Just-In-Time Delivery of Wheel Sets - Just-In-Time levering van wielstellen

(1)

Delft University of Technology

FACULTY MECHANICAL, MARITIME AND MATERIALS ENGINEERING

Department Marine and Transport Technology Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl

This report consists of 124 pages and 12 appendices. It may only be reproduced literally and as a whole. For

Specialization: Transport Engineering and Logistics

Report number: 2013.TEL.7798

Title:

Just-In-Time Delivery of Wheel

Sets

Author:

M.J. de Vries

Title (in Dutch) Just-In-Time levering van wielstellen

Assignment: Master Assignment Confidential: yes

Initiator (university): prof.dr.ir. G. Lodewijks

Initiator (company): ir. B. Huisman (NedTrain B.V.) Supervisor: dr. ir. Y. Pang

(2)

Delft University of Technology

FACULTY OF MECHANICAL, MARITIME AND MATERIALS ENGINEERING

Department of Marine and Transport Technology Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl

Student: M.J. de Vries Assignment type: Master Assignment

Supervisor (TUD): dr. ir. Y. Pang (TU Delft) Creditpoints (EC): 36 Supervisor (Company): ir. B. Huisman (NedTrain B.V.) Specialization: TEL

Report number: 2013.TEL.7798 Confidential: Yes

Subject: Just-In-Time Delivery of Wheel Sets

This Master’s thesis describes a study of forecast methods to predict future demand of wheel sets at Maintenance Depots of NedTrain. Wheels sets, located under trains, are subject to wear and have to be exchanged when they have worn below a certain limit. For maintenance on trains it is required to have a certain set (types and amount) of wheels sets in stock to prevent non-availability of trains. Currently stocks of wheel sets are located at different locations. NedTrain intents to concentrate wheel sets at a central depot, and to supply wheel sets to local Maintenance Depots by Just-In-Time principle. It is believed that by doing so the turnaround stock of Ready-For-Use wheel sets can be reduced.

As of 2007 the maintenance policy for wheel sets has been changed, which has led to a more deterministic way of wheel wear. This has opened possibilities to forecast failure dates, and thus demand for wheel sets, with higher accuracy than could have been achieved a several years ago with attempts based on probability.

The assignment is to provide insight how the current logistic chain is organized and how currently demand for wheel sets is predicted. A study has to be performed on internal and external causes of wheel wear, in order to seek for possibilities to include these causes into a demand forecast. Since demand fluctuates over time, a prediction method is required. The possibility of determining the required level of supply of wheel sets to Maintenance Depots by using forecast results has to be studied.

The report should comply with the guidelines of the section. Details can be found on the website. The professor,

(3)

Summary

NedTrain, a subsidiary of the Dutch Railways (NS), is responsible for maintenance of rolling stock of NS. Maintenance is performed at maintenance depots and involves wheel sets that are located under trains. Wheels are subject to wear due to usage and need to be exchanged at maintenance when they have worn below a certain limit. Therefore, it is key to have a certain set (types and amount) of wheel sets available for upcoming maintenance to prevent non-availability of trains. At this moment wheel sets are stocked at different locations. However, NedTrain intents to concentrate wheel set stock at a central depot, and to supply wheel sets Just-In-Time to local maintenance depots. It is believed that in this way the turnaround stock of overhauled wheel sets can be reduced.

Insight in the interaction between departments is provided with an overview of the wheel set chain. At this moment two forecasting methods for wheel set demand are used. Maintenance depots plan demand one week ahead by using information from NS Reizigers on planned train arrivals. The overhaul depot forecasts supply four weeks ahead, based on former overhaul levels. It is observed that little information on future demand is exchanged and demand beyond four weeks is unknown.

Wheel wear has shown a more deterministic behavior since a use-based maintenance policy for wheel sets has been introduced for intercity rolling stock of NS as of 2007. Since demand patterns have not been changed and current forecasts based on former overhaul/demand levels shows varying results an alternative forecasting approach has been developed. This forecast approach is based on actual wheel set conditions and wheel wear. Wheel wear at wheel lathes is 2 - 6 times as high than wear during operation. A selection of wheel wear causes is made, based on short-term influences, which are used as variables in the forecast method. An overview of causes is included in Appendix I.

In this report six different methods of forecasting wheel set demand are studied and validated with 2012 demand. Wheel set demand is forecasted based on wheel set exchanges due to passing a wear limit, which covered 86% of wheel set exchanges in 2012. The best performing method is selected on the lowest error between forecasted versus real demand in the first three months. Validation shows that wheel set demand can be forecasted with a Mean Absolute Deviation of 26 - 27 days, which exceeds the current planning horizon of maintenance depots. The best performing method includes fuzzy logic to account for temporary increased wear rates due to maintenance, stage of wear and seasonality. Forecasted failure dates of wheel sets are rounded down to future maintenance periods.

The current accuracy on short-term is considered as too low to serve as reliable basis for determining a required level of wheel sets to supply maintenance depots with an interval of one month and wheel set availability of 90%. The application of fuzzy logic and tuning has improved accuracy on monthly levels, but still demand for wheel sets is frequently overestimated. For this reason Just-In-Time delivery of wheel sets, based on current demand patterns and forecast accuracy, is currently not the designated solution to supply maintenance depots at NedTrain.

In this report is shown that diameters within a population of wheel sets are not always equally distributed, which results in troughs and peaks in demand. The forecast accuracy on mid-term is accurate enough to provide insight for NedTrain in these future demand patterns for at least 2 years ahead. This information might be used as guideline for mid or long-term decisions on future overhaul and stock levels of wheel sets. Improvements are proposed for the forecast model and with regard to NedTrain practices to help increase predictability of future wheel set demand.

(4)

(5)

Summary (Dutch)

NedTrain, een dochtbedrijf van de Nederlandse Spoorwegen (NS), is verantwoordelijk voor onderhoud aan rollend materieel van NS. Onderhoud wordt uitgevoerd bij onderhoudsbedrijven en omvat wielstellen. Wielen zijn onderhevig aan slijtage door gebruik en moeten worden uitgebouwd tijdens onderhoud wanneer een bepaalde slijtage limit is gepasseerd. Het is daarom belangrijk een bepaalde set (typen en hoeveelheid) wielstellen beschikbaar te hebben voor toekomstig onderhoud om niet-beschikbaarheid van treinen te voorkomen. Voorraden van wielstellen zijn momenteel verspreid. NedTrain heeft de intentie om voorraden te centraliseren en wielstellen Just-In-Time te leveren aan onderhoudsbedrijven. Hierdoor kan mogelijk de omloopvoorraad van wielstellen worden gereduceerd.

Met behulp van een overzicht van de wielstelketen is inzicht gegeven in de interactie tussen afdelingen. Op dit moment worden twee voorspel methoden gebruikt om de vraag naar wielstellen te voorspellen. Onderhoudsbedrijven plannen hun vraag één week vooruit op basis van informatie van NS Reizigers. Het revisiebedrijf voorspeld de vraag vier weken vooruit op basis van voorgaande productiehoogten. Daarbij wordt weinig informatie onderling gedeeld over toekomstige vraag.

