ToolBox, Decision Making Tool for Optimising …...Deliverable Nr 3 – ToolBox, Decision Making Tool for Optimising Lengths for Maintenance Due date of deliverable: 30.09.2013 Actual

This project was initiated by ERA-NET ROAD.

ToolBox

ToolBox, Decision Making Tool for Optimising Lengths for

Maintenance Deliverable Nr 3

September 2013

VTI

TRL

AIT

CETE

WSP

Review of Function Triggers, September 2013

Page 2 of 43

Project Nr. 832704

Project acronym: ToolBox

Project title:

ToolBox, selection of maintenance candidates

Deliverable Nr 3 – ToolBox, Decision Making Tool for Optimising Lengths for Maintenance

Due date of deliverable: 30.09.2013

Actual submission date: 04.10.2013

Start date of project: 20.10.2011 End date of project: 30.10.2013

Author(s) this deliverable:

Emma Benbow, TRL, UK

Mark Harrington, TRL, UK

Alex Wright, TRL, UK

Contributors:

Johan Lang, WSP

Leif Sjögren, VTI, SWE

Mohamed Bouteldja, CETE of Lyon, FR

Veronique Cerezo, IFSTTAR, FR

Version: draft 01

ToolBox, Deliverable 3

Page 3 of 43


Page 4 of 43

Executive summary

Toolbox is one project in the ERA-NET ROAD programme. “ERA-NET ROAD – Coordination and Implementation of Road Research in Europe” was a Coordination Action funded by the 6th Framework Programme of the EC. The partner countries in ToolBox are Austria, France, Sweden and UK.

Pavement managers deal with complex decisions when identifying lengths of their networks in need of maintenance and planning the appropriate maintenance treatments. Currently, they are heavily dependent on experience, even though many support systems exist, such as guidelines, monitoring and information systems.

This project aims to advance the development and implementation of practical strategies and tools to assist road authorities in optimising the maintenance of their road networks, whilst still addressing the key interests, new challenges and expectations of road users.

ToolBox aims at developing a “concept for proper maintenance planning” to assure the selection of adequate maintenance works (“schemes” or “objects”) to make effective use of the maintenance budget, based on available road condition data, to give minimal negative effects on road users, safety for road workers and the environment.

This report covers Work Package 3 of the ToolBox project, where a decision making tool has been developed in Microsoft Excel. This tool will be used within Workpackage 4, to demonstrate a new and innovative way for optimising lengths for maintenance.

The ToolBox tool firstly groups lengths of a similar condition into schemes, suggests what maintainance might be needed in order to return these schemes to good condition and can then be used to perform whole life costing on these schemes. Functional triggers are used to defined the overall condition of individual lengths and these are described in Chapter 2 of this report. These functional triggers cover comfort, durability, safety and environment.

Data requirements for Toolbox are discussed in Chapter 3, along with whether the data is generally available and how missing, or insufficient data could be dealt with. Chapter 4 contains a description of the spread sheet tool, developed for ToolBox, including the method used to select maintenance candidates. How treatments are then chosen for these maintenance candidates is also described in this chapter. The Life Cycle Cost Analyses used to prioritise candidates are then discussed and described in Chapter 5.

The goal of ToolBox is to have a working concept that can be demonstrated in the end of the project. If this will be successful it is possible to, in the future, improve the models or even add new models to the concept.


Page 5 of 43

List of Tables

Table 1: Example Comfort thresholds and functions that may be used in Demonstration ..... 13

Table 2: Example Durability thresholds that may be used in Demonstration ......................... 15

Table 3: Example Environment thresholds that may be used in Demonstration .................... 17

Table 4: Required input parameters to ToolBox ..................................................................... 20

Table 5: Parameters available, from partner countries .......................................................... 21

Table 6: Parameters used in the definition of the ALERTINFRA warnings ............................ 23

Table 7: Percentage of length with warnings in curves and straight lines .............................. 24

Table 8: Percentage of length with warnings in curves .......................................................... 26

Table 9: Percentage of length with warnings in straight lines ................................................ 26

Table 10: Warning availability with data measured every 100m ............................................ 27

Table 11: Example of scheme selection algorithm ................................................................. 32

Table 12: Treatments applied to road networks (taken from 2013R08EN) ............................ 33

List of Figures

Figure 1 Overview of process to enable selection of maintenance candidates (* the weight given to each functional trigger will be determined by individual road owners’ policies) . 11

Figure 2: Number of 50m-length segments in curves without warning .................................. 24

Figure 3: Number of 50m-length segments in straight lines without warning ......................... 24

Figure 4: number of Xm-length segments in curves without warning on the measurement path ................................................................................................................................. 26

Figure 5: number of Xm-length segments in curves without warning depending on the measurement path .......................................................................................................... 26

Figure 6: Screen shot of the tool, where the required data sampling interval is chosen ........ 28

Figure 7: Screen of the tool where the sampling interval of the original input data is provided ........................................................................................................................................ 28

Figure 8: Layout of sheets within Excel based maintenance candidate selection tool. .......... 30

Figure 9: Screen shot of part of the data sheet, showing columns for parameter value input 30

Figure 10: Screen shot of part of the data sheet showing columns containing calculated trigger values ................................................................................................................... 31

Figure 11: Accessing the ToolBox add-in ............................................................................... 31

Figure 12: Flow diagram of treatment choice ......................................................................... 34

Figure 13: Summary sheet provided by the tool. .................................................................... 36

Figure 14: Obtaining data from specific schemes .................................................................. 36


Page 6 of 43

Figure 15 ISO 15686 definition of WLC and LCC (ISO 2007). ............................................... 37

Figure 16 Illustration of the challenges in reflecting the life cycle consequences of doing maintenance later on one road section when having several technical parameters (TP), conditions indicators (CI, red and blue), maintenance criteria (blue TP1 and TP2) and maintenance strategies/options (A and B (red and blue cont and dash)). ...................... 39

Figure 17 Illustration of the technical parameter values used to reflect the effect of each maintenance treatments on service life and condition after treatment. ........................... 41


Page 7 of 43

Table of content

Executive summary .................................................................................................................. 4

List of Tables ............................................................................................................................ 5

List of Figures ........................................................................................................................... 5

Table of content ........................................................................................................................ 7

1 Introduction ....................................................................................................................... 9

1.1 Background ................................................................................................................ 9

1.2 The project ToolBox ................................................................................................... 9

1.3 Overview of this report ............................................................................................. 10

1.4 Work Package 3 ....................................................................................................... 10

2 Functional Triggers in ToolBox ....................................................................................... 11

2.1 Comfort .................................................................................................................... 12

2.1.1 Comfort trigger ..................................................................................................... 12

2.1.2 Comfort trigger defined for use on the demonstration network ............................ 12

2.2 Durability .................................................................................................................. 13

2.2.1 Durability trigger ................................................................................................... 13

2.2.2 Durability trigger defined for use on the demonstration network .......................... 15

2.3 Safety ....................................................................................................................... 15

2.3.1 Safety trigger ........................................................................................................ 15

2.3.2 Safety trigger defined for use on the demonstration network ............................... 16

2.4 Environment ............................................................................................................. 16

2.4.1 Environment trigger .............................................................................................. 16

2.4.2 Environment trigger defined for use on the demonstration network ..................... 17

2.5 Combined Condition Indicator .................................................................................. 17

3 ToolBox Data .................................................................................................................. 19

3.1 Data needed and available for ToolBox ...................................................................... 19

3.2 Dealing with unavailable or insufficient data ................................................................ 22

3.2.1 Unavailable data for Comfort trigger .................................................................... 22

3.2.2 Unavailable data for Durability trigger .................................................................. 22

3.2.3 Unavailable data for Safety trigger ....................................................................... 22

3.2.3.1 Effect of unavailable parameters ...................................................................... 23

3.2.3.2 Effect of data spacing ....................................................................................... 25

3.2.3.3 Using ALERTINFRA in ToolBox ....................................................................... 27

3.2.4 Unavailable data for Environment trigger ............................................................. 28


Page 8 of 43

3.2.4.1 Noise index ....................................................................................................... 28

3.2.4.2 Fuel consumption ............................................................................................. 29

4 Maintenance Candidates ................................................................................................ 30

4.1 What the tool looks like ................................................................................................ 30

4.2 How the tool selects maintenance candidates ............................................................ 31

4.3 Treatments .................................................................................................................. 33

4.4 Outputs from the tool ................................................................................................... 35

5 Optimising LCCA and user expectations ........................................................................ 37

6 Next steps in ToolBox ..................................................................................................... 42

Sources .................................................................................................................................. 43


Page 9 of 43

1 Introduction

1.1 Background “ERA-NET ROAD II – Coordination and Implementation of Road Research in Europe” was a Coordination Action funded by the 7th Framework Programme of the EC. The funding partners in ERA-NET ROAD (ENR) are the National road Administrations (NRA) of Belgium, Germany, Denmark, Finland, France, Netherlands, Norway, Sweden, Slovenia and United Kingdom (www.road-era.net). Within the framework of ENR this joint research project, ToolBox was initiated.

