Reliability performance and maintenance - A survey of failures in wind power systems

Transcription

1 Reliability performance and maintenance - A survey of failures in wind power systems Master Thesis by Johan Ribrant Master Thesis written at KTH School of Electrical Engineering, 2005/2006 Supervisor: Lina Bertling, KTH School of Electrical Engineering Assistant supervisor: Thomas Ackermann, KTH School of Electrical Engineering Examiner: Lina Bertling, KTH School of Electrical Engineering XR-EE-EEK 2006:009

2 ii Reliability performance and maintenance a survey of failures in wind power systems

3 Abstract The wind power industry has expanded during the past few years and the growth has mainly focused on a growing market and the development of larger wind turbines. Different designs have emerged and the technical knowledge makes it possible to put wind turbines off shore. The fast expansion of the wind power market has also come with some problems. The new designs are not always fully tested, and the designed lifetime of 20 years is typically never achieved until the next generation of turbines is erected. The extreme conditions and the high loads that a wind turbine is exposed to makes the coordination of maintenance an interesting issue. How much maintenance is needed? Are there any ways of minimizing the maintenance and yet have a good availability for the wind turbine? The technical availability of wind turbines is high, around 98%, but this is due to fast and frequent service and not just because of good reliability or maintenance management. The problem area for this thesis work, performed within the RCAM group at KTH School of Electrical Engineering, is focused on the reliability for the components of the wind power system. If the most critical components for the system can be identified, it will show in what areas to focus when planning the maintenance for the system. If the condition of these critical components can be supervised, the maintenance can be planned even further. Investigations of failure statistics from four different sources reveal the reliability performance of the different components within the wind turbine. The gearbox is found to be the most critical as the downtime per failure is high in comparison to the other components in the wind power turbine. The statistical data presented also show trends of higher and even increasing failure frequency for bigger turbines compared to small turbines which have a decreasing failure rate over the operational years. Causes for failures to the gearbox are discussed and one of the major contributors to the failure is alignment. If the alignment is incorrect the wear on the gear and the bearings will be excessive and the lifetime of the gearbox will be reduced. To reduce the risk of a failure, the monitoring of the gearbox is required. One way of monitoring the performance of the gearbox is by using a condition monitoring system. A Condition Monitoring System, CMS, is a tool for telling in what condition the components in a system are. CMS are used today in many other applications but in the wind power industry the CMS is relatively new. With CMS a prediction of impending failure is given for each component, and therefore maintenance and repairs can then be better scheduled. The CMS for the gearbox primarily measures vibrations but a supervision of the oil is also necessary. The CMS used today are capable of detecting failures well in time prior to a failure and they are even able to predict which component inside the gearbox is defective. As a conclusion of this thesis work, it has been found that the gearbox is one of the most critical components when it comes to which component that influences the downtime the most. It is also shown that condition monitoring systems of today are able to supervise the gearbox adequately. The theoretical implications of using condition based maintenance together with condition monitoring systems shows great benefits and the overall conclusion is that the use of CMS is beneficial when it comes to reducing the amount of failures of the gearbox and also when it comes to scheduling the preventive maintenance. iii

4 iv Reliability performance and maintenance a survey of failures in wind power systems

