Renewable Energy 36 (2011) 2154e2165

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Challenges in the reliability and maintainability data collection for offshore

wind turbines
Z. Hameed a, *, J. Vatn a, J. Heggset b
a
Norwegian University of Science and Technology, Production and Quality Engineering, N-7491, Trondheim, Norway
b
SINTEF Energy Research, Trondheim, Norway

a r t i c l e i n f o a b s t r a c t

Article history: Wind energy is abundantly available both onshore and offshore. As a response to the present climate
Received 25 June 2010 crisis focus on wind energy is increasing due to its renewable and environmentally friendly character-
Accepted 11 January 2011 istics. Due to social and political reasons the trend has been shifted largely from onshore to offshore wind
Available online 18 February 2011
farms. Offshore wind energy production faces a wide range of new challenges in design, development,
manufacturing, installation, and maintenance and operation. The need, objectives, method, benefits, and
Keywords:
application of a proposed reliability and maintainability database are identified in this paper. In the
Reliability
offshore oil and gas industry the OREDA concept for data collection has been running for more than 25
Maintainability
Failure
years. Therefore it will be briefly described what is considered to be the state of the art in this industry
Database when it comes to data collection. Potential challenges and issues pertaining to the reliability and
Offshore wind turbine maintainability data collection of offshore wind turbines are outlined and categorized. The architecture
of the proposed database is illustrated. The main building blocks of the database are briefly described
and their possible effects on the reliability and maintainability of offshore wind turbines are highlighted.
It is expected that the realization of the proposed database will open a new vista of knowledge in
understanding the real behavior of offshore wind turbines in the marine environment. Another expec-
tation is the benefits it will bring to the technological areas ranging from design to operation.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction offshore wind power capacity will double each 2nd to 3rd year until
2015, each 3th year from 2015 to 2020 and, finally 5th year from
Wind energy is the emerging source of power today and its 2020 to 2050 [1].
share of total energy production is increasing on a very rapid scale. Due to the rapid growth of offshore wind power (its share in the
Currently the installation of wind turbines is at its peak and focus is energy market is expected to rise exponentially within the coming
shifting from onshore to offshore locations. The projection of wind decades) has necessitated researchers and industry partners to
energy growth worldwide and the development of offshore incorporate novelty from the design phase to operation and
compared to onshore wind turbines are shown in Figs. 1 and 2 maintenance (O&M). With the emergence of this offshore wind
respectively. power arena, research is focused on how to have robust design in
It is expected that the share of offshore wind energy will a marine environment, what kind of vessels should be employed to
increase rapidly as the time passes and possibly may surpass other install and maintain the wind turbines, how the issues related to
traditional and renewable energy sources. One such scenario for transportation, logistics, and accessibility in a offshore environ-
the growth of offshore wind energy is shown in Table 1. ment will be addressed, and what the optimal O&M strategies will
Table 1 indicates that total offshore wind power production is be. These issues are of great concern for the people engaged in
calculated to be TWh 2,559 in 2050, offshore wind power will different areas of offshore wind turbine work today. There are new
supply a little more than 6% of global final electricity consumption concepts and thoughts to address all such kinds of challenges and
and will constitute approximately 18.4% of total wind power one possible solution to deal with them is to have methodologies
capacity. The assumed growth implies that the accumulated global that ensure the reliability of the wind farm as a whole.
One such factor to ensure the reliability of offshore wind turbines
(OWT) is to implement the RAMS (Reliability, Availability, Main-
* Corresponding author. Tel.: þ47 73 59 71 02; fax: þ47 73 59 71 17. tainability and Safety) engineering holistically, starting from the
E-mail address: [email protected] (Z. Hameed). design to the operational phases. As reliability and maintainability

0960-1481/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.renene.2011.01.008
 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165 2155
Table 1
Scenario for global offshore wind power development [1].

Year Offshore Yearly growth Offshore of total Production from Expected global electricity Penetration of
wind GW offshore wind/% wind power/% offshore wind/TWh/y consumption/TWh/y offshore wind, %
2006 0.9 1.2 3 15,500 0.0
2015 12.8 34 2.6 42 21,300 0.2
2020 42.4 27 4.0 140 23,800 0.6
2030 251.1 19-5 9.5 829 29,750 2.8
2050 773.8 5.5 18.4 2559 40,100 6.4

are integral part of the RAMS concept, one way to address the new coordinated and dedicated RAMS database for wind power equip-
challenges of OWT is to have a dedicated RAMS database which will ment in the Nordic countries. The typical situation is that the
underpin the implementation of RAMS engineering on the whole manufacturers of wind turbines retain the detailed data about
system. In order to realize this database, data should be gathered faults and other incidents through O&M contracts with the wind
from all available wind turbine manufacturers who are engaged in turbine owners. Such contracts usually define the requirements for
the OWT business. The database should contain all the necessary availability and reliability, and detailed reports about component
information for conducting RAMS engineering for having a reliable faults and power interruptions are typically neither required nor
design, decreased downtimes, and stable production from the made available to the wind turbine owner upon request. A chal-
system. lenge to solve in this context will be to work out a suitable
The concept of having a common database for the industry is not compromise between the data transparency required in the data-
a new idea. It follows the inception of reliability approaches and base and the protection of information that is considered to be
consequently their applications in engineering after World War I. sensitive/competitive by the owners.
A good repository of existing databases related to reliability and In order to enable the assessment of RAMS for critical compo-
accidents is summarized in [2] and the reader is strongly recom- nents, such a database is a necessity and the work to implement
mended to consult in order to gain an overview of the existing and make it operational should thus be given high priority. By the
databases and their nature. Databases related with reliability, involvement of several countries in accomplishing such a task, the
safety, and offshore accidents could be of direct use in formulating number of comparable components will be large and thus provide
the new database for the OWT industry. more reliable statistics. This means that the wind turbine owners
Today’s best-known wind turbine databases include WMEP [6] will get a standard to base their O&M contracts upon.
and LWK [23] from Germany, and the WindStats Newsletter pub-
lished in Denmark. Additional databases originate from Finland
2.2. Description of the objectives of a RAMS database
(VTT), and Sweden (Elforsk) [3]. Besides these data sources for wind
turbines, a dedicated RAMS database is available under the heading
The main objective of a reliability study should always be to
of “OREDA” (Offshore Reliability Data) for the oil and gas industry
provide decision support. Before a reliability study is initiated, the
which was the joint venture of different industries engaged in this
decision maker should clarify the decision problem, and then the
business. Although this database is related to the oil and gas
objectives and the boundary conditions and limitations for
industry it still may provide a good foundation for making a reliable
the study should be specified such that the relevant information
and accurate RAMS database for OWT and could be a good starting
needed as an input to the decision is at hand, in the right format
point for having a repository for a RAMS database. This database is
and on time [4].
dealing exclusively with reliability issues and it is expected that the
The objective of such an activity will be to develop the scientific
experience gained from this database could be translated into
foundation for a future RAMS database for critical components in
a dedicated RAMS database for OWT.
wind power farms. This could be achieved through the following
activities:
2. RAMS database for OWT
- definition of basic requirements including terminology, and
2.1. The need for a RAMS database stakeholders.
- definition of stakeholders’ needs regarding RAMS data.