Pavement managers deal with complex decisions when identifying lengths of their networks in need of maintenance and planning appropriate maintenance treatments. Currently, they are heavily dependent on experience, even though many support systems exist, such as guidelines, monitoring and information systems. Decades of road network monitoring and follow up projects, such as the Long Term Pavement Performance (LTPP) program, have generated a huge volume of empirical data on pavement condition, how this develops and how this affects road users that can be made useful if analysed in a structured and sound manner. To complement this information, several decades of research and development has accumulated a substantial volume of knowledge, models and tools that can use the information to assist in maintenance decisions, with an aim to assist in applying a strategy to deliver a sound road network with minimum cost (although most tools do not address user expectations). Even so, these support tools are not yet implemented to their full potential. The question arises - “Why?”

1.2 The project ToolBox The aim of the ToolBox project is to advance the development and implementation of practical strategies and tools to assist road authorities in optimising the maintenance of their road networks, whilst still addressing the key interests and expectations of road users.

ToolBox will develop a “concept for proper maintenance planning” to ensure the selection of adequate maintenance works (“schemes” or “projects”) to make effective use of the maintenance budget, based on available road condition data, and giving minimal negative effects on road users and safety for road workers.

ToolBox will develop a clear understanding of the concepts applied in the selection of lengths for maintenance (candidates), linked to comfort, safety, durability and the environment, including how the data is used, combined and weighted within current decision tools and models. Here, ToolBox will not aim to develop new models, but will identify and extract key tools from existing models used in Europe. Also, the work will develop an understanding of how existing knowledge (data) can be used to account for road user expectations in the selection of object lengths for maintenance. ToolBox will then take these existing and new concepts to establish a set of functional triggers for selecting lengths (candidates) for maintenance on the network that include road user expectations and combine them to make recommended prioritised treatment objects.

ToolBox will demonstrate the application of the concepts developed within the project via a prototype tool applied to a sample test network, to compare and contrast the approach proposed by the ToolBox tool with the approach proposed by current systems.

ToolBox will deliver its objectives via five work packages. The core activities are summarised in the following paragraph to show how they link together within the project. Although the


Page 10 of 43

work packages are led by a leader from one of the partners, the work will be done in a cooperative manner with close contact between partners.

The first Work Package (WP) will review and specify the current situation regarding the frameworks, tools and models used in current Pavement Management Systems (PMS). The second WP will adapt selected models to fit the ToolBox principle. This means specifying the necessary data and finding a common base for (at least) all partner countries. Since the focus is to develop a working framework, the third WP will commence the integration of weighting factors and functional triggers, and the selection of prioritisation models. Considerations of maintenance strategies and treatment methods will take place in this WP and a life cycle perspective included in the final results. Finally, a demonstration will be carried out on selected networks from the partner countries in the fourth work package. It is planned that this demonstration will be done using data collected during routine surveys.

The first two work packages have been completed, each with its own deliverable. This report is a summary of the progress achieved within Work Package 3 (WP3) and is the third deliverable of ToolBox. Two more deliverables are planned, one demonstration activity and one report (a final summary report).

1.3 Overview of this report This report presents the decision making tool for optimising lengths for maintenance, which has been produced within Work Package 3.

In Chapter 2 of this report, a summary is given of the models that will be used to obtain Condition Triggers and the Combined Condition Indicator, along with the thresholds, weights and functions that will be used for the demonstration.

Data requirements for ToolBox are discussed in Chapter 3, along with whether the data is generally available and how missing, or insufficient data could be dealt with.

Chapter 4 contains a description of the spread sheet tool, developed for ToolBox, including the method used to select maintenance candidates. How treatments are then chosen for these maintenance candidates is also described in this chapter

The Life Cycle Cost Analyses used to prioritise candidates are then discussed and described in Chapter 5.

The major steps remaining are testing and tuning of the models, separately, and then in combination, and then to demonstrate the use of the tool. The results from this will be validated and compared using a cost/benefit process. These steps are presented in Chapter 6.

1.4 Work Package 3 The main goal of ToolBox WP3 is to make a working framework (in Excel) that can select maintenance candidates by using higher level indicators as input, such as safety, durability, comfort etc. instead of only using technical parameters such as rut depth and IRI (International Roughness Index), as is the common approach currently. An overview of this process is shown in Figure 1.

To summarise: For each 100m length, technical parameters will be combined into functional triggers (using the models described in Chapter 2). These in turn will be combined with user-defined weightings, to form a Combined Condition Index. This Combined Condition Index will then be used by the spread sheet to identify potential maintenance candidates. The individual triggers, indices and parameters are then used to identify potential treatments for each maintenance candidate.


Page 11 of 43

These maintenance candidates can then be prioritised, by the user, based on life cycle cost analyses, described in Chapter 5.

Figure 1 Overview of process to enable selection of maintenance candidates (* the weight given to each functional trigger will be determined by individual road owners’ policies)

2 Functional Triggers in ToolBox

In an objective system, the choice of maintenance candidates must rely on the availability of quantitative condition information, which is typically provided in the form of condition data expressed as technical parameters. Within WP2 of the ToolBox project, the aim was to establish a set of functional triggers for selecting lengths for maintenance on the network and there was therefore a need to convert technical parameters into a single value for each of the triggers considered for ToolBox: Safety, Durability, Environment and Comfort. The ToolBox project has only used existing data, and did not aim to develop new data. It has also been assumed that the condition data provided is of good quality (even if it is known that it is not).

A review was carried out within WP2, to determine whether models have already been developed that combine parameters to obtain the ToolBox triggers. Where such models were found, a description of the model was given in D2 (Benbow & Sjögren, 2013).

Where such models were not identified, the following process was undertaken:

Identify which parameters should be included in the trigger;

Create an index for each parameter, to determine whether the road is in good, moderate or poor condition with respect to that parameter;

Combine each of the relevant parameter indices to form the trigger.

The minimum section length to be used is 100m and all triggers (whether based on existing models or developed in WP2) will have values of between 0 and 100 inclusive, where 0 means that the road is in good condition and 100 means that maintenance is needed.

In the future, new models could be added or the selected models could be improved.


Page 12 of 43

2.1 Comfort

2.1.1 Comfort trigger No model, that utilized all parameters influencing comfort, was identified during the review and therefore a model was suggested (Benbow & Sjögren, 2013). This takes the form:

a1*GC(IRINS) + a2*GC(IRIOS) + a3*GC(LRNS) + a4*GC(LROS)

+ a5*GC(RutNS) + a6*GC(RutOS) + a7*GC(Edge) Equation 1

This trigger will have a value between 0 and 100 (inclusive), and IRINS is the IRI measured in the nearside wheelpath (closest to the edge of the road), IRIOS is the IRI measured in the offside wheelpath (closest to the middle of the road); LRNS and LROS are the Localised Roughness parameters calculated in the nearside and offside wheelpaths, respectively; RutNS and RutOS are the rut depths calculated in the nearside and offside wheelpaths, respectively and Edge is the edge roughness parameter. The Gc are indices based on each parameter, whilst a1,…,a7 are the weights given to the indices.

GC(IRI) = 0 if IRI ≤ TLIRI mm/m

fIRI(IRI) if TLIRI mm/m < IRI < TUIRI mm/m Equation 2

100 if IRI ≥ TUIRI mm/m

GC(LR) = 0, if there is no localised roughness in the 100m length

fLR(localised roughness) if there is moderate localised roughness Equation 3

100, if there is a lot of localised roughness present

GC(Rut) = 0, if Rut depth ≤ TLRut mm

fRut(Rut depth) if TLRut mm < Rut depth < TURut mm Equation 4

100, if Rut depth ≥ TURut mm

GC(Edge) = 0, if the edge of the road is smooth

fEdge(Edge roughness), if the edge is moderately rough Equation 5

100, if the edge is very rough

fIRI, fLR, fRut and fEdge are functions that monotonically increase in value between 0 and 100 and the TL and TU are the lower and upper technical limits for each of the parameters.

2.1.2 Comfort trigger defined for use on the demonstration network The weighting factors to be used for the comfort trigger for the demonstration network are as follows:

Comfort trigger = 0.2*GC(IRINS) + 0.2*GC(IRIOS) + 0.1*GC(LRNS) + 0.1*GC(LROS) +

0.1*GC(RutNS) + 0.1*GC(RutOS) + 0.2*GC(Edge) Equation 6

Most weight is given to general ride quality, with all other parameters contributing equally to the trigger.