5 Sammanfattning De senaste åren har vindkraftsindustrin expanderat och tillväxten har främst varit inriktad mot en växande marknad samt mot utvecklingen av större vindturbiner. Olika typer av design har utvecklats och det tekniska kunnandet gör det nu möjligt att placera vindkraftverken till havs, offshore. Den snabba utvecklingen av vindkraften har även bidragit med en del problem. The nya turbinerna är inte alltid fullt testade och den förväntade livslängden på 20 år är aldrig uppfylld innan nästa generation vindkraftverk är installerade. De extrema förhållandena och de höga laster som ett vindkraftverk är utsatt för gör samordningen av underhållet till en intressant fråga. Hur mycket underhåll behövs? Finns det möjligheter att minimera underhållet och ändå ha en god tillgänglighet för vindkraftverket? Den tekniska tillgängligheten för vindkraftverk är hög, omkring 98%, men detta är främst beroende på snabbt och frekvent underhåll och inte på grund av bra tillförlitlighet eller bra underhållsplanering. Problemområdet för detta examensarbete, som är utfört inom RCAM-gruppen vid KTH skolan för Elektroteknik, fokuserar på tillförlitligheten hos komponenter inom vindkraftsystem. Om de mest känsliga kritiska komponenterna för systemet kan identifieras, visar detta inom vilket område som underhållet behöver fokuseras. Om tillståndet för dessa kritiska komponenter kan övervakas, kan det förebyggande underhållet planeras ytterligare. En undersökning av felstatistik från fyra olika källor avslöjar tillförlitligheten hos de olika komponenterna i vindkraftverket. Växellådan visar sig vara den mest kritiska komponenten då hindertiden vid varje fel är hög i jämförelse med the andra komponenterna i vindkraftsystemet. De statistiska data som presenteras visar också på trender av högre och ökande felfrekvens för större turbiner jämfört med mindre turbiner som istället har minskande felfrekvens över driftåren. Orsaker till fel på växellådan diskuteras och ett av de stora bidragen till fel är upprikting, dvs. att alla komponenter är injusterade mot varandra. Om uppriktningen är felaktig kommer slitaget på växellåda och lager bli mer än förväntat och livslängden på växellådan blir förkortad. För att förhindra risken för fel behövs övervakning av växellådan. Ett sätt att övervaka växellådan är att använda ett tillståndsbaserat övervakningssystem, CMS (Condition Monitoring System). Ett tillståndsbaserat övervakningssystem, CMS, är ett verktyg för att avgöra i vilket tillstånd komponenterna i systemet befinner sig i. CMS används idag inom flera områden men är relativt ny inom vindkraft. CMS gör det möjligt att förutsäga när och vilken komponent som är på väg att fela och därmed kan underhållet planeras bättre. CMS för växellådan mäter huvudsakligen vibrationer men övervakning av växellådeoljan är minst lika viktig. De CMS som används i dag är kapabla att detektera fel i god tid och de kan även förutsäga vilken komponent i växellådan som är defekt. Sammanfattningsvis har detta examensarbete funnit att växellådan är en av de mest kritiska komponenterna när det gäller vilken komponent som påverkar hindertiden mest. Det har också visats att dagens tillståndsövervakningssystem kan övervaka en växellåda. De teoretiska resonemangen av att använda tillståndsbaserat underhåll tillsammans med dagens övervakningssystem visar på stora fördelar och den sammanfattande slutsatsen är att användandet av CMS är fördelaktigt när det gäller att reducera antalet hindertimmar för växellådor samt när det gäller att planera det förebyggande underhållet. v

6 vi Reliability performance and maintenance a survey of failures in wind power systems

7 Acknowledgments I would especially thank the following people for giving me valuable input and information: Nils-Eric Carlstedt at Swedpower AB, for letting me use the information found in their statistical database. Anders Andersson at Vattenfall AB, for giving valuable information and for the guided tour at Näsudden. Per Erik Larsson at SKF Luleå, for useful information. Hannele Holtinen at VTT in Finland, for valuable information and for the translation of the incident report. The personnel at Smöla, for providing information about their windfarm. And finally the colleagues at RCAM, KTH school of Electrical Engineering. vii

8 viii Reliability performance and maintenance a survey of failures in wind power systems

9 Table of contents Introduction Background and problem discussion Thesis background Problem background Problem discussion Approach Thesis Overview... 4 Theory The basics of a wind power plant Modeling of the system Choice of components The components of the wind power system Reliability theory Definitions Probability distributions and their applications The Alternating Renewal Process Measurements Maintenance methods Corrective maintenance Preventive maintenance Comparison of maintenance methods Maintenance strategy Analysis Survey of failures for wind power turbines Access to statistical data Failure statistics Statistics from Sweden Statistics from Finland Statistics from Germany Discussion about the reliability of the statistic data Conclusions on the findings in the statistical survey Overview of the Gearbox Gearbox design Gearbox operating conditions Gearbox development Gearbox wear and failures Causes for gearbox failures Conclusion on gearbox failures Condition Monitoring Systems Benefits of a Condition Monitoring System Insurance and CMS Condition monitoring in general Condition monitoring for gearboxes Conclusions about condition monitoring for gearboxes ix

10 Closure Conclusions and future work Conclusions Future work References Literature Interviews Appendix 1 Incident report from Sweden Appendix 2 - CMS suppliers x

11 Introduction 1 Background and problem discussion In the last 20 years turbines have increased in power by a factor of 100, the cost of energy has reduced, and the industry has moved from an idealistic fringe activity to the edge of conventional power generation. - European Wind Energy Association, Wind Energy - The Facts, 2005 [4] The wind power industry has expanded during the past few years and the growth has mainly focused on a growing market and the development of larger wind turbines. Different designs have emerged and the technical knowledge makes it now possible to put wind turbines off shore. The fast expansion of the wind power market has also come with some problems. The new designs are not always fully tested, and the designed lifetime of 20 years is typically never achieved until the next generation of turbines is erected. Some manufacturing failures have been so extensive that turbine manufacturers nearly went bankrupt. Consider that a modern wind turbine operates for about 13 years in a design life of 20 and is almost always unattended. A motor vehicle, by comparison, is manned, frequently maintained and its design life of about 150,000 kilometres is equivalent to just four months of continuous operation. - European Wind Energy Association, Wind Energy - The Facts, 2005 [4] The wind turbine is in several ways a unique power generating system as the power train components are subject to highly irregular loading from turbulent wind conditions, and the number of fatigue cycles experienced by the major structural components can be far greater than for other rotating machines [4]. 1.1 Thesis background This thesis work is a part of the pre-study on reliability-centered maintenance for wind power systems with focus on condition monitoring systems [1] performed within the RCAM group at KTH, School of Electrical Engineering on behalf of Elforsk. Elforsk is an organization owned by the Swedish power industry that encourages the industry to perform joint research and development within electrical generation and distribution. The long-term goal for that research project is to identify problem areas and possible solutions for optimal maintenance management. The focus of the research project is condition monitoring systems that could support the maintenance. This thesis work will examine one of the underlying problems with the availability of the turbine and suggest a possible solution for this kind of problem. 1