The rapid growth of the wind power has compelled the need for - analysis and definition of input data required to fulfill the user

a database for the reliability and safety of wind turbines. The needs.
situation in the wind power industry today is that there is no

Fig. 1. Global wind energy growth scenario [20]. Fig. 2. Development of onshore and offshore wind energy [20].
2156 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165

- development of data collection schemes and corresponding systems, as well as for customers that will ask for a database
guidelines. according to the specification.
- specification of a RAMS database including data tables and The wind power industry is relatively young with continuous
formats. technological evolvement. There has for example been a steady
- recommendations for future implementation and operation of increase in production capacity and size has been observed, and an
the database. increasing number of turbines are mounted offshore with rougher
- motivation and acceptance from the stakeholders for data climate and other failure mechanisms than for onshore turbines. As
collection by implementing processes that secure anonymity a consequence, there are a lot of new actors in this area, and there is
of the data, and emphasize the utilitarian values of such a considerable need to make the experience available. Thus the
a standardized data collection regime. collection of operational history in a structured way will make it
possible for the actors to take advantage of operational experience
The degree of innovativeness could be substantial compared at an early stage, and reveal possible design or construction-related
with the situation today, where it is acknowledged that there is problems early.
a problem caused by the lack of open and available data of sufficient The improvement of equipment reliability is dependent on
detail relating to the wind power industry. Existing data collections experience from real-life usage. The collection, analysis and feed-
are usually not detailed enough. For example, in addition to the back of data to equipment designers and manufactures are, there-
information “Gearbox failure occurred”, it is essential to know, fore, paramount. Also, when purchasing new equipment, RM
which part of the gearbox failed due to which mechanism, what type (reliability and maintainability) data are key parameters to take
of maintenance was conducted and what will be the effect on future into account [18].
failure behavior to be of use for maintenance optimization purposes. A typical feedback loop for potential uses of the data is shown in
The proposed database will provide the stakeholders with high Fig. 3.
quality RAMS data for a variety of applications (maintenance plan-
ning and optimization, basis for condition based maintenance (CBM) 2.5. Application
introduction, availability analyses, life cycle cost and profit (LCC &
LCP) analyses, exchange of experience between different wind farm It is expected that the realization of the proposed RAMS data-
owners, etc). The realization of such a task will provide a solid basis base may provide the foundation for the improvement from design
for initiatives towards other European and internationally for an to planning O&M strategies of OWT. The reason behind such a shift
even broader cooperation within this field in the future. could be to think and act in advance to ensure the reliability of the
whole OWT farms which are expected to be quite complex and
2.3. Method intricate compared to onshore wind turbines.
The possible application areas of the proposed RAMS database
The proposed methods and work plan to realize the objective of are identified and listed below. For a detailed study, the reader is
a RAMS database could be: referred to consult [4].

- literature study for establishing the state of the art. 2.5.1. Engineering design
- interviews with operators of RAMS databases in other Reliability is considered to be the one of the most important
countries and other businesses to secure the transfer of quality characteristics of technical products. Reliability assurance
experience. should therefore be an important topic during the engineering
- questionnaires to various stakeholders in order to reveal user design. Many industries like nuclear power, aviation, automobile,
needs and to assess the critical components, their dominant and offshore, have realized this and have integrated the concept of
fault modes and the corresponding consequences. reliability in design. Such integration may be accomplished through
- complementary analysis and evaluation of existing data concepts like concurrent engineering and design for X.
investigating critical components and failure modes.
- pilot study for testing the data collection schemes and 2.5.2. Quality
guidelines. The concept of quality and reliability are closely linked. Reliability
- several statistical methods to define the link between input may in some respects be considered to be a quality characteristic
data and statistics that meet the user needs. (perhaps the most important characteristic). Complementary systems
are therefore being developed and implemented for reliability
2.4. Benefits management and assurance as part of a total quality management
(TQM) system.
The main advantages after realizing this database could be
a detailed specification of a RAMS database complete with a data 2.5.3. Verification of quality/reliability
model comprising data tables and their relations, and data formats. In A number of official bodies require that the procedure and/or
addition, the activity will provide data collection schemes with cor- the user of technical systems are able to verify that their equipment
responding guidelines and means for quality assurance of the whole
process from data collection to the generation of statistics. The key
words in this context are anonymity, transparency and traceability.
This work will provide a standardized framework (terminology,
data formats, guidelines, etc.) that facilitates the exchange of
experience between the stakeholders, as well as benchmarking
between companies and countries. As part of this work the project
will ensure that the database will be defined in accordance with
international standards regarding RAMS and risk initiatives. The
detailed database specification delivered by the scheme will be Fig. 3. Typical feedback of analysis from collected reliability and maintenance data
a valuable document for software vendors to offer data collection [18].
 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165 2157