Table 1 shows the thresholds and functions that may be used for the indices, GC (defined in section 2.1.1) that will be applied to the data in the Demonstration (Note: These may change


Page 13 of 43

when the demonstration is implemented).

Table 1: Example Comfort thresholds and functions that may be used in Demonstration

Parameter TL TU f

IRI* 3 mm/m 4.5 mm/m fIRI(IRI)=100*(IRI-3)/1.5

Localised Roughness 0 1 N/A

Rutting 10 mm 20 mm fRut(Rut depth)=10*(Rut depth -10)

Edge roughness

SWE 20mm 40mm fER(edge roughness)= 5*(edge roughness-20)

UK 0.2 0.7 fER(edge roughness)= 200*(edge roughness-0.2)

2.2 Durability

2.2.1 Durability trigger No model, that utilized all parameters influencing durability, was identified during the review and therefore a model was suggested (Benbow & Sjögren, 2013). This takes the form:

b1*ISI + b2*IDef + b3*IRQ + b4*ISurf Equation 7

Where ISI is an index for structural strength

IDef is an index for pavement deformation

IRQ is an index for ride quality

ISurf is an index for the visual condition of the surface.

And b1,…,b4 are the weights given to each index.

Structural index, ISI

Where deflection measurements have been collected and Residual Life calculated for the pavement, it was suggested that the following be used for the structural index:

ISI = 0, if Residual Life ≥ TLRL years

fRL(Residual Life) if TLRL > Residual Life > TURL years Equation 8

100, if Residual Life ≤ TURL years

fRL is a monotonically decreasing function with values between 100 and 0, based on values of Residual Life and TLRL and TURL are thresholds applied to the data.

Where deflection measurements are not available a modified version of a structural index used in Sweden was suggested (Lang et al., 2013):

ISI=MAX(0, MIN(100, Imax + p*(Irut+Iiri+Ied-Imax)/200))

Where Imax=MAX(Irut,Iiri,Ied)

0≤p≤20

Irut=MIN(100, MAX(Iarut,Ibrut)+p*MIN(Iarut,Ibrut)/100) Equation 9


Page 14 of 43

Iarut=MAX(0, MIN(100, 100*(RD-TLrut)/(TUrut-TLrut)))

Ibrut=MAX(0, MIN(100, 100*(∆RD-TL∆rut)/(TU∆rut-TL∆rut)))

Iiri=MIN(100, MAX(Iairi,Ibiri)+p*MIN(Iairi,Ibiri)/100) Equation 10

Iairi=MAX(0, MIN(100, 100*(IRI-TLiri)/(TUiri-TLiri)))

Ibiri=MAX(0, MIN(100, 100*(∆IRI-TL∆iri)/(TU∆iri-TL∆iri)))

and Ied=MAX(0, MIN(100, 100*(Edge Rough.-TLed)/(TUed-TLed))) Equation 11

RD is the maximum rut depth, over the whole width of the road, ∆RD is the yearly change in rut depth, IRI is the maximum IRI measured in either wheelpath and ∆IRI is the yearly change in IRI. TL and TU are thresholds applied to the data.

Irut is only calculated when the distance between the wheel paths is >1700mm and the road width>6.5m. Where this is not the case, Irut =0.

Ied is only calculated when the road width ≤6.5m. Where this is not the case, Ied =0.

Deformation index, IDef

IDef=0.5*G(RutNS) + 0.5*G(RutOS) Equation 12

Where RutNS and RutOS are the rut depths measured in the nearside and offside wheelpaths, respectively, and

G(Rut) = 0, if Rut depth ≤RutL mm

fRD(Rut depth) if RutL <Rut depth<RutU mm Equation 13

100, if Rut depth ≥RutU mm

Ride quality index, IRQ

IRQ = 0.5*G(IRINS)+0.5*G(IRIOS) Equation 14

Where IRINS and IRIOS are the IRI values measured in the nearside and offside wheelpaths, respectively and

G(IRI) = 0 if IRI ≤ TLIRI mm/m

fIRI(IRI) if TLIRI < IRI < TUIRI mm/m Equation 15

100 if IRI ≥ TUIRI mm/m

and fIRI is a monotonically increasing function, based on IRI values.

Since very few countries have a network-level measure of potholes, these were omitted from this index.

Visual condition index, ISurf

ISurf = 0 if Surface deterioration ≤ TLSurf %

fSurf(Surf det’n) if TLSurf < Surf deterioration < TUSurf % Equation 16

100 if Surface deterioration ≥ TUSurf %

and fSurf is a monotonically increasing function, based on surface deterioration values.

Some countries perform manual surveys and have the % of each length affected by visual deterioration i.e. their measure will include many forms of surface deterioration. However, this is not the case for all countries e.g. the UK has automatic crack data and network-level fretting data only. Thus, the thresholds used for the demonstration may vary from country to country.


Page 15 of 43

2.2.2 Durability trigger defined for use on the demonstration network

The weighting factors to be used for the durability trigger are as follows:

Durability trigger = 0.5*ISI + 0.2*IDef + 0.2*IRQ + 0.1*ISurf Equation 17

The largest weighting has been given to the structural index, since structural failure can result in the road being unusable. Longitudinal and transverse unevenness have been given equal weighting, with the smallest weighting given to the visual condition index. This is since the presence of these defects does not always imply problems with structure e.g. fretting is usually just an issue affecting the surface layers.

Table 2 shows the thresholds that may be applied to the data in the Demonstration (Note: These may change when the demonstration is implemented).

Table 2: Example Durability thresholds that may be used in Demonstration

Index Parameter TL TU f

Structural

Residual life (years) 10* 0* fRL(Res Life) =

100*(e(10-res life)/4 -1)/(e2.5 -1)

Rutting (mm, Equation 9) 10 20

p=20 Change in rut (mm/year, Equation 9)

0.6** 2.1

IRI (mm/m, Equation 10) 3 4.5

p=20 Change in IRI (mm/m/year, Equation 10)

0.07** 0.10

Edge deformation (Equation 11)

SWE 20 40 N/A

UK 0.2 0.7 N/A

Deformation Rutting 10 20 fRut(Rut)=10*(Rut depth -10)

Ride Quality IRI (mm/m, Equation 14) 3 4.5 fIRI(IRI)=100*(IRI-3)/1.5

Visual Condition

Cracking, fretting, bleeding, patches, potholes, surface homogeneity

AUT To be defined

Cracking and fretting

UK

* Values taken from guidance on UK Deflectograph data from HD30/08 and HD29/08 (DMRB, Volume 7, Section 3).

** Values taken from (Lang et al., 2013)

2.3 Safety

2.3.1 Safety trigger The ToolBox safety trigger is based on ALERTINFRA, which is a software tool that automatically detects dangerous areas of infrastructure. The tool was developed by the Technical Centre of French Ministry of transports (CETE) and IFSTTAR (ex. LCPC)and


Page 16 of 43

provides up to 15 warnings on curves and 4 warnings on straight roads (19 warnings among which one warning in crossroads and four warnings in straight lines). A detailed description of the software is presented in D2 (Benbow & Sjögren, 2013). The Safety trigger is defined as

ISafety = 0 if no safety risk detected by ALERTINFRA

f(Risk_level) if safety risk detected on a curve by ALERTINFRA Equation 18

YY if safety risk detected on a straight road by ALERTINFRA

Where YY is a constant value <100 and the function f is a monotonically increasing function with a value of 0 when the safety risk=0, 100 when the safety risk = 15.

2.3.2 Safety trigger defined for use on the demonstration network When a risk is identified on curves by ALERTINFRA, the level of risk estimated takes into account 15 individual warnings. The safety index is defined as follow:

Safety trigger = (W * Iwarning) * W Equation 19

With W: weight of the various warnings and

W = 4.34 (sum of the weights of all the warnings)

Iwarning : index taking the value 0, 1 or 2 (0: warning not detected on the area; 1: warning detected on the area, 2: warning detected and connected to another warning).

The individual weights (W) for the warnings have been defined by statistical analyses, and have a value between 0 and 1, with the actual value depending on the importance of the warnings.

Thus, the global safety index is a linear combination of the various risk observed on the infrastructure.

When a risk is identified by ALERTINFRA on a straight road, a value of 35 will be used for the Safety trigger. This value is obtained by considering the accident rates on straight lines and curves and the various geometrical configurations of the infrastructure, where accidents are occurring.

2.4 Environment

2.4.1 Environment trigger The external noise, particulates and fuel consumption are included in the environment trigger:

Environment trigger = a1*INoise + a2*IFuel + a3*IPart Equation 20

Where INoise is an index for external noise

IFuel is an index for fuel consumption

IPart is an index for particulates and other emissions.