Sinds gebruiks-afhankelijk onderhoud wordt toegepast voor wielstellen onder intercitymaterieel vanaf 2007 vertoont de slijtage van wielen meer deterministisch gedrag. Het vraagpatroon voor wielstellen is onveranderd gebleven. Omdat huidige voorspellingen een variërend resultaat geven is een nieuwe voorspelmethode voor de vraag naar wielstellen opgesteld. Deze voorspelmethode is gebaseerd op huidige condities van wielstellen en wielenslijtage. Een selectie van oorzaken van wielenslijtage is gemaakt op basis van invloeden op korte termijn en gebruikt als variabelen in de voorspelmethode. Als voorbeeld: wielenslijtage op de kuilwielenbank is 2 - 6 maal hoger dan slijtage tijdens gebruik.

In dit rapport zijn zes verschillende voorspelmethoden voor de vraag naar wielstellen bestudeerd en gevalideerd op basis van de werkelijke vraag in 2012. De vraag is voorspeld op basis van het passeren van de slijtage limiet, dit representeert 86% van de uitbouwen. De meest nauwkeurige methode kan de vraag naar wielstellen voorspellen met een Mean Absolute Deviation (MAD) van 26 - 27 dagen. Deze afwijking is groter dan de huidige voorspeltermijn van onderhoudsbedrijven. De voorspel-methode maakt gebruik van fuzzy logic voor tijdelijke hogere slijtage t.g.v. onderhoud, stadia van slijtage en seizoensinvloeden. De uitvaldata worden afgerond naar toekomstig onderhoud.

De voorspelnauwkeurigheid op korte termijn is momenteel te laag om de grootte voor leveringen van wielstellen aan onderhoudsbedrijven te bepalen, in perioden van één maand bij 90% gevraagde beschikbaarheid. Toepassing van fuzzy logic heeft de voorspelnauwkeurigheid op maandniveau verhoogd, maar niet voorkomen dat de voorspelde vraag structureel te hoog is. Om deze reden is Just-In-Time aanlevering van wielstellen, gebaseerd op huidige voorspelnauwkeurigheid en vraagpatroon, op dit moment niet de beste oplossing om wielstellen lokaal aan te leveren.

In dit rapport is aangetoond dat de distributie in diameters van een wielstellen populatie niet altijd gelijk is, wat resulteert een variërend vraagpatroon. De voorspelnauwkeurigheid op middellange termijn is nauwkeurig genoeg bevonden om NedTrain inzicht te verschaffen in toekomstige vraagpatronen voor ten minste 2 jaar vooruit. Deze informatie is geschikt als richtlijn voor beslissingen voor toekomstige revisie en voorraad levels van wielstellen. Verbeteringen voor de voorspelmethode en voor NedTrain zijn gepresenteerd om te helpen bij het vergroten van de voorspelbaarheid van de vraag naar wielstellen in de toekomst.

(6)

(7)

Nomenclature

ATB Automatische Trein Beïnvloeding DDM DubbelDeks Materieel

DST Draaistel (Bogie) EMU Electrical Multiple Unit FIFO First In First Out

GTU Gebruiks Toestandsafhankelijke Uitbouw GUI Graphical User Interface

ICE InterCity Express ICM InterCity Materieel ICR InterCity Rijtuigen

IL&T Inspectie Leefomgeving & Transport (former IVW) HIT Highperformance Individual Trainprofile

KPI Key Performance Indicator LBM Landelijk Bureau Materieel LCM Long-Cyclic Maintenance LLC Landelijk Logistiek Centrum

MBN Materieel Bijsturingscentrum NedTrain MDBM Mean Distance Between Maintenance MTBM Mean Time Between Maintenance MD Maintenance Depot

NS Nederlandse Spoorwegen NSR NS Reizigers

R&O Refurbishment & Overhaul (Haarlem) RCF Rolling Contact Fatigue

RLW Revisie LoopWerken SCM Short-Cyclic Maintenance SCO Supply Chain Operations SGM StadsGewestelijk Materieel SLA Service Level Agreement STG Systeem Technische Grens UIC International Union of Railways

VBBK Veiligheid, Beschikbaarheid, Bedrijfszekerheid, Kosten (NedTrain KPI’s) VIRM Verlengd InterRegio Materieel

WSP Wheel Slip/Slide Protection WST Wielstel (Wheel Set) WRI Wheel-Rail Interface

(10)

(11)

1. Introduction

This Master Assignment focuses on the logistic processes, wear mechanisms and demand of wheel sets at NedTrain maintenance depots, with the aim to develop a tool to predict future wheel set demand for Just-In-Time delivery of wheel sets.

1.1 NedTrain organization

NedTrain is a subsidiary of the Dutch Railways (NS), owned by the government of the Netherlands, and is responsible for maintenance, service, cleaning, repair and overhaul of rolling stock of Dutch Railways, which contains around 3000 car bodies. Besides NS, NedTrain also performs service and small maintenance on rolling stock of other (international) rail operators.

There are 35 different train operators active on the rail infrastructure in the Netherlands. The majority of them are freight transport and regional operators. NS accounts for about 80% of all rail transport in the Netherlands (Lewis & Olofsson, 2009). European regulation has obliged NS to separate some departments as of 1992. The maintenance division ‘Materieel & Werkplaatsen’ was separated as subsidiary ‘NS Materieel’ in 1992. As of 1999 the name has been changed in NedTrain B.V. Currently NedTrain has over 165 years of experience in maintaining rolling stock (NedTrain, n.d.).

Safety, Availability, Reliability and Costs are the performance indicators for maintenance at NedTrain. Maintenance is performed at 30 service

depots, 4 maintenance depots and 2 refurbishment workshops located alongside the track throughout the Netherlands. The headquarters of NS and NedTrain are located in Utrecht. NedTrain has over 3000 employees and a turnover of 450 million euros. The organogram in Figure 1.1 shows the position of NedTrain as subsidiary in the organization of NS.

1.2 Problem statement

Maintenance of trains is performed at maintenance depots about every 75.000 kilometers (or 112 operational days). This includes maintenance of wheel sets under trains. A wheel set is an assembly of two wheels, an axle, bearings and (dependent on the type) brake disks and/or gearbox. With different assemblies different types of wheel sets exist.

Wheels are subject to wear due to usage on the track and removal of tread defects. These tread defects (incurred during operation) cause vibrations and large forces as wheels are not ‘perfectly round’. For this reason wheel sets are machined on a so-called ‘wheel lathe’ (in Dutch: kuilwielenbank). As of 2007 all wheel sets of intercity trains are preventively machined during maintenance, regardless whether a defect has occurred. Machining a wheel set causes the wheel diameter to decrease.

