TRAFFIC NOISE MODELING USING ARTIFICIAL NEURAL 4 NETWORK

Abstract

Noise generated from vehicular traffic is a major source of environmental pollution. WJ JO has recognized environmental noise to be a harmful environmental pollutant which has been reported to have adverse psychosocial and physiological eflbcts on human health [169]. In the rapidly developing countries like India, transportation sector is growing swi lily. This has led to overcrowded roads and pollution. Vehicular emissions of dust particles, smog and noise have reached or even exceeded levels of those from industrial production or private households, and are harmful to the environment and human health. The recognition of traffic - noise as one of the main sources of environmental pollution has led to the development of models that enable the prediction of traffic noise level from fundamental variables like traffic how, speed and density, distance from roads and condition of roads. Trallic noise prediction models are required as aids in the planning and design of urban proiects and roads, and also in the assessment of existing or envisaged changes in traffic noise conditions. These models are commonly needed to predict sound pressure levels, specified in terms of ç,, 111 . i. etc., at selected locations and in the analysis of mitigation measures during road construction and in idcntit\'ing the variables with the highest noise incidence. Since traffic, environmental and geospatial characteristics, emission level of vehicles, noise surveying methods and road traffic conditions differ from region to region and from one country to another. Different countries have developed noise prediction models according to their traffic, environmental and geospatial characteristics. The most popular ones include Federal l-lighway Administration (Fl-I WA) model in United States. Calculation of Road Traffic Noise (CoRTN) model in United Kingdom, Acoustical Society of Japan (ASi) model in Japan. Stop and Go model in Bangkok, MITHRA in France. Geographic Information System (GIS) model in china and Richtlinien fur den Iarmsehutz an Straf3en (Guidelines Ibr Noise Protection on Streets) i.e. RLS-90 in Germany etc. Steele [l57J, Rajakumara and Mahalinge Gowda [139j presented detailed review of various noise prediction models including above mentioned models. Several attempts have been made to predict and model road traffic noise statistically by different researchers. These models are based on theoretical factors that are applicable S based on statistical relationships and on macroscopic traffic variables such as traffic flow and average speed. Their results have been very good related to roads and highways where traffic prevails, and flow conditions are relatively homogeneous. tinder heterogeneous traffic flow and speed conditions these models gives poor performance. To overcome problems associated with heterogeneous traffic conditions researchers applied stochastic models for traffic noise prediction. Despite mathematical, statistical and stochastic models, models based and soft computing techniques such as Artificial Neural Network (ANN), Genetic Algorithm GA and advanced engineering tool such as GIS were developed. Noise level depends on many factors such as number of vehicles, speed of vehicles, background noise, meteorological and geospatial conditions. These parameters together make noise modeling a complex task and highly nonlinear phenomena, which turns out conventional deterministic models inappropriate. To overcome the limitations associated with the conventional deterministic, statistical and empirical models a new paradigm is required. Study of ANN has been done for that purpose. ANNs are appropriate soft computing tools lbr modeling multifunction, nonlinear and complex data related problems. An ANN is an intbrmation processing paradigm that is inspired by the way a biological nervous system, such as the brain processes information. In this information processing system, the elements called neurons, process the information. To create a noise free environment, noise abatement techniques and equipments are required so that the noise level along a highway can be minimized up to an acceptable value. The installation of noise barriers between noise source and noise sensitive areas along major roads and freeways is another way to combat traffic noise. ANN has been successfully applied to determine the optimized height of a highway noise barrier. Traffic congestion remains a major societal and economical problem across the world, with no visible sign of substantial reduction in future. Traffic control systems arc based on the concept of avoiding traffic instabilities and of homogenizing the traffic flow in such a way that the risk of accidents is minimized and mean velocity or the traffic flow is maximized. Intelligent Transportation Systems (ITS) applications are those which improve the efficiency of surface transportation systems and solve transportation problems by using modern information and communication technologies. ITS like Advanced Traveller information Systems (NITh) and Advanced Traffic Management systems (ATMS) have been deployed in few Indian cities like Pune and Hyderabad. To fulfil the increasing traffic demand, there is need to implement ITS for efficient utilization of transport infrastructure. One of the most important requirements of these systems is the ability to predict the nature of the traffic stream accurately. An attempt has been made to apply ANN for the prediction of traffic flow 4 four lane divided national highway networks in heterogeneous conditions. 'thus present thesis entitled raffle Noise Madding Using Artificial Neural Network" deals with development of traffic noise prediction model, barrier height determination, effects of traffic noise on human health and traffic volume prediction for mixed type traffic in Indian conditions. Chapter wise summary of the thesis is given below: Chapter 1 is introductory in nature and gives a brief account of general theory of road traffic noise modeling and prediction along with highway noise barrier. The basic introduction about ANN along with its working mechanism is also presented. At the end of the chapter, summary of the whole work is embodied. In chapter 2, a road traffic noise prediction model for Indian conditions is developed using regression analysis which is based on Calixto model [33]. A statistical model of road traffic noise in an urban setting was developed by Calixto et al. [33] which is based on the Iuict that percentage of heavy vehicles plays an important role over road traffic noise emission. Keeping this in mind, the weighting factor that represents the weightage of presence of heavy vehicles over road traffic noise emission was calculated in Indian road conditions. Developed model is then checked for validation by using actual data which were measured at selected location on Ni 1-58 and ibund suitable luir Indian road conditions. In chapter 3, ANN has been applied to predict noise pollution level in Chandigarh. a planned city of India. The motive behind this study was to investigate if a neural network can be used in a statistically sound manner to model traffic noise in the case of Indian conditions. where most of the traffic is of mixed type. Models based on back-propagation neural network were trained, validated and tested using data collected through field studies. it was found that ANNs have better capability to reduce the error in traffic noise prediction as compared to linear regression and modified Federal I Iighway Administration (FHWA) model. In chapter 4, optimized height of a highway noise barrier has been determined using ANN. Field measurements were carried out to collect traffic volume, vehicle speed, noise level and site geonwtry data. Barrier height was varied from 2 to 5 mcters in increments of 0.1 meter for each measured data set to generate theoretical data for network design. Barrier attenuation was calculated for each height increment using FHWA model. For neural network Iii design purpose classified traffic volume, corresponding traffic speed and barrier attenuation data have been taken as input parameters, while barrier height was considered as output. ANNs with diflèrent architectures were trained, cross validated and tested using this theoretical data. Results indicate that ANN can be useful to determine the height of noise barrier accurately which can effectively achieve the desired noise level reduction, for a given set ol traffic volume, vehicular speed, highway geometry and site conditions. In chapter 5, the physiological effects of traffic noise on the people living in the vicinity of highway in a medium sized city have been investigated. A noise survey questionnaire was prepared lollowing the international guidelines. Questionnaire based survey at ten selected locations was carried out in the city. Traffic flow, speed of vehicles and noise level data were recorded at the ten selected locations in the city. Results of the study demonstrate that vehicular road traffic is the major source of noise pollution which creates annoyance among people. Regression analysis was performed between various noise descriptors and percentage of highly annoyed population which shows a strong correlation bet•-een them. In chapter 6, the problem of short terril prediction of traffic volume using past traffic data is studied. Besides traffic volume, speed and density, the model incorporates both the time and the day of the week as input variables. The model has been validated using actual rural highway traffic flow data collected through field studies. It was concluded that ANN has produced good results even iIspeeds of each category of vehicles were considered separately as input variables. In chapter 7, concluding observations with a significant analysis of the work presented in earlier chapters of this study is carried out. In addition, a brief discussion on the scope for further work is described.