Page 17 of 43

INoise = 0, if CPX ≤ TLCPX db

fNoise(CPX) if TLCPX < CPX < TUCPX db Equation 21

100, if CPX ≥ TUCPX db

fNoise is a monotonically increasing function with values between 0 and 100, based on values of CPX.

IFuel = 0, if Fuel consumption ≤ TLFuel l/10km

fFuel(Fuel consumption) if TLFuel < Fuel consumpt’n < TUFuel l/10km Equation 22

100, if Fuel consumption ≥ TUFuel l/10km

fFuel is a monotonically increasing function with values between 0 and 100, based on values of Fuel consumption.

During the review carried out within WP2, it was found that no model existed to predict the level of particulates emitted by vehicles travelling on a road and that emissions are specific to individual vehicles. For example, a Ford Focus with a 1.6l petrol engine may be in the same Euro classification as a 1.6l petrol Volkswagen Golf, but the two cars may have different emissions, due to different engine efficiencies and filters etc. Vehicle age will also have a large effect on the volume of particulate matter in a vehicle’s exhaust fumes. Thus to model this effectively would require a large amount of information and also need a very detailed and complicated model. Developing such a model is not within the remit of the ToolBox project and thus it was felt that fuel consumption should be used as a proxy measure for particulate emissions.

2.4.2 Environment trigger defined for use on the demonstration network

The weighting factors to be used for the durability trigger are as follows:

Environment trigger = 0.5*INoise + 0.5*IFuel

Table 3 shows the thresholds that may be applied to the data in the Demonstration (Note: These may change when the demonstration is implemented).

Table 3: Example Environment thresholds that may be used in Demonstration

Index Parameter TL TU f

Noise CPX (db) 102.4 104.5 fNoise(CPX) =

100*(CPX-102.4)/2.1

Fuel Fuel consumption (l/10km)

3.0* 3.5* fFuel(Fuel) = 200*(Fuel-3)

* These values are large, due to the way that it is planned to calculate fuel consumption for the demonstration. This is discussed further in section 3.2.4.2.

2.5 Combined Condition Indicator The Comfort, Durability, Safety and Environment triggers will be combined, to produce the Combined Condition Indicator (OCI) as follows:

CCI = wC*Comfort Trigger + wD*Durability Trigger + wS*Safety Trigger

+ wE*Environment Trigger Equation 23


Page 18 of 43

Where wC + wD + wS* + wE = 1 and are values chosen by the ToolBox user, depending on individual road owners’ policies.


Page 19 of 43

3 ToolBox Data

In the demonstration and tests, as well as for the evaluation work, ToolBox will only work on available data and all proposed models have been adjusted to existing data, where possible. The data needed for the ToolBox triggers, along with whether it is available for all partner countries, is discussed in section 3.1. Suggestions have been made in section 3.2 as to how to deal with missing data e.g. substitute with dummy data, substitute with another technical parameter or calculate a proxy value.

3.1 Data needed and available for ToolBox

In Table 4 the data needed for the ToolBox triggers can be seen and the minimum length over which these parameters are required is 100m.

The tool, ALERTINFRA, which we propose to use to generate the safety trigger, was developed with data collected every 1 meter. However, this level of precision is not always possible to obtain, especially since the plan for ToolBox is to use the data collected during routine road monitoring. This issue is detailed in section 3.2.3.

Table 5 shows whether these parameters are available for the partner countries.


Page 20 of 43

Table 4: Required input parameters to ToolBox

Trigger

Comfort

Durability

Safety

Environment

Index Structural

1 Structural

2 Deform-

ation Ride

Quality Visual

condition Noise 1 Noise 2 Fuel

Selected Model ToolBox specified

ToolBox specified

Lang et al. 2013

ToolBox specified

ToolBox specified

ToolBox specified

ALERTINFRA* ToolBox specified

McRobbie et al. 2003

Simplified VETO

Crossfall x*

Curvature x*

Gradient x*

Position x*

Longitudinal Unevenness

IRI IRI, ∆IRI IRI IRI

Longitudinal Unevenness

Localised roughness

Vertical acceleration

x*

Transverse Unevenness

Rut depth Rut depth, ∆Rut depth

Rut depth

Edge roughness x x

Texture MPD* x† MPD

Skid resistance x*

Residual life x

Visual condition x

Posted Speed x*

CPX x

* ALERTINFRA requires data to be input for every meter in the original version. This requirement is discussed further in section 3.2.3. † Texture profile required


Page 21 of 43

Table 5: Parameters available, from partner countries

Technical parameter/ Country Austria France Sweden UK

Crossfall x x x x

Curvature x x x x

Gradient x x x

Position x x x x

IRI x x (on some

national roads) x x

Yearly change in IRI (∆IRI) x x

Localised roughness x x (on some

national roads) x x

Vertical acceleration x

Rut depth x x (on some

national roads) x x

Yearly change rut depth (∆RD) x (on some

national roads) x x

Edge roughness x x

MPD x x x x

Texture profile x

Skid resistance x x Project level x

Residual life (deflection) Project level Project level x x

Visual condition

Cracking, fretting,

bleeding, patches, potholes, surface

homogeneity

Cracking, fretting,

bleeding, patches, potholes

All defects at a project level

Cracking, fretting

Speed limit x x (on some

national roads) x x

CPX x


Page 22 of 43

3.2 Dealing with unavailable or insufficient data

3.2.1 Unavailable data for Comfort trigger Where IRI or localised roughness data is not available from the offside wheelpath, set G(IRIOS)=G(IRINS) and G(LROS)=G(LRNS)

If a measure of localised roughness does not exist, set G(LRNS)=G(IRINS) and G(LROS)=G(IRIOS).

If a measure of edge roughness does not exist, set G(Edge)=G(RutNS).

3.2.2 Unavailable data for Durability trigger It may be the case that historical data is not available to all road authorities and therefore, where this is the case, we would set Irut=Iarut in Equation 9 and Iiri=Iairi in Equation 10 above.

Where the road ≤6.5m wide and Edge Roughness data is not available, set Ied=IRut in Equation 10 above.

Where ride quality data is unavailable from the offside, set G(IRIOS)=G(IRINS) in Equation 14 above.

3.2.3 Unavailable data for Safety trigger As discussed in section 2.3, the software ALERTINFRA has been chosen to generate the Safety trigger.

The standard inputs to ALERTINFRA include Radius of Curvature, Skid resistance, Vertical acceleration, Gradient, Crossfall and Macrotexture and, in order to make accurate safety studies, the original version required these inputs for every metre along the road, whilst a newer version of the tool requires data at 5m spacing.

The tool then uses this data to produce warnings, which have been described in Deliverable 2, and Table 6 indicates which parameter is used for the definition of the different warnings. As can be seen from this table radius of curvature is used in all of the warnings but its importance is variable: In the case of warnings in curves, the values of the radius of curvature allows detection of a risky area, whereas in straight lines, other than identifying that the length lies on a straight line, this value is not used.

Numerous countries collect data from road monitoring surveys but these data are different from one country to another due to the devices used, the sampling, the final use in the national indices, etc. Whilst most road owners will have access to most of the parameters needed, by ALERTINFRA, at a spacing of 100m, they do not have access to data spaced at 1 or 5m.

Thus, for the safety trigger, we have needed to consider the effect of not only there being a lack of data for certain parameters but also the effect of the data being provided over much larger sampling lengths than the 1 or 5m ideally needed for ALERTINFRA.


Page 23 of 43

Table 6: Parameters used in the definition of the ALERTINFRA warnings

Geometry Surface characteristics Length Events

Radius of curvature

Crossfall Slope SFC MPD Vertical

acceleration V1 x x x Location of the last town V2 x x Location of the last town V3 x x V4 x x V5 x x V6 x x V7 x x V8 x x V9 x x V10 x x V11 x x V12 x x V13 x x V14 x x x I1 x Intersection S1 x x S2 x x S3 x x S4 x x

3.2.3.1 Effect of unavailable parameters Whilst most parameters are collected routinely in all countries, not all countries measure skid resistance at a network level (e.g. Sweden). Where this is the case, one warning cannot be detected by ALERTINFRA (V7). The maximum safety index that can be calculated in curves becomes:

Max (Isafety) = 94 (instead of 100).

Considering the relatively small gap between this value and the theoretical maximum value of 100, it was felt that it would be better to avoid using an acceptable skid resistance value in the tool and instead set the skid resistance value to be equal to 1. This will result in the warning for skid resistance never being triggered.

In straight lines, one of the four warnings available will not be triggered, if skid resistance is not available. The consequence is again a decrease of the maximum safety index, which becomes:

Max (Isafety) = 28 (instead of 35).

