Life Cycle Cost Analysis for Turnouts: A comparison between straight and bent turnouts

,

STOCKHOLM SWEDEN 2020

Life Cycle Cost Analysis for

Turnouts

A comparison between straight and bent turnouts

VASILIKI-ROUMPINI ARGYRI

TRITA-ABE-MBT-2045

Life Cycle Cost Analysis for turnouts

A comparison between straight and bent turnouts

Vasiliki-Roumpini Argyri

Master’s thesis April 2020

School of Architecture and Built Environment KTH Railway Group

KTH Royal Ins tute of Technology Stockholm, Sweden

“Τhe true courage is to admit that the light at the end of the tunnel is most likely the   headlights of another train approaching us from the opposite direction”  

Acknowledgements

I would like to address a big thank you note to Arne Nissen from Trafikverket for providing me the data for the research, but most importantly for his unlimited availability and valuable guidance through the whole process of this thesis. Without his experience, exper se, older research but mainly his willingness to help this project would have never been completed. Special thanks to Andreas Bäckman, my supervisor from Sweco for trus ng me with this challenging research which he set the founda ons for and helped me get back on track whenever necessary.

A big thank you to Anders Lindahl, my supervisor from KTH, for his constant support since the very first day of classes in KTH un l today. He has been an inspiring and caring teacher making students excited about railways.

I want to thank Jan Dahlberg, my boss in Sweco for providing me with the me and work flexibility needed in order to complete this project.

Finally, I want to thank the people closer to me for their pa ence and support when I needed it the most.

Abstract

In a world with unlimited travel op ons, railways play a key role in transporta on. In order to serve the demand at a sa sfactory level, it is important that the infrastructure quality remains high and safe. Maintenance is then the most important aspect of railway infrastructure.

This project’s aim is to develop a tool that would evaluate the cost differences and maintenance needs during the life cycle of turnouts, bent with different radii to straight, as a crucial part of the infrastructure, not only technically but also financially. When the cost over a life cycle is provided then design decisions can get more efficient.

Maintenance history of seven years of preven ve and correc ve maintenance data from databases Bessy and 0felia for single turnouts across the Swedish rail network were studied, analysed and evaluated. Along with informa on from interviews with key informants the cost driving parameters were specified. The calculator was developed in Microso Excel, giving results for bent turnouts in 4 different radii categories and the respec ve straight turnouts. An EV-UIC60-760-1:14 turnout was used as a case study for different radii categories and 3 different scenarios were run in order to test the robustness of the tool. The results showed that bent turnouts have a higher life cycle cost than straight in the order of 1 to 3 mkr depending on the radius, the bigger share of which is usually the preven ve maintenance cost, with the specifics to vary between the categories and different scenarios tested. The way maintenance data are registered and classified plays an important role in the analysis.

Sammanfa ning

I en värld med obegränsade resemöjligheter spelar järnvägen en vik g roll för transporter. För a llgodose transporte erfrågan på en llfredsställande nivå är det vik gt a infrastrukturen har en god och säker kvalité. Underhållet är då den vik gaste delen för en fungerande järnvägsinfrastruktur.

Denna rapports mål är a utveckla e beräkningsverktyg som ska utvärdera kostnadsskillnaderna och underhållsbehovet under livscykeln för böjda spårväxlar med varierande olika radier samt raka, som en avgörande del av infrastrukturen, inte bara tekniskt utan också ekonomiskt. När kostnaden över en livscykel beräknas kan beslut om u ormning bli bä re.

Underhållshistorik för sju års förebyggande och korrigerande underhållsdata från databaserna Bessy och 0felia för enstaka spårväxlar i det svenska järnvägsnätet har studerats, analyserats och utvärderats. Tillsammans med informa on från intervjuer med sakkunniga specificerades kostnadsparametrarna. Beräkningarna har genomförts i Excel, vilket gav resultat för böjda spårväxlar med fyra olika radier samt raka spårväxlar. En spårväxel, EV-UIC60-760-1: 14 har använts för en fallstudie för de olika radiekategorier och 3 olika scenarier skapades för a testa beräkningsmodellens robusthet.

Resultaten visade a böjda spårväxlar har en högre livscykelkostnad än raka i storleksordningen 1 ll 3 mkr beroende på radie, där den största andel vanligtvis är den förebyggande underhållskostnaden, där specifika onerna varierar mellan hur de olika kategorierna och olika scenarier testas. Även hur underhållsdata registreras och klassificeras spelar en vik g roll i analysen.