Wheel sets have to be exchanged when they have worn below a certain limit. Exceeding this limit is not allowed. For NedTrain it is of high importance to know which wheel sets need to be exchanged

(12)

before a train enters the maintenance depot. When the right wheel sets are not available this results in non-availability of the train for operation by NS Reizigers. The availability of trains, and thus throughput time in a maintenance depot, is a Key Performance Indicator for NedTrain. For this reason it is required to have a certain set (amount and types) of wheel sets available for the upcoming maintenance during a certain time period.

The amount of wheel sets in ‘the chain’ is fixed, which is illustrated in Figure 1.2. A central depot (R&O) in Haarlem is responsible for overhauling wheel sets according to the demand in the chain. From this depot wheel sets are supplied to maintenance depots. Currently, wheel sets are stocked at different locations, for example at maintenance depots and the central depot (R&O) in Haarlem. NedTrain intents to concentrate wheel set stock at the central depot in Haarlem, and to supply wheel sets to local maintenance depots by Just-In-Time principle. It is believed that by doing so the turnaround stock of Ready-For-Use wheel sets can be reduced.

In order to have different wheel sets delivered in time at maintenance depots, supply chain management has to anticipate to future demand. Since the demand fluctuates over time, a prediction method is required. Both internal influences (processes) and external influences (weather, track conditions, etc.) have an effect on the demand for wheel sets. It is important to study the degree to which these influences have a contribution to demand, and how could be accounted for this.

In order to estimate future demand levels for wheel sets, a suitable method has to be found. A forecasting model that accounts for internal and external influences on demand of wheel sets could be a solution. By studying different methods and their accuracy, a recommendation can be given on the applicability of Just-In-Time deliveries of wheel sets and the required level to supply within a certain time period.

(13)

1.3 Research questions and approach

The problem statement as described in the former section leads to the flowing main research question:

How could the demand for (overhauled) wheel sets be predicted, accurately enough for the purpose of Just-In-Time delivery to maintenance depots?

To be able to answer the research question sufficient knowledge should be gained about the current logistic and planning processes at NedTrain and the causes that might influence wheel set demand. In order to provide a satisfying answer to the main research question, four sub-questions have been composed.

1. How is the current logistic chain of wheel sets organized?

Answering this question provides insight to the current supply chain and logistic system. It reveals the relations (coordination) between different departments, the information used at the different locations and the method(s) of maintenance and demand planning. An overview will be made of the wheel set chain and the maintenance process of wheel sets. This overview is useful at a later stage to show how a forecast model could be applied in the current organization.

2. What are the internal and external causes of wheel set wear?

The answer of the question gives insight in the roots and causes of wheel set wear and demand. Secondly, it determines to which extent these factors influence wear rates. A literature study will be performed on the basis of relevant papers and articles on wheel set wear, wear rates and causes.

3. How could internal and external causes of wheel set wear be used to make a forecast of wheel set demand?

This question tries to find possibilities to accurately predict demand for wheel sets. This can be achieved by making a forecast model and including appropriate wear causes as found by answering the former question. To aim is to study accuracy, possibilities and potential of the forecast model by using data of 2010 and 2011, and validate the model by using data of 2012.

4. Could the required level to supply a maintenance depot be determined, given the forecast results, supply interval and wheel sets availability for maintenance?

This question seeks for a possibility to minimize the stock of the local maintenance depot. When the accuracy of the forecast has been validated, the required level of wheel sets to supply a maintenance depot could be determined. The required level to supply is a measure of the optimal order quantity of wheel sets. Given the accuracy of the forecast model and required availability of wheel sets, it can be investigated how wheel sets can be delivered Just-In-Time.

Concluding, in this assignment it will be researched how the current supply chain and logistic system at NedTrain is organized, which causes of wheel set wear can be measured/predicted and how these can be included into a proposed forecast model. Given this accurate forecast, the optimal order quantity to supply to Maintenance depots will be estimated, in order to save costs on the parts supply chain, in which wheel sets are included. The accuracy of the forecast will be validated with appropriate data. The potential of improving process control in the Delft Systems Approach will be discussed as well.

(14)

1.4 Boundary conditions

A several of boundary conditions are set to improve the feasibility of the study. For example two out of four maintenance depots have been studied and analysis is restricted to one particular type of rolling stock of NS. An overview of NedTrain locations is given in Section 2.1 in the next chapter.

Boundaries for the logistic system analysis and forecast model:

• Information, data and processes from maintenance depots in Onnen and Eindhoven, Refurbishment & Overhaul Haarlem and Supply Chain Operations department in Tilburg will be analyzed and included

• Data analysis will be restricted to VIRM (‘Verlengd InterRegio Materieel’) double-decker rolling stock of NS

• The proposed forecasting model will be based on the data as retrieved from the points above

Boundaries for the optimal order quantity calculation: • The supply interval is fixed

• The required availability of wheel sets for maintenance is fixed at 90% • The maximum stock level at a depot may vary

• Cost factors (holding, ordering, etc.) are not included • Stock-allocation is outside the scope

Assumptions of the forecast method and model will be listed in Chapter 5, as some require understanding of some processes and/or terminology at NedTrain, which will be clarified in the following chapters.

1.5 Structure of report

The chapters in this report are based on the four sub-research questions as mentioned in Section 1.3. At the end of each chapter a summarization and conclusion is drawn.

This first chapter gave a general introduction to NedTrain, the problem statement and research questions. The next chapter gives an introduction to the different types of maintenance NedTrain uses, introduces wheel sets in more depth and presents motivation for this assignment.

The third chapter contains an analysis of the current logistic system of wheel sets at NedTrain. In this chapter the main processes of maintenance and overhaul as far as wheel sets are concerned are presented. Also current planning and forecasting methods and practices are presented together with their application. The Delft Systems Approach is used to draw a system overview of these processes. Finally, a conclusion is drawn based on all findings.

The fourth chapter gives an extensive overview of the causes and influences of wheel wear. A categorization is made to group these causes based on their origin. Also wheel defects that occur frequently are presented. The chapter is concluded with a summary of causes that influence wheel set lifetimes on short-term, and thus wheel set demand.

The fifth chapter shows results from data analysis. Machining policies and monitoring wheel sets are discussed. Furthermore, wear rates of the wheels and the influence of different wear causes are

(15)

presented. Former forecast approaches are presented, and a framework for the forecast model will be proposed with different methods for forecasting wheel set demand.

The sixth chapter presents validation of the forecasting methods, followed by a conclusion on the accuracy and suitability of the chosen forecast method. With results from the forecasting method and current practices a proposal will be given for an order quantity calculation based on fixed delivery intervals and Just-In-Time deliveries will be discussed. A proposal for application of the forecast model will be presented as well.

The report concludes with an evaluation of the proposed forecasting method, supply levels and Just-In-Time deliveries of wheel sets. Recommendations based findings in this research and on information on wheel set demand at different departments are given. The last section presents possibilities for further research.

(16)

(17)

2. Maintenance and wheel sets

2.1 Introduction of maintenance process

Rolling stock is the collective term for tram-, metro cars, locomotives, passenger coaches, wagons, etcetera, which are able to move on a track. A train set (asset) is a combination of the vehicles above operated in a fixed setting. Each train set consists in turn of a several car bodies.