As for curves, the difference between the theoretical maximum value for the safety index and the maximum achievable value, is relatively small. Therefore, it is again proposed that, where skid resistance data is unavailable, a value of 1 is usedand indicate in the final results that skid resistance is not taken into account in the choice of the sections to maintain.

Despite the small difference between the maximum theoretical value of the safety index and the attainable maximum value, the impact on the percentage of the route’s length without warning, both on straight lines and curves, should be investigated and this is discussed in the following paragraphs.

Simulations of the effect of unavailable data on the results obtained from ALERTINFRA have been carried out using data from a road in the French national network. The characteristics of this road represent the main geometrical configurations of road infrastructure observed in France. The section is 30km length and the data have been collected at a longitudinal spacing of 1m.


Page 24 of 43

Five cases have been considered:

Reference case with the whole data available

Without macrotexture (MPD)

Without skid resistance (SFC)

Without vertical acceleration (unevenness)

Without skid resistance and macrotexture.

First of all, the number of 50m-length segments without warnings is estimated on the reference route (Figure 2) for all of the different cases. Comparing Figure 2 with Figure 3, we can notice that the impact is not the same on straight lines and on curves for the same unavailable data. Warnings in straight lines seem to be more sensitive to the lack of data, which is consistent with the fact that these warnings are mainly defined from surface characteristics. This can also be seen in Table 7, which summarises the percentage of the total length of the route where warnings are detected in the five cases considered.

Figure 2: Number of 50m-length segments in curves without warning

Figure 3: Number of 50m-length segments in straight lines without warning

Table 7: Percentage of length with warnings in curves and straight lines

Normal (all data

available)

Without skid

resistance

Without MPD

Without SFC

Without surface

characteristics

With warning (both curves and straight lines)

61% 51% 60% 48% 31%

% length with warnings in curves

23% 23% 23% 22% 22%

% length with warnings in straight lines

38% 28% 37% 26% 9%

As can be seen from Table 7, when warnings in both curves and straight lines are considered, the percentage length with a warning is halved, when there is no surface characteristic data available. However, when they are considered separately, it can be seen that the percentage of the route presenting warnings in curves is rather stable and not affected by the lack of data concerning surface characteristics. This fact can be explained by

0

100

200

300

400

500

Normal(tired with all data)

WithoutMPD

WithoutSkid resistance

WithoutSFC

WithoutSurface characteristics

0

100

200

300

400

500


WithoutMPD


WithoutSFC


0

50

100

150

200

250


WithoutMPD


WithoutSFC


0

50

100

150

200

250


WithoutMPD


WithoutSFC



Page 25 of 43

the definition of the warnings in curves (see Table 10). Indeed, only three warnings among the fourteen take into account the surface characteristics. On the contrary, the impact on straight lines is very important (Table 7), and the percentage of the route with warnings is less than a quarter when there is no surface characteristics data. Nevertheless, considering the fact that most of the countries do measure skid resistance, we decide in Toolbox to keep safety index provided by Alertinfra but to warn the user that the safety diagnosis is incomplete.

3.2.3.2 Effect of data spacing The original ALERTINFRA tool required data at 1m spacing. Work was done in France two years ago to check the impact of using data with a longer longitudinal spacing (5m for geometry). A comparison between the results, using 1m data, and those using 5m data was done on two real sites. The first site was a 3.4km long route with dual carriageways, presenting radius of curvature ranging between 30 and 600 m. The second was a test track presenting high variation in the longitudinal slope, and small values for the radius of curvature. On these two sites, 13 dangerous areas were detected by using 1m and 5m spaced data. Moreover, the total length of dangerous areas represents around 40% of the routes, when using data with either a 1m or 5m spacing. Lastly, the gap in the location of the risky areas when using the two data spacing options was less than 10m. Thus, it was concluded that data at 5m spacing gives consistent enough results, to be able to use this lower resolution data.

Concerning skid resistance, values with a spacing of 10 or 20m can be used without too much effect on the detection of the risky areas. Nevertheless, increasing the data spacing mainly results in an increase of the safety index values. This fact is favourable to ensuring safety but caution is needed when comparing safety indicators, calculated using lower resolution data, directly with other indicators.

The use of data with a spacing of 100m will have an even larger effect on the safety index values. Indeed, some warnings cannot be activated with this spacing due to their definition. For example, warning V10 matches to a succession of two curves separated by a distance less than 100m. By considering the definition of each warning, one can suppose that some of them will not be activated when using a set of data with a spacing of 100m, and this may lead to a decrease in the maximum safety index.

Simulations of the effect of data spacing on the results obtained from ALERTINFRA have been carried out using data from the same road used for the simulations in section 3.2.3.1. The data for this route has been collected at 1m spacing, and for the purpose of the simulation, has been resampled to 10m, 50m and 100m spacing. The number of sections without warnings for these three spacings have then been compared to that for the reference 1m data. The results of this are presented in Figure 4 and Figure 5. It can be seen that the use of data at 100m spacing leads to a dramatic reduction in the number of segments with warnings compared with the reference. This would be expected, in that the number of 100m lengths in any route is obviously 100 times less than the number of 1m lengths. Therefore, we have also considered the percentage of the route with warnings, for all data spacings. The results of this are given in Table 8.


Page 26 of 43

Figure 4: number of Xm-length segments in curves without warning on the measurement path

Figure 5: number of Xm-length segments in curves without warning depending on the measurement path

Table 8: Percentage of length with warnings in curves

Normal (with all data

available)

Without Skid

resistance

Without MPD

Without SFC

Without Surface

characteristics

1 m 94% 94% 93% 94% 92% 10 m 49% 49% 41% 40% 26% 50 m 60% 60% 50% 48% 30% 100 m 67% 67% 59% 55% 36%

Table 9: Percentage of length with warnings in straight lines

Normal (tired with all data)

Without Skid

resistance

Without MPD

Without SFC

Without Surface

characteristics

1 m 97% 97% 97% 97% 99%

10 m 69% 69% 78% 78% 92% 50 m 63% 63% 72% 74% 91% 100 m 61% 61% 69% 73% 91%

We can notice that the increase of spacing entails a decrease of 1/3 of the total length presenting warnings when all the data are available, both on curves and straight lines. Moreover, the impact of unavailable data is not the same considering the spacing. However, the rate of detection of dangerous areas stay correct and gives an usable information for choosing the maintenance candidates. Moreover, a lot of the warnings are detected with threshold values applied on the radius of curvature data. Thus, the question is to know whether a using a different value of radius of curvature, instead of the average, would improve the results given above, for data measured at a spacing of 100m. Thus, use of average radius of curvature over a 100m length, and minimum radius of curvature has been considered. Table 10: Warning availability with data measured every 100m indicates the warnings that can/cannot be detected with these two different types of data.

0

500

1000

1500

2000

2500

3000


WithoutMPD


WithoutSFC


1 m

10 m

50 m

100 m

0

500

1000

1500

2000

2500

3000


WithoutMPD


WithoutSFC


1 m

10 m

50 m

100 m

1 m

10 m

50 m

100 m

0

200

400

600

800

1000

1200


WithoutMPD


WithoutSFC


1 m

10 m

50 m

100 m

0

200

400

600

800

1000

1200


WithoutMPD


WithoutSFC


1 m

10 m

50 m

100 m

1 m

10 m

50 m

100 m


Page 27 of 43

Table 10: Warning availability with data measured every 100m

Alert Definition

Available with average value of radius?

Available with minimum value of radius?

V1 Curve with a high speed change (> 20 km/h) No Yes

V2 Curve with a too long introduction and after an «easy» section

Yes Yes

V3 Curve with a high direction change No No V4 Long curve (> 225 m) Yes Yes V5 Curve with reducing radius No No V6 Curve with radius < 150m after an «easy» section Yes Yes V7 Curve with poor adherence Yes Yes V8 Curve with poor macrotexture Yes Yes V9 Curve with poor unevenness (short wavelength) Yes Yes

V10 Inconsistent curves No No V11 Curve with radius < 120m and low superelevation No No V12 Curve with radius < 200m and low superelevation No No V13 Left curve with reverse cross fall No No V14 Curve with steep slope Yes Yes I1 Intersection in a curve Yes Yes S1 Straight line with bad unevenness Yes Yes S2 Straight line with low skid resistance Yes Yes S3 Straight line with low macrotexture Yes Yes S4 Straight line with high value of longitudinal slope Yes Yes

Then, the maximum value of safety index Isafety in both situations is:

With the average value of radius, Max (Isafety) = 67 (instead of 100)

With the minimum value of radius, Max (Isafety) = 67 (instead of 100).

Lastly, on straight lines there is no impact on the safety index Isafety values.

3.2.3.3 Using ALERTINFRA in ToolBox The work focuses on solving two issues:

Unavailable data due to larger sampling interval than the one needed in ALERTINFRA;

Unavailable data due to unavailability of the measurements (surface characteristics, etc.).

