Table 2: Comparison of Pavement Life Predictions

The major reason why raw comparisons of the kind made in Table 2 are invalid is that each design method uses its own representation of the pavement structure (called a pavement model) to compute the strains and to derive the subgrade failure criterion.

The method by which the AUSTROADS rutting criterion was derived will be described to illustrate the process. Neither measurements of rut depths nor measurements of subgrade strains was involved. For many years many unbound road pavements had been designed in Australia using the empirical CBR design chart shown in Figure 1, which reproduces Figure 8.4 of the AUSTROADS design guide. A survey of users indicated that roads designed using the chart appeared to perform satisfactorily in about 80 to 90% of cases. It was decided, therefore, to accept Figure 8.4 as a record of pavement performance and to derive a rutting criterion from it. The origins of Figure 8.4, and the method used in the desk-top study to produce the rutting criterion have been detailed by Jameson (1996).

Briefly, twenty five ‘test pavements’ were selected from Figure 8.4. For example, a 600mm thick pavement constructed on a CBR 3 subgrade and which sustained 4 million ESAs could be taken as one pavement for analysis. The pavements were selected to span a wide range of subgrade strengths and traffic levels as represented by repetitions of standard axles. The pavement layers were then divided into sub-layers, and modulus values assigned to the sublayers and subgrade in accordance with the systematic procedure detailed in the AUSTROADS design guide. For example, the subgrade vertical moduli were assumed to be 10 times the CBR in MPa.

Figure 1: AUSTROADS design chart for granular pavements
with thin bituminous surfacing

The maximum vertical compressive strain at subgrade level caused by a standard axle was then calculated for each of the twenty five pavement structures using CIRCLY. The strains were plotted against repetitions. The failure criterion is simply the equation of the line of best fit to the plotted points.

Had a different system of modulus assignment been used, the AUSTROADS rutting criterion would have been different. Each failure criterion listed in Table 1 is associated with its particular system of assigning stiffnesses to the pavement layers. For example, both Woodman and Wardle/Rodway used a system devised by the Corps of Engineers (Barker and Brabson, 1975) which is very different from that adopted by AUSTROADS. It is essential that the system used to establish the failure criterion is also used when analysing or designing pavement structures. For example, if the subgrade modulus had been assumed to be 12 CBR in deriving the failure criterion, the designer must also use 12 CBR.

But if the rutting criterion from one method is used in conjunction with the layering system of another method, the design outcome is totally invalid. The vital empirical link between the design and the performance data obtained from test pavements or in-service roads is broken.

Returning now to the raw comparisons of rutting criteria made in Table 2; some of the different rutting criteria may produce similar design thicknesses provided their own pavement models are used to calculate the load-induced subgrade strain values.

This does not mean, however, that the material properties within the pavement model do not matter provided they are used with the correct criterion. The assigned modulus values must also be realistic in the sense that they must reflect the true stiffnesses of pavement materials. Unless the models are realistic, the design method cannot be used to analyse or design pavement structures that contain materials that are different from those used in the test pavements.

In addition to pavement model inconsistency, there are other impediments to directly comparing failure criteria. These are discussed in the following sections.

Subgrade strain repetitions per vehicle

Each design method has within it a procedure for converting the number of vehicle passes along the pavement to repetitions of maximum subgrade strain. The number of strain repetitions per pass depends upon tyre width, the arrangement of tyres, the degree of vehicle wander and the method used to quantify the effect of the wander. Successive vehicles moving along a wide pavement such as an aircraft runway or a container-loading area do not follow the same path. The degree of channelisation will be much less than that of roads where vehicles tend to move in narrow marked lanes and vehicle widths are relatively uniform. The method used to handle wander will affect the failure criterion that is developed from performance data. Consequently the same method must be used when using the failure criterion to design pavements.