12 1.2 Problem background The extreme conditions and the high loads that a wind turbine is exposed to makes the coordination of maintenance an interesting issue. How much maintenance is needed? Are there any ways of minimizing the maintenance and yet having a good availability for the wind turbine? These are issues that are discussed today within research and development as well as within operations and maintenance for wind power plants. The reasons for the underlying problems within the wind industry were somewhat cloudy when the work with this thesis began. The technical availability of wind turbines is high, around 98%, but this is due to fast and frequent service and not just because of good reliability or maintenance management [1], [31]. It has also been known that the manufacturers seldom reveal data about their products and even more rarely do they share information about their failures, which is quite understandable. A report from Elforsk shows that in recent years the amount of damage claims has increased according to a report from a German insurance company. It is suggested that the reasons for this development are the following [2]: 1. Technical reasons o Insufficient prototype testing o Excessively fast development o Insufficient dimensioning and wrong selection of components 2. Operation and Maintenance reasons o Bad documentation o Lack of appropriate maintenance o Lack of quality control o Insufficient stocking of spare parts The suggestion that some components are not fully developed and that the maintenance is not appropriate is interesting, and motivated this work. New methods of how to predict the maintenance that is needed have evolved with the introduction of condition monitoring systems. A Condition Monitoring System, CMS, is a tool for telling in what condition the components in a system are. With this tool it is possible to predict when a component is likely to fail and therefore it is also possible to schedule its replacement in advance. Condition monitoring systems are used today in many other applications, e.g. the pulp and paper industry [30]. In the wind power industry the CMS is relatively new. In 2001 a project for examining the possibility of using a condition monitoring system was undertaken by Elforsk in cooperation with Göteborgs Energi and SKF Nova. At that time no monitoring system was available for a wind turbine and SKF Nova developed a system that was tested on a Vestas V44 turbine. [3] Today there are different condition monitoring system options available such as integrated systems sold by the manufacturer or separate systems sold by companies such as for example manufacturers of bearings. A typical owner of wind power turbines can have different turbines from different manufacturers and needs a CMS that will be functional for all of them. 2

13 The effect of these different systems is not yet thoroughly researched and this paper is an effort to clarify some of the issues of condition monitoring within wind power systems. 1.3 Problem discussion The problem area for this thesis work is focused on the reliability of the components of the wind power system. If the most critical components for the system can be identified, it will show in what areas to focus when planning the maintenance for the system. By doing an in depth study of the failures one can find out which components fail, how often they fail and if it is possible to measure the wear of the component and from this measurement decide when to perform the maintenance. The wind power systems usually have a high rate of availability but this is because of frequent maintenance [31]. Frequent maintenance however, is obviously not a good and optimal solution. Preventive maintenance at the right moment will save money for the owner of the wind power plant. Especially since some wind power plants are situated at remote sites, for example offshore. The problem discussion can be narrowed down into two major questions that will be clarified and given an answer in this thesis work. These questions are: 1. What component or components are most critical in the wind turbine when it comes to number of failures and the resulting downtime caused by these failures? 2. Is it possible to use a CMS to supervise these critical components and is CMS a suitable tool for decreasing the amount of maintenance for the wind power system? 1.4 Approach This thesis uses a quantitative approach as well as a qualitative approach. In a pre study phase, the state-of-the-art and basic fundamentals are investigated through books and course material used in wind power courses at the School of Electrical Engineering, KTH. The main findings about the lifetime and failures of the components are based upon statistical data which is analyzed with measurements used in reliability theory. The findings are then supported by information found in articles, books and interviews related to the area of failures within wind power systems. Quantitative analysis - based on statistical data from Sweden, Finland and Germany. Qualitative analysis - based on articles, internet resources, field trips, interviews and e- mail correspondence. 3