Considering the data sampling issue, an interpolation of the data is proposed to generate an input file with a more suitable data spacing for use in ALERTINFRA. For this project, a linear interpolation is proposed. Other types of interpolation, e.g.polynomial interpolation, could be tested in the future. Thus, a specific tool has been developed, which can take data files from the different European countries involved in ToolBox (variable file format and different sampled data) and create a suitable input data file for ALERTINFRA.


Page 28 of 43

Figure 6: Screen shot of the tool, where the required data sampling interval is chosen

Figure 7: Screen of the tool where the sampling interval of the original input data is provided

Then, the resampled data are analyzed with ALERTINFRA to obtain the safety index. The methodology of calculus of safety index is detailed in Deliverable 2.

To conclude, we observed that the lack of data or the differences in the path measurement have consequences on the safety index calculated by ALERTINFRA but it still provides usable information by detecting around 2/3 of dangerous areas (in percentage of total route length). A tool was developed by CETE of Lyon to resample the data collected with a path of 100m and to inject them in ALERTINFRA instead of modifying directly the software developed with 1m-path data.

3.2.4 Unavailable data for Environment trigger

3.2.4.1 Noise index Some countries have network level measurements of CPX values but this is not common. A method for predicting CPX, from texture data, has been developed in the UK and this can be used to provide proxy values for this measure (McRobbie et al., 2003). To use this would require a road authority to have access to the raw texture profile, or be able to implement this model in their PMS.

The method either needs to be provided with the surface type, or can predict what this might


Page 29 of 43

be from raw texture data, and, if available, luminosity data (this is obtained from image intensity, is dimensionless and scaled from 0 to 255, with 0 being an entirely black image and 255 entirely white).

The surface type method uses the following parameters:

A’ = 10th percentile of surface texture depth filtered at 200 cycles per metre

B’ = 10th percentile of surface texture depth filtered at 40 cycles per metre

C’ = Skewness of texture depth distribution filtered at 20 cycles per metre

D’ = B’/A’

E’ = B’ – C’

The method then predicts the probability that the surface is one of four types: HRA, thin surfacing, brushed concrete and grooved concrete, based on the parameters D’ and E’.

The noise prediction method then uses the parameters S and R, defined as

, with B’ and C’ defined above and the constants a=-1.26 and

b=-1.1 and R=A’/B’.

The noise prediction method is limited to predict CPX on the same surface types produced by the surface type algorithm i.e. HRA, thin surfacing, brushed concrete and grooved concrete.

3.2.4.2 Fuel consumption Fuel consumption is not something that is currently measured on road networks. Therefore, a model was needed to provide a proxy value for this measure.

The amount of fuel consumed by a vehicle travelling over a road is dependent on the type of vehicle, its mass, the type of fuel it runs on, the wheel bearing resistance, air resistance, the alignment of the road surface, the rolling resistance etc. A National Road Administration (NRA) has no (or very little) control over the number and types of vehicles travelling on its road network, and therefore would be limited to making changes to the road alignment and rolling resistance. Road alignment (i.e. gradient, crossfall, curvature) is generally set when the road is constructed and is very costly to change. Thus, of the contributions to fuel consumption, the aspect over which an NRA has most control, and the greatest chance of changing, is the rolling resistance of the pavement. Thus, we will use a model for rolling resistance as a proxy for fuel consumption within ToolBox.

The following model is a simplified model based on the developments in a MIRIAM related project (Hammarström et al., 2012) and will be used for estimating the fuel consumption that can be considered generated by the road surface:

FC= 2.75 + 0.0682 x IRI + 0.198 x MPD

Where IRI is the maximum IRI value and MPD the maximum mean profile depth measured in either wheelpath.

As discussed in D2 (Benbow & Sjögren, 2013), this function represents the amount of fuel consumed by a truck and trailer, travelling at a constant speed of 22.2 m/s (80 km/h) on a flat and straight road, with no prevailing wind. Thus, it will be a much larger value than that expected for an average vehicle travelling on the road and cannot be used to estimate the consumption of a normal vehicle fleet (i.e. a combination of HGV, cars etc.).

The use of the truck and trailer model should enable easier classification of surfaces as either good or poor for fuel consumption, as there is increased sensitivity to changes in the surface for this particular vehicle type.


Page 30 of 43

4 Maintenance Candidates

A spread sheet tool has been developed in Excel that combines technical parameters into functional triggers (using the models described in Chapter 2) and then combines these in turn, with user-defined weightings, to form a Combined Condition Index (section 2.5). This Combined Condition Index is then used by the spread sheet to identify potential maintenance candidates. This Chapter contains a description of the tool and how it works.

4.1 What the tool looks like

The maintenance candidate selection tool consists of 5 sheets within an Excel workbook (see Figure 8) and utilises Excel macros (the user will need to enable these when opening the workbook). The majority of the tool is implemented in the data sheet. This sheet contains a table where input data can be inserted, from which the functional triggers, Combined Condition Index and treatment options are automatically calculated. The results of these calculations are dependent on the parameters that can be found (and, if necessary, altered by the user) on the ‘Trigger Weighting Parameters’, ‘Trigger Thresholds’ and ‘Treatment Thresholds’ sheets.

Figure 8: Layout of sheets within Excel based maintenance candidate selection tool.

A section name or description can be entered by the user into column A of the data sheet, whilst the end chainage of the length is entered in column B (see Figure 9). Measured condition parameters (e.g. IRI, Rutting), can be input into columns C to P, whilst the outputs from ALERTINFRA (used for the safety trigger) are input in columns Q to AI.

Figure 9: Screen shot of part of the data sheet, showing columns for parameter value input

The calculated Safety, Comfort, Durability and Environment triggers are shown in Columns AJ, AY, BV and CB respectively, with the Combined Condition Index given in column CC (Figure 10). These values are automatically calculated and cannot be edited by the user.


Page 31 of 43

Figure 10: Screen shot of part of the data sheet showing columns containing calculated trigger values

Once the user is satisfied with the input data, the ToolBox Excel add-in can be used to group the lengths into schemes. This add-in can be accessed via the add-in menu (Figure 11).

Figure 11: Accessing the ToolBox add-in

This tool groups together lengths of similar condition into schemes. All lengths are assigned to be in a scheme, whether they would be considered in need of maintenance by an engineer or not. Two parameters affect this selection process, the ‘Look ahead distance’ and the ‘Max percentage difference from mean’, both of which can be found and adjusted by the user on the ‘Scheme Creation Parameters’ sheet. The selection process is described in more detail in the following section.

4.2 How the tool selects maintenance candidates

Once the Combined Condition Index has been calculated the lengths can then be grouped into maintenance candidates (schemes). The aim of this process is to group contiguous lengths of similar condition (in terms of their need of maintenance work), which in ToolBox is defined by the Combined Condition Index.

To achieve this grouping, the tool needs to know the location referencing of each row of data provided i.e. where each length can be physically found on the network. Commercially available pavement management systems (PMS) have access to a description of the network, which enables them to fit the data to the network. However, this network information is not always widely available and the fitting process quite complicated. Thus it was felt beyond the scope of ToolBox to implement something similar. Therefore, it has


Page 32 of 43

been assumed that the data provided is for a continuous route i.e. data provided in row 3 of the data sheet is for the length following that given in row 2 etc.

The task of dividing a continuous length of pavement into schemes based on a measure of its condition is fundamentally an exercise in clustering (specifically constrained 2D clustering, in which the algorithm is constrained to cluster only lengths that are next to one another). There are many algorithms that can be used to perform this, and consideration was given as to which might be most appropriate to use for ToolBox. Three of the key areas considered were: accuracy (the algorithm’s ability to find an optimal solution), speed and complexity. It was found that there were benefits and disadvantages to each algorithm considered and none stood out as being clearly best. Therefore, the algorithms used by current UK Pavement Management Systems (PMS) were researched. This found that these used a simple procedural sweep of the data from one end of the route to the other, adding a length to the current scheme if the condition of that length falls within a defined range, centred on the mean condition of the current scheme, and also lies within a certain distance of the scheme. Allowing a “look ahead” to lengths not contiguous with the current scheme reduces the premature ending of a scheme if a single length fails to qualify.

An example route is shown in Table 11 which contains the Combined Condition Index for each 100m length and also the average condition for the scheme, as each length is added to it.

Chainage

Combined Condition Index Scheme Running scheme average

100 50 1 50 200 52 1 51 300 54 1 52 400 56 1 53 500 63 1 55 600 55 1 55 700 62 2 56 800 64 2

Table 11: Example of scheme selection algorithm

The following is a step-by-step list of the decisions made at each length by ToolBox, using a condition range of 10% and a look ahead distance of 1 length:

1. Starting at chainage 100. This is the first length of the route, so is automatically added to Scheme 1. Since the Combined Condition Index of this length is 50, the scheme average is thus 50.