NedTrain performs four different types of maintenance on assets at three distinctive types of locations, namely service depots, maintenance depots and at the Refurbishment & Overhaul workshop. The locations are shown in Figure 2.1. Maintenance can be categorized in daily-, short cyclic-, long cyclic maintenance and revision. Periods when performing service, maintenance and/or revision is set in the Rolling Stock Maintenance Manual, managed by the Fleet Services department of NedTrain. Each asset type has its own maintenance manual. The periods of maintenance are based on mileage or time an asset is operated on the track and is monitored on a daily basis by the operator NS Reizigers (NSR). It is

not allowed to exceed the terms. The operator is also responsible for scheduling assets to the appropriate maintenance location when required, according to the maintenance periods.

Daily maintenance or serviceis performed repetitively on assets each 2 or 8 days, respectively known as technical check B and A. It involves small maintenance on safety aspects, daily inspection, cleaning and repair of small defects. Daily maintenance is performed at service depots located at 30 locations close to the track. At these locations assets are temporary stored when not in operation, e.g. during night. If at the service depot a problem is detected which requires more comprehensive maintenance, the asset will be scheduled to enter a maintenancedepot.

Short Cyclic Maintenance (SCM) is performed following a repetitive scheme based on mileage or days. For the majority of asset types of the Dutch Railways a period of 77.000 kilometers or 112 days applies, whatever comes first. This type of maintenance involves inspection and more comprehensive reliability- or safety related activities, e.g. repair of defects or exchange of parts. SCM is performed at maintenance depots located in Onnen, Maastricht, Leidschendam, Amsterdam and partially in Eindhoven. The throughput time of SCM in a maintenance depot is generally between three to five days. Periodic maintenance on asset parts is performed between a several months up to a several years, dependent on the specific part. Both preventive and corrective maintenance is performed, dependent on the system or part it concerns. Preventive maintenance consists mainly of use- and condition based maintenance. Corrective maintenance consists mainly of failure-based maintenance. Besides planned arrivals also unplanned arrivals occur due to various failures during asset operation.

(18)

Unplanned in this context means that an asset requires maintenance not following its normal maintenance scheme (mileage or time).

NedTrain decomposes assets into groups of parts. At the first level a separation is made into car body and main parts. These are further separated at second level into construction parts, repairable parts and consumable parts. Consumable parts are replaced by a new part after becoming obsolete or defective. Repairable and main parts are both re-usable by means of overhauling the particular part. Main parts are seen as separate category because it is either legally obligated by law to monitor its condition (legal restrictions for investigation) and/or is economically viable (e.g. high investment or maintenance costs) and or required by the owner (NS Reizigers) of the parts (Vuijk, 2012; Haverkamp, 2005). Main parts are characterized by relative high investment costs, high costs for overhauling, life-cycle maintenance scheme, unique identification number (main parts are traceable), tracked performance and required managing (Vuijk, 2012). NedTrain has indicated 15 groups of main parts, some examples are: bogies, wheel sets, traction motors, gearboxes and compressors. These main parts have their own maintenance schedule based on mileage or time, which are set in a rolling stock maintenance manual. The overhaul interval of main parts is generally in terms of years, but might be inspected or monitored frequently.

Long cyclic maintenance (LCM) is performed on specific components and main parts, in periods of 2 up to 30 years. Parts that require overhaul are exchanged at maintenance depots and send to the overhaul workshop in Tilburg or Haarlem.

Revision is performed on main parts and complete assets. Small main parts are overhauled at the component workshop in Tilburg. Examples are compressors, air-conditioning units and traction motors. Bogies and wheel sets are overhauled at the overhaul workshop in Haarlem. In Haarlem also complete assets are overhauled, this is called modernization. Spare parts are stored at the individual depots and centrally at the central warehouse in Tilburg. Supply Chain Operations is a department linked to this warehouse and responsible for purchasing, logistic planning and transportation of parts.

In this report only wheel sets are studied, which has its their own maintenance schedule and overhaul interval up to five years, an identification number and is subject to an use based maintenance policy with periodic monitoring. Wheel set lifetime with a number of overhauls approximately equals the average asset lifetime and is about 30 years (Haverkamp, 2005).

2.2 Introduction of wheel sets

Under each car body two bogies at both ends are installed. A bogie is the structural framework for wheel sets, contains the suspension of the train and its main function is to reduce the axle load on the track. Furthermore, thanks to bogies the amplitude of vertical motion is halved (Nijland, 2010). Axle loads are restricted due to limitations of the infrastructure to prevent subsidence and high loads on structures such as bridges (Bonnett, 2005). Each bogie (Figure 2.2) contains of two wheel sets (Figure 2.3).

(19)

Figure 2.2 Bogie (courtesy: Alstom)

A wheel set consists of an axle, bearings, two wheels, two railway tyres and (depending on the configuration) brake discs, gearbox and/or other equipment. A basic distinction is made between driven (motor) or non-driven (running) bogies and wheel sets. So, different types of wheel sets exist, each with its own configuration. In total 32 different types of wheel sets are in use at the entire fleet of the Dutch Railways. This report focuses on VIRM rolling stock, which has 6 different types of wheel sets.

Figure 2.3 Wheel set (Nijland, 2010)

The picture above shows a plain drawing of a wheel set, containing two wheels. Basically there are three classifications of wheels in the railway industry: solid, tyre and assembly types of wheels (Iwnicki, 2006). The Dutch Railways have been using tyred wheels over 60 years as solution to save metal and thus costs. In comparison to solid wheels, tyred wheels have a removable outer layer, the

railway tyre.

The railway tyre is mounted on an inner hub, which is mounted on the axle (Iwnicki, 2006). This is shown in Figure 2.4. Railway tyres are kept in place by interference fits, as the diameter of the railway tyre is slightly smaller than the disc diameter. Shrink fitting is another common term for this kind of assembly. The railway tyre is provided with an inner ring (shoulder) and during the assembly process a snab ring is applied as external fastener to prevent lateral movement. Fastening to prevent rotation is held by friction, and introduces tensile or compressive forces in both parts. The outer layer of a wheel has a certain wheel profile and is subject to wear. A railway tyre has the advantage that it can be removed when it has worn out below a certain limit. The opposite type of wheels is a solid wheel (in Dutch: volwielen), which has to be replaced entirely when it has worn out.

(20)

Figure 2.4 Wheel set with inner hubs (left), overhauled wheel set (middle) and railway tyre (right)

In order to maintain a certain quality of the wheels, especially the wheel tread, it has to be machined when material defects start to initiate or propagate. This is done at one of the 4 maintenance depots during Short Cyclic Maintenance (SCM) on an underflow wheel lathe. On this wheel lathe two wheel sets, i.e. four wheels, can be machined simultaneously when still mounted under an asset, see Figure 2.5. The asset is shunted to subsequently machine all wheels. In 2007 a use-based maintenance policy of wheel sets was introduced to lower the effect of surface defects with the aim to extend the potential lifetime of wheel sets. It means machining wheels (re-profiling into the original shape) such that any surface-defects (as flat spots, small holes or damages) are removed to prevent propagation of cracks. It should be performed each time a train visits a maintenance depot (about every 90 days). Re-profiling should extend the lifetime of wheel sets, since defects at wheels are smaller because of shorter intervals of maintenance.