14 1.5 Thesis Overview Chapter 1 - Background and problem discussion This chapter gives an overview and a presentation of this thesis work and stipulates two questions that are subject to investigation. Chapter 2 The basics of a wind power plant The fundamentals of the components used in a wind power plant are presented. Different ideas on how to solve the problem with power regulation is explained briefly. Chapter 3 Reliability theory In this chapter main concepts and measurements within reliability theory are explained. Ways of modelling the lifetime of a component are introduced along with ways of modelling the wear and repair process of a system. The findings from reliability theory are later used as a tool to extract failure rates and important key figures from the statistical data. Chapter 4 Maintenance methods This chapter deals with different strategies for repair and maintenance. Different ways of how to perform maintenance were encountered during the work within this thesis and these are explained in this chapter. The basics from reliability theory are applied and aid in explaining the differences between the strategies. The concept of condition monitoring is briefly introduced to complement these strategies. Chapter 5 Survey of failures for wind power turbines The theoretical measurements found within reliability theory are applied to statistical data from three different countries. The findings from the statistical survey are analysed and presented along with conclusions about the frequency of failure and the downtime. With the conclusions found in this chapter, the first questions for this thesis work will be answered: which component fails the most and which has the longest downtime? Chapter 6 Overview of the Gearbox The statistical survey confirms that the gearbox is one of the most critical components for the wind power turbine. This chapter explains more about the gearbox and about some design terminology related to gearboxes. Chapter 7 Condition Monitoring Systems In this chapter an explanation on how the monitoring system works is given along with a description of techniques on how to monitor the gearbox. The chapter explains what these systems are capable of and what type of measurements they perform. No advice will be given as to which of the available systems is the best one, as a more thorough investigation would be needed. Chapter 8 Conclusions and future work A summary of all the findings together with topics and ideas for future work is described. 4

15 Theory 2 The basics of a wind power plant The function of a wind power system is to transform the kinetic energy in the wind into electric energy. This is accomplished by letting the wind energy force an aerodynamic rotor to turn. The wind energy is thus transformed into mechanical energy. The mechanical energy in the form of a slow turning rotor shaft is geared up to a high-speed shaft which is connected to a generator. Inside the generator the rotational mechanical energy is transformed into electrical energy. The electric power output is then connected to the grid. The basic function of the wind power system may sound easy but the system is still very complex. The development within wind power has been extensive in recent years and different concepts and construction designs have evolved. There has been a constant drive for higher performance and a higher power output. In addition to the complexity of the business, each manufacturer has basically chosen their own way of designing a wind turbine system. The evolution process within the wind power business has changed the features of some of the components, but the basic idea of turning wind energy into electrical energy via a generator is still the same. Many developments and improvements have taken place since the commercialization of wind technology in the early 1980s, but the basic architecture of the mainstream design is little changed. Most of the wind turbines have upwind rotors and are actively yawed to preserve alignment with wind direction. European Wind Energy Association, Wind energy, the facts, 2005 [4] The three-bladed rotor proliferates and, typically, has a separate front bearing with a low speed shaft connected to a gearbox which provides an output speed suitable for a four-pole generator [4]. 2.1 Modeling of the system The wind power system is a complex system and to do a better analysis a certain level of modeling has to be made. When modeling a complicated system, a good approach is to divide the system into smaller parts such as subsystems or components. In this case the whole plant including structure and all electrical parts up to the grid connection will be viewed as the system. The system consists of several complex parts that ought to be modeled as subsystems, but as a first approach all the subsystems are modeled as components of the main system. 2.2 Choice of components The selection of components for the description of the main system is not just an arbitrary choice but a choice of what is useful in practice and where available data can be found. The choice of which component should be used for modeling the whole system is based on function and available information. 5

16 2.2.1 A choice based on function When describing a wind power system, a common way is to explain the main function by dividing the system into a set of different components with different features, for example brakes, tower, rotor blades etc. The different components are manufactured differently and are easy to replace as modules in the system, hence it is convenient to view them as separate components in the system A choice based on information The second choice for which components to be used in the modeling of the system is based on what information that is available. When statistics of failures are reported, it is inconvenient to have reports sheets with every component down to the smallest bolt, instead they are grouped according to a set of components based on their function. Failure reports from Germany, Sweden and Finland are divided into the same set of components and they are basically based on their function within the system. 2.3 The components of the wind power system The names of the components are general and apply to almost all designs of wind turbines. The terminology used for the components comply with the same terminology used within the wind power industry. The system components described here are for a common system with the basic features. Rotor blades Wind Hub Nacelle Wind Tower Foundation Figure 1: Overview of different parts of a Wind Power Plant Rotor blades and Pitch system The wind makes the rotor blades turn, thus making the shaft inside the wind turbine turn. There are different designs of the blades but lightweight and sturdy are the basic features. The blades are generally made from glass fiber reinforced plastic. The reinforcement can also be carbon fiber or laminated wood. Some blades have advanced techniques for lightning protection built into the blade. Another feature of some blades, is heating inside the blades to be used in arctic climates. 6