2. Move to chainage 200. This has a Combined Condition Index of 52, which is within the acceptable range (50±10%). So, add the length to Scheme 1.

3. Move to chainage 300. This has a Combined Condition Index of 54, which is within the acceptable condition range of the scheme (51±10%), so add the length to Scheme 1.

4. Move to chainage 400. This has a Combined Condition Index of 56, which is within the acceptable condition range of the scheme (52±10%), so add the length to Scheme 1.

5. Move to chainage 500. This has a Combined Condition Index of 63, which is not within the acceptable condition range (53±10%). So look ahead to the following length.


Page 33 of 43

6. Move to chainage 600. Since the current length’s value (55) is within the mean of scheme 1 (53±10%), add both the look ahead length and the current length to scheme 1. Reset the look ahead counter to 0.

7. Move to chainage 700. This has a Combined Condition Index of 62, which is not within the acceptable condition range of the scheme (55±10%), so look ahead to the following length.

8. Move to chainage 800. Since the current lengths value (64) is not within the mean of scheme 1 (55±10%), and the look ahead limit has already been reached, create a new scheme and add the look ahead length and the current length to it.

4.3 Treatments

In order to estimate costs for a maintenance candidate, one needs to know what treatment might be suitable for that candidate and also how such treatment may be best carried out.

Table 12 lists the treatments that are generally applied to European road networks.

Table 12: Treatments applied to road networks (taken from 2013R08EN)

Pavement type Deterioration Maintenance

Flexible and semi-rigid

Surface

Retexturing

Surface treatment

Bituminous wearing course replacement

Bituminous inlay

Structural

Bituminous overlay

Replacement

Concrete overlay/inlay

Rigid

Surface

Retexturing

Surface treatment

Thin asphalt overlay

Thin concrete overlay

Structural

Bituminous overlay

Replacement

Concrete overlay

Choosing a treatment for a specific maintenance candidate, or scheme, requires knowledge of, not only the condition of the road, but also the road’s age, construction, its use, the traffic volume and composition and also details of any historical maintenance. This is data that any engineer considering the candidate should have access to, however, it was not found to be readily available by the ToolBox project team. Thus, the choice of treatment has been carried out using a simplified method (based on SWEEP.S, software used by the UK Highways Agency (SWEEP, 2008)) and we have assumed that all pavements have flexible construction. This could be improved on, or changed, in future versions of the tool.

For each 100m length, lying within a maintenance candidate, the individual triggers (Safety, Comfort, Durability, Environment) and some parameters are used to identify potential treatments. The flow diagram in Figure 12 shows how this is achieved.


Page 34 of 43

Figure 12: Flow diagram of treatment choice

Structural index≥SI1 and

Rut≥R3 or Structural index≥SI2 and Rut≥R2?

Yes

Structural index≥SI3 and Rut≥R1?

No

Full reconstruction [SS1]

Thick overlay [S1]

Partial reconstruction [SS1]

Structural index≥SI2 and Rut≥R3 or Structural

index≥SI2 and IRI≥IRI2?

Structural index≥SI3 and Rut≥R2 or Structural

index≥SI3 and IRI≥IRI2?

No

No

No

No

Yes

Yes

Yes

Yes Plane off and overlay [S1]

Thin overlay [S1] Structural index≥SI3 and Rut≥R3?

No

No

Yes

Yes

Yes

100mm inlay [SS2]

50mm inlay [SS2]

Thin surface [S2]

Rut≥R1 or IRI≥IRI1 or visual defects ≥V1 or Edge ≥E1 or localised roughness ≥LR1 or

(Fuel ≥F and IRI≥IRI2)?

Rut≥R3 or IRI≥IRI2 or visual defects ≥V2 or Edge ≥E2 or localised roughness ≥LR2 or (Fuel ≥F and IRI≥IRI3) or

Safety trigger≥S (V9 or S1)?

Noise ≥N or visual defects ≥V3 or (Fuel ≥F and

IRI≥IRI2) or Safety trigger≥S (V7, V8, S2 or S3)?


Page 35 of 43

Where: SI1>SI2>SI3 are thresholds applied to the Structural index

R1>R2>R3 are thresholds applied to rut depth

IRI1>IRI2>IRI3 are thresholds applied to the IRI parameter

V1>V2>V3 are thresholds applied to Isurf, the visual defects index

F is a threshold applied to the fuel consumption

E1>E2 are thresholds applied to the edge deformation parameter

LR1>LR2 are thresholds applied to the localised roughness parameter

N is a threshold applied to the Noise index

S is a threshold applied to the Safety trigger.

In addition to the treatments, suggested in the flow chart above, a further category of “redesign” is needed, where there are safety issues with road alignment:

Treatment Requirements Category

Redesign Safety trigger ≥S, and ALERTINFRA outputs of V1 to V6, V10 to V14 or S4

SS1

4.4 Outputs from the tool

The tool divides the whole route, provided by the user, into potential maintenance candidates (or schemes), ready for the user to select which candidates to carry out cost-benefit analysis on. To aid this selection process, the tool summarises (see Figure 13) the schemes created, including the number of lengths included in the scheme and the number of lengths with a minimum suggested treatment of each of the treatment options. The summary also orders the schemes by the number of lengths with each of the minimum suggested treatment from most severe (redesign) to the least severe (thin surface).


Page 36 of 43

Figure 13: Summary sheet provided by the tool.

Once the user has selected the schemes that they intend to carry out cost-benefit analysis on they can quickly extract the required data by filtering the data table (on the data sheet) by the chosen scheme number (Figure 14).

Figure 14: Obtaining data from specific schemes


Page 37 of 43

5 Optimising LCCA and user expectations

5.1 Background to LCCA and its application to ToolBox

Life cycle cost considerations includes the projects’ initiation, planning, design, construction, service life, as well as disposal. In ISO 15686 (ISO 2007) the term whole-life costing (WLC) is used to define the methodology for economic evaluation of products’ or assets’ whole life benefits and costs during a predefined period of analysis. Often the term cost-benefit analysis (CBA or BCA) is used synonymously for the methodology to evaluate different alternative WLC. While WLC considers both costs and benefits, the life cycle cost (LCC) can be considered to be a subset of WLC for which only the asset’s costs are included. CBA is normally used in planning stages which emphasise societal benefits and costs, as opposed to WLC and LCC which focus on decisions in the later stages with more emphasis on road manager costs.

Figure 15 ISO 15686 definition of WLC and LCC (ISO 2007). Usually in traditional LCC or WLC all of the asset’s costs or benefits are included in the analysis. For large projects, such as infrastructure investments, this can result in complex calculations. However, the complexity can be significantly reduced when comparing two or more design alternatives, if the costs or benefits that are equal among the alternatives are ignored. In this case the term life-cycle cost analysis (LCCA) is usually used to define the methodology to find the optimal design alternative among several alternatives with respect to the LCC. The analysis can, in these cases, focus on costs that differ between the alternatives. In the case of deciding on optimum solutions for maintenance the calculations can be significantly reduced by just looking at the differences. However, it should be kept in mind that a comparison between alternatives with long term consequences, such as rehabilitation or maintenance adding to bearing capacity, and maintenance with shorter service life needs to be taken into account in some way. Large differences in service lives could be included either by including the consequences in the LCCA or by explicitly pointing out the pros and cons.

The effect of maintenance on road workers is another example that is difficult to include in LCCA. In this case, the difficulty lies in reflecting their situation in monetised units and including this on the same basis as the other costs. Instead, it has been suggested that for Toolbox, decisions related to road worker safety should be taken in a multi-criteria decision tool (STARS, 2013) where road worker safety is rated using a score system. This rating will then be compared to the LCC and WLC for each scheme.

The components of WLC suggested for inclusion, to describe the consequences of maintenance schemes are:


Page 38 of 43

- Road owner costs

Investment cost

Maintenance costs

- Road user costs

Delay costs

o Reroute time

o Queue time and reduced speed

o Speed reduction due to inferior performance

Vehicle operation costs

- External costs

Emissions

-Road workers

Safety

One objective of ToolBox is to be able to support prioritisation between different maintenance schemes. This means that some schemes will be selected for treatment during the coming season and others will programmed in the years to come or not considered for treatment. Of major importance is the question of consequences of postponing maintenance. Figure 16 below seeks to illustrate some of the challenges in the analysis of timing of maintenance on individual objects. Postponing maintenance will likely have an effect on both road owner costs and society costs. Road owner budgets can be reallocated to be used on more urgent schemes but costs will still appear later when maintenance is decided upon. An important question is if the costs of maintenance will be higher due to deterioration or not. An intermediate decision can also be to propose mitigating measures such as ensuring drainage and sealing. The consequences on society costs are related to the level of function that can be maintained during the period up to the next maintenance treatment and during the actual treatment itself. Complicating matters are for example the fact that postponing maintenance may lead to the need for more extensive treatment and that other condition indicators will be of relatively larger importance to the need for maintenance. Finally, it needs to be pointed out that different maintenance treatments differ much in their abilities to improve different condition indicators and the way technical parameters change during treatment and its service life. More extensive treatments are more likely to improve durability and bearing capacity related deterioration during and after treatment. The latter is to a large extent taken into account during the prior scheme selection process in which treatments are suggested based on condition and performance.