satisfies the specified requirements. Such requirements usually reliability information, development of guidelines and software for
have a basis in safety and/or environmental protection. Moreover, data collection and analysis and publication of reliability data and
during the last few years it has become more and more common know-how. Two examples reported by the participating oil
that buyers of technical equipment require a quantitative assess- companies refer to total savings of USD 70 million for alternative
ment of the quality and reliability as part of the total system designs in comparison with the original designs. Similar savings
documentation. have also been reported by other companies. Reliability data were
not the only basis on which the decisions were made, but they did
2.5.4. Risk analysis provide better grounds for selecting a good design.
The risk analysis is based on the identification and description of Application of the OREDA database ranges from Availability
potential accidental events in the system, the potential causes of studies, Risk Analysis, LCC, Benchmarking, Maintenance planning
each accidental event are identified by a causal analysis, and most and optimization, and operations. The OREDA data structure is
well-designed systems include various barriers and safety func- shown in Fig. 4.
tions that have been installed to stop the development of accidental Main OREDA achievements are:
events or reduce the consequences of accidental events. Reliability
analysis is a main part of any quantitative risk analysis and several - existed for 28 years (2009) as an international joint industry
methods are common for risk and reliability analysis. project with 7-11 international oil and gas companies from
5 countries as members.
2.5.5. Optimization of maintenance and operation - established as the forum for the collection and exchange of

Many industries have fully realized the important connection reliability data, and utilization of reliability methods in the oil
between maintenance and reliability and have implemented the & gas industry.
reliability centered maintenance (RCM) approach. The RCM - a comprehensive databank of E&P (exploration and

approach is the main tool to improve the cost-effectiveness and production) reliability data collected worldwide for:
control of maintenance in all types of industries, and hence to  Offshore topside and subsea equipment
improve availability and safety. Reliability assessment is also an  Onshore E & P equipment
important element in the following applications: life cycle cost - developed guidelines and software for data collection and

(LCC), life cycle profit (LCP), logistics support, spare part collection, analysis, topside and subsea.
and manning level analysis. - issued 4 comprehensive reliability data handbooks (’84, ’92,

’97 and ‘02) that have been sold worldwide (>50 countries).
2.5.6. Environment protection - established a collaboration forum with subsea system

Reliability studies may be used to improve the design and suppliers.

operational regularity of antipollution systems like gas/water - developed an ISO Standard based on the OREDA concept

cleaning systems. Many industries have realized that most of the (ISO 14 224).
pollution from their plants is caused by production irregularities - w 40 papers at various seminars and conferences.

and that consequently the production regularity of the plant is the

most important factor in order to reduce pollution. Reliability and
regularity studies are among the most important tools to optimize
3.2. Offshore WMEP database [6]
production regularity.
The Offshore-WMEP (Scientific Measurement and Evaluation
3. Currently available databases
Programme) database will on the one hand support the wind
energy sector by improving technology and O&M procedures in
There are number of databases available which directly or
order to obtain greater reliability and availability. On the other
indirectly deal with reliability and maintainability concepts
hand, WMEP will monitor developments in terms of techniques,
onshore and offshore with different types of technical systems.
electricity yields and cost.
OREDA is one of them which deals with the RAMS aspects of
Due to its size, the generation of a common database will enable
offshore oil and gas facilities and WMEP is in its beginning stage for
statistically reliable predictions concerning the success of
collecting data from OWT. One of the objectives is to address the
O&M issues. These two databases are briefly described now to
provide know-how about how they work. It is expected that the
proposed RAMS database will include the benefits of these two
databases. OREDA will provide knowledge about the implementa-
tion of RAMS concepts and WMEP will give understanding about
OWT’s operation and behavior in an offshore environment.

3.1. OREDA database [5]

Being a RAMS database, the introduction and familiarization

with OREDA is of paramount importance. In this way one could
easily comprehend how the structure of this database works, how
the data are collected, what kind of information is extracted from it.
Moreover it is important to know how companies use this database
to plan their O&M strategies and lot of other advantages involved
with using such a kind of database.
OREDA has provided a sound basis for boosting reliability
technology and collecting and exchanging reliability data through
the establishment of common standards and references for Fig. 4. OREDA database structure-topside [5].
2158 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165

Fig. 5. Structure of WMEP database [6].

operational concepts. Furthermore, based on anonymous bench- 4. Data collection for OWT
marking and weakness analyses, operators and manufacturers have
the opportunity to test and, if necessary, optimize the performance The format and classification of any database is of paramount
of their offshore wind farms. importance in the data collection activities. Due to less available
The Offshore-WMEP will gather and evaluate data and provide experience with offshore wind turbines, it is necessary to get on-
collaborating operators, manufacturers, scientists, and other hand information about the existing models being developed to
involved parties with the opportunity to gain basic insights into ensure reliability. It must be identified what kind of input data and
offshore wind energy in terms of reliability and availability and to in which format and what types of results are produced. In addition
address larger political questions. what is the effectiveness of these results from the application point
In order to get a large statistical basis for evaluations and of view. For instance, one model is developed in [8] where they
therefore results with strong validity, it is planned to include as have modeled the failures by using Monte Carlo simulation and its
many German wind farms as possible and even wind farms illustration is shown in Fig. 6. A very detailed modeling of avail-
further abroad. The structure of the WMEP database is shown in ability with reliability block diagrams must be conducted. Based on
Fig. 5. such modeling, the effects of different maintenance strategies such
In order to get a large statistical basis for evaluations and there- as the Failure-Based, Time-Based and Condition-Based on the wind
fore results with strong validity, a common datapool provided by the farm model can be assessed to determine the most cost-effective
participant of the program is planned. Manufacturers and operators strategy by taking into account the costs and availability of main-
will get scientifically substantiated analyses of their offshore wind tenance crew and stocks of spares.
turbines and offshore wind farm performance in return. In order to obtain the proper data structure for an input to the
The authors consider that the WMEP database is not clear reliability models, outline of the format and classification for
regarding how it will address the RAMS requirements for the collected data have been identified by [7], these are briefly
purpose of analysis. The block containing “Detail findings, Opera- described here.
tion &Maintenance” has been highlighted in Fig. 5. It is not The “driving force” to establish a Failure and Maintenance
mentioned what could be included in the detail findings and how database should be collected by the maintenance organizations.
this database will be helpful in formulating the O& M strategies Also the companies involved during preparation, installation and
because the measurements regarding factors such as failure rate erection activities of the wind farms and the cable companies can
and failure modes are not there. also provide relevant data. The format roughly consists of three

Fig. 6. Modeling wind turbine failures [8].