17 The most common design is a three-bladed rotor. The two-bladed rotors are used commercially but most manufactures prefer to produce three bladed rotors. A two bladed rotor spins faster than a rotor with three blades and might appear less appealing to the eye [4]. Closely interconnected to the rotor blades is the pitch system. The objective of the pitch system is to regulate output power at high operational wind speeds. This involves turning the blades about their long axis (pitching the blades) to regulate the power extracted from the rotor. Pitch regulation changes the rotor geometry and this involves active control of the system to sense blade position, measure output power and to instruct changes of the blade pitch [4]. The pitching angle is controlled by the control system and is usually regulated by a hydraulic system but electrical motors for pitching the blades are also available. Not all wind turbines use the pitching technique; some rely on other techniques to regulate the power output. Pitch regulation also makes it possible to smoother start up the wind power turbine as wind increases. Since pitching offers a better output, these are favored among larger turbines. The thrust of the rotor on the tower and foundation is lower for pitch-regulated turbines and this allows for reduction of material and weight Hub The hub is seldom separately defined in failure statistics but is categorized as a part of the structure. For the complete understanding of the structure of the wind power plant the hub is shown separately in Figure 1. The hub is the centered construction, which connects the blades to the main shaft. The hub is usually made out of cast iron [5]. Inside the hub is electrical and mechanical equipment for controlling the blades Structure Tower, Foundation and Nacelle The structure consists of the tower, and the nacelle and the rotor that it carries. Generally, it s better to have a high tower, as wind speeds increase further away from the ground. When examining failure statistics one finds that the component structure usually includes the foundation beneath the tower and the nacelle. The nacelle is the housing for the gears and the electric generator at the top of the tower, see Figure 1 and also Figure 2. 7

18 Drive train Electrical system Gearbox Generator Figure 2: View inside the nacelle Drive train The drive train basically consists of the shaft and the bearings and occasionally a clutch between the gearbox and the generator. In Figure 2 the drive train is represented by a single box, but in reality it is the interconnecting shafts between the hub, the gearbox and the generator. The shaft goes into the nacelle from the hub, where the blades are connected, and connects to the gearbox. The shaft rotates with low speed and needs to be geared up, which is done in the gearbox. On the other side of the gearbox the high-speed shaft exits into the generator. When examining the different components within the wind power system one finds many sets of bearings at different locations where there are rotating machinery. The bearing that is mentioned as a part of the drive train is present if the turbine is constructed with a main bearing. Another way of designing the turbine is by implementing the main bearing directly into the gearbox Gearbox The gearbox transforms low-speed revolutions from the rotor to high-speed revolutions. To transform the low rotational speed of about 30 rpm to 1500 rpm, usually three stages are needed. The design of the gearbox is subject to constant changes. At the moment a common solution is to use a planetary stage gear which has a feature of being very compact. Via a high-speed shaft the gearbox is then connected to an electric generator. A high speed revolution of about 1500 rpm is a requirement for transforming rotational energy to electrical power of good frequency. Less rotational speed is needed if the generator has more pole pairs Generator The type of generator used in the wind turbine varies, but usually it is an induction generator or a double fed induction generator, DFIG. The generator transforms the rotational energy into electrical energy. The generator is connected to the electrical system and supplies the transformed energy to the electrical system. 8

19 2.3.7 Electrical system This is basically all equipment required to deliver and control the electrical energy that follows from the generator to the grid. The electrical energy usually has to be controlled in different ways depending on amount of active and reactive power, voltage and phase. Modern designs let the power output from the generator pass through a set of power electronic components to control the power and the frequency before supplying it to the grid. The boundary for the wind power system in this thesis work is between the electrical system and the grid Control system The control system is made out of a main computer inside the nacelle or in the tower structure. The control unit surveys the power output, wind and wind direction and controls the settings so that the pitch and the yaw can be optimized. The control system is connected to several sensors within the wind power structure. This control system is not to be confused with condition monitoring systems. The function of the control system is only to supervise the system so that performance at the moment is optimized, the safety of the system is maintained and alarms are reported in case some sensor signal is above a set parameter limit value. In larger wind farms the control systems from different turbines are monitored by a centre of operations Sensors In a typical turbine there are about 30 to 50 monitoring sensors; more modern turbines have more sensors, about These sensors include wind measurement equipment as well as sensors for temperature, wind direction, vibrations, revolutions, cable twist etc. The sensors are connected to the control system. If a CMS is installed in a turbine, some sensors can be shared and some need to be independent. For the most basic condition monitoring, i.e. vibration monitoring of the gearbox, only about eight measuring points are needed but modern CMS integrate measuring values from other parts of the system just like the control system, e.g. temperature, wind direction etc, hence more measuring points are needed Mechanical brakes Mechanical brakes are essential for safety reasons. During high winds and repair it is crucial that these brakes are functional. The wind power system can utilize both aerodynamic brakes and mechanical brakes. Aerodynamic brakes are when the blades are pitched into a position where as less wind force as possible is absorbed. The mechanical brake system consists of a disc break in conjunction with the gearbox Hydraulic system Hydraulic components are used in the turbine. Pitching, braking and yawing are features within the turbine that commonly rely on hydraulic systems. 9