Furthermore, LCCA on its own cannot be the only basis for decisions. If the blue lines in Figure 16 below are an indicator of, for example, bearing capacity (having consequences mainly for the road owner) and the red lines are an indicator related to safety with absolute values that need to be maintained, optimum LCCA might be modelled to appear in violation of safety requirements. For example, for safety reasons there should be a specified threshold similar or equal to the worst trigger value for safety. The same should apply to all triggers. Then, LCCA might render an answer that increasing safety can be done cost effectively and consequently the optimum level of safety can be far better than the trigger threshold.


Page 39 of 43

Figure 16 Illustration of the challenges in reflecting the life cycle consequences of doing maintenance later on one road section when having several technical parameters (TP), conditions indicators (CI, red and blue), maintenance criteria (blue TP1 and TP2) and maintenance strategies/options (A and B (red and blue cont and dash)). LCCA decisions can be based on a number of indicators derived from life cycle costs. There are two basic principles:

Net present value. All costs in the calculation period are discounted to a specific year, usually in respect to planning years, construction dates or traffic opening.

Annual costs. Net present values of all costs are inverse discounted to all years in the calculation period, such they become evenly distributed and mimic other annual costs of operation and maintenance.

This introduces calculation period and discount rates as two important aspects of LCCA. Calculation periods and discount rates are treated in cost benefit analysis of public infrastructure investment assessment where usually the benefit per spent cost is of interest. Analysing benefits of maintenance would also be of interest but complicates the LCCA. Instead the analysis is based on the fact that a road is present and has pre-determined standard requirements. This allows skipping the benefits side and focus on the costs associated with maintaining a road in compliance with these requirements. In LCCA of products the costs of maintenance can be calculated over the life span of the product. For roads the equivalent to product can be individual layers and features such as wearing course, binder layer or drainage facilities, each with some sort of technical service life. Figure 17 also illustrates the added complexity of two maintenance alternatives with respect to service life and calculation period. This means that a maintenance treatment can address one specific or several of these components, each with a different service life. Individual handling of each component become evident if for example a thin overlay is selected as treatment with no effect on subsequent layers and roadside features. The backlog for meeting deterioration of for example bound base or binder layer will still remain and probably increase during the thin overlay service period. Discount rates are introduced in cost benefit analysis of public infrastructure investment assessment to reflect the willingness to gain utility/benefits of expenditure today or in the future, or for that matter a reduced utility/benefit if not spending (social discount rate). In LCCA it becomes important to also include any known changes in costs over time that is not reflected in common inflation but a consequence of for example changing requirements, equipment costs, availability of resources and changes in efficiency.

CI, TP

Year

A.

B.

TP1

TP2


Page 40 of 43

5.2 LCCA on selected ToolBox schemes

The outcome of the ToolBox scheme selection phase is a number of road sections grouped into schemes with similar road properties and ordered by for example a number quantifying the urgency for maintenance. For each section a treatment is automatically suggested. The purpose of the LCCA phase is then to optimise maintenance schemes trying to meet different objectives such as budget, LCC, safety, environment and comfort. This is done by adjusting suggested maintenance treatments and checking the resulting output in LCC and WLC terms. Thus the overall concept is to identify how the capability of current approaches could be expanded by developing a framework to create maintenance candidates that will both optimise benefits to road users and the environment and deliver adequate road owner life cycle costs.

The LCCA is done scheme by scheme since additional information is needed that is not available in the scheme selection and since an operator is needed to adjust the suggestions made by the system. A more automated system excluding the operator step would be too complicated and rigid.

Each treatment needs information about:

Effects during maintenance

o Road closure design, including night/day shift and closure length

o Road closure duration or treatment progress speed

o Costs for road owner per m2

o Width of treatment

Effects after maintenance, during service life

o Improvement of trigger parameters

o Service life of each pavement layer, including quality aspects, and the resulting average trigger parameters during the service life

Furthermore, each scheme needs information about possible reroutes and the probable maximum delay per vehicle. Maximum delay is a mechanism to catch effects of that road users will find alternative routes without any planning efforts by the road manager as well as avoiding problems with unrealistically long delays coming from simplified models. However, rerouting may cause problems if for example traffic is finding its way through areas with high accident risk or being environmentally sensitive.

Calculation is based on the treatment service life, even though the effects of the treatment may last longer. After this period another treatment is selected and a new decision can be made and LCCA conducted. To include effects on components with a life span longer than the treatment service life a separate handling is necessary.


Page 41 of 43

Figure 17 Illustration of the technical parameter values used to reflect the effect of each maintenance treatments on service life and condition after treatment.

Delays during the life cycle can be caused by maintenance and by inferior performance. Delays due to inferior performance are based on the fact that road users reduce speed when the condition is inferior or function is reduced. Inferior condition in ToolBox is expressed by models using IRI, which is then transformed into trigger based models. Delays during maintenance are calculated using the STARS tool in which the total number of delay hours can be extracted. To avoid unrealistic delays, which is not appearing in real, a maximum delay per vehicle is introduced. The maximum delay is the same as the extra time estimated by individual drivers when choosing reroutes.

Vehicle operation costs are usually very high compared to the costs of maintenance and a truncation error like problem will appear in the minds of decision makers if they do not look at the actual differences between alternative maintenance treatments. In ToolBox a reference level for all triggers is used as comparison. Only the vehicle costs relevant to choosing maintenance alternatives are included in the analysis.

Emissions are similar to fuel consumption in that the difference between alternatives is the relevant piece of information and that models are related to surface condition indicators such as IRI and MPD. Local emissions, noise and particulates also need information about the exposure situation such as number and location of residents.

CI, TP

Year

TP1

TP2

Average TP

Expected life

Average TP


Page 42 of 43

6 Next steps in ToolBox

The next steps are to finalise the ToolBox tool and to consider whether the thresholds suggested within this report should be adjusted for the Demonstration.

Demonstration networks have been chosen and these are a combination of the road types motorways and primary roads and cover a range of road quality. Between 1000 and 2000 km of data has been obtained for each partner country and this will be used in the demonstration.

The project will end with a demonstration of the tool (Deliverable D4) and a final summary report, D5.

.


Page 43 of 43

Sources

2013R08EN: “Maintenance methods and strategies”. Technical committee D.2 – road Pavements. ISBN 978-2-84060-323-8. PIARC publication http://www.piarc.org.

Benbow E and L Sjögren: “ToolBox, Review of Functional Triggers”. Deliverable 2 of the ToolBox project (2013).

Hammarström U, Eriksson J, Karlsson R and Yahya M-R: “Rolling resistance model, fuel consumption model and the traffic energy saving potential from changed road surface conditions”. VTI report 748A. (2012)

HD29/08: “Data for Pavement Assessment” Design Manual for Roads and Bridges (DMRB), Volume 7, Section 3, Part 2, HD29/08. http://www.dft.gov.uk/ha/standards/dmrb/vol7/section3/hd2908.pdf (2008)

HD30/08: “Maintenance Assessment Procedure”. Design Manual for Roads and Bridges (DMRB) Volume 7, Section 3, Part 3. http://www.dft.gov.uk/ha/standards/dmrb/vol7/section3/hd3008.pdf Lang J (WSP), J Berglöf (WSP) and L Sjögren (VTI): “Index för att beskriva vägars tillstånd”. WSP report (2013). (Indices to describe road condition, only in Swedish)

McRobbie S G, H Viner & M A Wright: “The use of surface texture measurements to predict pavement surface type and noise characteristics”. TRL project report PR CSN/32/03, Transport Research Laboratory, Crowthorne, UK (2003).

Swedish Transport Administration (Trafikverket): “The maintenance standard for paved roads 2011” (2012:074).

SWEEP: “HAPMS project: Technical description of SWEEP.S”. Highways Agency document, owned by the HAPMS Project Office (The Highways Agency, 29th Floor, Euston Tower, 286 Euston Road, London NW1 3AT). 2008.

ToolBox, Decision Making Tool for Optimising …...Deliverable Nr 3 – ToolBox, Decision Making Tool for Optimising Lengths for Maintenance Due date of deliverable: 30.09.2013 Actual

Documents