Περίληψη

Σε έναν κόσμο με απεριόριστες επιλογές μετακίνησης, ο σιδηρόδρομος παίζει πρωταγωνιστικό ρόλο στις μεταφορές. Η διατήρηση υψηλής ποιότητας των υποδομών καθώς και υψηλών στάνταρ ασφαλείας αποτελούν σημαντικές προϋποθέσεις για την ικανοποίηση της ζήτησης σε ικανοποιητικό επίπεδο. Στόχος αυτής της εργασίας είναι η ανάπτυξη ενός εργαλείου που θα αξιολογεί τις διαφορές κόστους και αναγκών συντήρησης των σιδηροδρομικών αλλαγών (ψαλιδιών), συγκρίνοντας όσα είναι τοποθετημένα σε ευθυγραμμία με αυτά που βρίσκονται σε καμπύλη. Τα τμήματα αυτά θεωρούνται ζωτικής σημασίας στοιχεία των σιδηροδρόμων σε τεχνικό και οικονομικό επίπεδο. Όταν το κόστος κύκλου ζωής είναι διαθέσιμο, τότε οι επιλογή μεταξύ των διαθέσιμων εναλλακτικών κατά τη σχεδίαση μπορεί να γίνει πιο αποτελεσματικά. Ιστορικό προληπτικής και διορθωτικής συντήρησης από τις αντίστοιχες βάσεις δεδομένων Bessy και 0felia για σιδηροδρομικές αλλαγές από όλο το σουηδικό δίκτυο αναλύθηκε και αξιολογήθηκε. Οι πιο σημαντικοί οικονομικοί παράγοντες προσδιορίστηκαν μέσω της ανάλυσης των δεδομένων καθώς και μέσω συνεντεύξεων με πληροφορητές-κλειδιά. Το υπολογιστικό εργαλείο αναπτύχθηκε στο Excel, δίνοντας αποτελέσματα για σιδηροδρομικές αλλαγές τοποθετημένες σε καμπύλες χωρισμένες σε 4 διαφορετικές κατηγορίες μεγέθους ακτίνας όπως επίσης και για τις αντίστοιχες περιπτώσεις ψαλιδιών τοποθετημένων σε ευθυγραμμία. Στα αποτελέσματα παρουσιάζεται μία μελέτη περίπτωσης από κάθε κατηγορία μεγέθους ακτίνας και 3 διαφορετικά σενάρια προκειμένου να ελεγχθεί η ευαισθησία του εργαλείου στις αλλαγές των διαφόρων παραμέτρων που υπεισέρχονται στους υπολογισμούς. Τα αποτελέσματα έδειξαν ότι οι σιδηροδρομικές αλλαγές που είναι τοποθετημένες σε καμπύλη έχουν μεγαλύτερο κόστος κύκλου ζωής από αυτές που βρίσκονται σε ευθυγραμμία, της τάξης των 1 έως 3 εκατομμυρίων σουηδικών κορωνών αναλόγως με την ακτίνα της καμπύλης. Το μεγαλύτερο μέρος του κόστους συνήθως προκύπτει από την προληπτική συντήρηση με επιμέρους διαφοροποιήσεις μεταξύ των κατηγοριών και σεναρίων που ελέγχθηκαν. Ο τρόπος καταγραφής και κατηγοριοποίησης των εργασιών συντήρησης στις βάσεις δεδομένων παρατηρείται να παίζει σημαντικό ρόλο και να επηρεάζει την ανάλυση.

Table of contents

Acknowledgements 5

Abstract 7

Sammanfa ning 8

Περίληψη 9

List of Abbrevia ons 13

1 Introduc on 15

1.1 Background 15

1.2 Hypothesis 16

1.2.1 Scope and research objec ves 16

2 Literature Review 19

2.1 Turnout types 19

2.1.1 Naming 19

2.1.2 Geometrical Sor ng 20

2.1.3 Group of Standard turnouts (Standardsor ment) 22 2.1.4 Group of turnouts under management (Sor ment förvaltning) 22

2.1.5 Se lement group of turnouts 24

2.1.6 Bent turnouts 25

2.2 Turnout geometry 27

2.3 Turnout wear and tear 29

2.4 Maintenance 30

2.4.1 Maintenance principles 30

2.4.2 P-F Interval 32

2.4.3 RAMS 33

2.5 Sta s cal analysis and assump ons 34

2.5.1 System reliability 34

2.5.2 Bathtub curve 35

2.5.3 Mean system down me 36

2.5.4 Clusteriza on of repairable systems 36

2.5.5 Iden fying outliers 37

2.6 Life Cycle Cost Analysis 37

2.6.1 Basic stages and concepts 38

2.6.2 Terms and formulas 40

3 Methodology 43

3.2 Interviews 43 3.3 Assump ons 44 3.4 Data analysis 45 3.5 Clusteriza on 46 3.6 Delimita ons 47 3.7 Cost drivers 49

3.8 Maintenance cost calcula on 50

3.8.1 Correc ve maintenance 51

3.8.2 Preven ve maintenance 52

4 Results 55

4.1 Failure sta s cs 55

4.1.1 Sta s cal significance 55

4.1.2 Correc ve maintenance failure rate 55

4.1.3 Preven ve maintenance need-inspec on remarks 63

5 LCC calculator 71

5.1 Model development basis 71

5.2 Case study 72

5.3 Sensi vity analysis 74

5.3.1 Life length/MGTs 74

5.3.2 Different turnout type-more frequent large replacements 76

5.3.3 Longer PM me 76

6. Discussion and Conclusion 79

6.1 Commen ng on the results 79

6.2 Uncertainty 81

6.3 Conclusion 82

7. Future Studies and sugges ons 83

8. References 85 Appendix 1 89 Appendix 2 90 Appendix 3 99 Appendix 4 101 Appendix 5 109

List of Abbrevia ons

ANN -Annuity

CE -Cost Effec veness

CM -Correc ve Maintenance

ERRAC -European Rail Research Advisory Council EV -Enkel Växle, Single Turnout

HPP -Homogeneous Poisson Process idd -independently iden cally distributed IQR -Interquar le Range

LCC -Life Cycle Cost LRT -Light Rail Transit MDT -Mean Down me MGT -Million Gross Tons MTTR -Mean Time to Repair MTTF -Mean Time to Failure MS -Maintenance Strategy MUT -Mean Up Time

NHPP -Non-Homogeneous Poisson Process NPV -Net Present Value

PM -Preven ve Maintenance R -Radius

RAMS -Reliability, Availability, Maintenance, Safety RQ -Research Ques on

RO -Research Objec ve SE -System Effec veness TBF -Time Between Failures TPV -Total Present Value

TRV -Trafikverket, Swedish Transport Administra on λ -failure rate

1

Introduc on

This chapter includes the background, hypothesis, scope and research ques ons of the topic studied. The methodology followed during the research process and the delimita ons of it are also presented in this sec on. Chapter 2 contains the theore cal background of the subjects studied and the defini ons of the terms and concepts involved. In chapter 3 the assump ons made in the study, the data used and their analysis procedure are presented. This chapter also includes the LCC calcula on model development and the different scenarios tested. The discussion of the results takes place in chapter 4. The conclusions are summarized in chapter 5 and in chapter 6 future studies and development are proposed.

1.1 Background

According to ERRAC (ERRAC 2017) rail (commuter, metro, tram and LRT) represents 45% of public transport in Europe, while rail’s freight transport share is 18,3%. Increasing urbanisa on urges the need for efficient transporta on. This demand can be served by rail and requires cu ng edge and sustainable technological solu ons. Digitaliza on of railways’ assets can maximize their produc ve u liza on, reducing the total cost of ownership over their life cycle. Working towards ERRAC’s vision for 2050, higher capacity is required and more traffic over the exis ng network. In such a case, it is important to use assets with fewer maintenance requirements and service interrup ons.