Railway tyres have some important dimensions following from a certain wheel profile, see Appendix (Nijland, 2010).

• Flange thickness and height: these dimensions have to secure that a wheel set remains on the track and does not derail

• QR-dimension: distance between two points on the flange, basically describes the wheel flange angle

• Railway tyre thickness: measured at the center of the wheel tread, determines the wearable thickness

• Other dimensions are related to the fitting of the railway tyre on the central hub.

During SCM is checked whether these dimensions are within the tolerances as set in the rolling stock technical standard. When these tolerances are exceeded the wheel has to be machined into the original shape, this is where the term ‘re-profiling’ originates.

Re-profiling has to be performed within limits compared to other wheels, bogies and car bodies. For each type of asset different tolerances within and between wheel sets, between bogies and between car bodies apply. These tolerances result from consequent angles of bogies and car bodies when different wheel diameters are applied. Without these tolerances the wheel wear (patterns) and running behavior of the bogies will be influenced negatively, several traction related systems will not function properly, buffers and drawbars between car bogies will not be aligned correctly and some asset types might exceed the gauging profile (Beekmans, 2013; Nijland; 2010). Especially modern vehicles allow for smaller tolerances (Railway Technical, n.d.). For VIRM assets the following tolerances according to the rolling stock technical standard apply:

• Tolerance between two wheels of one wheel set: 0.3 mm

• Tolerance between two wheel sets in one bogie: 3 mm (motor) or 50 mm (running) • Tolerance between two bogies within one coach: 50 mm

(21)

Figure 2.5 Underflow wheel lathe (left) and Wheel drop (right)

For wheels under operational assets a minimal diameter applies, the rejection limit, which basically describes the thickness of the railway tyre. When a railway tyre has reached the rejection limit (which is generally 35 mm), the wheel set has to be exchanged in a maintenance depot. Bogies and wheel sets can be exchanged on a wheel drop. This is shown in Figure 2.5. At this machine the bogie or wheel set under an asset is lowered below the depot floor level and exchanged with another unit. A wheel set that is disassembled can only be replaced by a wheel set of the same type. Mutual exchange with different types of wheel sets is not possible. This also applies for bogies.

Currently around 12.000 wheel sets are in operation and 1.300 wheel sets are in stock as spare parts. The investment cost of a bogie is approximately 10.000 euro, while overhauling a wheel set costs 2600 or 4300 euro for running or motor wheel sets respectively. Based on these costs is calculated that one millimeter of wearable railway tyre costs 100 and 150 euro for running or motor wheel sets respectively. Overhauling a wheel set is 10 times as expensive as exchanging a wheel set (Nijland, 2010).

2.3 Motivation

The theoretical lifetime as described in the rolling stock maintenance manual of a new or overhauled railway tyre is 1,2 million kilometers, i.e. the failure limit, which is seldom achieved. The rolling stock maintenance manual for VIRM assets prescribes 1,7 million kilometers as ultimate limit for overhauling wheel sets, but is far lower in reality. Figure 2.6 on the next page shows the average lifetime of wheel sets for all rolling stock of Dutch Railways combined as measured from 1991. Between 1991 and 2003 a decrease in wheel set lifetime was observed by NedTrain but at that time no technical explanation was found for this significant decrease (Nijland, 2010). The downward trend was an incentive for starting research projects to wheel set performance, maintenance and profiles with the aim to reverse this trend. This is illustrated in Figure 2.6.

A major motivation for starting this current Master Assignment were results from an earlier project, called ‘Lean on Wheels’ in 2010, which focused on the chain of wheel sets at NedTrain. This project was intended to introduce improvements on both technical and process (Geertman, 2013). Technical measures as lower rejection limits, training for lathe operators and improvements of the wheel slip/slide protection system were introduced to potentially enhance the lifetime of wheel sets by 40% (Geertman, 2013). The project was stopped halfway during execution. This was the result of different priorities with respect to other projects. For this reason the improvements in process aspects were not implemented. This included measures to redesign (logistic) processes in the wheel set chain and

(22)

the project has resulted in an unclear situation about current practices in the wheel set chain. All departments have drawn their own conclusions from this project, they kept, meanwhile changed or scraped measures that were introduced. On the other hand, the project has shown benefits and potential of lean methods such as Just-In-Time deliveries.

In the last decade many projects have been executed and measures have been applied that have influence on wheel set lifetimes, wear rates, and logistic processes. Amongst these are:

• The Gotcha Monitoring System for detecting wheel defects along the track as of 2002/2003 • Installation of a double-headed wheel lathe at the depot in Eindhoven in 2004

• A forecast model from the logistics organization of NedTrain in 2004

• Improvement of the wheel slip/slide protection systems on certain asset types (for example ICM in 2005 and VIRM in 2010)

• Introduction of another wheel profile under intercity trains as of 2007 • Introduction of use-based maintenance policy for wheel sets as of 2007

• Monitoring wheel set performance with the ‘Wielenwacht’ tool from the TopWiel project as of 2007

• Monitoring machining quality with the ‘Quality Analizer’ installed on all underflow wheel lathes • Application of the ‘DAAN-tool’ to translate rather complicated rolling stock technical standards

of wheel set diameter tolerances into a user-friendly software interface (GUI) • The stopped project Lean on Wheels as described in this section in 2010 • A new Supply Chain Operations department as of 2012

These are some examples that have had positive influence on the predictability of wheel set lifetimes, but have changed current practices, which does not all have been evaluated yet.

Figure 2.6 Overview of wheel set lifetimes between 1991 and 2013. The data from 1991 up to 2004 is taken

from (Vermeij et al., 2008), data from 2004 up to 2013 has been gathered from analysis of W. Nieuwehuize (NedTrain).

(23)

3. Wheel set supply chain and logistic process

In order to provide insight in the supply chain of wheel sets and the logistic process, interviews are performed with planners and stakeholders at the maintenance depot in Onnen, Eindhoven, Refurbishment & Overhaul (R&O) workshop in Haarlem and Supply Chain Operations (SCO) in Tilburg. In this chapter an overview of the wheel set chain is modeled by using the Delft Systems Approach. With this method a distinction is made between the executing process and control function. The model shows what information is used as input for making a planning, forecasts and setting standards for the executing process. This model is on the aggregation level of the subsystem wheel sets. Both maintenance (at maintenance depots) and overhaul (at R&O Haarlem) are included. The interviewed employees and others concerned with wheel sets have verified the structure (collection of relationships) on correctness.

