1
Developing a Risk Analysis and Decision Making
Strategy for an Offshore Wind Farm
Iraklis Lazakis
1
, Maria A Kougioumtzoglou
2
Abstract
The renewables sector and particularly offshore wind energy is a fast developing industry over the last few years.
Especially activities related to the Installation, Operation and Maintenance (O&M) of offshore wind turbines becomes a
challenging task with inherent risks. This paper assesses the risks related to the above stages of a wind farm lifecycle
using the FMECA (Failure Mode, Effects and Criticality Analysis) and HAZID (Hazard Identification) methods. All
works, from installation to O&M are considered together with the wind turbine main components. An integrated risk
analysis methodology is presented addressing personnel Safety (S), Environmental impact (E), Asset integrity (A) and
Operation (O). The above is supplemented by a cost analysis with the aid of BBN (Bayesian Belief Networks) method in
order to assist the decision making process related to installation and O&M tasks. All major risks and critical wind
turbine components are identified as well as measures are suggested in order to prevent or mitigate them. Moreover,
inspection and maintenance plans are elaborated in general for the mentioned activities.
Keywords
Risk analysis (O&M); Offshore wind farm; FMECA; HAZID; BBN
Introduction
Wind power is known to humans since ancient times. It is a form of energy that not only has no time or place restrictions
but it also contributes in reducing greenhouse gases emission and boosting the economy of countries that depend on oil
and gas imports for the energy coverage [1]. These characteristics makes it appealing to industry that tries to exploit it by
developing more and more onshore or offshore wind farms [2].
The rapidly expanding number of wind farms makes quantifying and managing the different elements of risk that are
present in each of the installation, operation and maintenance stages of a wind turbine necessary. In this respect, risk
analysis and decision making can be a key that will enable fast growth, investments, further technological development
and reasonable cost of energy.
This paper presents the study regarding the investigation and assessment of the risk and reliability features of offshore
wind turbines at different stages of its lifetime and identification of the critical components in terms of their operation in
order to increase their availability and operability characteristics. A lot of risk analysis methods formerly or currently
used in the offshore renewables and oil and gas sectors is examined as shown in section 2. The description of wind
turbine and the demobilization of its components is demonstrated comprehensively in section 3, as well as the overall
risk analysis methodology, including the HAZID and FMECA approaches, which are complemented with risk matrices
for various consequence categories. Also, the cost benefit analysis with BBNs is presented in the same section. In Section
4 the outcomes of the analyses and the simulations are submitted highlighting the possibe high-risk areas and the most
costly components. Finally, conclusions and recommendations for future research on the current study are shown in
section 5.
1
Iraklis Lazakis, Department of Naval Architecture, Ocean and Marine Engineering, University of Strathclyde, 100
Montrose Street, Glasgow G4 0LZ, Scotland
2
Maria A. Kougioumtzoglou, Hellenic Tankers, Kifisias 349, Kifisia, Athens, Greece, 14561
Corresponding author:
Maria A. Kougioumtzoglou, Hellenic Tankers, Kifisias 349, Kifisia, Athens, Greece, 14561,
Email:
marakikgl@gmail.com
2
Literature review
Risk is defined in a different way by each one of us so there is not a universal definition of “risk”. Generally it includes a
combination of probabilities of occurrence and consequences of an unwanted outcome [3, 4, 5, 6, 7, 8]. Consequences
can be loss of life, injuries, environmental, social and economic impacts [8, 9, 10, 11, 12]. Other views focus on both
positive and negative aspects of risk[7, 8,13,14] and argue that one should not eliminate the other.
Hazard, is associated with risk but they are not the same. According to WHO [5] hazard is the “Inherent property of an
agent or situation having the potential to cause adverse effects when an organism, system, or (sub) population is exposed
to that agent”. Similarly, [8, 11, 12, 15, 16, 17] describe hazard as situation likely to cause harm, injuries and damage. So
while “hazard is any source of potential damage, harm or adverse health effects on something or someone under certain
conditions”, “risk is the chance or probability that something or someone will experience an adverse effect if exposed to
a hazard” [18].
A probability is the way we have to express quantitatively the likelihood of an event or consequence to happen.
According to past papers [6 , 7, 8, 11] probability can be either a subjective measure of uncertainty if it comes from
expert’s judgment, or a classical statistical approach. Determination of a probability and decision making in each case
involves a certain degree of uncertainty that derives from our lack of important information [8, 10, 18, 19]. Decision
makers face uncertainty when either the probabilities or the consequences are unknown or there are multiple outcomes
for each alternative and that is the difference with risk, since risk exists when we know all the consequences but not
which will definitely occur [8, 10, 19, 20, 21, 22].