Large part of railway cost consists of maintenance and track renewal expenses. Close monitoring of track condi ons through the whole length of the railway network and especially on the most sensi ve parts of it is important for fast and effec ve maintenance ac ons. Such points are turnouts which have a complicated structure consis ng of expensive elements compared to the rest of the track. They require a high investment cost, that can be even four mes higher than that of track, and large-scale maintenance. The produc vity and the speed of the line depends on the number and the type of turnouts (Lichtberger 2005) . Alongside with digitaliza on which would save valuable me by repor ng failures along the network or performance records that could lead to preven ng failure itself and reduce maintenance-related down me (Bergquist & Söderholm 2015) , an interpreta on of this reported performance in financial terms is a very important aspect. Following a life cycle cost analysis approach using the reported data of failure would give useful insights of the financial aspect of installing, maintaining and replacing turnouts on the railway network. Such a tool would aid the assessment of different types of turnouts, evalua ng the performance and economy of their different parts, assis ng respec ve stakeholder’s maintenance and design decisions.

A PhD study on the Life Cycle Cost of turnouts was conducted by Arne Nissen (Nissen 2009) covering a wide range of turnout types and related parameters. The study includes data from databases BIS, 0felia, Bessy, TFÖR and Agresso, but also data from track recording cars to iden fy irregulari es on the tracks. In this research it is concluded that the proper data

are available in Banverket databases, but the way these are being reported and stored has to be improved in order to be effec vely used for analysis and decision making. An LCC model is considered by this disserta on to be a suitable tool primarily for comparing different alterna ves and traffic scenarios instead of a cost calculator for the whole life cycle of an item.

1.2 Hypothesis

Based on empirical observa on of increased maintenance need in the case of bent turnouts, wear and tear is considered to be higher compared to that caused in the case of straight turnouts. This results in a higher overall life cycle cost of the first.

1.2.1 Scope and research objec ves

In order to quan fy the increased wear of bent turnouts, an LCC calcula on tool is developed, suitable to compare the cost over the life cycle of different types of turnouts, bent over straight, bent with different curvature and geometrical characteris cs. At the end of the research, the following research ques ons should be answered.

RQ 1 How much can a turnout be bent un l it gets financially not efficient? RQ 2 Which are the parts of the turnout that are financially crucial?

RQ 3 What type of failure is more common? How does the failure type differ between the different types of turnouts?

RQ 4 Which part of a turnout’s life cycle is the most costly?

The following research objec ves are set in order to form the framework within which the above men oned hypothesis can be verified or rejected:

RO 1 Review the type of turnouts and their components, the maintenance strategies followed and the Life Cycle Cos ng procedure that has to be followed while making the tool.

RO 2 Gather and analyse the required maintenance history and cost data in order to iden fy the cost driving parameters that primarily affect the overall cost of the life cycle.

RO 3 Develop the life cycle cost calcula on tool.

RO 4 Check the robustness of the model by tes ng a case study chosen to highlight the importance of the different parameters involved in the analysis.

V.R . A rg yri | L ife C yc le C os t A na lys is f or Tu rn ou ts KTH| R oy al Instut e of Technology 2020 1 7

2

Literature Review

In order to develop a Life Cycle Calcula on tool assessing the cost over the life cycle of different types of turnouts, it is necessary to set the theore cal background where the model is based on, using relevant literature. In this chapter, RO1 is applied. Ma ers concerning the geometry and the characteris cs of different types of turnouts, such as straight, fully and par ally bent turnouts, the different groups these belong to and the various geometrical and other parameters of each are presented. Furthermore, this chapter con nues by reviewing the different maintenance techniques and rou nes followed in railways and the way these are targeted on turnouts’ wear and tear. Finally, in the last part of this sec on, the concept and guidelines for an LCC analysis are presented and similar studies and tools found in literature are listed along with their contribu on in the development of the tool of this thesis. Through this process the framework for the development, the assessment and the discussion of the results of the tool is set.

2.1 Turnout types

The use of turnouts is essen al for the undisturbed run of trains on the track while changing course. They divide the track into two or even three tracks running on the same level and in both direc ons. (Esveld 2001)

According to the Swedish Transport Administra on’s (Trafikverket 2016b) requirements, turnouts in Sweden are divided in three main categories as described further below. The naming standards of the turnouts are also presented following.

2.1.1 Naming

The new swedish standards (Trafikverket 2015a) for turnout naming are as demonstrated in the following figure.

Figure 2: Turnout naming

The radius at the above turnout naming corresponds to R ₀ as explained in paragraph 2.1.5. The first le ers of a turnout name describe the turnout type. The different types of a turnout and a schema c representa on of each can be found following.

2.1.2 Geometrical Sor ng

EV: Single turnout (enkel växel) is a turnout that has one diverging track and can be right or le hand depending on the direc on of the diverging track from the beginning of the switch towards the point of crossing (Hay 1982) .

Figure 3: Right hand EV turnout (Trafikverket 2015a)

EVR: Single turnout with movable point frog (enkel växel med rörlig korsningsspets) is a turnout with one diverging track and a movable point frog. The frog point moves to close the gap of the frog depending on the direc on the switch is thrown. It is used on heavy tonnage lines where the jump of the wheel over the frog gap would otherwise cause high wear and tear on the frog point and the flangeway (Ruppert 2018) .

Figure 4: EVR turnout (Trafikverket 2015a)

KRYSSVX: Double hand crossover (kryssväxel) A crossing connec on between two tracks.

Figure 5: KRYSSVX turnout (Trafikverket 2015a)

SPK: Diamond crossing (spårkorsning) is a crossing which allows trains to cross another track in the same eleva on at any angle. The trains cannot change tracks. (Ruppert 2018)

Figure 6: SPK turnout (Trafikverket 2015a)

DKV: Diamond crossing with double slips (dubbel korsningsväxel) is the type of turnout where two tracks are crossing each other and all four different crossing paths have connec on with each other.

Figure 7: DKV turnout (Trafikverket 2015a)

EKV: Diamond crossing with single slip (enkel korsningsväxel) is the type of turnout where two tracks are crossing each other and two crossing paths have connec on with each other.

