PERFORMANCE BASED MAINTENANCE MANAGEMENT FOR RURAL ROADS

Abstract

India's 3.3 million kilometers of road network is the second largest road network in the World after USA, carrying 85 percent of passenger and 70 percent of freight traffic and eighty-five percent of its total network comprises of Low Volume Roads (LVR). In India, rural roads are defined as the LVR having traffic volume less than 450 Commercial Vehicles per Day (CVPD) and district roads and village roads are categorized as LVR. Current replacement of rural roads in India can cost to the tune of Rs. 200,000 millions (US $ 4445 millions). Additional length of 0.37 million km of rural roads are being presently constructed/upgraded under the Government of India sponsored programme viz; Pradhan Mantri Gram Sadak Yojana (PMGSY), costing about Rs. 760,000 million (US $ 16,889 millions) and by the end of year 2009, the construction of majority of rural roads will be completed. Hence, there would be an immediate need for shifting the emphasis towards maintenance and rehabilitation activities of these huge network of rural roads constructed under PMGSY and other different schemes. The main problems associated with rural roads in India are : lack of historical pavement performance database for developing pavement performance models; maintenance guidelines suggested in IRC:SP:20,2002 are subjective in nature based upon the past experience of field engineers; and lack of timely maintenance. Thus, there is an immediate need to carry out the pavement performance study for rural roads, which will be useful in developing pavement deterioration models and in suggesting optimum maintenance strategies for the preservation of huge network of roads. This scientific approach will also be helpful in optimal utilizing the constrained maintenance budget provisions for a developing country like India. The broad objective of the present study is to carry out the pavement performance study for the period of about three years on the rural road sections constructed under different rural road schemes and to develop combined pavement condition prediction models as well as to calibrate the World Bank developed HDM-4 pavement deterioration models for local conditions of India by utilizing the collected pavement performance data in the study. The scope of the study also consists of suggesting the optimum maintenance strategies for rural roads, based upon the developed pavement performance models. The experimental matrix has been designed for categorizing the rural road sections into nine cells and variables considered for defining experimental matrix are as : terrain types .climatic condition types, traffic variations and age of the rural road since its construction or latest rehabilitations. The individual cell represents the rural road sections in any particular category and in this study an attempt has been made to develop combined pavement condition prediction models as well as calibration of HDM-4 based various pavement deterioration models for rural roads representing each individual cell category. Based upon the designed experimental matrix for carrying out the pavement performance study, sixty one in-service rural road sections with a traffic varying between 0-450 CVPD in seven districts of Uttaranchal state of India selected. "Window" methodology for development of pavement performance models has been used in this study as this methodology permits section distress measurements to be made in a brief period of time and consequently allows for the evaluation of a greater number of sections and categories. In terms of "Window" methodology applications, the first step consists of defining homogeneous rural roads sections according to their most representative variables (structure, traffic, geometry, and climate) as a single family. Each of these sections is then considered as a window in a pavement's performance curve and together with other windows, represent a pavement's performance of that family. Ifthe sections in a chosen family have different ages, the pavement performance curves for particular family can be obtained. The pavement performance data collected in this study are then used for the development of composite of PSI/ PCI based pavement condition prediction models as well as calibration of HDM-4 pavement deterioration models for each of the individual cell category. The road network and inventory data for each of the identified rural road section has been collected from the construction and maintenance records of in the concerned highway division's in-charge of the maintenance of these roads as well as by carrying out the actual field studies. Each sample road section comprised of 1 km in length and 100 m length of representative test section per sample road section was considered for distress measurements. The type and extent of distresses (i.e. cracking, raveling, and pothole and edge break area) have been measured by visual condition survey for all selected test sections, during winter, pre-monsoon and after monsoon season in the years 2004, 2005 and during pre-monsoon season of 2006 and the pavement performance database in 'Microsoft Access' are designed and developed for arranging the pavement performance data in easy assessable format. The rut depth for all selected rural road sections are measured with a 2 m straight edge and the pavement roughness on each selected rural road section in plain and rolling terrain are measured with 'Fifth Wheel Bump Integrator', towed by the mobile van as per the standard procedure in terms of Unevenness Index and the same has been subsequently converted into International Roughness Index (IRI). The average motorized and non-motorized traffic volume/ traffic composition in plain, rolling and mountainous terrains are determined after carrying out 24 hours traffic counts on all selected rural road sections in the study area. The normal annual traffic growth rate has been adopted as six percent for motorized traffic and three percent for non-motorized traffic as per the guidelines given in Indian Roads IRC: SP:20, 2002 for rural roads. The basic as well as economic characteristics of vehicle fleet plying on rural road sections in Uttaranchal State has been collected from the secondary source such as Road User Cost Study data (MORT&H, 2001). Present Serviceability Index (PSI) has been calculated for each selected rural road section by using the equation as given in AASHO Road Test and the model parameter estimations for different form of models are carried out by carrying out regression analysis and taking 'PSI' as dependent variable and the 'pavement age' as an independent variable. The model corresponding to minimum root mean square error (RMSE) and maximum goodness of fit (R2 value) and satisfying boundary conditions have been suggested for each cell category as well as average models for plain, rolling, and mountainous IV terrains. These models have been validated on different rural road sections by carrying out various statistical tests and it shows a significant relationship between the observed versus model predicted PSI values for all the three types of terrains. Present Condition