The Wardle/Rodway rutting criterion shown in Table 1 is utilised by the pavement design computer program APSDS (Airport Pavement Structural Design System). There is a fundamental difference between APSDS and all the other pavement design methods listed in the table. These methods calculate only the maximum value of subgrade strain caused by the vehicle and empirically relate this to failure repetitions by a process of calibration, as described in the AUSTROADS example given earlier in this paper. APSDS calculates the subgrade strain at all points across the pavement in order to capture all the damage contributions of all the vehicle wheels in all their wandering positions. The calibration procedure becomes more complex and is detailed in Wardle and Rodway (1995, 1998). For the same performance data, the failure criterion obtained in this way will be different from that obtained from a simpler calibration based on single strains. The primary reason for APSDS’ different approach is that it enables the designer to nominate the degree of vehicle wander that is appropriate to the particular pavement being designed. For example, the standard deviation of wander of large aircraft from taxiway centreline is typically taken as 780 mm whereas that for aircraft parking bays is approximately 200 mm. These differences in the degree of chanelisation produce significantly different design pavement thicknesses.

Design reliability

In the context of the current discussion, a design reliability of 50% means that pavements designed using the failure criterion are as likely to last for more than the design repetitions as they are to fail before all the design repetitions have been applied. The inclusion in Table 1 of the Shell failure criterion at two levels of reliability, 50% and 95%, illustrates that criteria can only meaningfully be compared if their degrees of reliability are the same.

Pavement condition at end of design life

‘Failure’ means the degree of rutting or pavement roughness that was deemed to be operationally unacceptable on the test pavements or roads used to develop the failure criterion. This must also be the pavement condition at the end of its design life. Several methods have been used to define this condition, including maximum rut depth, height of heave adjacent to the rut, measures of rideability by test vehicles, and various indices such as the Pavement Serviceability Index. The PSI was utilised in the AASHO Road Test. Direct comparisons between failure criteria are only possible if the end of life pavement conditions are similar.

USE OF ROAD DESIGN METHODS FOR HEAVY DUTY PAVEMENTS

It has been reasoned by Woodman (1992) and others that a rutting criterion based on road pavement performance will not be fully applicable to the design of industrial and airfield pavements that cater for much heavier wheel loads. It is argued that the relationship between subgrade vertical strain and permanent deformation of the pavement surface is not independent of wheel load. The reasoning is as follows.

Pavement design methods are commonly based on controlling permanent surface deformation (rutting) by limiting the elastic strain at the top of the subgrade to empirically-determined allowable levels. This approach relies on two concepts. Firstly it is assumed that most surface deformation comes from the subgrade rather than from within the overlying pavement layers. Secondly it is assumed that the subgrade material’s plastic (permanent) deformation is related to the magnitude of its elastic (temporary or recoverable) strain. Given these assumptions, the integral of the elastic strain with depth into the subgrade should provide a measure of the surface rutting that occurs.

The concept is illustrated in Figure 2. The CIRCLY-calculated vertical strains beneath a 20 tonne and a 4 tonne wheel have been plotted against depth. Although not important to the argument, both tyre pressures are 1 MPa. The heavier wheel represents those typically used on large aircraft, mine-haul vehicles and industrial equipment of the kind used on docks and container terminals. The 4 tonne wheel represents the heaviest road-truck wheel. The two pavement thicknesses have been chosen so that both tyres produce the same vertical compressive strain, 0.008, at subgrade level.

It can be seen that the vertical strains within the subgrade are greater in the case of the heavier wheel. The permanent deformation at the pavement surface due to the heavier wheel will be greater by an amount related to the area between the two curves. That is, equal compressive strains at the top of the subgrade should not produce equal surface rutting.

Figure 2: Effect of wheel load on pavement rutting for equal subgrade strains

Figure 3: Effect of wheel load on the shape of the load pulse at subgrade level

In addition to the above argument, Figure 3 shows that the shape and extent of the strain pulses at the top of subgrade are very different for the two loads. This provides further reason to doubt that pavement rutting should depend only on the maximum value of subgrade strain. The question of the effect of the shape of the load pulse upon pavement damage is beyond the scope of this paper but has been discussed elsewhere (Rodway, 1995).