Figure 8: EKV turnout (Trafikverket 2015a)

NOS: Non-symmetrical single turnout (något osymmetrisk enkel växel) where the diverging

and the through track are diverging from the symmetry line going through the crossing with several different radii.

Figure 9: NOS turnout (Trafikverket 2015a)

3V: Combined turnout (tredelig växel) is one compound turnout consis ng of two single turnouts.

Figure 10: 3V turnout (Trafikverket 2015a)

IBV: Single turnout (Innerbåge) where the main and the diverging track are bent towards the same direc on.

Figure 11: IBV turnout (Trafikverket 2015a)

YBV: Single turnout (y erbåge) where the main and diverging tracks are bent towards different direc ons .

Figure 12: YBV turnout (Trafikverket 2015a)

SYM: Equilateral turnout (symmetrisk växel) is the type of bent turnout where the tracks are diverging with the same radius from the symmetry line that passes from the point of crossing.

As men oned in chapter 1, it was decided that only single turnouts would be used for the analysis in this project. Other categories with more complex geometry possibly have different wear and tear levels and mechanisms and a comparison between mixed turnout types would lead to less safe conclusions.

2.1.3 Group of Standard turnouts (Standardsor ment)

This category includes all the turnout alterna ves available to choose from when a new turnout has to be placed or when an exis ng one has to be changed or rebuilt. These alterna ves for single turnouts in the main track (huvudspår) are listed below and consist of concrete sleepers and manganese crossing for 60E (table 1) and wooden sleepers for BV50 turnouts (table 2).

Table 1: Standard 60E single turnouts with concrete sleepers and manganese crossing (Trafikverket 2016b) Turnout type (main track)

single turnout (enkel växel)

single turnout with movable point frog (enkel växel med rörlig korsningsspets) EV-60E-208-1:9 EVR-60E-300-1:8,21 EV-60E-300-1:9 EVR-60E-300-1:9 EV-60E-500-1:12 EVR-60E-760-1:14 EV-60E-760-1:14 EVR-60E-760-1:15 EV-60E-760-1:15 EVR-60E-2500-1:26,5 EV-60E-1200-1:18,5 EVR-60E-2500-1:27,5 EV-60E-580-1:13 (exception needed) EV-60E-580-1:15 (exception needed)

Table 2: Standard BV 50 single turnout with wooden sleepers and steel crossing (Trafikverket 2016b) Turnout type (main track) Cross material

single turnout (enkel växel) EV-BV50-225/190-1:9 steel

There are some addi onal turnout types for side tracks. These are the ones included in marshaling yards and are not listed here since they are considered to be used under special circumstances and condi ons where the trains run with a maximum speed of 30 km/h and the wear and tear of the tracks and turnouts is expected to be different.

2.1.4 Group of turnouts under management (Sor ment förvaltning)

This group includes the largest number of turnouts that can be found in use. The turnouts of this category are no longer produced, but there are s ll spare parts to maintain them, the documenta on and exper se required to support them. Single turnouts under this category are presented in the following table (Trafikverket 2015b) .

Table 3: Single turnouts of managed group with UIC60 rail profile, concrete sleepers and steel or manganese crossing (Trafikverket 2015b)

Turnout type (main track) single turnout (enkel

växel)

single turnout with movable point frog (enkel växel med rörlig

korsningsspets) EV-UIC60-300-1:9 EVR-UIC60-760-1:14 EV-UIC60-500-1:12 EVR-UIC60-760-1:15 EV-UIC60-760-1:14 EVR-UIC60-2500-1:26,5 EV-UIC60-760-1:15 EVR-UIC60-2500-1:27,5 EV-UIC60-1200-1:18,5

Table 4: Single turnouts of managed group with BV50 and SJ50 rail profile (Trafikverket 2015b) Turnout type (main track)

EV-BV50-190-1:8,1 EV-SJ50-5,9-1:9 EV-BV50-300-1:9 EV-SJ50-7,85-4,8-SYM EV-BV50-225/190-1:9 EV-SJ50-8,4-1:6,28 EV-BV50-225/480-1:12 EV-SJ50-8,4-1:7,5 EV-BV50-600/365-1:12 EV-SJ50-8,4-1:8,1 EV-BV50-600-1:13 EV-SJ50-8,4-1:9 EV-BV50-600-1:15 EV-SJ50-11-1:9 EV-SJ50-300-1:9 EV-SJ50-11-1:12 EV-SJ50-12-1:13 EV-SJ50-12-1:15 EV-SJ50-12-1:15-SYM

2.1.5 Se lement group of turnouts

Turnouts that have stopped being produced or supported with documenta on and drawings from trafikverket are included in this category. These turnouts are gradually being replaced with others from the previous two groups. (Trafikverket 2015b)

Table 5: Single turnouts of the se lement group with BV50, SJ50, SJ43 and S49 rail profiles (Trafikverket 2015b) Turnout type (main track)

EV-BV50-190-1:7,5 EV-SJ43-4,5-1:9 EV-SJ50-5,9-1:9 EV-SJ43-5,9-1:9 EV-SJ50-5,9-1:12 EV-SJ43-5,9-1:10 EV-SJ50-7,85-1:4,8-SYM EV-SJ43-5,9-1:12 EV-SJ50-7,85-1:7,47-SYM EV-SJ43-6,1-1:9 EV-SJ50-8,4-1:9 EV-SJ43-7,85-1:4,8-SYM EV-SJ50-11-1:9 EV-SJ43-8,4-1:6,28 EV-SJ50-11-1:12 EV-SJ43-8,4-1:6,6 EV-SJ50-12-1:9 EV-SJ43-8,4-1:7,5 EV-SJ50-12-1:12 EV-SJ43-8,4-1:8,1 EV-SJ50-12-1:13 EV-SJ43-8,4-1:9 EV-SJ50-12-1:15 EV-SJ43-11-1:9 EV-SJ50-12-1:15-SYM EV-SJ43-11-1:12 EV-S49-8,4-1:4,8-SYM EV-SJ43-10-1:9 EV-S49-11-1:7,5/6,6 EV-SJ43-10-1:12 EV-S49-20,667-1:18,5 EV-SJ43-10-1:15

60E turnouts are the most contemporary ones but since they were rela vely recently introduced (2014) not only their use is less extended yet but also the period that they have been used is very limited compared to their life expectancy. Turnouts have a long life cycle that can be considered up to 40 years, thus a long me period is needed un l all the old turnouts will be replaced by the new ones included in the first above men oned category. Nevertheless, consultancies currently design for the new type of turnouts which are the ones that the industry produces.