Index (PCI), an index reflecting the composite effects of varying distress types, severity level, and extent upon the overall condition of pavement has been calculated for each rural road section in all cell categories. The parameter estimation for different forms models are carried out by carrying out regression analysis and taking 'PCI' as dependent variable and the 'pavement age' as an independent variable. The model corresponding to minimum root mean square error (RMSE) and maximum goodness of fit (R2 value) and satisfying boundary conditions have been suggested for each cell category as well as average models for plain, rolling, and mountainous terrains. These models are validated on different rural road sections by carrying out different statistical tests and it shows a significant relationship between the observed versus model predicted PCI values for all the three type of terrains. The pavement deterioration models, i.e., initiation and progression of total cracking, raveling, potholing and progression of edge break, rut depth and roughness as incorporated in World Bank developed "Highway Development and Management Tool (HDM-4)" have been calibrated for local conditions of rural roads in plain, rolling and mountainous terrains of India. Calibrating HDM-4 pavement performance models to the local condition is very important because of its economic impact and the direct influence of the initiation and progression of pavement distress processes makes it necessary to adequately adjust them to perform an accurate economic evaluation of the road studies. The procedure proposed for the statistical calibration of pavement performance models have been based mainly on determining the factors that allow for a more precise or better adjustment of the simulated distress curves to the real performance data. Two different procedures have been followed depending on whether the factor to be determined corresponds to the initiation or progression phase of the distress. The procedures proposed for calibrating surface distress initiation factors are based on obtaining the coefficient between the observed year of occurrence of the distress to the year of occurrence as predicted by the un-calibrated models. In the case of the progression factors, the calibration has been done by minimizing the squares of the differences of the estimated and observed data or sum of squared differences (SSD).The statistical indicators of average absolute error(AE), root mean square error(RMSE) and goodness of fit measure(R2 values) are calculated and the calibration factors corresponding to minimum RMSE and maximum goodness of fit (R2 values) has been suggested for rural roads in each cell category. The validity of the suggested pavement deterioration models has also checked on different rural road section, by using statistical parameters like coefficient of determination value (R2) and carrying out't' test. It shows a significant relationship between the observed and HDM-4 model predicted distress values for all the three type of terrains. Four Artificial Neural Network (ANN) based unified pavement deterioration models for the prediction of progression of cracking, raveling, rut depth, and roughness of rural roads in different terrains have been developed in this study by using the collected pavement performance data. Twelve different architectures of ANN models have been attempted by varying the number of hidden layers and number of neurons per hidden layer for each of the four models and the model corresponding to minimum root mean square error (RMSE) during training and testing stage has been suggested. The suggested ANN models have been validated for rural road sections (different from used for training and testing) and comparison has been made between the observed distresses versus ANN model predicted and calibrated HDM-4 model predicted distresses for each of the validation rural road section. From the study results it is found that the suggested ANN models give better predictions than of HDM-4 models. Based upon combined pavement condition prediction models, age for triggering different PSI/PCI values are determined for rural roads representing cell-1 to cell-8 and accordingly the appropriate broad maintenance strategies along with the various maintenance alternatives options for each broad maintenance strategy has been suggested. VI Using the calibrated HDM-4 pavement deterioration, life cycle cost analysis have been carried out for rural road sections constructed under different schemes representing cell-1 to cell-8 and under PMGSY schemes by following the design specifications as specified in IRC:SP:20,2002. Various scheduled and responsive maintenance standards consisting of applications of Single Bituminous Surface Dressing (SBSD), Double Bituminous Surface Dressing (DBSD), Fog seal, Rejuvenations, and Pre-mix carpet have been described. Ten different maintenance policies consisting of various maintenance standards have been formulated. All the maintenance alternatives in each maintenance policy have been compared with the base option of "do minimum". The benefits derived from implementation of other options have been calculated by comparing the predicted economic cost streams in each year (against that for the respective year of the base case option) over a specified analysis period of 25 years. Optimum maintenance alternatives for each of the maintenance policies have been suggested based upon the maximum of Net Present Values (NPV) and Internal Rate of Returns (IRR) amongst the all maintenance alternatives. For rural roads constructed under different schemes as well as PMGSY schemes, net present values (NPV) comparison have been made between optimum maintenance alternatives as suggested from HDM-4 life cycle cost analysis versus corresponding maintenance provisions of IRC:SP:20,2002. Increase in NPV has been worked out by providing the HDM-4 based optimum maintenance alternatives. Typical details of Total Transportation Cost (TTC) composition i.e. Road Agency Costs (RAC) and Road User Costs (RUC) have been worked out for rural road sections under PMGSY schemes and design curves A, B, C, and D. The details of decrease in TTC against the corresponding increase in RAC have been worked out for each rural road section for providing optimum maintenance alternatives as suggested by HDM-4 against the corresponding maintenance provisions in IRC: SP: 20. Optimum time and type of maintenance alternatives have been worked out for rural roads constructed under different schemes in the present study for VII each defined maintenance policy. World Bank tool HDM-4, after calibrating to local Indian conditions has been used for carrying out life cycle cost analysis and the rural roads considered for analysis consisting of varying design traffic volume and composition and located in different types of environmental conditions of plain, rolling and mountainous terrains. From the study results it is found that HDM-4 based maintenance provisions suggested in this study are more beneficial than the corresponding maintenance provisions as suggested in IRC:SP:20, 2002 for the maintenance of rural roads in India.