In the next section a brief overview of the layout of the models and a short introduction to the Delft Systems Approach and its terminology is given. The subsequent sections describe how a planning is made at each department at NedTrain, how stocks are controlled, and how wheel set demand arises from the maintenance process on assets. Special attention is given to what information is being used and exchanged. The aim is provide insight in the possibilities to forecast with the existing structure, as well later on in the report showing how current structure might be simplified when departments collaborate to a higher level. The first model (wheel set chain model) is explained in Sections 3.2 (executing process), 3.3 (function control) and 3.4 (process control). The second model (maintenance process model) is explained in Section 3.5. The structure of models is shown in Figure 3.1. Note that some systems and/or concepts in this chapter will be discussed in later chapters in more depth or with special attention to it’s contribution to forecasts.

3.1 Introduction to the Delft Systems Approach

A brief introduction of the Delft System Approach model is presented in this section to introduce the elements that are being used and their purpose. It is important to realize that the Delft System Approach model is built with elements that represent particular functions, rather than the means in which a function is fulfilled or realized. In this way (small) adjustments could be made into any process just by looking and considering the function that must be fulfilled. Secondly, it allows for redesigning processes with regard to their functions.

(24)

The wheel set chain model can be vertically separated into two parts. The left side represents the overhaul department in Haarlem, which is the supplier of overhauled wheel sets; the right hand side represents basically each maintenance depot, the supplier of worn wheel sets.

Horizontally three layers can be distinguished. The lower layer represents the executing process; this is the actual throughput of wheel sets in either the overhaul or maintenance depot. The top layer represents the function control; this is the translation of various inputs into a planning and/or standards, which are fed to the processes below. The middle layer represents the process control; which acts to disturbances in the executing process by using standards as set in the function control.

The executing process, process control and function control contain five different elements. These elements are shown in Figure 3.2 below.

Figure 3.2 Elements used in the Delft Systems Approach

The encoding/decoding function ensures that input/output of the transformation function (e.g. maintenance or overhaul) is suitable as input/output. The filter function acts as a test to a certain quality or quantity to feed to or from the transformation function. The buffer function is a stocking point that might be regularly controlled on quality (e.g. when stock is kept for longer periods). The transformation function is the main function in the model. In the current models it represents the maintenance- or overhaul process. Other functions that can be seen in the model are initiating- and evaluating functions, which transform various inputs into standards for the executing process. Examples are a planning or standard that is initiated. Also comparing- and controlling functions are present, which act in the process control layer to disturbances in the executing process.

In the model two different arrows are used. Thick arrows show the actual throughput of wheel sets in the executing process. Thin arrows show the flow of information between functions. The numbers between parentheses in the sections below correspond to the numbers as listed in the model. The process as presented in the model will be explained per layer, starting with the executing process (Section 3.2), and followed by function control(Section 3.3) and process control (Section 3.4).

3.2 Wheel set chain: executing process

This section describes the flow of wheel sets in the executing process (lower layer). The elements of the executing process at R&O Haarlem (left hand side of the wheel set chain model) and the executing process at maintenance depots (right hand side of the wheel set chain model) are clarified briefly below. Note that R&O Haarlem supplies overhauled wheel sets but demands worn wheel sets, and maintenance depots demand overhauled wheel sets and supply worn wheel sets.

(1) Encode: Wheel sets received at R&O Haarlem are first checked on defects, wheel diameters are measured and the reason for receiving is registered in the ERP system of R&O Haarlem.

(25)

(2) Filter: At this location a quantity control (are the right wheel sets delivered to the production line?) and quality control (what are the wheel diameters and/or defects?) is performed. The wheel set diameters are checked and compared to the rejection limits. The wheel set department and bogie department use different rejection limits, which will be discussed in Subsection 3.3.1 (STG-limits). A wheel set will be stocked at location (6) when it has not passed the rejection limit yet.

(3) Buffer: This work-in-progress stock contains rejected wheel sets or with defects and is ready to be overhauled. This buffer ensures a continuous flow as input for the overhaul process (transformation) and should match the levels of production planning.

(4) Transformation: This is where the overhaul of wheel sets is performed. The jobs are already prescribed at (1) and might contain: railway tyre overhaul, gearbox overhaul, brake disks or bearings replacement, etcetera. New railway tyres at overhaul are re-profiled at a wheel lathe.

(5) Filter: At this point a quality check (correct profile, diameter, alignment, fit tolerance etc.) of the overhaul process is performed. The completed jobs are inspected before the wheel sets are stocked. (6) Buffer: This stock contains Ready-For-Use wheel sets until they are requested at a maintenance depot, or internally at the bogie department when wheel sets have been removed from an overhauled bogie. The stock contains a combination of overhauled and half-worn wheel sets. The latter are wheel sets that have wheel diameters above the rejection limit.

(7) Decode: When wheel sets are requested at a maintenance depot, they are picked from the stock for transport. The trigger for arranging these transports is given by SCO.

(8) Encode: Wheel sets are received from either R&O Haarlem or another maintenance depot.

(9) Filter: At arrival a quantity check is performed, to be sure the right amount and types are received according to the request.

(10) Buffer: At this locations Ready-For-Use wheel sets are stocked. As with location (6) the stock contains a combination of overhauled and half-worn wheel sets. The stock might be used for maintenance at location or re-located to another maintenance depot (11) whenever they are required at another location. Relocations are performed in agreement with SCO.

(12) Transformation: This is the actual maintenance process in which wheel sets are exchanged and demanded. The input of this process comes from the buffer stock (10) and the output is half-worn or rejected wheel sets. A second model of has been made to show how demand arises from the maintenance process, this model is explained in Section 3.5 (Maintenance process).

(13) Filter: A quality check of the wheel sets is performed during the maintenance process and the reason for exchange (for example defects or passing the rejection limit) is registered in the data-system R5. This filter can be seen as part of the transformation process.

(14) Buffer: This temporary stock contains either half-worn wheel sets that are still usable or rejected wheel sets that have to be overhauled at R&O Haarlem.

(15) Decode: Wheel sets that require overhaul are picked from location (14) and sent to R&O Haarlem. A trigger for arranging these transports is given by either SCO or the maintenance depot.

3.3 Wheel set chain: function control

This section describes the input and output of the function control (top) layer in the wheel set chain model. First, subsequently exchange limits (Subsection 3.3.1), the DAAN-tool (Subsection 3.3.2) and Min-Max stock levels (Subsection 3.3.3) are discussed. These subsections are followed by planning demand at maintenance depots (Subsection 3.3.4), and forecasting demand at R&O Haarlem (Subsection 3.3.6).

(26)

3.3.1 Rejection and exchange limits

The rejection limit for railway tyres is set at 35 mm and obliged by rolling stock technical standards. This limit is measured as shown in Figure 3.3. Exceeding this limit is not allowed. This is why maintenance depots use a higher limit for wheel set exchange to secure this rejection limit is not passed during asset operation between short cyclic maintenance. This limit is set at 37 mm (Beekmans, 2013). For convenience this will be called exchange limit.

