Road User Effects Modelling in HDM-4 Christopher R. Bennett Highway and Traffic Consultants
HIGHWAY DESIGN AND MAINTENANCE STANDARDS MODEL (HDM-III) u Developed by the World Bank and released in 1987 u Used in over 100 countries for different types of investment studies u Predicts pavement performance over time and under traffic and effects of maintenance on pavements u Predicts the effects of pavement and operating conditions on vehicle operating costs (VOC) u Fundamental relationships based on research conducted in Kenya, the Caribbean, India and Brazil
DEVELOPMENT OF HDM MODEL
HDM TECHNICAL RELATIONSHIPS STUDY (HTRS) u Funded by ADB u Led by N.D. Lea International Ltd. (Canada) u Hosted and supported by Institut Kerja Raya Malaysia (IKRAM) u Other participants: –Bill Paterson (World Bank) –Works Consultancy Services (New Zealand) –Department of Transport (South Africa) –Van Wyk and Louw (South Africa) –TRL (U.K.) –University of Auckland, University of Pretoria, Michigan Technical Institute –Snowy Mountain Engineering Corporation (Australia) –Various Individuals
HTRS Approach u Key areas for attention identified at HDM-4 UK and Malaysian workshops u Virtually no primary research u Primarily consisted of reviewing existing research and implementing/adapting results u Working papers circulated to a large number of reviewers and comments incorporated into final report
REVIEW OF PREVIOUS EXPERIENCES WITH HDM u Contacted academics, consultants, governments, lending agencies u Identified studies in over 100 countries u Used the results to identify the key areas requiring attention in HDM-4 and for preparing draft specifications u Very few studies undertook rigorous calibration/ adaptation of HDM u Summarised parameter values by region and study u Identified alternative models and relationships to those in HDM-III u VOC and RDME results presented in two internal reports
Components of RUE
RUE Research u Most models in use draw on HDM-III u No major RUE studies since HDM- III u Several studies addressed HDM-III calibration or investigated single components - e.g. fuel
Key Changes to HDM-III u Unlimited number of representative vehicles u Reduced car maintenance costs u Changes to utilisation and service life modelling u Changes to capital, overhead and crew costs u New fuel consumption model u New oil consumption model u Changes to speed prediction model u Use of mechanistic tyre model for all vehicles
New Features in HDM-4 u Effects of traffic congestion on speed, fuel, tyres and maintenance costs u Non-motorised transport modelling u Effects of roadworks on users u Traffic safety impact u Vehicle noise impact u Vehicle emissions impact
Factors Influencing RUE
Motorised Transport
Non-Motorised Transport
Maintenance and Repair Cost Modelling
Parts and Labour Costs u Usually largest single component of VOC u In HDM-III user’s had choice of Kenya, Caribbean, India and Brazil models u All gave significantly different predictions u Most commonly used Brazil model had complex formulation u Few studies were found to have calibrated model
Brazil Parts - Roughness
Brazil Parts - Age
Observations on Brazil Model u there are inconsistencies in the Brazil predictions between vehicles u users believe that HDM-Brazil often over- estimates parts consumption u users found the model difficult to calibrate u because of the non-linearity of the parts consumption relationships, assuming that all vehicles are midway through their life gives a distorted estimate
Observations Continued: u some analysts (eg RTIM) prefer the use of a logit model over a continually increasing roughness model u the use of standardised parts results in a distortion of the costs u significant regional variations in maintenance practices
Ratio of Parts to Labour
Data Sources u Small studies conducted in: –Botswana –New Zealand –Pakistan –South Africa –St. Helena –Sweden u NO major studies identified
NDLI Proposals u Replace HDM-III Brazil model with linear model u Standardise predictions to 100,000 km u Eliminate roughness effects below 3 IRI
1995 RUE Workshop Proposals u Linear model definite improvement over HDM-III u Significantly reduce the light vehicle parts consumption u Increase the heavy bus parts consumption u Slightly reduce truck parts consumption u Modify coefficients to account for survivorship bias and technical improvements
Proposal for HDM-4 u Linear models: –PARTS = {K0pc [CKM^kp (a0 + a1 RI)] + K1pc} (1 + CPCON dFUEL) –LH = K0lh [a2 PARTS^a3] + K1lh u Adjusted roughness: –RI = max(IRI, min(IRI0, a0 + a1 IRI^a2))