Even though the engineering, financial and market interest is targeted in the standard assortment of turnouts, the research cannot be limited just in 60E turnouts, but requires the addi on of the rest of the groups in order to get a large enough sample for the analysis. The above listed 60E turnouts share characteris cs and can be replaced by the directly corresponding turnouts of the other two groups as shown in table 6 below.

Table 6: EV turnouts with corresponding geometry (Trafikverket 2018)

Simple Turnouts (EV) with corresponding geometry

EV-60E EV-UIC60 BV50 SJ50

EV-60E-208-1:9 EV-BV50-225/190-1:9 EV-SJ50-11-1:9

EV-60E-300-1:9 EV-UIC60-300-1:9 EV-BV50-300-1:9 EV-SJ50-300-1:9

EV-60E-500-1:12 EV-UIC60-500-1:12

EV-60E-760-1:14 EV-UIC60-760-1:14

EV-60E-760-1:15 EV-UIC60-760-1:15

EV-60E-1200-1:18,5 EV-UIC60-1200-1:18,5

EV-60E-580-1:13 EV-BV50-600-1:13 EV-SJ50-12-1:13

EV-60E-580-1:15 EV-BV50-600-1:15 EV-SJ50-12-1:15

with movable frog EVR-60E-300-1:8,21 EVR-60E-300-1:9 EVR-60E-760-1:14 EVR-UIC60-760-1:14 EVR-60E-760-1:15 EVR-UIC60-760-1:15 EVR-60E-2500-1:26,5 EVR-UIC60-2500-1:26,5 EVR-60E-2500-1:27,5 EVR-UIC60-2500-1:27,5 2.1.6 Bent turnouts

Table 7: Bent turnouts Bent turnouts

IBV Innerbåge

YBV Ytterbåge

SYM Symetrisk

The characteris cs of bent turnouts are as explained in sec on 2.1.2. The following rules apply on them (Trafikverket 2016a) :

● Bending is not allowed for turnouts with movable frog

● Turnouts should be bent with a radius larger than the design limit for each turnout type as presented in the following table for the standard turnouts.

Table 8: Radii limits for bent turnouts (Trafikverket 2016a) Minimum allowed radius (m)

turnout type (originally) straight track (stamspår) min R1 diverging track (grenspår)min R2 EV-60E-208-1:9 2000 188,702 EV-60E-300-1:9 1000 230 EV-60E-500-1:12 600 273 EV-60E-580-1:13 (exception needed) 600 294,559 EV-60E-580-1:15 (exception needed) 600 294,559 EV-60E-760-1:14 600 335 EV-60E-760-1:15 600 335 EV-60E-1200-1:18,5 600 400 EV-BV50-225/190-1:9 2000 173,5

The horizontal radius of the devia ng track of a bent turnout is calculated using the following formula (Trafikverket 2016a) :

, where:

2 R1 0) (R1 0)

R = ( _{* R}

/

+ R

R0-the radius of the devia ng track of the straight turnout

R1-the radius of the originally straight track of the curved turnout R2-the radius of the devia ng track of the bent turnout

Basic principles and characteris cs of the bent turnouts can be found in TDOK 2013:0478 (Trafikverket 2018) . It is stated there that if possible turnouts should be avoided to be placed in curves or have cant. This is due to several reasons as stated below:

● Turnouts on a curve have a higher risk of interfering with train traffic. ● They limit the speed in the original track.

● They need more maintenance.

● The me required for their maintenance is 2 to 3 mes more than that required for straight turnouts.

● They have shorter life expectancy.

● If spare parts for such kind of turnouts have to be replaced, the delivery me for them takes several days while in the case of straight turnouts it only takes a few hours.

The heavier the traffic on the turnout, the more cri cal the above men oned factors can get. If it is necessary for a turnout to be placed on a curve then the largest possible radius and the lowest possible cant should be used. A be er solu on in these cases would be the BKS to be placed on a straight while the rest of the turnout would be on a curve. For turnout connec ons with speeds up to 70 or 80 km/h between main tracks (huvudspår) on curve, turnout connec ons with angle 1:18,5 should be used, while in case of limited radius and cant, 1:15 turnout connec ons may be used (Trafikverket 2018) .

Turnouts with movable point frogs cannot be bent according to the regula ons, but in the databases used such turnouts were found to be bent as well.

2.2 Turnout geometry

The basic components and terminology of the turnout parts and the geometry of it are presented in this sec on.

Figure 14: Turnout components terminology in english and swedish (Pålsson Göteborg, Sweden 2011)

There are differences when naming the different components of a turnout from country to country. Following a list with the different components name and use at the turnout is given.

1. The switch guides the wheel in one of the two direc ons of the turnout. It is the moving part of the turnout layout which is made from two long blades, the switch

rails , moving from one stock rail, which is the non-moving part of the switch , to another guiding the train. The edges of the switch blades are called switch toes . 2. The crossing is the non-moving part of the turnout layout. It creates a gap in the rail

for the flange to pass through in either direc on. On either side of the crossing nose ,

wing and check/guardrails are provided to assist the guidance of the wheel sets through the crossing (Network Rail 2012) .

3. The point machine is the device moving the points of the switch (Network Rail 2012) . 4. The sleepers maintain the appropriate gauge width, distribute the wheel loads from the rails to the ballast and keep the track stable against lateral longitudinal and ver cal movement (Edwards 2018) .

5. Ballast is crushed stone in several different sizes. It reduces the sleepers pressure to the subgrade by spreading them to a wider interface. It stabilizes the sleepers by resis ng the movements caused by the forces from the train movement. The most important func on of ballast is that it provides adequate drainage maintaining the track structure (Dersch 2018) .