The subject of risk analysis, risk assessment and risk management in general is a relatively new but extensively explored
area with various studies contributing to its thorough examination. Effective risk mitigation is desirable by all individuals
and companies, and risk management is or should be applied to all stages of a project lifetime [67]. Especially in the
maritime and offshore industry the aim is to reduce the risks from major hazards that could jeopardize the integrity of the
offshore structure and the health and safety of the workforce and ensure the protection of the environment [23, 68]. The
correct identification of the hazards and their consequences is a key issue in providing information to aid decision
making and increase the level of a project success. Thus there are many tools, processes, techniques and methodologies
developed nowadays to cover this need.
Some of the main standards for risk management are those: Australian Standards/New Zealand Standards: 4360 2004,
Association for Project Management [16], Project Risk Analysis & Management (PRAM) Guide 2nd edition [57],
Project Management Institute [61], Guide to the Project Management Body of Knowledge (PMBoK): Chapter 11 [61],
ISO/ IEC 31010:2009 [63] Standards and many more.
More specifically for offshore oil and gas industry, HSE introduced the Safety Case approach in 1992 [24], in which
guidelines are given to operators of each offshore installation field for “reducing the risks from major accident hazards to
the health and safety of the workforce employed on offshore installations or in connected activities”. After that, many
standards and codes have been established the last years as guidelines for this purpose. Although they refer mainly to oil
industry they can have a good application on wind industry. The most important of them are from ISO [15], HSE [24],
DNV [25, 26], IMO [27], ABS [28], OREDA [29] and Norsok [30].
Apart from standards there are various software tools for risk analysis valuable for the industry based on a quantitative
approach of risk assessment. These are RBM (Risk Based Management) II released from the Dutch Government, PHAST
and Synergi Life Risk Management from DNV GL, SHEPHERD a software property of Shell Global Solutions,
RISKCURVES that is an integrated QRA software from TNO, EFFECTS that is a consequence analysis and damage
calculation software from TNO, HAMSAGARS which is a QRA software from HAMS-GPS, RISKAN and many more
[62].
The most known techniques of hazard identification are Expert Judgment, Check Lists and the structured techniques
HAZID (Hazard Identification), PHA (Process Hazard Analysis), What-IF Method, FTA (Fault Tree Analysis), ETA
(Event Tree Analysis), FMEA/FMECA (Failure Mode, Effects and Criticality Analysis), HAZOP (Hazard and
Operability), Monte Carlo Simulation and Risk Ranking Matrix [3
1]. All of them can be applied in our area of interest;
3
offshore installations and more specifically offshore wind farms, with FMECA and HAZID, the two methods that are
used in this study, being two of the most popular.
After all the necessary information about possible risks is gathered, risk evaluation is executed. The most well-known
method of risk evaluation is ALARP (As Low As Reasonably Practical). The idea in this method is that the risk should
be minimized to a point where it is acceptable but without expending grossly disproportionate cost, time and effort [32].
Regarding decision making, in the frame of risk mitigation and ALARP, one of the strongest tools that decision makers
have to deal with the problems raised is Bayesian Belief Networks (BBNs) a tool for modeling under uncertainty by
using conditional probabilistic calculations and graphical representation of the logical relationships between variables.
Risk management process in general, includes setting up the context, assessing the risk (hazard identification, risk
analysis, and risk weighing), handling the risks, monitoring, communication and consultation, as well as the connection
between these procedures. The efficiency of Risk management depends on the selection of the risk method. The
suitability of each method depends on its strengths and weaknesses and on the needs of the project. Usually, two or more
methods are combined in order to cover all stages or needs of a project’s lifecycle and each other’s weaknesses and
flaws. This procedure has also been followed here. HAZID cannot support much detail and it is usually used for
operational procedures. FMECA’s complexity and the ever-increasing list of possible failure modes of the components
make it difficult to be widely applied. A spherical and general overview though of both mechanical and operational
aspects of an offshore wind turbine can be obtained when combining these two methods. Additionally, a financial
perspective on the cost of critical components and their failure probabilities can be assessed through BBN analysis.
FMEA/FMECA-HAZID
FMEA is one of the first systematic techniques used to identify problem that may originate from system malfunctions.