Adjusted Roughness
Parameter Values u Estimated from HDM-III Brazil model u Exponential models converted to linear models which gave similar predictions from IRI u Roughness effects reduced 25% for trucks u For cars, roughness effects same as for trucks u For heavy buses, roughness effects reduced further 25%
Implications of Changes
Age Effects u Parts modelled at 0.5 of vehicle life u User will be able to enter an age distribution and have this used in calculations
Congestion Effects u Parts consumption is assumed to increase under congested conditions u Use equation: –PARTS = PARTS (1 + CPCON dFUEL) u Default value for CPCON is 0.10 indicating that a 100% increase in fuel results in a 10% increase in parts
Utilisation and Service Life
HDM-III u Contained three utilisation methods: –Constant Kilometreage –Constant Hours –Adjusted Utilisation u Contained two service life methods: –Constant Service Life –de Weille’s Varying Service Life
Adjusted Utilisation u Predicted utilisation as function of speed and ‘elasticity of utilisation’ u Default elasticity values derived from Brazil study u Some Brazilian vehicles had unusually high utilisations u Analysts tended to adopt default values
Elasticity Values Applied
Effect of Speed on Utilisation
Service Life Modelling u de Weille’s method based on the assumption that the faster the vehicle travels the shorter the life u No empirical data to support method u Made costs very sensitive to speed
Speed on Service Life
Methods Applied
Implications of Methods
Recommendations - Service Life u NDLI recommended use of ‘Optimal Life’ (OL) model u 1995 RUE Workshop recommended OL model u Proposed that OL model be adopted for HDM-4
Recommendations - Utilisation u NDLI proposed modified adjusted utilisation method u 1995 RUE Workshop did not support method u TRL have proposed alternative method for utilisation u Recommended that TRL method be adopted
Capital Costs
Modelling Approach u Comprised of depreciation and interest costs u HDM-III used a simple linear model u Affected by operating conditions through the effects of speed on utilisation and speed on service life (de Weille’s method) u HDM-4 will use ‘Optimal Life’ method
Optimal Life Method u Proposed by Chesher and Harrison (1987) based upon work by Nash (1974) u Underlying philosophy is that the service life is influenced by operating conditions, particularly roughness u Relates life -- and capital costs -- to operating conditions
OL Method
Implementation u NDLI found that the OL method had problems when applied with ‘typical’ field data u Chesher (1995) proposed addressing problems by adjusting age effects for survivorship bias
Implicaitons of Age Modification
HDM-4 Implementation u User defines ‘target’ OL in km at low roughness (3 IRI) u User defines financial replacement value and utilisation characteristics u The age exponent is calibrated u Effect of roughness on service life established u Depreciation calculated
Roughness on Life
Roughness on Depreciation
Fuel Consumption
Fuel Model u Replaced HDM-III Brazil model with one based on ARRB ARFCOM model u Predicts fuel use as function of power usage
Forces Opposing Motion u Calculates: –aerodynamic resistance (Fa) –rolling resistance (Fr) –gradient resistance (Fg) –curvature resistance (Fcr) –inertial resistance (Fi) u Uses more detailed equations than HDM-III
Modifications u Made modifications to ARFCOM approach to improve predictions of engine and accessorypower u Replaced engine speed equations with speed based function from simulation
Model Parameters u Two basic model parameters for use: –idle fuel rate –fuel conversion efficiency factor u Parameters can be readily derived from other fuel models u Expect to provide a range of values for different vehicle types from various published sources
Implications of New Model u Lower rates of fuel consumption than HDM-III for many vehicles u Effect of speed on fuel significantly lower for passenger cars u Considers other factors -- eg surface texture and type -- on fuel u Model can be used for congestion analyses
Speeds
Speed Model u Minor changes to HDM-III probabilistic model u Same model form: u Refinement of some constraining speeds
Congestion Effects
HDM-4 Congestion Modelling u HDM-III did not consider congestion u HDM-95 considered effects of congestion on speeds but not on other VOC u HDM-4 expanded the HDM-95 approach to consider other VOC components