Figure 16: Schema c representa on of different lengths of EV turnouts (Trafikverket 2018)

The pictures above give an overview of the turnout geometry, assis ng in comprehending the way the different components contribute to its opera on.

L3 has 0 length when BKS2 is placed on a curve and some variable length depending on the type of the turnout if placed on a straight (see table 1 of the appendix).

In the picture below the different cri cal measurements along the turnout are named and shown. Specific tolerances depending on the type of the turnout apply for each one of these points of measurement. If the findings while measuring these during preven ve maintenance inspec ons, are outside of the acceptable margin, there is the need for adjustment.

Figure 17: measurements’ placement along a simple turnout, source (Trafikverket 2019)

2.3 Turnout wear and tear

There is extensive research on wheel-rail interac on at turnouts, generated by the observa on that the failure report rate on these areas is high, such as the resul ng maintenance need. Important factor causing these faults is the disturbance of the nominal wheel-rail contact condi ons at various places across the turnouts. Some of the disturbances are caused by the existence of double rail where the transi on from stock rail to switch rail in the switch panel is performed but also in the crossing part where the wheel transfers between the wing and nose rail. These discon nui es of the turnouts cause mul ple wheel-rail contacts and high impact loads. What is more, the lack of transi on curves and the differences of the rail cross-sec ons at various parts of the turnout significantly affect the vehicle dynamics (Kassa et al. 2006) . Transi on areas in turnouts are built-in irregulari es, which along with the insula on joints are sensi ve areas of the track. What is more, unstrained rail end is an important geometric irregularity (Steenbergen 2008) . There are weak rail cross-sec on areas in turnouts, such as the switch rail, where the rail head is smaller but they have to carry the same load. This results in higher wear in these areas, such as fa gue crack (Pålsson 2014) .

Based on simula ons, it is observed that even if the geometry of the turnout is in an ideal condi on, there is a large ver cal and lateral dynamic loading of the wheel-track system at the crossing nose (Alfi & Bruni 2009) . Furthermore, for a given combina on of wheel profile and axle load, the speed of the train has low impact on the contact pressure, in contradic on to an increase in axle load. High axle load and low speed would lead to an increased wear measurement for the worn wheel profile (Kassa & Johansson 2006) .

Figure 18: Illustra on of two point contact situa ons in the switch panel at the le picture and in the crossing panel at the le picture, source: (Kassa et al. 2006)

Concerning degrada on, it is observed that for large radius curves there is higher rolling contact fa gue (RCF), which is defined as wear leading to spalling, on the high but also the

inner rail, when there is an increased level of lateral track irregulari es. On the other hand, for small radius curves wear, which is defined as the loss of material from a contac ng surface, is dominant in the high rail (Kartunnen 2015) . Wear, rolling contact fa gue and plas c deforma ons are the most common damage mechanisms on turnouts that cause turnouts to have the highest propor on of all maintenance costs in the Swedish network (Casanueva et al. 2014) .

In the picture below the geometry around the crossing of a turnout is visible. While the wheel passes over the crossing it first comes in contact with the wing rail and the contact point will gradually move towards the outside of the wheel profile. The lateral mo on of the wheelset is restrained by the check rail and possible contact with the crossing nose is prevented (Pålsson 2014) .

Figure 19: Crossing geometry of a turnout

2.4 Maintenance

Maintenance is the “ combina on of all technical, administra ve and managerial ac ons during the life cycle of an item intended to retain it in, or restore it to, a state in which it can perform the required func on” (Norrbin & Stenström 2017) . It is a very important and demanding aspect of railways. Close monitoring of the infrastructure is required in order to fulfil the constantly increasing demands in punctuality, efficiency, but also higher axle loads and speeds. Track renewal is considered to be 70% of the annual total expenses for manual and mechanical maintenance and track renewal therefore good informa on and data availability is important for efficient expense reduc on and cost distribu on (Esveld 2001) .

2.4.1 Maintenance principles

The basic principle of the track maintenance procedure is demonstrated in figure 20. The green line demonstrates the deteriora on of the railway track in case no maintenance ac on is taken. Phase a occurs soon a er the construc on or replacement of the track component due to its se lement and is quite unpredictable. Phase b is the main phase of analysis which takes place during the track’s life- me. Phase c should not be allowed to happen since rapid deteriora on takes place which may affect the safety of the infrastructure (Esveld 2001) .

Each peak reduces with maintenance ac ons which keep the quality of the track within the acceptable limit.

Figure 20: Maintenance principle for track deteriora on (Esveld 2001)

There are two main maintenance categories, preven ve maintenance (PM) which is planned and takes place while the system operates and correc ve maintenance (CM) which takes place a er a failure has occurred. PM is further divided into simple preven ve maintenance (1P) which changes the system reliability to some newer me and preven ve replacement (2P) which is correc ve replacement that restores the reliability curve to the new one. CM is dis nguished in minimal repair (1C) which makes no change in system me and restores the system reliability to it when it has failed and correc ve replacement (2C) which renews the system me to zero and the reliability curve is that of a new system. (Yuo-TernTsai et al. 2001) The different reliability changes depending on the maintenance type applied are demonstrated in figure 21. Depending on the state of the infrastructure, a re-investment or new investment may be required. In these cases the quality level gets higher than originally but larger expenses are required (Lindahl 2018) . The subcategories of each case are schema cally presented in figure 22.

Figure 21: Reliability vs me for different maintenance types (Yuo-TernTsai et al. 2001)

Figure 22: Schema c representa on of maintenance procedures (CENELEC 2010)

Concerning turnouts, maintenance is more manual compared to maintenance performed to improve track irregulari es on the line, since the components of the turnout are special and more delicate (Karis 2018) . Condi on based maintenance is performed aiming to the iden fica on of poten al failures in order to prevent func onal failures to occur and consecu vely to decrease train delays (Stenström et al. 2016) .

2.4.2 P-F Interval

The preven ve failure-func onal failure (P-F) interval is the interval between the detec on of a poten al failure and the degrada on in a func onal failure. In the case of preven ve maintenance P-F interval a predic on method used is the one described in figure 23 below, when the necessary monitoring condi ons apply. The Point P on the diagram indicates when the failure can be first detected and the slope of the curve a er this indicates the degrada on speed. The point F is where the component experiences a func onal failure and cannot fulfill its purpose anymore. Maintenance ac on must be taken within the interval between points P and F including inspec on and reac on me (Stenström 2015) .