 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165 2159

main parts. Parts 1 and 2 will be available for all clients. Part 3 will 5.1. Specific Challenges
be different for each individual client only.
5.1.1. RAMS database as a basis for self maintenance machines
The concept of Self Maintenance Machines (SMM) is here and
4.1. Database and Population
the RAMS database is one input for this concept. There is a great
challenge in how to collect and provide the data to have an effective
In this first part, a general description will be given of the turbines
and reliable SMM.
(the “population”) that are in the database. Various cross sections
The Self Maintenance Machine [9] concept has already been
will be made through the data, for instance to rank the types of
applied to other technology fields such as the photocopier industry.
turbines, to show trends in installed power or installation costs. In
The idea of SMM is to design and build a machine using embedded
this part, discussions will also be held about the quality of the data
intelligence together with sensors and actuators that can operate so
and the completeness of the provided data. This is necessary to
that it is not fully functional but at a designated level (e.g. 50%),
interpret the results of the analyses as good as possible.
thereby adding robustness and fault tolerance. The aim is to make
use of the existing functional redundancies to keep the wind turbine
4.2. General Results fully operational or, if necessary operate at a reduced capacity. The
realization of the SMM concept is based on the Function-Behavior-
This part will show the results of the turbine performances in State (FBS) model and Functional Redundancy Designer (FRD).
terms of energy output, availability, maintenance, etc. The turbines Input from RAMS database will play a significant role in deter-
will be categorized per turbine type. mining and designing the FBS and FRD respectively. The FBS model
is part of the Functional Redundancy Designer, which is responsible
for systematically providing the designer with information on
4.3. Client Specific Results
potential existing redundancies of the system. This is intended to
help with the decision making process, with respect to the
This part will show the similar results as mentioned in the
component selection. In this way, the designer knows what func-
second part, however in this case the analyses will only be per-
tional redundancies are included in the system (Refer to Fig. 7).
formed for the population of the individual clients. Clients can
benchmark the performances of their own turbines with the
5.1.2. Efficiency enhancement tool
generic performances in Part II.
It will be expected that after the realization of this proposed
There are number of factors which are necessary to know about
RAMS database, the efficiency of the wind farms might be
OWT in this context.
enhanced where the production will be higher, resource utilization
will be maximum, and economics will improve. In [10], a model has
i) Region
been developed to achieve these objectives and it is based on the
ii) Wind farm
reliability data, maintenance data, scheduled activities, weather
iii) Wind turbine (manufacturer, type, characteristics)
impacts. A schematic presentation of the model is shown in Fig. 8.
iv) Main systems
The challenge is how effective the RAMS database will be to be
v) Components
used in such kinds of models to enhance the production and reduce
vi) Sub-components (The lowest level of detail to be considered
the downtimes of the offshore wind turbines.
has to be decided, e.g. maintainable items)
vii) Date in use
5.1.3. Failure rate estimation
viii) Wind climate
The kind of method used to calculate the failure rate will
ix) Terrain characteristics
significantly alter the accuracy and reliability of the collected data.
x) Marine characteristics
If the proper technique is not employed for the estimation of failure
xi) Investment costs
rate then the collected data might become pessimistic or over
xii) Initial costs for service contracts, insurance, warranties, etc.
optimistic and this is not good from the application point of view.
xiii) Maintenance history
An example is MIL-HDBK 217F reliability data which is too pessi-
mistic for commercial applications [4].
5. Challenges in reliability and maintainability data
collection 5.1.4. Prediction of reliability
There are number of ways to predict the reliability of engi-
The major challenges which are being faced today are related to neering systems like rotary equipment where the pattern recog-
O&M for offshore wind farms. These have a significant influence on nition tools, machine vision tools like support vector machines are
the formulation of the RAMS database. So at the beginning, one extensively used for the prediction and identification of incipient
must be familiar with the challenges and issues pertaining to O&M
before starting to collect the data.
The main challenges regarding O&M of deep sea offshore wind
turbines are related to accessibility, corrosion and costs. Moreover
an important aspect of future wind turbine development is today’s
easy solution of adapting existing onshore designs to cope with
harsh offshore marine environments.
Challenges related to RAMS data collection are divided into two
main categories namely specific and general. The former are related
to factors such as technological development and improvement,
introducing novel concepts, qualification of new technologies, and
optimizing O&M strategies. The latter is linked with issues like the
management of the database and enhancement of data quality. Fig. 7. Desired knowledge for building FBS and FRD tools [9].
2160 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165

Fig. 8. DNV model for risk analysis [10].

faults. Similar tools could be used for the prediction of the reli- exposed to the exactly the same environmental and operational
ability of offshore wind turbines. Another challenge how this conditions. Therefore the samples should be carefully checked that
database will cope with this aspect and how much it will be helpful they are homogeneous before merging them together. In many
in the formulation of the reliability prediction procedures for databases, the samples are merged without any checking [4].
different systems of wind turbines like the gearbox and generator. Thus the homogeneity of the collected data must be ensured for
an accurate description of the failure rates and other important
5.1.5. Observation window reliability parameters and estimates. Also either existing methods
System failure data are usually collected from a rather limited or new ways should be taken into account to handle the inhomo-
time period that may be called the observation window. If the (real) geneous data and proper checks should be established to identify
life distribution is Weibull distribution, with an increasing failure when such kinds of data are entering the system and what kind of
rate function and a constant failure rate estimate is used, the failure treatment has been carried out.
rate is overestimated in the early phase of the component’s life and It is important to know how these samples will address the
the same is underestimated in the last part of its life. The results implementation of Cox model where the relationship between the
will be even more incorrect if the extrapolation is based on the survival of a component and several explanatory variables will be
estimated constant failure rate beyond the time interval from established. Interpretation of the Cox model involves examining
where the data was collected [4]. The accuracy of data is heavily the coefficients for each explanatory variable. A positive regression
dependent on the selection of a suitable observation window which coefficient for an explanatory variable means that the hazard is
poses another challenge in this exercise. higher and conversely, a negative regression coefficient implies
a less hazard state [22].
5.1.6. Renewal process
People who analyze life data are not always aware of the 5.1.8. Technical Innovation
difference between the concepts of failure rate function (FOM) and The offshore wind turbine area is in its nascent state and lot of
rate of occurrence of failure (ROCOF). Assume there is a system with new concepts and ideas are emerging to cope with the marine
an increasing ROCOF w(t). If the failure data is collected in an environment and it is necessary to have the robust design due to
observation window in an early phase of the system’s life, the challenges involved in the accessibility and transportation. It is an
resulting “average failure rate” is often different from what it will be uphill task for the developers of this RAMS database how to
obtained in a later observation window. This effect has been seen in incorporate adequate flexibility and adaptability to provide accu-
several offshore data collection projects, for example, for downhole rate statistics for such type of wind turbines. A few of the novel
safety valves. When a valve has failed and it has been replaced with concepts are shown in Fig. 9.
a new valve of the same type, and it has been (erroneously) believed
that a renewal has been done. The environmental conditions in the 5.1.9. Qualification of new technologies
well have, however, changed with time and produced a more Based on the emerging new concepts and ideas, new technol-
hostile environment [4]. ogies are under development and more are expected to be flooded
Incorrect thinking regarding the incorporation of the renewal in the market with the growth of offshore wind industry. For the
process in the system could severely harm the system and could qualification of new technologies, RAMS data will play a pivotal role
end in the total breakdown of the whole set up. This aspect is in declaring its reliability level.
serious and it must be addressed meticulously. DNV-RP-A203 qualification procedures for new technology
are employed for declaring the fitness of new technology [11].
5.1.7. Homogeneous samples The other existing approach is the TRL (Technical Readiness
The individual components that are classified within the same Level) where the assessment of new technology is based on 9
generic item do not need to be identical and do not need to be levels [12].
 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165 2161

Fig. 9. Novel concepts in offshore wind design.

5.1.10. Suitability for maintenance optimization database (related with O&M part) and then the pros and cons of
The quality of data should be reasonable so that it is possible to both the databases could be comprehended and further improve-
get effective and efficient results from maintenance optimization ments could be suggested in a better way. Such of kind of compa-
because such kind of optimization is the important underpinning in rability was conducted in [14] where the accuracy of four different
ensuring the reliability of the whole engineering system. In [13], types of failure data was judged.
some of issues have been highlighted to ensure reliable maintenance In Fig. 10, the failure frequency information from four different
optimization. Maintenance actions will only be effective and effi- databases has been summarized and compared. Because of the
cient if they specifically address the most relevant deterioration and different classifications used in the databases this can only be done
failure mechanisms. Global statistics, like an average failure rate, will over a limited number of component groups. Despite that a fairly
therefore be inappropriate for most maintenance actions, and apply good agreement can be seen in the failure levels. The most striking
only to complete replacements, which tend to be costly. Analyzing difference can be seen in the data from WindStats. There is a very
data without knowing the underlying failure mechanisms can lead significant difference between the registered failure frequency of
to totally wrong results (consider failures caused by wearout and the Danish and German wind turbines. But when the data is clus-
operator error). Hence data have to be collected under strict rules, tered on a more global level, one may conclude that the Danish
defining failure and individual maintenance actions precisely, and results in WindStats seem to underestimate the actual number of
using a well-defined system component structuring and indicating failures significantly.
what events have happened e.g. the indentified events are helpful in
defining the direction for maintenance optimization in the form of 5.1.12. Data availability
giving cost figures for preventive, corrective, and condition-based The quality and availability of the data must be ensured by
maintenance, set up costs because such figures are necessary for the setting continuous data collection and consequently regular avail-
grouping of different activities together. ability of that data to the database. The data must be sufficient to
One of the main objectives of any RAMS database is to find ways plan viable and effective maintenance strategies.
to determine the periods of preventive maintenance. The quality
and quantity of data should be sufficient to define the accurate 5.1.13. Data quality
times of preventive maintenance. It has been noted that in some It is necessary to identify the interaction of different compo-
cases preventive maintenance has an adverse effect on the reli- nents with each other which can contribute a lot in determining the
ability. To address such issues, the quality of data should be good failure rates and failure modes. For example, it must be clearly
enough to tackle this problem. defined and be incorporated in the data collection that the speed of
The main objective of this step is to determine the type and the wind and the operation conditions of the rotor are significant in
frequencies of maintenance and inspection tasks. In principle each the determination of failure rates of the wind turbine. Correct
failure mode/failure cause in the FMEA (Failure Mode and Effect
Analysis) should be combated by a maintenance task. The RCM
(Reliability Centered Maintenance) logic of an RCM analysis will be
a starting point for identifying relevant maintenance tasks. To deter-
mine optimal frequencies of maintenance tasks it is usually required
to establish a cost model to optimize. Life cycle costing will be a central
part of such model. The use of so-called influence diagrams will very
often help the communication between the personnel such as
analysts and maintenance engineers and economists.
When the maintenance tasks are identified, and frequencies set
it will usually be natural to group these activities into maintenance
packages, each package describing what to do, and when to do it. It
is a challenge to establish such an optimal grouping strategy.

5.1.11. Comparability of data

To check the accuracy, it is recommended that the collected data
should be compared with other existing or similar databases of the
same type. By doing this way, differences and similarities in the Fig. 10. Monitored yearly failure frequencies from different public databases. LWK:
data could be established in a coherent way. For example the pro- data from Schleswig-Holstein [15], MWEP: data from the German monitoring program
posed RAMS database output could be compared with WMEP [16] and data from the WindStats journal [17].
2162 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165

determination of failure modes and failure rates underpin the data software will be evident after using it for some time. To minimize
quality enhancement. the effect of such kinds of errors, utmost care must be taken while
developing the mainframes of such software. Moreover, it is rec-
5.1.14. Specifications of complex systems ommended that the lessons should be learnt from the experience
In the case of offshore wind farms, many concepts are in their with other data collection regimes like OREDA for having accurate
nascent stage for simplification of design. But still no concrete results and reliable data collection software.
are available at hand. So keeping in mind these aspects, the speci-
fication of complex systems of wind turbines like rotor, gearbox and 5.2.8. Communication
generator should be described in a simple but realistic manner. Effective, efficient, and reliable tools of communication among
the different data suppliers and persons engaged in such activities
5.2. General Challenges is important. Proper forums should be established where the staff
could share their experience and ideas for improvement of the
5.2.1. Cost effectiveness current data collection regime.
The data collection regime should be cost-effective in the sense
that small resources of the data providers should be engaged in this 5.2.9. Built-in consistency checks in data collection
exercise. An attempt could be incorporated for automated collec- It may be possible that the collected data could contain wrong or
tion of data. In addition, it should be guaranteed and assured to the incomplete information. The inclusion of such kind of information
supplier of that data that this analysis will be of benefit to their in the database will directly affect the accuracy of the database and
company. consequently the analysis and planning done based on that data-
base. So to avoid such kinds of malpractice, there must be provi-
5.2.2. Avoid dubious and incomplete data sions in the data collection mechanism to address them.
Identification of what kinds of parameters are required should
be clearly marked to avoid any inconsistency in the data. By doing it 5.2.10. Spot checks on collected data
this way, the interpretation of data will be easy and there will be To evaluate the accuracy of collected data, spot checks from time
less chance of incorrect or irregular use of that data. to time are imperative. This can be accomplished by conducting the
analysis of the collected data and then compare the results with
5.2.3. Competence of data collector those of the expected results. By doing this it can be determined
The selection of suitable personnel for the data collection is of whether the data collection regime is functioning in a proper way
great importance especially in the case of offshore wind turbines and if it is not, what kinds of corrective measures are necessary. Such
where accessibility and availability are of serious concerns kind of spot checks will open new vistas for brainstorming sessions
compared to onshore wind farms and even the offshore oil and gas with the different experts who will be engaged in this project.
industry. So in this scenario, good understanding of this system is
vital for the data providers. They must be fully aware about the 5.2.11. Clear definition of data boundaries
consequences of any error or misjudgment in the data collection Before starting the data collection, it must be clearly identified
and then its transmission to the database. which kind of information is required and what type of analysis will
be done to ensure the reliability of the operating system. The
5.2.4. Typographic errors definition of different data boundaries could be based systemwise
Due to the intricacies of offshore wind turbines, the typographic on the main elements like the rotor, gearbox, generator, control,
errors should be avoided as much as possible. One way to reduce tower, and foundations. Other ways of defining the data boundaries
these irregularities is to use standard templates for data collection could be based on the number of failure rates, failure modes and
and determining the values of upper and lower limits of each failure their effects, state of the component and system based on the
rationally. The advantage of this measure will be that if someone failure rate. Clear definitions of data boundaries are directly helpful
tries to enter higher values in the standard database template or in enhancing the accuracy of the collected data.
otherwise, there must be an alert message to them asking for the
justification of that value. In this way, personnel will get a chance to 5.2.12. Automated data transfer
rethink and correct the value before entering it in the database. To maintain an efficient database, there must be provision for an
automated data transfer from different data suppliers to the main
5.2.5. Missing compulsory and non-compulsory information database. While realizing it, there will be fewer loads on the data
Clear boundaries must be defined regarding the classification of suppliers and in the latter case it may be possible that more suppliers
what information is compulsory or not. Such kind of demarcation will be interested to share and enter their data into this main database.
must be introduced in the data collection template. Moreover,
provision should be there to shift the parameters from one 5.2.13. Feedback to operating department
boundary to another keeping in view the knowledge of the system Proper mechanisms should be designed where necessary feedback
and accuracy of the database. must be shared with the operators. Due to this feedback, the operators
will be motivated to supply more high quality data after knowing the
5.2.6. Questionable information effectiveness of their previous contributions in the database.
There is a chance that sometimes questionable information will
be present in the database. One reason for such kind of information 5.2.14. Combination of coded data and free text
may be the presence of outliers in the database and there must be To provide the flexibility in sharing the data, it should be
proper checks in the data collection system to address such kind of possible to share the data in a combination of coded and free text.
unjustified information.
5.2.15. Harmonizing the collected data
5.2.7. Software errors There will be different suppliers of data so it must be ensured
The data collection will be carried out by dedicated software that the collected data are harmonized in a viable statistical way to
accessed by different data suppliers. It is likely that errors in the give reliable results.
 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165 2163

5.2.16. Taxonomy of the collected data - Maintenance and reliability aspects: The emphasis is put
The description of the collected data should be based on system mainly on obtaining insight into the failure behavior, down
level, equipment level, component level, etc,, to have a proper time (availability) and the maintenance aspects of the
taxonomy of the data. Using such taxonomy makes it easy to different wind turbine types and its components. A certain
perform an analysis on any selected level for planning O&M strat- level of detail for data collection is therefore required.
egies and carrying out other tasks based on that analysis.

5.2.17. Anonymity of the data suppliers 6.2. Equipment data

Means must be incorporated during the data collection so that
no supplier is traceable, thus keeping the anonymity of each and The description of an equipment unit is characterized by the
every supplier. The main reason behind this activity is that manu- following [18]:
facturers do not like to share information about their system with
other groups. So there must be checks in the data collection system - Classification data, e.g. industry, plant, location, system.
to ensure that no supplier of the data can be identified. - Equipment attributes e.g. manufacturer’s data, design
characteristics.
5.2.18. Aggregation of data - Operation data, e.g. operating mode, operating power,
Aggregation of data for each reporting period for the entire environment.
population of wind turbine must be avoided. The main disadvantage
of such type of data aggregation is the inability to make a comparison These data categories are to be general for all equipment classes.
between different wind turbine configurations or concepts. It is Additionally, some data are required that are specific for each
preferable that data should be collected in a segregated fashion. equipment class (e.g. number of stages for a compressor).
The classification of equipment into technical, operational and
6. Architecture of RAMS database for OWT environmental parameters is the basis for the collection of RM data. This
information is also necessary to determine whether the data are suitable
There is an exclusive RAMS database available for the offshore or valid for various applications. Some data are common to all equip-
oil and gas industry under the umbrella of OREDA. But no such type ment classes and other data are specific to a particular equipment class.
of dedicated database is at hand for offshore wind turbines. Some
databases are available like WMEP where part of the database is 6.3. Failure data
reserved for planning O&M strategies but details are not available
about how such data are collected and how RAMS concepts will be These data are characterized by the following [18]:
implemented based on it. Moreover the details are not yet known
about who will contribute in this database. So bearing in mind the - Identification data, e.g. failure record number and related
ongoing rapid developments in offshore wind turbines, it has equipment that has failed.
become imperative to have an exclusive database like OREDA for - Failure data for characterizing a failure, e.g. failure date, items
offshore wind turbines. failed, failure impact, failure mode, failure cause, failure
In order to define such a database, it is vital to learn from the detection method.
experience of the offshore oil and gas industry and those compiling
onshore data collections for wind turbines. Based on the short- A uniform definition of failure and a method for classifying
comings and weaknesses of previously available databases for failures are essential when it is necessary to combine data from
RAMS purposes, an accurate and reliable database could be estab- different sources (plants and operators) in a common RM database.
lished which would reduce the weaknesses and difficulties faced in The identification of what type of data that is required has
the previous ones. One way of establishing an effective database is a significant effect on the reliability of the analysis. The data will
that the challenges highlighted in the last section should be accurately describe the failure rate, failure mode and help in
addressed in a realistic and viable way. defining the root cause of failure. This will give inputs for making
The RAMS data are to be collected in an organized and struc- a predictive tool for describing the state of the operation and is one
tured way. The major data categories for equipment, failure, of the critical factors which can contribute a lot to the reliability of
maintenance and state information data are given below. Note that data analysis and set future trends based on that analysis.
the OREDA concept does not include state information data. In
Fig. 11 the inclusion of state information is explicitly demonstrated
6.4. Maintenance data
in addition to operational data.
In the ongoing activity, the areas of focus are highlighted as
These data are characterized by the following [18]:
rectangles in Fig. 11. At this level, the decision of what kind of data
should be included will be carried out in detail. The applications of
- Identification data, e.g. maintenance record number, related
the proposed RAMS database is chalked out at this moment and
failure and/or equipment record.
then later on it will be decided about the data analysis method to be
- Maintenance data, parameters characterizing a maintenance
used for these purposes.
action, e.g. date of maintenance, maintenance category,
maintenance activity, impact of maintenance, items maintained.
6.1. Operational data - Maintenance resources, maintenance manhours per discipline

and total, utility equipment/resources applied.

Operational data might consist of the following [7]: - Maintenance times, active maintenance time, down time.

- Energy production information: Besides the amounts of The type of failure and maintenance data is normally to be
wind energy produced per year or per month, the wind speed common for all equipment classes, with exceptions where it is
distribution and wind rose can be provided when available. necessary to collect specific types of data e.g. subsea equipment.
2164 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165

Fig. 11. Schematic presentation of the proposed RAMS database.

Corrective maintenance events are to be recorded in order to database for offshore wind turbines. Although the environment is
describe the corrective actions following failure. Preventive main- offshore, there are some similarities for logistics, transportation, and
tenance records are required to retain the complete lifetime history accessibility but the operational conditions might be different for
of equipment. offshore wind turbines compared to the equipment and worksta-
tions established for exploiting and exploring oil and gas reserves
6.5. State information offshore. Also the economic profit gained from OWTs is on a different
scale than oil and gas installations, leading to more focus on costs.
State information (condition monitoring information) may be Offshore oil and gas platforms are standalone and the operation of
collected in the following manner: one platform is not likely to have a considerable effect on the neigh-
boring one. But in the offshore wind turbine case, a number of wind
- Readings and measurements during maintenance. turbines are installed in clusters where the number of wind turbines
- Observations during normal operation. and their layout are optimized. This generates wake phenomenon
- Continuous measurements by use of sensor technology. which is of a critical nature. Brainstorming is required to extract
information about how this wake effect could contribute to the failure
After the inclusion of state information, the structure of RAMS rates of other wind turbines in the same layout or in other layouts in
database (based on OREDA as shown in Fig. 4) will look like that proximity. A visual example of wake phenomenon is shown in Fig. 13.
shown in Fig. 12. It is well-known that there are plenty of differences between the
The final goal of any data collection regime is to depict the two systems in contrast to having only some similarity. But still it is
realistic state of the system under measurement. This objective recommended that the experience from OREDA and other existing
could be realized when proper parameters for measuring the data onshore and offshore wind turbines databases should be fully
are selected. The selected parameters should describe the realistic exploited before the initiating any activity for RAMS data for
or close to realistic state of the operation. In case any data do not offshore wind farms.
fulfill the required objective, then corrective measurements must Moreover, it is equally important to learn from the experience
be considered to enhance the effectiveness of the collected data. and expertise from the organizations which were engaged in the
The OREDA database is an established set-up for the offshore oil data collection and analysis of onshore wind turbines e.g. condition
and gas industry. It is important to learn lessons from the experience monitoring techniques, fault detection and isolation. The WMEP
of OREDA but at the same time it is imperative to understand the database is aimed at producing the data which will address the
possible differences between the OREDA and the proposed RAMS O&M issues of OWT. So it is important to learn how this database
 Z. Hameed et al. / Renewable Energy 36 (2011) 2154e2165 2165

used in the whole system. For an offshore wind farm however the
accessibility of the site for O&M hardware equipment as well as the
adopted maintenance strategy are of equal importance to the ac-
hieved availability level [19]. Thus planning O&M strategies
including maintenance optimization, LCC, and LCP studies could be
based on the output obtained from this database.