HDM-95 Speed-Flow Model
Recommended Model Parameters
HDM-4 Congestion Model u 3-Zone model predicts as flows increase so do traffic interactions u As interactions increase so do accelerations and decelerations u Adopted concept of ‘acceleration noise’ -- the standard deviation of acceleration
Acceleration Noise
u Modelled with two components: traffic induced and ‘natural’ noise u Traffic noise function of flow u Natural noise function of: driver’s natural variations road alignment roadside friction non-motorised transport roughness
Traffic Noise u Modelled using sigmoidal function u Integrated with Three-zone Model u The maximum traffic noise and ratio Q0/Qult governs predictions u Easy to calibrate
Natural Noise u Driver and alignment noise combined u Side friction, non- motorised transport and roughness assumed to be linear u Maximum values of 0.20, 0.40 and 0.30 m/s/s respectively
Calculation Approach u Run as calibration routine once unless vehicle characteristics changed u Uses Monte Carlo simulation of a vehicle travelling down a road with different levels of acceleration noise u Determines additional fuel as function of noise u Results in matrix of values of dFUEL vs Mean Speed vs Accel. Noise
Typical Simulated Accel. Profile
Simulation Results - Small Car
Simulation Results - Artic. Truck
Flow on Additional Fuel
Tyre Consumption
HDM-4 Tyre Model u Did not prove possible to locate any major new tyre research since HDM-III study u Swedish team recommended simple procedure for adapting HDM-III parameters as function of tyre life u This was applied and parameters estimated for light vehicles to allow for consistent modelling
Mechanistic Model u Tyre consumption proportional to forces on tyre u Increase with 4th power of speed u Does not consider ablative wear or surface material properties
Oil Consumption
Oil Consumption Model u HDM-III only function of roughness u Recommended by NDLI to eliminate from HDM-4 u 1995 RUE Workshop indicated should be included u Model contains two components –Fuel use due to contamination –Fuel use due to operation
Heavy Vehicle Trailers
Modelling u User defines trailer to be associated with a towing vehicle u Trailer leads to higher mecahnistic forces u Use standard HDM-4 speed, fuel, tyre, capital co models
Maintenance and Repair Costs u Based on unpublished NZ study u Original research did not relate costs to roughness u Assumed linear increase of 20% between 3 and 7 IRI
Additional Costs Due to Speed Changes
Speed Change Cycle u Two principal components –Deceleration from initial to final speed –Acceleration from final to original (or other) speed u May include idling or travel at reduced speed u Important for work zones and other specific traffic interruptions
Speed Cycle Model u Used ARRB Polynomial Acceleration model u Time to accelerate and decelerate from NZ research
Example of Speed Cycle
Acceleration Profile
Model Development u Used NZVOC Model u Predicts additional fuel and time due to speed changes u Defined as: ADDCST = (DECCST + ACCCST) - UNICST u Costs calculated as function of initial and final speed for acceleration and deceleration by vehicle class
Models u Developed regression models for the additional time and additional fuel to accelerate/decelerate u Parameter values function of initial and final speed
Work Zones
Bias Due to Use of Means
Use of Means u HDM-III uses mean speeds in calculations u For non-linear functions (eg fuel, time) this leads to bias in results u 1995 RUE Workshop requested this be considered in HDM-4
Findings u Used simulation model to predict the fuel and time as function of COV u Bias for travel time less than 1% so recommended it be ignored u Fuel consumption bias more significant and data used to develop correction equation: FUELBIAS = COV COV^2
Time Versus Space Speeds
Speeds u Time Mean Speed - mean speed of all vehicles passing a point u Space Mean Speed - mean speed of all vehicles over a section over a time period u HDM predicts time speed but space speed is correct measure
Speed Corrections u Ran simulation to calculate space speeds as function of time speed and COV u Error generally less than 2% but since easy to correct for proposed equation: SPEEDBIAS = COV COV^2
Emissions
HDM-4 Model u Developed by VTI in Sweden u Conducted statistical analysis of emissions as function of fuel use u Developed simple linear model
Noise
HDM-4 Model u Proposed by NDLI to adopt UK CRTN model for HDM-4