20 Yaw system The yaw system is the system for controlling how the tower turns, because as the wind turns the nacelle needs to adjust itself so it faces the wind properly. This system contains bearings, gearwheels, brakes and a yaw motor. 10

21 3 Reliability theory The main objective of a reliability study should always be to provide information as a basis for decision. - Rausand and Hoyland, 2004 [6] The results provided by a reliability study will not tell us exactly what to do, but in what direction to look. For example, a reliability study can be useful in areas of risk analysis, optimization of operations and maintenance. The risk analysis is a way of identifying causes and consequences of failure events, and the optimization is a way of telling how failures can be prevented and how to improve the availability of a system. One can see reliability theory as a tool for analysing and improving the availability of the system. 3.1 Definitions A definition of reliability is: the ability of an item to perform its required function under given conditions for a given time interval [10]. This ability can be described in terms of probability and the probability distribution may be used to model the lifetime of a component. Examples of probability distributions is shown in chapter 3.2 and examples of measurements of reliability and availability is shown in chapter Probability distributions and their applications To model the lifetime of components probability distributions may be used. There are several different types of distributions suitable for different kind of applications. In this thesis only the Weibull and the exponential distribution will be considered The Weibull distribution and the exponential distribution The Weibull distribution is a widely used life distribution in reliability analysis. The distribution is very flexible and can through an appropriate choice of parameters model many types of failure rate behaviours. [6] The Bathtub-curve can be modelled easily with three different sets of parameters respectively for the three different phases. The distribution for the useful life period is a special case of the Weibull distribution. This special case of Weibull distribution is equal to an exponential distribution. Hence for the useful life period, the exponential distribution is used. Equation 1: The Weibull distribution f ( t) = λ β β β 1 ( λt) t e β for t > 0 Equation 2: The exponential distribution λt f ( t) = λ e for t > 0 The use of exponential distribution for lifetimes comes with a number of important side effects. 11

22 The failure rate is constant which means that it is independent of time. [6] The exponential distribution has no memory, so an item is always viewed as good as new as long as it is functioning. [6] When estimating the reliability function, the mean time to failure and so on, it is sufficient to collect data on the number of hours of observed time operation and the number of failures. The age of the component is of no interest in this context. [6] Bathtub curve and other shapes of curves Normal mechanical failure modes degrade at a speed directly proportional to their severity. Thus, if the problem is detected early, major repairs can be prevented in most instances. - Davies, 1998 [7] According to Davies [7] one needs to find the right time for the failure to prevent major repairs, but before trying to find the time for a failure one needs to examine and learn more about the lifetime of the component. The failure rate of a component is often high in the initial phase of its lifetime. This can be explained by the fact that there may be undiscovered defects in the components [6]. When the component has survived the initial period, the failure rate stabilizes at a level where it remains for a certain time until it starts to increase again as the component begin to wear out. The shape of the curve depicting the failure rate of the component, is similar to that of a bathtub, hence the expression bathtub-curve. Figure 3 shows the bathtub curve with the three typical phases. The initial phase is called burn in period, the stable phase is called useful life period and the end phase is called wear out period. Other examples of names for these three periods are break in, operations and breakdown. This terminology varies in literature but the main concept of three different stages in the life of the component or system are still the same. Number of failures Burn-in period Useful life period Wear out period time Figure 3: The Bathtub curve Figure 3 gives one example of a possible shape for the failure function. There are other failure functions with other shapes, but the bathtub curve appears as a good choice for mechanical components such as gearboxes, which later on will be studied further (see chapter 6). For the 12

23 majority of mechanical items the failure rate function will usually show a slightly increasing tendency during the useful life period, because of the wear on the mechanical components [6]. 3.3 The Alternating Renewal Process When a component fails, immediate repair is undertaken and when the repair is done, the component is put back into the system and is considered as good as new, hence the expression renewal. status of system Failure occurs as good as new 1 0 Wear time Repair time time Figure 4: Alternating Renewal process Wear model To be able to understand and to apply theoretical tools to a physical component models are used. One way of modelling the system is by setting it to one of two states: up or down, failure or no failure, see also Figure 4. We can picture the state of the system as a binary process. The statistical data used in this thesis is only based on the stages; up or down, hence only a model with two states will be used Improved wear model It is also possible to look at models with intermediate states between completely new and completely failed. In this type of model, failure is a damage accumulation process [8], see Figure 5. A good example is mechanical deterioration, where there are several states between brand new and failed. Wear is defined as the progressive loss of substance resulting from mechanical interaction between two contacting surfaces. - Davies, 1998 [7] A model with several states appears suitable for systems with monitoring equipment. The wear model with different stages of deterioration is applicable when analysing specific components where the different stages of wear have been well defined. 13

24 status of system 1 Damage accumulating process as good as new 0 wear time repair time time Figure 5: Damage accumulating process Note: For this thesis work, no information of intermediate stages of wear is available thus the process used in the thesis is an alternating renewal process, described in Chapter Repair time The repair time can be modelled similarly to the lifetime of operations. There is a suitable distribution for repair time, the lognormal distribution, which for example takes into account that some repairs can be made quickly while other repairs rely on spare parts that are not available at the moment. It is also common to use the exponential distribution for repair time. The repair time is of course important when detailed models of the maintenance are considered but as we will later find out it is difficult to find data concerning repair of wind power turbines and yet more difficult to find out the exact amount of time spent on repair. The information that may be available is the amount of time that the system was unavailable, but this time may consist of scheduled maintenance and stoppages caused by other events not connected to any failure. In this thesis a model of exponential distribution for repair time will be considered. 3.4 Measurements To be able to acquire useful information about the performance of a system or component, some measurements of the reliability and availability have to be used. Later in the analysis of data form the wind power plants these measurements will be used in order to compare different components and different systems Measurements of reliability performance The reliability can be measured in many ways depending on the particular situation, for example as: Mean time to failure or number of failures per time unit or failure rate. [6] The mean time to failure, MTTF, is defined as the mean time between initial operation and the first occurrence of a failure or malfunction, as the number of measurements of such time on many pieces of identical equipment approaches infinity. When a failure has occurred the item is repaired and put back into operation and the item is then considered as fully functioning. The mean down time, MDT, is defined as the average time that the system is not functioning when a component is being repaired, and is basically the time it takes to repair a failure. The 14

25 mean time between failures, MTBF, takes into account the mean time to failure and the mean down time. The down time is usually much shorter than the time of operations and then the two measurements can be viewed as: MTTF MTBF, see Figure 6. Operating status Up Down MTTF MDT time MTBF Figure 6: Measurements of reliability Measurements of availability performance The availability performance is defined as: the ability of an item to be in a state to perform a required function under given conditions at a given instant of time or during a given time interval, assuming that the required external resources are provided Maintenance terminology, SIS 2001 [10] By using the measurements of reliability performance, i.e. MTBF and MTTF, the availability for the system can be described as the portion of operational time, MTTF, over a nominal period of time, in this case MTBF, given that the time t approaches infinity. In Equation 3 the equation for such a measurement of availability is shown. Availabili ty Equation 3: Availability [6] = MTTF MTTF = when MTTF + MDT MTBF t The measurement of availability differs within wind power. A commonly used measurement of availability is the amount of operational time divided by the nominal time, see Equation 4. The nominal time is usually a period of one year and then the availability is presented as percentage of operational time per year. This type of definition is used in Sweden and within the WMEP research in Germany. Equation 4: Availability [9] Nom. time Downtime Availablity = Nom. time Another way of expressing the availability is to eliminate downtimes not caused by the wind power plant, such as external failures of the grid, see Equation 5. This type of definition is used in Finland. 15

26 Equation 5: Availability with regard to grid disturbances [16] Nom. time ( Downtime Downtime caused by gridfailures) Availablity = Nom. time Note: These two different definitions of the availability have been used in the sources for the statistical data. The effect these differences have on the result is not investigated, but assumptions say that the two different definitions will not influence the result. A third option to use for availability is to not use the nominal time of one year but the actual available operational time. E.g. The available operational time is only when the wind is blowing and not when the plant has stopped due to low winds or to high winds. Unavailability is the period which the plant is not functioning. This can be scheduled downtime (maintenance) or unscheduled downtime (malfunction or failure). 16

27 4 Maintenance methods Maintenance is required for almost all types of machinery and applies also to the wind power system. The type of maintenance that is performed can be defined as either preventive or corrective maintenance. Preventive maintenance is carried out at predetermined intervals or according to prescribed criteria and is intended to reduce the probability of a failure. Corrective maintenance is carried out after a failure and is intended to repair the system. [10] In other words, preventive maintenance is performed before a failure and the corrective is preformed after the failure occurs. An ideal maintenance strategy meets the requirements of machine availability and operational safety, at minimum cost. - Rao, 1996 [11] Consequently the challenge in planning the maintenance is to decide on when to perform preventive maintenance. In this chapter an explanation of three different methods for maintenance is presented; corrective maintenance and two types of preventive maintenance; scheduled maintenance and condition based maintenance, see Figure 7. Maintenance Preventive Maintenance Corrective Maintenance Condition based Maintenance Scheduled Maintenance Figure 7: Classification of maintenance types [6] 17

28 4.1 Corrective maintenance Corrective maintenance is defined as [10]: Corrective maintenance - Maintenance carried out after fault recognition and intended to put an item into a state in which it can perform a required function. This type of maintenance is often called repair and is carried out after the failure of a component. The purpose of the corrective maintenance is to bring the component back in to a functioning state as soon as possible, either by repairing or replacing the failed component. [8] To only use corrective maintenance is seldom a good solution. This means that you will run you system until a breakdown occurs and in some literature this is referred to as a breakdown strategy. [23] With a breakdown strategy the preventive maintenance is reduced to a minimum and the system will be operated until a major failure of a component occurs which will result in a shutdown of the wind turbine. This strategy is risky, since failures of relative small and dispensable components can lead to severe consequential damages. Another aspect of such a strategy is that most component failures are likely to be related to the actual load condition of the wind turbine and is also likely to happen during high load conditions. This means that the shutdown of the turbine is related to high wind periods. Downtime in such periods will lead to higher production loss. If the wind turbine is situated offshore, the accessibility is likely to be bad during high wind periods. [23] Another drawback of this strategy is that when repair is needed the downtime can be extensive since logistics gets more complicated and delivery periods for spare parts can be long. A breakdown strategy minimizes the cost for repair and maintenance during operation. With no knowledge of the consequence of a failure until it occurs makes it impossible to calculate the costs of replacements. The lifetime of the component is unpredictable and only once the component has failed can an assessment of the cost and lifetime be made. [11] Scheduled maintenance Condition [%] Corrective maintenance Breakdown Time Figure 8: Corrective Maintenance compared to Scheduled Preventive Maintenance 18

29 4.2 Preventive maintenance Preventive maintenance is defined as [10]: Preventive maintenance Maintenance is carried out at predetermined intervals or according to prescribed criteria and intended to reduce the probability of failure or the degradation of functioning of an item. The preventive maintenance is performed regularly to postpone failures or to prevent failures from occurring. There are two different types of preventive maintenance; the scheduled maintenance and the condition based maintenance. What differs between these two are the way of deciding when to perform the preventive maintenance Scheduled maintenance Scheduled maintenance is defined as [10]: Scheduled maintenance - Preventive maintenance carried out in accordance with an established time schedule or established number of units of use. Scheduled maintenance means that preventive maintenance is carried out in accordance with an established time schedule [10]. The time-schedule for the preventive maintenance can be either clock-based or age-based maintenance. Clock-based maintenance means that the preventive maintenance is carried out at specified calendar times and age-based maintenance means that the maintenance is carried out when a component reach a certain age. The age does not need to be calendar time, but measured in for example revolutions or operational time etc. [6] Preventive maintenance performed at scheduled intervals should be designed to reduce the probability of failures. Maintenance cycle times will be matched to the requirements of the system. The system will be inspected and maintained periodically, see Figure 8. The components that first show sign of wear and fatigue will be maintained and replaced. This type of maintenance strategy means that components exposed to wear will be replaced regularly even if they are not at the end of their lifetime. Scheduled maintenance requires regular access to the system and a big share of the costs for the maintenance will stem from the supply for cranes and maintenance personnel. Transport of personnel and spare parts to the wind farm can also be cost intensive with this preventive maintenance strategy. The advantage of preventive maintenance is that it can be scheduled ahead of time and the coordination of logistics can be made easy. [23] 19

30 4.2.2 Condition based maintenance Condition based maintenance is defined as [10]: Condition based maintenance Preventive maintenance based on performance and/or parameter monitoring and the subsequent actions. Performance and parameter monitoring may be scheduled on request or continuous. Condition based maintenance is a type of preventive maintenance that is based on the performance and monitoring of parameters from the system. With this type of preventive maintenance, monitoring equipment collects machine data. The condition monitoring may be scheduled, on request or continuous. The collected machine data can indicate required maintenance prior to predicted failure. Maintenance is initiated when a condition variable approaches or passes a threshold value. The system components will be operated to a defined condition of wear and fatigue. When this condition is reached, the component needs be maintained or replaced. [23] Examples of condition variables that the system monitors are vibration, temperature, number of particles in the lube oil etc. The ability to monitor the condition of components facilitates planning of maintenance prior to failure and will minimize downtime and repair costs. The components will be used closer to their lifetimes and the coordination of spare parts will be easy. Another benefit of implementing a condition based system is that trends and statistical data such as mean time to failure can be provided.[11] The statistical data from monitoring system is important for getting reliable data for remaining lifetime of components in the system. With site specific data the prediction of remaining time for the components can be more precise. Figure 9 shows an example of condition based maintenance along with corrective and scheduled maintenance. Scheduled maintenance Condition [%] Condition based maintenance Corrective maintenance Breakdown Figure 9: Condition based maintenance compared to scheduled and corrective maintenance Time 20