A NEURAL NETWORK NOISE PREDICTION MODEL FOR TEHRAN URBAN ROADS

Mansourkhaki, Ali; Berangi, Mohammadjavad; Haghiri, Majid; Haghani, Mohammadreza · 2018 · Crossref

DOI: 10.3846/16486897.2017.1356327

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the need for accurate traffic noise prediction models in Tehran, Iran, where rapid urbanization and vehicle growth have exacerbated noise pollution. Existing classical models, such as CoRTN and RLS90, rely on linear regression and fail to account for the non-linear nature of traffic flow and the specific acoustic effects of urban structures, particularly building facade reflections. The authors developed a multilayer perceptron artificial neural network (ANN) model to predict the equivalent continuous sound level ($L_{Aeq}$) more accurately by incorporating these complex variables. The methodology involved collecting 51 data samples from 34 locations across Tehran over a one-month period. Field measurements were taken using a sound level meter at 1.2 meters height, with simultaneous video recording to analyze traffic characteristics. Input parameters included total hourly traffic volume, average vehicle speed, percentages of four vehicle categories (cars, vans/pickups, heavy vehicles, motorcycles), road gradient, building density, and a novel "Building Reflection Factor" (BRF). The BRF was calculated using panoramic photography and satellite imagery to quantify the height and distance of adjacent buildings relative to the receiver. The dataset was randomly split into 80% for training, 10% for validation, and 10% for testing. The ANN was trained using the Levenberg-Marquardt algorithm in MATLAB, with various network architectures tested to optimize performance. The results demonstrated that the optimal ANN architecture (6-10-1 structure) achieved a correlation coefficient of 0.9915 with measured data, significantly outperforming a proposed multiple linear regression model ($R = 0.837$) and classical models. The prediction error for the ANN model ranged narrowly between –1.41 and 1.34 dB(A), whereas the regression model exhibited errors between –4.63 and +3.61 dB(A). A statistical paired t-test confirmed that the ANN predictions did not differ significantly from field measurements, validating the model's high efficiency and goodness-of-fit. The inclusion of the Building Reflection Factor was identified as a critical factor in reducing mean square error and improving predictive accuracy. The study concludes that ANN models are superior to traditional linear regression methods for predicting urban traffic noise, particularly in environments with complex geometric features like building reflections. The findings highlight the importance of incorporating structural parameters such as BRF into noise prediction models for cities with dense urban fabrics. Given that all measured sites exceeded permissible noise limits, the authors emphasize the urgent need for noise mitigation strategies, such as noise barriers and improved traffic management, to protect public health in Tehran.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction