Anna-Carin Brink, ARUP, Perth, Australia
Since the 1980’s billions of dollars have been invested in developing methods to make pavement design and construction more sustainable through the re-use of existing resources. This includes insitu recycling and re-use of existing pavement layers for flexible and rigid pavements. This technology has evolved over the years in recognised design processes captured in, for example: TG2 . Specifications have been developed by various entities such as the Roads and Maritime Services in New South Wales  and Transport and Main Roads in Queensland . WIRTGEN and BOMAG have also contributed by providing sophisticated insitu recycling equipment.
Yet, in this day and age, there are still Government Agencies writing in their project Scope and Performance Requirements (SPR) that: “where pavements are required to be lowered from their existing levels by more than 10 mm, provide new full depth pavement”. This goes hand-in-hand with the fact that there are Engineers who do not recognise the fact that, especially in the urban environment an existing road pavement is an asset that has to be maintained, it already contains a compacted/improved subgrade, imported subbase and/or basecourse materials and a relatively expensive wearing course.
The aim of this paper is not to promote insitu recycling as such, but to demonstrate that pavement rehabilitation design is not new pavement design. The compacted, consolidated subgrade underneath an existing pavement structure can have an insitu modulus that is higher than that of the subgrade in the adjacent “virgin” ground and should therefore not be capped to the same low value that may exist in the undeveloped subgrade during rehabilitation design. Analysis of FWD test results, correlated with visual observations and geotechnical testing, can display a clear distinction between a failed pavement structure in need of rehabilitation and an adequate structure that only requires milling and repaving of the existing asphalt wearing course.
The moduli back-calculated for granular pavement layers often exceed new pavement Guide values and must be capped during modelling. However, it should be recognised that over the years, asphalt overlays constructed during maintenance activities have protected the pavement structure (as intended), thereby creating perpetual pavement structures.
The above will be demonstrated in this paper by describing how to obtain relevant data to conduct the pavement investigation, analyse the results, calculate the pavement remaining life and perform the rehabilitation design. The data used for illustration purposes in this paper, was selected from physical test results obtained from in-service Arterial Road (AR) and Local Road (LR) pavements that are currently under rehabilitation construction.
The aim of pavement investigation is to obtain as comprehensive a picture as possible of the existing pavement structure and insitu conditions to determine its adequacy/inadequacy for pavement rehabilitation design purposes. This includes visual observations, a geotechnical investigation and FWD testing, as described below.
The easiest and often most valuable method to obtain information on a pavement condition is by conducting visual observations. Visual defects with photo and even video evidence should be recorded in a systematic way relative to a reference system.
There are various documents presenting comprehensive and detailed descriptions of the most common types of pavement distress and methods to record it, such as the Austroads Guide to Asset Management Part 5 Series of documents  and AGPT05-11 .
The crux of the matter is to get an accurate record of what the existing pavement looks like. Visual observation is the first step in demarcating pavement sections and selecting positions for test pits during the geotechnical investigation.
An experienced Pavement Engineer will be able to conduct an assessment while walking a particular road section and have a clear picture of what pavement rehabilitation measures are required, which will generally be confirmed by the geotechnical investigation and FWD test results.
The geotechnical investigation should focus on the top 1.5 m (relative to final design levels) of the pavement structure. Simply coring through the pavement structure, recording the type and thickness of every pavement layer and then sampling subgrade material for testing at a depth of 1.5-2.0 m, provides close to useless information.
Irrespective of whether it is a rigid or a flexible pavement being examined, the investigation should focus on recording the thickness of every pavement layer and also obtaining samples for testing of every layer in the profile, as follows:
- Asphalt modulus or concrete compressive strength from cores extracted from the pavement;
- Dynamic Cone Penetrometer (DCP) testing to a depth of 1.5 m (or refusal);
- Insitu density and moisture content, Atterberg limits, shrinkage, grading and CBR (4-day soaked) of every granular pavement layer; and
- Insitu density and moisture content, Atterberg limits, shrinkage, grading and CBR (4-day or 10-day soaked) of subgrade.
If necessary, the original asphalt mix design can be determined by reconstitution of the mix during laboratory testing. Thin slabs of asphalt, carefully removed across a pavement lane can also be used in rutting studies. More details as to what is required can be obtained from .
Intrusive investigations are usually limited in trafficked roads due to the high cost of managing safety risks associated with digging and reinstating test pits in an existing pavement; i.e. the high cost of accommodation of traffic while ensuring a safe environment for the road user as well as for the geotechnical investigation team. FWD testing has specifically been developed to minimise this risk as well as to supplement “destructive” testing methods.
FWD testing should be conducted at the following intervals:
- 25 m in the left and right wheel paths (alternate) of slow lanes;
- 50 m in the left and right wheel paths (alternate) of lanes other than the slow lanes; and
- Nominal 10 test points in the intersection with local roads to the extent covered by the construction works.
Analysis of Results
The information obtained during the pavement investigation all contribute to obtaining uniform pavement sections, developing pavement models for specific road sections, determining the remaining life of each and also superimposing the pavement rehabilitation measures in order to meet the required design life.
The method used to establish uniform pavement sections , build various pavement models and back-calculate pavement layer moduli to be used in further analyses are described in this chapter. Back-calculation can either be through a calculation “engine” (software programme) - or manually using Benchmarking Methodology . Of particular importance is the interpretation of the back-calculated granular layer moduli and choice of subgrade design CBR to be used when calculating the remaining life of the existing pavement.
It is assumed that appropriate measures will be implemented during design and construction with regards to pavement drainage, and that it is therefore not a factor to be considered in this paper.
Uniform Pavement Sections
The main reason for identifying uniform pavement sections is that the structure/capacity of the pavement can vary both longitudinally and transversely along the alignment.
In accordance with  homogeneous/uniform sub-sections should generally exceed 100 m in length and will be considered homogeneous if the deflection values have a coefficient of variation (CV) of 0.25 or less. A simple and effective method to numerically define uniform sections is using the AASHTO  cumulative difference (CD) approach. The maximum deflections measured for the AR and LR pavements are illustrated graphically in FIGURE 1 and FIGURE 2, respectively. Data for the left wheel path (LWP) and right wheel path (RWP) was plotted separately. The CD plots for the deflection results are also included on the Figures. A uniform section is based on a marked change in grade of the slope of the plot.
The relevant Design Equivalent Standard Axle (DESA) loading to be used in the rehabilitation design is also indicated on the Figures.
The FWD maximum deflections in the LWP and RWP of both road sections and the uniform pavement sections chosen using the AASHTO  method with the corresponding CV results are summarised in TABLE 1 and TABLE 2.
Most of the CV values calculated were less than 0.25. Where the CV results were higher than 0.25, it was generally due to limited results over a short distance with a large standard deviation in for example Section 2 on the LR (refer TABLE 2).
The pavement profiles obtained from the geotechnical investigation were used to compile models of the pavement structures. Pavement structures can overlap uniform pavement sections. The pavement models obtained for the AR and LR are shown in TABLE 3 and TABLE 4, respectively. The subbase layer thickness was capped to 300 mm for back-calculation purposes.
Back-calculation of Pavement Layer Moduli
Methods of Back-calculation
As there is no analytical solution to the back-calculation problem, a number of solution techniques have been devised:
- Table Lookup: Numerous deflection predictions for a range of layer thicknesses and modulus values have been pre-calculated. Back-calculation then involves a table search to find the “best fit” to the solution
- Global searching: All modulus values are treated simultaneously as unknowns, with iterations to achieve the “best fit” between the deflections predicted from the current set of modulus values and the measured deflections (e.g. EFROMD )
- Incremental searching: Similar to global searching but the modulus values are incrementally fixed which in theory improves the numerical stability of the problem. This works because, for example, for flexible pavements the deflections at 900 mm and greater offsets are most controlled by the subgrade modulus. (e.g. Rubicon Toolbox  and ELMOD )
- Other techniques relying on more mathematically sophisticated approaches such as neural networks
- A semi-mechanistic semi-empirical analysis technique developed by Horak and Emery  whereby deflection bowl parameters, measured with the FWD, are used in a relative benchmarking methodology. This method will be discussed briefly below
In practice, the best way to decide if a solution is sufficiently close is to visually compare the predicted and measured deflections. As this is part of the output in Rubicon, it is one of the reasons why it was selected.
All back-calculation methods rely on a calculation “engine” to predict the deflections for a given set of modulus values. For example, in  the “engine” is the CIRCLY32  software programme and in  and  it is the WESLEA program, as discussed in Paragraph 3.3.3.
It must be noted that back-calculation is an approximate method, which at best gives an indication of the likely stiffness behaviour of the different pavement zones, provided that the implied assumptions of linear elastic and lateral homogeneity broadly apply.
Manual Back-calculation using Benchmarking Methodology
This manual back-calculation methodology is based on deflection bowl descriptors or parameters and their correlations with structural layers or zones within the pavement structure. The basis for the methodology is described with the help of FIGURE 3 which shows that the deflection bowl measured under a loaded wheel can be described in terms of three distinct zones.
In Zone 1, close to the point of loading (300 mm radius), the deflection bowl has a positive curvature. Zone 2 is where the deflection bowl switches form a positive to a reverse curvature and is often referred to as the zone of inflection. The point of inflection depends on specific pavement layer compositional factors which will generally be from about 300-600 mm from the point of loading. Zone 3 is where the deflection bowl has switched to a reverse curvature and extends to the road surface (approximately 600-2,000 mm); i.e. where the deflection reverts back to zero.
The deflection bowl parameters are summarised in TABLE 5.
It should be noted that the benchmarking methodology utilises most of the deflection bowl test results and does not only focus on the maximum deflection and curvature as in . Furthermore, in  the “curvature” is simply the mathematical difference between the maximum deflection and the deflection at 200 mm; i.e. D0 - D200. The resulting “curvature” in  is therefore a linear difference, and not the actual curvature as calculated in .
TRH 12  uses a colour-coded rating system for visual condition survey classification of pavements. The behaviour state classification as expanded by  has been incorporated in  and the relative structural strength of zones of layers in the pavement structure can be linked to the visual condition rating system as shown in TABLE 6. Similar colour-coding will also be used to categorise back-calculated pavement layer moduli.
In the absence of sophisticated software packages, FWD test results can therefore be manually analysed in a simple and efficient manner.
Summary of Rubicon Characteristics
Rubicon  uses a search process which is similar to the way that manual back-calculation is performed, starting with the subgrade and outer deflections and working inwards towards the load point and upper layer moduli. The search process uses a numeric method to search for an optimal configuration that would maximize the fit of the bowl. The program has been in use internationally for more than 10 years.
The underlying model used in Rubicon is linear elastic, isotropic, with homogeneous layers extending laterally to infinity. The calculation engine is the WESLEA algorithm developed by the US Corps of Engineers which has been validated in many comparisons with other internationally known programs . WESLEA is also widely used in the United States and elsewhere.
Rubicon automatically includes an upper and lower subgrade layer, 1 200 mm and infinitely thick, respectively. The thickness of the upper subgrade is greater than the 200-400 mm recommended in . For back-calculation to be effective, it is recommended that the upper subgrade layer should be between 800 and 1,500 mm, unless evidence of a shallow subgrade exists, such as a bedrock layer close to the surface . This is in line with international best practice, where it has been found that a too thin upper subgrade does not correctly model the stiffness of the subgrade, resulting in increased pavement layer stiffnesses. This did not inhibit identification of areas of low strength subgrade.
For a given set of moduli values the predicted deflections obtained using Rubicon and EFROMD will depend on the accuracy of the calculation engine. As noted above, studies indicate that WESLEA and CIRCLY32 are of comparable accuracy. For confirmation, a comparison of the predicted deflections under a 750 kPa FWD load for the derived modulus values at a chainage along the project was chosen as shown in TABLE 7. The table shows acceptably close values, with the CIRCLY predictions between 5 and 7 microns less than the Rubicon predictions.
Back-analysis of FWD Deflections
Before conducting the Rubicon analysis, Pavement Layer Category Sets (see TABLE 8) were developed to fit the layer moduli data and visualise the condition of the various pavement layers. The Category Sets do not necessarily follow the limits proposed by , but were set wider for certain layers to enable automatic back-calculation by the software with minimal manual input.
Once the layer moduli were determined, the average, standard deviation and 10th percentile modulus values were calculated, as summarised in TABLE 9 and TABLE 10.
Wearing course moduli were corrected for temperature and speed afterwards. If the resulting 10th percentile modulus was less than 2,500 MPa, the layer was considered cracked and the modulus of 840 MPa (indicated in brackets) for the Sydney region which has a WMAPT of 28oC was used in the pavement models.
Basecourse and subbase moduli were capped at 500 MPa and 350 MPa, respectively, with the capped values indicated in brackets.
When conducting the back-calculations, the data was manipulated to obtain an average data fit error of less than 0.5% for every test. The 10th percentile back-calculated moduli for the models obtained in this way were then used in CIRCLY to calculate the remaining life of each model.
Colour-coding the pavement layer moduli in accordance with the Category Sets provides a visual aid to quantify the condition of the existing pavements. From TABLE 9 it is obvious that Section 1 of the AR is in a good condition and this road section would likely require milling and repaving of the asphalt wearing course only, while, depending on existing surface levels relative to final design levels, Sections 2 and 3 would need rehabilitation. Similarly, from TABLE 10 it is obvious that the LR wearing course, basecourse and subbase layers are in a poor condition, and that pavement rehabilitation will be required.
Another aspect of the analysis that needs mentioning is that it is important to obtain accurate pavement profiles during the geotechnical investigation, but that with this pavement rehabilitation design methodology the errors that could be made due to incorrect layer thicknesses during the back-calculation process are cancelled out by the remaining life analysis.
To explain: Each FWD deflection bowl contains a balanced set of results for a particular pavement structure. If a specific pavement layer is modelled too thin, the moduli will approximate towards the upper limit of the relevant category set, and vice versa. However, during the CIRCLY analysis to determine the remaining life of each road section, the same layer thicknesses and moduli will be used. Although a back-calculated modulus may be high because of a thin layer, in the remaining life analysis the thinner layer will make a lesser contribution to the remaining life of the structure despite a high modulus thereby retaining the balance and cancelling out errors.
Furthermore, note the Fair to Good moduli back-calculated for the upper subgrade as this is imperative to choosing the subgrade design CBR when determining the pavement remaining life and the rehabilitation design of each road section. Although calculated, the lower subgrade results are not used in remaining life calculations.
The choice of granular layer moduli and subgrade design CBR need further clarification and are described in Paragraphs 3.3.5 and 3.3.6, respectively.
Back-calculated Granular Layer Moduli
As mentioned, where the existing pavement is in a good condition, the moduli back-calculated for granular pavement layers underneath existing thick asphalt overlays exceeded the vertical modulus suggested for new pavements and had to be capped during modelling (refer TABLE 9). Tables 6.4 and 6.5 of  suggest reduced moduli for granular material overlain by bound material. However, “very good” basecourse and subbase moduli were obtained underneath relatively thick asphalt layers which also had “very good” moduli for all sections across various roads on the project. This is often the case where subsequent maintenance asphalt overlays have resulted in pavement structures approximating full depth asphalt.
On the other hand, pavement sections that were in a poor condition did have reduced basecourse moduli similar to those suggested in . This raises the question whether “failed” moduli are suggested for new pavement design purposes leading to unnecessarily expensive pavement designs, while the materials do not actually behave like that in practice.
Subgrade Design CBR
The subgrade design CBR was determined using a combination of methods, as follows:
- Soaked CBR tests (1 point or 4 points) for samples compacted to 100% of Maximum Dry density (MDD) under standard compaction of Optimum Moisture Content (OMC);
- Inferred from DCP test results;
- Inferred from FWD back-calculated moduli; and
- Presumptive subgrade design CBR values .
To put it in perspective; the DCP and FWD not only test the material in a stable and contained state, but it also has the advantage of reflecting the pavement system’s response; i.e. not a single separated layer material in isolation. Pavements respond as a system and the “pavement strength balance” shifts and settles with time.
It is well known that existing pavements have strength in depth due to consolidation and therefore a residual strength especially in the subgrade as most pavements fail in the top 300 mm, if not just on the surface. Furthermore, when considering the physical location/depth of the existing pavement subgrade, it is unlikely that it will require unearthing or exposure.
Samples excavated and tested from adjacent, Greenfields areas earmarked for new lanes and/or pavement widening where no subgrade improvement has been done yet, is also quite likely to give relatively low CBR results. However, this is not a problem. It is logical that the subgrade of the new pavement sections will be improved during construction and therefore brought to the same, already improved level as the existing pavement subgrade.
It is therefore incorrect to cap the subgrade CBR of the existing pavement to the same value as that obtained for the adjacent new pavement during rehabilitation design, or to a lesser value that may have been inferred from CBR testing. Pavement rehabilitation design is not new pavement design.
Pavement structures underlain by sandstone generally give results where all the methods used to determine the subgrade design CBR of existing and adjacent new pavement correlate well.
Once pavement layer moduli and subgrade design CBR were confirmed, the remaining life of the pavement sections was calculated as described in paragraph 3.3.7.
Pavement Remaining Life
The pavement models obtained through back-calculation of the FWD results were further analysed using the 10th percentile moduli in CIRCLY to quantify the remaining life of each uniform pavement section identified. Further manipulation of the layer moduli, limiting values to new pavement design moduli or sub-layering, was not considered warranted as the Rubicon back-calculated modulus is the average for a particular layer and not the top of the layer (the value normally used in CIRCLY).
It must be noted that in this instance, due to the fact that existing pavements structures had relatively thick asphalt wearing courses, using CIRCLY to determine the remaining life and conduct pavement rehabilitation designs was possible. However, as CIRCLY does not contain transfer functions to analyse granular materials, for granular pavements with thin bituminous wearing courses, the layered elastic tool in Rubicon Toolbox will be more appropriate.
The back-calculated moduli were used when determining the theoretical remaining life; summarised in TABLE 11.
A design life of 20 years was required for existing pavements. Road sections with a remaining life of less than 20 years therefore required rehabilitation. The pavement rehabilitation design method is described in Chapter 4.
Pavement Rehabilitation Design
In the urban environment there are constraints such as buildings with accesses, existing kerbing and footpaths, etc. which restrict pavement surface levels. Asphalt inlays or insitu recycling with the appropriate wearing course are generally considered.
The process to produce the pavement rehabilitation design for each section is relatively easy and is illustrated in TABLE 12 by comparing the original pavement models with the rehabilitation pavement models. If milling and repaving is required, the rehabilitation pavement model shows a reduced existing wearing course thickness with a 50 mm new asphalt wearing course. Where rehabilitation is required this can involve milling off the existing asphalt wearing course and even some of the existing basecourse to achieve a pavement structure that meets the requirements in terms of design traffic and life.
The LR sections were in such a poor condition that the rehabilitation design involved removal of 225 mm of existing pavement material. This comprised the entire wearing course, basecourse and a portion of the subbase. A further stipulation in the design was that the remaining granular material had to be ripped and recompacted to a depth of at least 150 mm before paving the asphalt base and wearing courses. The LR was a perfect candidate for insitu recycling which would have avoided removal of a substantial quantity of material. Even though the crossfall had to be corrected and some areas needed regrading, obtaining a workable “mix design” meeting  requirements would have been possible.
Note that the existing subgrade was not disturbed in any of the rehabilitation pavement models which emphasises the fact that the back-calculated upper subgrade modulus should be used in the design and need not be capped to the modulus inferred from CBR test results obtained on disturbed samples reconstituted in the laboratory, or from the adjacent new subgrade.
As mentioned, the rehabilitation pavement models in TABLE 12 are for the situation where final road levels will remain close to existing. However, cut or fill can be modelled in the same way, and in some instances, considering the design traffic, a cut of up to 100 mm can be accommodated while still achieving the design life of the pavement.
The preceding paper discusses what is required from the pavement investigation and analysis of results to perform a pavement rehabilitation design.
It demonstrates that the moduli back-calculated for granular basecourse and subbase layers underneath thick asphalt wearing courses for road sections that are visually in a good condition, far exceed the moduli suggested to be used for the design of new pavements. Pavement sections that were in a poor condition did have reduced basecourse moduli similar to suggested values. This raises the question whether “failed” moduli are suggested for new pavement design purposes leading to unnecessarily expensive pavement designs, while the materials do not actually behave like that in practice.
Furthermore, the confined subgrade beneath an existing pavement, even though it may consist of relatively poor quality material can have a modulus that is higher than the modulus inferred from CBR test results obtained on disturbed samples reconstituted in the laboratory, or from the adjacent new subgrade.
In the context of pavement design reliability, the probability concepts applied in estimating traffic loading and uncontrolled environmental conditions, Pavement Design is not such an exact science that a 10 mm cut in final levels will “break” the pavement. The requirement by some to provide new pavements where pavements are required to be lowered from their existing levels by more than 10 mm is unnecessary and unsustainable.
Pavement rehabilitation design is not new pavement design; it requires experience, knowledge of the materials that pavements are constructed with, the understanding of how those materials behave in service, under various loading and environmental conditions, and sometimes just common sense.
Acknowledgement: This article has been reproduced from ‘AAPA International Flexible Pavements Conference, Sydney, New South Wales, Australia’ with due permission of author.
- Technical Guideline (TG2): Bitumen Stabilised Materials – A Guideline for the Design and Construction of Bitumen Emulsion and Foamed Bitumen Stabilised Materials. 2009.
- RMS QA Specification R75. Insitu Pavement Stabilisation using Slow Setting Binders. 2015.
- RMS QA Specification R76. Insitu Pavement Stabilisation using Foamed Bitumen. 2018.
- Queensland Department of Transport and Main Roads Specification MRTS35 Recycled Materials for Pavements. 2017.
- Guide to Asset Management Part 5A-5H (AGAM05A-H), Austroads, Sydney. 2007-2009.
- Guide to Pavement Technology Part 5: Pavement Evaluation and Treatment Design (AGPT05-11), Austroads, Sydney. 2011.
- Guide to Pavement Technology Part 2: Pavement Structural Design (AGPT02-17), Austroads, Sydney. 2017.
- American Association of State Highway and Transportation Officials (AASHTO) Guide for Design of Pavement Structures. Appendix J: Analysis Unit Delineation by Cumulative Differences. American Association of State Highway and Transportation Officials, Washington D.C. 1993.
- Vuong, B.T. EFROMD2 User’s Manual: A Computer-based Program for Back-calculating Elastic Material Properties from Pavement Deflection Bowls, Version 1, Australian Road Research Board, Vermont South, Victoria. 1991.
- Rubicon Toolbox, Tools for Pavement Design and Analysis Version 3.1.0, Modelling and Analysis Systems, Rubicon Solutions, Cullinan, South Africa.
- ELMOD Software for Pavement Analysis, Dynatest North America, California.
- Horak, E. and Emery, S. Falling Weight Deflectometer Bowl Parameters as Analysis Tool for Pavement Structural Evaluations. Proceedings of the 22nd Australian Road Research Board Conference, Canberra, Australia. 2006.
- CIRCLY – Version 7.0, Mincad Systems. 2018.
- Committee of State Road Authorities (CSRA). Flexible Pavements Rehabilitation Investigation and Design. Draft Technical Recommendations for Highways 12 (TRH 12), department of Transport, Pretoria, South Africa. 1997.
- Maina, J.W. Denneman, E. and De Beer, M. Introduction of New Road Pavement Response Modelling Software by Means of Benchmarking. Council for Scientific and Industrial Research, Pretoria, South Africa. 2008.
- Long, F. Personal Communication. Rubicon Solutions, Cullinan, South Africa. 2016.
- Supplement to Austroads Guide to Pavement Technology Part 2: Pavement Structural Design (AGPT02), RMS NSW, Sydney, Australia. 2015.