The maintenance depot in Onnen uses, besides the exchange limit (25), a secondary slightly higher limit as threshold when a wheel set is likely to reach its exchange limit at the next short-cyclic maintenance. It means that failures are expected one maintenance period ahead (Kempenaar, 2013). This limit is set at 38 mm and called GTU-limit (in Dutch: Gebruiks Toestandsafhankelijke Uitbouw). When this limit is passed a GTU-order, a wheel set request, is sent

as preliminary notification to R&O Haarlem that a replacement wheel set will be required at next maintenance (Kempenaar, 2013).

R&O Haarlem on the other hand uses rejection limits that are even higher. Wheel sets built out of bogies and separately arriving wheel sets are checked against these limits at R&O Haarlem. Surprisingly, the wheel set department and bogie department use different economic rejection limits. They are based on economic considerations. The idea was that costs involved for transporting and exchanging wheel sets with little remaining lifetime were too high. Hence, they are called economic rejection limits (18) (Kempenaar, 2013; Stroomer, 2013). Initially the economic rejection limits were set at 55 mm, but have been lowered to 42 mm in 2010.

The exact economic considerations and evaluation (17) of these limits have not been determined during the research. However, there seems to be potential to further lower these levels. For example, costs associated with transports are calculated per hour, regardless of the effective load of the truck. As the effective load of transports differs per week, there is capacity to transport these wheel sets to maintenance depots without additional costs. The different rejection limits cause that rejection of a wheel set is dependent on the particular location rather than wearable thickness. An overview of the limits that are being used throughout the wheel set chain is presented in Table 3.1.

Asset type Wheel set type Rejection limit Exchange limit GTU limit Economic rejection limit (wheel set department) Economic rejection limit (bogie department) VIRM 1 315 35 mm 37 mm 38 mm 42 mm 47 mm 316 35 mm 37 mm 38 mm 45 mm 50 mm 317 35 mm 37 mm 38 mm 42 mm 47 mm VIRM 2/3/4 325 35 mm 37 mm 38 mm 42 mm 47 mm 326 35 mm 37 mm 38 mm 42 mm 47 mm 327 35 mm 37 mm 38 mm 42 mm 47 mm

Table 3.1 Rejection limits (Applicable to BVV supplier, bogie department levels before machining)

Figure 3.3 Measuring the railway

(27)

3.3.2 DAAN-Tool

As of 2010 the maintenance depots have access to a stand-alone software program, called DAAN-tool, which translates rather complicated requirements on tolerances between wheel sets under an asset into a user-friendly software interface. A screenshot is of the tool is shown in Figure 3.4. It uses information from rolling stock technical standards. It allows one to enter diameters of all wheel sets under an asset into the program, which responds with a message if wheel sets are within tolerances and what tolerances are allowed. It simplifies correct use of the rolling stock technical standards, which otherwise had to be calculated by hand. Before the DAAN-tool was introduced, decisions

were based on a standard rule that a deviation between diameters of two wheel sets of 50mm was allowed, which turned out to provide varying (often non-economical) decisions. A more in depth study showed that these standards were more complex than just a difference in wheel set diameters (e.g. the angle of bogies and coaches are the most important) and allows more flexible combinations of diameters. This has led to the development and introduction of the DAAN-tool. The DAAN-tool takes angles into account, and allows for wider tolerances between bogies and coaches in comparison to applying the separate norms one by one (Nijland, 2010). Lathe operators use the program to check wheel diameter tolerances after machining wheels. Planners use the tool is to see in advance what tolerances are possible to exchange wheels. The program as used by NedTrain is currently developed and used for VIRM and ICM rolling stock only.

3.3.3 Min-Max levels

Min-Max levels (21) are intended as tolerance between which R&O Haarlem should keep the amount of overhauled main parts. The Min-level is estimated to ensure that enough ready-for-use wheel sets are available at all stock locations to perform maintenance. The Max-level is estimated to prevent building up excessive stock levels of wheel set types that are not necessarily required. The Min-level is estimated as safety-stock. The Min-Max levels (21) apply to the total Ready-For-Use stock of wheel sets, regardless of the location and state. There are no Min-Max levels for local stocks at maintenance depots (Van Wegen, 2013).

The idea of restricting stock levels between reasonable levels is a possibility of controlling stocks, however the current application is questionable in terms of efficiency. At this moment it is unclear who manages these levels (22). Also it is unclear in which way and at what time these levels have been set. Documentation reads that Min-Max levels should be adjusted to the demand of the former 2 years (i.e. moving average), it should include a ratio of planned/unplanned demand, and should be adjusted on weekly basis, with the idea to adapt the levels when the demand rises (Ten Have, 2010).

At time of writing, the levels have not been changed for at least 5 years and, possibly because of that, used as a standard or given fact of which no one exactly knows on what basis they have been set (Stroomer, 2013; Van Wegen, 2013). (22) Another belief is that current Min-Max levels are just a summation of average stock levels from all stock locations. Surprisingly Supply Chain Operations and R&O Haarlem uses different Min-Max levels for controlling stock (27) and production (29) respectively. This applies to VIRM wheel set types 315 (motor wheel set) and 327 (running wheel set), but to

(28)

wheel sets from other rolling stock types as well. In which way both departments coordinate the total stock with different Min-Max levels is unclear. An overview of these levels is shown in Table 3.2.

Asset type Wheel set type

Min-level (SCO) Max-level (SCO) Min-level (R&O) Max-level (R&O) VIRM 1 315 10 20 14 27 316 17 34 17 34 317 16 32 16 32 VIRM 2/3/4 325 12 20 12 20 326 13 30 13 30 327 30 45 23 45

Table 3.2 Overview of Min-/Max levels for VIRM wheel sets

3.3.4 Planning wheel set demand at maintenance depots

Once per week LBC (Landelijk Bijsturings Centrum) as part of NS Reizigers (NSR) provides maintenance depots with a list of asset arrivals that require maintenance at the depots next week, categorized in order of mileage (Kempenaar, 2013). The mileage is important since the short-cyclic maintenance (SCM) period, in time or mileage, is not allowed to be exceeded. Based on this list each maintenance depot makes a maintenance schedule for the upcoming week. From practice it turns out that this asset maintenance planning list and exact moment of asset arrival has a limited reliability. Which asset will actually enter a maintenance depot is known only 24 hours on beforehand (Kempenaar, 2013). Secondly scheduled asset arrivals might be rescheduled to another day, or even next week. This result in the fact that guaranteed asset arrival is only known a short time before (Beekmans, 2013).

The maintenance depot planner will consult the data-system R5 to check which main parts are assembled on assets as listed in the asset maintenance planning list. R5 is a database system that contains the configurations of assets and data required for maintenance. The planner uses obliged periods, in time or mileage, as set in the rolling stock maintenance manual to check whether these main parts (such as bogies, wheel sets, gearboxes, compressors, traction motors, etc.) require maintenance or overhaul at the next planned SCM. When this is the case, the main part has to be exchanged at next maintenance. Otherwise it will exceed the obliged maintenance limit, which is not allowed. Wheel set exchanges might also be the result from axle bearing-, gearbox-, traction motor- or bogie revision periods. The planner checks these periods for each expected asset and lists the required main parts in a separate document, the main parts requirement list (23). This document is made on a weekly basis, based on the list of asset arrivals of NSR, and sent to Supply Chain Operations (28). A maintenance planning is made to reserve the required material and parts, and evaluated (24) with the available personnel and equipment for performing maintenance. The end result is a maintenance schedule (23) with required main parts based on the list of asset arrivals of NSR.