7. Conclusions

The need, method, benefits, and possible areas of application for

the proposed RAMS database have been identified. Both the tech-
nical and managerial challenges were outlined which could be
encountered during the realization of this database.
The structure for the database is suggested keeping in view the
implementation of RAMS concepts in an easy and efficient way. The
architecture of the data structure could be improved while adding
Fig. 12. Modified RAMS database structure after the inclusion of state information as new information or deleting existing information.
compared to OREDA The different components contained in the database have been
carefully discussed. It has been found that the application of this
will address issues such as failure rates and failure modes for
database is not limited to only one given structure rather there
formulating the O&M strategies.
could be many more.
A number of areas are mentioned where the output from this
It is expected that after the realization of the proposed RAMS
RAMS database could be employed and it is not limited only to
database, there will be an improvement in planning O&M strate-
these application arenas as shown by the arrow in Fig. 11. The
gies, implementing novel concepts like SMM, qualification for new
designers and manufacturers will be able to see the behavior of
technologies to be used in the OWT area.
OWT in a real environment and then based on this database, they
The availability of offshore wind turbines might be established
will be able to evaluate the strengths and weaknesses of their
with reasonable accuracy when all the strategies related to
concepts and ideas which are being employed during the design
logistics, transportation, and accessibility are based on this
and manufacturing phase. Based on this database, FBS and FRD
database.
concepts could be realized and implemented to incorporate the
redundancy in the critical components with more accuracy.
Due to factors such as the marine environment, accessibility, References
transportation and lifting mechanisms there are new ideas for OWT
like floating, bottom fixed, tubular structure and gearless and all [1] Twidell, J., Gaudiosi, G., Offshore Wind Power, ISBN 978-0906522-639.
[2] http. //, www.ntnu.no/ross/info/data.php.
these involve to some extent the introduction and realization of [3] http. //, www.we-at-sea.org/index.php?keuze¼r162.
new technical concepts. This process could be completed by per- [4] Marvin Rausand, Arnljot Høyland, System Reliability Theory: Models, Statis-
forming the techniques and methodologies for the qualification of tical Methods, and Applications, 2nd Edition ISBN: 978-0-471-47133-2.
[5] http://www.oreda.com/.
new technologies. The concept of qualification of new technologies
[6] http://offshorewmep.iset.uni-kassel.de/.
is based on RAMS data .So from this database, new technical [7] Curvers, A.P.W.M., Rademakers, L.W.M.M., Standarisation of collecting failures
concepts will be evaluated and then declared qualified or not based and maintenance data, Report ECN-Ce04e108, November 2004.
[8] Jesse A, Andrawus, Watson, John, Kishk, Mohammed. Wind Turbine Mainte-
on the input obtained.
nance Optimisation: principles of quantitative maintenance optimisation.
The availability of a wind farm, defined as the percentage of time Wind Engineering; 2007.
it is able to produce electricity, is a function of the reliability, [9] Echavarria E, Tomiyama T, van Bussel GJW. Proc. European Wind Energy Conf,
maintainability and serviceability of the hardware and software The concept of self-maintained offshore wind turbines. Milan: Italy, May 7th-
10th; 2007b.
[10] Le Gal, Pascal: Asset Management Software Tools - Safeti Performance Fore-
casting 2010.
[11] http://exchange.dnv.com:6389/dynaweb/offshore/rp-a203/@ebt-link;-
cs¼default;ts¼default;pt¼167?target¼%25N%14_2457_START_RESTART_N%25.
[12] Mankins John C. Technology readiness assessments: A retrospective,. Acta
Astronautica 2009;65:1216e23.
[13] Rmmert Dekker, Applications of Maintenance optimization models: a review
and analysis, Journal of Reliability and System Safety 51, (1996) 229e240.
[14] G.J.W van Bussel, ’Offshore wind energy, the reliability dilemma’; Proceedings
of the International Conference, 2002; Berlin.
[15] W. Eggersgl;üb, Wind Energie IX-XI, Praxis- Ergebnisse 1995-2000, Land-
wirtschaftskammer Schleswig-Holstein, Osterrönfeld, Germany.
[16] Jahresauswertung MWEP. ISET. Kassel: Germany, ISBN 3-9805896-6-8/5-X/n-
n; 1996-2000.
[17] WindStats Newsletters 1998 -2001, Vorlaget Vistoft, Denmark, ISSN 0903-5648.
[18] International Organization for Standardization. Petroleum and Natural Gas
Industrious e Collection and Exchange of Reliability and Maintenance Data for
Equipment, ISO 14424. Second Edition; December 15, 2006.
[19] G.J. W. van Bussel, A. R. H., C.A. Morgan, K. Argyriadis, (2001). State of the Art
and Technology Trends for Offshore Wind Energy: Operation and Mainte-
nance Issues. Offshore Wind Energy Special Topic Conference. Brussels.
[20] http://www.emerging-energy.com.
[21] http://www.popsci.com/technology/article/2010-01/wind-turbines-leave-
clouds-and-energy-inefficiency-their-wake.
[22] http://www.medicine.ox.ac.uk/bandolier/painres/download/whatis/COX_
Fig. 13. Clouds form in the wake of the front row of wind turbines at the Horns Rev MODEL.pdf.
offshore wind farm near Denmark [21]. [23] http://lwksh.de/cms/index.php?id¼2875.