, and multi-year intervention scheduling, particularly in resource-constrained environments where comprehensive pavement monitoring systems are unavailable.This study establishes a structured approach to optimize ANN for IRI prediction, enhance the effectiveness of Pavement Management Systems (PMS), and support sustainable transportation infrastructure through improved maintenance scheduling.