For wheel sets a slightly different approach is used because wheel sets are subject to an use-based maintenance policy. The planner uses the most recent wheel set dimensions as documented in the data-system R5 to check whether wheel sets needs to be exchanged. Planners might use the old-fashioned 50mm diameter ratio as a first check on wheel set tolerances (Beekmans, 2013) or use the DAAN-tool to check allowable tolerances in which wheel sets might be machined or mutually

(29)

exchanged. When a wheel set is likely to pass the exchange limit at machining it is added to the main parts requirement list(23) as well. These requests for wheel sets might include half-worn wheel sets, otherwise mutual exchanges are often necessary. The maintenance depot in Onnen provides a GTU-order when during maintenance or scheduling of maintenance the GTU-limit (38 mm) is passed.

Summarizing, the time horizon of maintenance schedules is about one week ahead and based on the list of asset arrivals of NS Reizigers. In this way is accompanied for (planned) wheel set demand, repetitively for the asset arrivals each week. They plan maintenance to reserve parts and capacity based on expected entry of assets every week (23). Short-term planning and vision is common for maintenance depots. Mid and long-term plans or forecasts are rarely, or even not, done (Kempenaar, 2013; Beekmans, 2013).

3.3.5 Gotcha disturbance

Besides planned arrivals of assets following short-cyclic maintenance, also unplanned arrivals between regular maintenance are present. The majority of severe wheel set defects is reported via the Gotcha Monitoring System and requires (unplanned) machining of these defects. Unplanned asset arrivals are called extra arrivals (EBK). When a wheel set under an asset in operation has exceeded the Gotcha 2 or 3 threshold, a service request for an extra arrival will be generated by MBN (Materieel Bijsturingscentrum NedTrain). MBN arranges a replacement wheel set, when the wheel set is likely to reach its rejection limit when machining it (Kempenaar, 2013). When an asset enters the maintenance depot as extra arrival only the wheel set having a defect is machined (Kempenaar, 2013). However, there is a possibility of combining an extra arrival with already planned short-cyclic maintenance.

Gotcha 2 and 3 alarms mean a disturbance at the lathe-process on scheduled assets at maintenance depots (Kempenaar, 2013). The schedule for machining planned wheel sets should be adapted for the unplanned Gotcha wheel set in agreement with MBN and only a short time in advance (Beekmans, 2013). Besides the available capacity of maintenance depots, also the availability of an asset for an extra arrival has to be decided by or consulted with NSR. It might happen that a train with Gotcha alarm is still required for operation due to capacity reasons of NedTrain and/or NSR, or immediate routing to a depot is not possible. An agreement with NSR is that after two Gotcha alarms a train should enter a maintenance depot as soon as possible. Different periods, in days, apply for each separate Gotcha level. Practice turns out that this could take up to 21 days (Van Klaveren, 2013).

3.3.6 Planning wheel set demand at overhaul depot

The department RLW (Revisie LoopWerken) at R&O Haarlem is responsible for overhauling bogies and wheel sets. The Delft Systems Approach model shows that R&O Haarlem has access to three stock locations for wheel sets:

• Buffer stock (3) in front of the overhaul process

• Ready-For-Use stock (6) as output of the overhaul process • External stock (16) at Lage Weide in Utrecht

The buffer stock (3) is the input for the overhaul process (4). The Ready-For-Use stock (6) is output of the overhaul process (4) and ready for transportation to maintenance depot stock locations (10) and (11) when they are requested. The external stock (16) is located at Lage Weide in Utrecht, which is separated from the working stock and contains spare main parts or main parts that are temporary not required (Stroomer, 2013). R&O Haarlem has the ability to check stock levels and arrange transports between the external stock and R&O Haarlem itself. This stock is also used for spare work in quiet periods, as troughs in demand, to level production capacity.

(30)

The Ready-For-Use stock (6) of wheel sets in Haarlem contains both half-worn and overhauled wheel sets. With half-worn is meant wheel sets that have been previously used under an asset, but are still above the exchange limit and thus do not need overhaul yet. Maintenance depots request these wheel sets frequently. For example when one wheel set has been machined below the exchange limit while the adjacent wheels sets are still above this

limit. In this case a new wheel set cannot be fitted, because of the large deviation in diameter. Hence, a half-worn wheel set in the appropriate size range might be a solution (Van Klaveren, 2013). In this way a replacement or adjacent wheel set do not have to be machined. A wrong delivery might lead to the decision to machine an overhauled wheel set to meet tolerances. An example is shown in Figure 3.5 at which at least 20 mm of wearable thickness was lost. Notice that 1 mm costs between 100-150 euro.

At time of the Lean on Wheels project in 2011 and 2012 R&O Haarlem pro-actively searched for wheel sets that matched these required range of diameters (half-worn wheel sets), but since the end of 2012 wheel sets are not delivered on requested diameter anymore. Current delivery of wheel sets is done based on FIFO principle from the Ready-For-Use stock (6). With this policy the diameter of the wheel set delivered to the maintenance depot has become unclear, which might give difficulties for a maintenance depot to comply with tolerances (Kempenaar, 2013; Stroomer, 2013).

Current forecasts at R&O Haarlem (19) are used for reserving required materials and personnel for their production, the overhaul of wheel sets. This forecast is carried out each 4 weeks as control function for their production. The input of this forecast is former production levels from the past two years as moving average. The output of this forecast is fed into the ERP system, called BAAN IV, once per 4 weeks. The forecasting model compares results from different Time-Series Analysis methods, the best curve fitting method is selected to forecast with (Stroomer, 2013). For the overhaul production of wheel sets a deliberation is held each Thursday with the internal material planner and production supply planner (Stroomer, 2013). In the model can be seen that there is no interaction between R&O Haarlem and maintenance depots about future demand. The maintenance depot in Onnen is an exception because they provide GTU-orders (requests for main parts) on average 90 days in advance, but it has not became clear how this information is used in the production planning. Requests for wheel sets at R&O Haarlem are received at random (Stroomer, 2013). For production planning the Min-Max levels of total wheel set stock are taken into account. Overhauls above the Max-levels do occur, for example when R&O Haarlem experiences a quiet period and keeps their personnel and production capacity up to a certain level (20). SCO states that exceeding these Max-levels is basically not acceptable (Van Wegen, 2013).

Peaks in wheel set overhaul are smoothened, or anticipated to, by manually re-scheduling overhauls earlier in time. This is also done when temporary defects or retrofit actions occur (Van Wegen, 2013). The current way of forecasting is applied since end of 2010. The planning tool they use for curve fitting is based on a tool from Gordian, an external consultancy organization in logistics, which was used in a project for ‘stock reduction’ (Stroomer, 2013).

Figure 3.5 A machined overhauled wheel set (right)