Figure 23: P-F interval (Dann n.d.)

2.4.3 RAMS

Maintenance prac ces applied widely in railway infrastructure and specifically in Sweden are based on RAMS studies (Norrbin 2016) . The four piles of RAMS according to the european standards (CENELEC 2019) are described below:

Reliability: The probability that an item can perform a required func on under given condi ons for a given me interval.

Availability: The ability of a product to be in a state to perform a required func on under given condi ons at a given instant of me or over a given me interval, assuming that the required external resources are provided.

Two defini ons of availability are given below (Rausand & Høyland 2003) : Availability at me t: A(t)=Pr (item is func oning at me t)

Average availability when an item is repaired to an “as good as new” condi on every me it fails: av A = MT T F (MT T F

_/

+ MT T R)

Where:

● MTTF-mean me to failure is the mean func oning me of an item

● MTTR-mean me to repair is the mean down me a er a failure (not only the mean ac ve repair me)

Maintainability: The probability that a given ac ve maintenance ac on, for an item under given condi ons of use, can be carried out within a stated me interval when the maintenance is performed under stated condi ons and using stated procedures and resources.

RAMS targets depend on the significance of the failure that occurs which is taken into account by the infrastructure manager (Patra 2009) . The different possible failures are classified below.

Figure 24: RAMS failure categories (CENELEC 2019)

Up to 25% of the annual maintenance budget is spent on turnouts and crossings. Condi on-based maintenance management is considered to be the key for reducing the overall cost of maintenance, while decreasing preven ve maintenance. What is more, taking into considera on the Life Cycle Cost of the equipment used and the different maintenance ac vi es, the most appropriate maintenance ac on and P-F interval can be defined, reducing the cost of track maintenance. These ac ons may lead to improved returns on railway companies’ investments in RAMS. In accordance with the above, the cri cal elements and important parameters defining the condi on of the turnouts have to be clarified. (Esveld 2001)

2.5 Sta s cal analysis and assump ons

2.5.1 System reliability

When reliability data is studied, the concept of MTBF-mean me between failures-or failure rates of parts operated under similar condi ons, considering that the mes between failures (TBF) follow the i.d.d. (independently iden cally distributed) assump on for homogeneous Poisson process (HPP) (Navas et al. 2017) . This assump on has to be tested, searching for possible correla on in the available data by plo ng the MTBF to iden fy any possible trend during me or between the different operated parts. In the case of iden fica on of a trend at the failure rates, then a nonhomogeneous Poisson process (NHPP) is used to model the data (Rausand & Høyland 2003) .

r(N(t) ) n!) e P = n = (λn

_/

_* −λ

where: n=0, 1, 2, 3, …. the number of failures in the interval (0, t] and λ the mean number of failures in this interval

It is hard to accurately predict the me of failure of one component, thus sta s cal regularity of such data is studied. The not always accurate simplifica on that failures of the different components composing the studied system are independent events is made for this study. What is more, the different parts of the turnout are being replaced a er failure and they are considered as repairable components. If a component of the system fails and gets replaced within a short me by a similar component then it causes minimal down me. As a small part of the whole system is replaced then it is realis c to assume that the system reliability a er the repair is the same as before the failure occurred. This is the assump on of minimum repair or that the system is “as bad as old” a er the repair. This assump on is not en rely accurate, since it does not take into considera on the influence of the imperfect repairs to the reliability of the system. NHPP assumes that the repairs do not affect the reliability of the system which may apply for minimum repairs but not for larger scale repairs. In an a empt to adjust the failure rate due to imperfect repairs, adjustments such as decreasing parameters of the reliability index or a scale parameter of the failure rate were used in different models (Sun et al. 2007) .

2.5.2 Bathtub curve

Concerning failure data, increased failure rates are o en observed at the beginning and the end of the life me of a set of items studied. At the ini al phase of the life cycle there may be unpredicted issues coming out of the manufacturing of the item that caused this infant mortality. At the end of the life cycle the wear of the item is higher and more failures occur. In between these two life periods, there are the normal life failures that occur randomly due to “stress exceeding strength”. The above behaviour and phases of the life me are presented in the figure below demonstra ng the so-called bathtub curve of the failure rate of mechanical items (ReliaSo Publishing n.d.) .

Figure 25: The bathtub curve of the failure rate of mechanical items (Rausand & Høyland 2003)

UIC60 turnouts were first introduced in 1983. For the above men oned reasons and since, it is assumed that a er α five year period the manufacturing procedure is more standardized and the burn-in period has ended. Considering a 40 year life cycle for turnouts, the exis ng ones on the infrastructure have theore cally not reached the end of their life. In reality it is hard to know when the turnouts are replaced since no such data is saved on the databases.

The older ones from the EV60 turnouts that can be found around the network are around 7 years old, thus data for them cannot be excluded from the analysis. The analysis is performed during recent years when according to the manufacturer (Vossloh n.d.) the produc on, tes ng and installa on procedure is standardized thus this infant period of the life cycle is considered to be minimized. For older turnout types, there is the chance that their behaviour is studied while they are in the wear-out period of their life cycle.

2.5.3 Mean system down me

According to literature (Rausand & Høyland 2003) concerning unplanned down me caused by failures, a series structure of n independent components where each component i has constant failure rate λi, when i fails then the system has a down me MDTi for i=1,2,...,n. The probability that the system failure is caused by component i is pi=λi/

_∑

n j and a good

j=1

λ

approxima on of the mean down me for an unspecified failure is . The scheduled down me can be es mated from the DT ≈ i DT i j

M

∑

n

i=1λ * M

/

∑

n j=1λ opera onal plans.

For a repairable system star ng opera ng at t=0, when a failure occurs, a repair ac on is ini ated. If X(t) is the state variable of the item then for X(t)=1 the item is func oning at me t and for X(t)=0 the item is not func onal. The MDT includes the me needed to repair the failure MTTR but also the me to detect and diagnose the failure, logis c me and me to test and startup of the item. Mean me to failure MUT is equal to mean me to failure MTTF. The above are demonstrated in figure 26 below.