Limitation of empirical methods

All of the empirically-derived subgrade failure criteria listed in Table 1 state that rutting depends only on the strain at the top of the subgrade, at least for the range of performance data used to develop each criterion. A fundamental limitation of any empirical method is that extrapolation beyond the data base used to calibrate the method cannot be made with confidence. This is especially true in the case of pavement design because the pavement model is acknowledged to be a much simplified representation of the real pavement structure. Also the design method does not purport to deal comprehensively with the actual process of rut formation pavement that is occurring; subgrade strain is simply accepted as an indicator of the rate at which rutting develops. Therefore extrapolations to different pavement materials, larger wheel loads, different wheel arrangements and thicker pavements will always be problematic.

A major extrapolation is involved in using a criterion based on performance under 4 tonne loads to design pavements loaded with 20 tonne wheels. The argument outlined above is not claimed to be conclusive. However it does suggest that, for the design of heavy duty and industrial pavements it is preferable to use design procedures that were developed from relevant performance data. Full-scale test pavements have been used to develop aircraft pavement design methods to cater for high wheel loads, deep pavement structures, a variety of wheel arrangements and large and variable degrees of vehicle wander. Further test pavements are currently under construction in the United States. These methods would seem to be more appropriate than the road design methods that have usually been used to design pavements for industrial loadings.

Pavement designers have many layered elastic design methods available to them. Each method contains an empirical failure criterion that has been derived by calibrating the method against pavement performance data. These criteria cannot, however, be directly compared with each other in isolation from their design methods.

The use of one failure criterion in conjunction with a design method other than its own will produce an invalid design.

A fundamental limitation of any empirical design method is that extrapolation beyond the data base used to calibrate the method cannot be made with confidence. This is especially true in the case of pavement design. Pavement design methods that have been developed for highway loadings will not be fully applicable for the much higher wheel loads used on airfields, docks, container terminals and mine haul roads. Methods based on pavement performance data that is more relevant in terms of loads and loading patterns should be used if they are available.

REFERENCES

AUSTROADS (1992). A guide to the structural design of road pavements. Austroads, Sydney.

Barker, W. and Brabston, W. (1975). Development of a structural design procedure for flexible airport pavements. Report No. S-75-17. US Army Corps of Engineers, Waterways Experiment Station, Vicksburg, Miss.

Brown, S. and Brunton, J. (1984). Improvements to pavement subgrade strain criterion. Journal Trans. Eng., ASCE Vol. 110, No 6.

Jameson, G.W. (1996). Origins of AUSTROADS design procedures for granular pavements. ARRB Transport Research Report ARR292, Melbourne.

Knapton, J. (1983). The structural design of heavy duty pavements for ports and other industries. Second edition. British Ports Association. London, UK.

Road Research Laboratory (1965). A guide to the structural design of flexibly and rigid pavements for new roads. Ministry of Transport Road Note 29. London.

Rodway, B. (1995). Design of flexible pavements for large multi-wheeled aircraft. Second International Conference on Road and Airfield Pavement Technology, Singapore.

Rodway, B. (1997). Going ‘round in Circlies. Australian Geomechanics Society Pavement Symposium. Sydney

Shell International Petroleum Company (1985). Addendum to the Shell pavement design manual. London.

Wardle, L.J. (1996). CIRCLY Users’ Manual, Version 3.0, MINCAD Systems Pty Ltd, Richmond, Australia.

Wardle, L.J. and Rodway, B. (1995). Development and application of an improved airport pavement design method. ASCE 1995 Transportation Congress, San Diego.

Wardle, L.J. and Rodway, B. (1998). Recent developments in flexible aircraft pavement design using the layered elastic method. Third Int. Conf. on Road and Airfield Pavement Technology, Beijing.

Woodman, G. (1992). Failure criteria for flexible pavements. PSA report for BAA Technical Services Division. Surrey, UK.