Figure 26: Average behaviour of a repairable item and main me concepts, source: (Rausand & Høyland 2003)

Consecu vely, the unavailability of a repairable item is the probability that the item is out of func on at me t: (t)A_′ = 1 − A = P(t) r(X(t))= 0 )

2.5.4 Clusteriza on of repairable systems

When the number of the items to be analyzed is quite high the reliability analysis can be a complex task, thus iden fica on of clusters of these items is an effec ve approach. Graphical representa on of the failure rate values λ(t) over me is one way to iden fy qualita ve differences and create groups of these items. In the NHPP case this approach can

lead to not balanced clusters in terms of the number of items that each would include, due to the trends iden fied in λ values (Navas et al. 2017) .

P = λ n! n

/

_{* e}−λ

2.5.5 Iden fying outliers

Outliers are observa ons with extreme values that can influence the result of a sta s cal analysis especially when the mean value is used. When performing such an analysis it is profitable to exclude these values from the sample to obtain more representa ve results. This is performed using the percentage points that separate the data into quarters, the quar les. The first quar le or 25th percen le is the point below which lies the 25% of the data, below the second quar le lies 50% of the data and so on. The interquar le range (IQR or fs) is the difference between the first and the third quar le and represents the data spread (Karla is et al. 2011) . Observa ons lower than 1.5*IQR from the 1st quar le or higher than the 3rd quar le are outliers. Observa ons respec vely lower or higher than 3*IQR are extreme outliers (Jenelius 2017) . This procedure was followed when calcula ng correc ve maintenance costs based on records of the databases as described in later stages.

Figure 27: Box Plots, percen les and outliers, source: (Jenelius 2017)

2.6 Life Cycle Cost Analysis

The performance of an element during its opera on extensively depends on the decisions made at the early phases of its life cycle. This is the reason why maintenance policy should be considered at the beginning of the life cycle, so that the effec veness of maintenance ac ons will not be limited in a later stage (CENELEC 2008) . “Life cycle cos ng is the process of performing an economic analysis to assess the cost of an item over a por on, or all, of its life cycle in order to make decisions that will minimize the total cost of ownership while s ll mee ng stakeholder requirements.” (CENELEC 2017)

2.6.1 Basic stages and concepts

The steps followed in the process of an LCC analysis are schema cally presented in figure 28. Depending on the scope and objec ves of each analysis, these steps are adjusted to its needs.

Figure 28: Schema c representa on of LCC procedure steps (CENELEC 2017)

Following the procedure described in the european standards for Life Cycle Cos ng (CENELEC 2017) , three main stages, translated to costs, are examined when conduc ng an LCC analysis. The first stage is the one including the concept of the project, its development and realiza on, all included within the acquisi on cost. The second stage includes the u liza on and enhancement of the project which in the case studied is the different types of maintenance of the turnouts needed and the maintenance cost related to it. The third and last stage is the re rement which means the turnout disposal or reuse cost. At this stage, the turnout is either disposed of or reused as a whole or as separate components. This choice is made based on the condi ons of the material and the service life that is expected to be le in each case in order for a reuse to be financially profitable (Swärd 2006) These stages/costs are summarized in the figure below.

Figure 29: Schema c representa on of turnout life cycle stages/costs

For a be er understanding of the cost breakdown structure in step 3 of the LCC process, the shape of figure 28 is given. Depending on the level of detail wanted in each level of the system’s breakdown and the cost categories (labour, material, etc) recognised in each life cycle stage as they were presented above, the size and the elements of the cube are shaped. This structure assists in construc ng systema c and analy c work, taking into considera on all the parameters involved.

Figure 30: Cost breakdown structure concept cube (CENELEC 2017)

The great influence of the design stage at the final life cycle cost is demonstrated in the diagram of figure 29. The uncertainty in the cost appears to be much larger at the early stages compared to the final ones of the life cycle. The design decisions which occur during development seem to have an important margin between the commi ed and the actual

costs, while a er the installa on and the use of the system the flexibility of the commi ed cost decreases significantly.

Figure 31: Commi ed and actual costs example (CENELEC 2017)

2.6.2 Terms and formulas

According to the above the acquisi on cost is the ini al (capital) cost while the ownership cost is the running cost. The second can be dis nguished in annual, related to maintenance and intermi ent costs in case of major renewals. Furthermore, maintenance and acquisi on costs are called tangible costs and are measurable and instantly paid, while the non-monetary ones, without physical substance such as quality, comfort and safety loss, are named intangible costs, they are hidden and hard to evaluate. During the life cycle of a railway component, such as a turnout, different cash flows occur in different years and are discounted to year zero using a discount rate. (Esveld 2001; CENELEC 2017)

Following the terms implemented in the LCC analysis are explained (Esveld 2001; CENELEC 2017) :

● Net present value (NPV): Sum of all discounted cash flows P VN =

∑

n (Ct) (1 ) ,

y=1

/

+ i

t

where:

n-the period of analysis i-the yearly discount rate

t-the number of year from year 0

● Total Present value (TPV): Sum of all the discounted costs during the years analysed including the capital cost (in year zero). P VT = NP V _{− C} 0

● Annual equivalent or annuity (ANN): The sum of interest and amor sa on, which has to be paid every year to finance an investment or maintenance strategy. Can be used

to compare projects with different life expectancy. NN (1 ) (1 ) ) P V

A = ( + i n_{* i}

_/

+ i n_{− 1 * T}

The assump ons made while construc ng the model are leading to uncertainty. A er making the best possible es ma on of the values that will be included in the model depending on the stages of the project and the level of detail preferred, it is considered important to perform a sensi vity analysis in order to evaluate the robustness of the model (CENELEC 2017) .

There are two methods of including uncertainty to the model. The first is by doing a sensi vity analysis through varying the input values by 10 to 30% and checking the robustness of the outcome. The second is using the Monte Carlo simula on where several different parameters can be varied at the same run of the model, using a random generator of normal distribu ons of the annuity. (Esveld 2001)