The concept of FMEA is reviewing all the components of a system and the causes or the ways in which a system can fail
(Failure Mode) and then the consequence of these failures. The consequences can be categorized in terms of safety,
reliability and environmental effect [8, 11, 17, 33, 34]. FMECA (Failure Modes, Effects. and Criticality Analysis) is the
extended model of FMEA, so that criticality is taken into account. It is mainly used to rank the failure mode based on the
severity of their consequences. FMEA/ FMECA are usually qualitative or semi-quantitative, and can be applied at any
phase of project life cycle preferably at the early stages of a project since the designers can have the ability to change the
probabilities of the critical failures [8, 33]. FMECA was originally part of risk management techniques developed for
defense and nuclear industries in the 1940’s. It was formally developed and applied by NASA in the 1960’s to guarantee
reliability of space program hardware and was quickly adopted by aerospace, petroleum, chemical and automotive
industries [35].
FMEA and FMECA typically consist of several stages. Definition of the system components is the first step of the
analysis. Then, identification of each component’s failure mode as well as their effects is the next crucial level to the
FMEA/FMECA approach. Next important step is analyzing the criticality of each failure and also estimating their rate.
Ranking of failure modes and determination of critical items is another important stage of the procedure that due to the
subjectivity of the applier a lot of attention and thorough review of the parameters need to be implemented. Design
process then absorbs the method’s results and helps identifying means of future reviewing and suggesting improvements
in design.
Depending on the analysis we want to conduct a proper FMECA worksheet has to be formed. A representative worksheet
is presented below where the name, function and operational mode of each element is mentioned and also all potential
failure modes for each function and operational mode. Also, the failure mechanisms (corrosion, erosion, fatigue etc.) for
each failure mode have to be listed and their acceptance criteria have to be chosen. Before the final decision of the
criteria categories, a lot of reviews and papers were taken into consideration. Main references were IMO and ISO sources
on the development of risk matrices, FSA applications and nuclear projects, where number of cases is small and
technology may be obsolete but due to catastrophic consequences their outcomes must be taken into serious
consideration. After thorough review we finalized our matrices as below [15, 23, 27, 36]:
The likelihood of their detection was evaluated with a ranking that is usually divided into five categories: 1 (almost
impossible), 2 (low), 3 (moderate), 4 (high), 5 (almost certain). The effects that a failure may have on the subsystem itself
4
or on other components as well as failure rates should also be listed and classified most commonly to a five-level
ranking: 1 (higly unlikely), 2 (remote), 3 (occasional), 4 (probable) and 5 (very frequent). The severity of the failure
modes regarding the global effects has to be evaluated and ranked into five categories that in most cases, for
computational reasons, are represented from numbers 1 to 5 representing: 1 (minor), 2 (marginal), 3 (major), 4 (critical)
and 5 (catastrophic). It can be assumed here that categories are related with 1 fatality to equal 10 major injuries and 100
minor ones. Finally, mitigation measures that could prevent failure should be mentioned.
The risks linked to failure modes is a function of frequency of the failure mode and consequences of the outcomes and
can be presented in the form of a risk ranking matrix to prioritize those that need immediate management.
An alternative to risk ranking matrix is the risk priority number (RPN) which is defined as:
RPN=S∙O∙D (1)
Where
S is the rank of severity of the failure mode taken from the severity matrix
O is the rank of occurrence of the failure mode taken from the occurrence matrix
D is the rank of detection of the failure mode taken from the detection matrix
The RPN is not a measure of risk, but of risk priority. The smaller the RPN is the better since you can deal with this
hazard later. Based on these two tools the responsible team should decide whether the system is acceptable or not and
propose improvements that will reduce the likelihood of occurrence of failure, reduce the consequences of failure or
increase the failure detection probability. After the improvements the FMECA worksheets and RPN have to be revised
and updated [37].
The main drawbacks of FMECA method is the limitation in examination of human and other external factors, as well as
focusing on a single initiating event and on the mode of operation. Furthermore, analysis is mainly based on team
experience on evaluating the failure modes of the components inducting subjectivity in the procedure [38].
HAZID is one of the most common and frequent used techniques for hazard identification being carried out at the first
stages of a project where not much detail is required [8, 11]. In HAZID the process is divided into nodes and with the aid
of pre-defined guidewords for hazard identification, all undesirable consequences associated with the defined node are
identified. Consequences are divided into broad categories such as human impacts, environmental impacts, and economic
impacts that are then divided in subcategories based on the type of consequence. Checklists from previous similar
HAZID can be used to assist the procedures. The same methodology as the FMECA is followed and risk matrices are
constructed as well. Its application is wide: from marine and offshore industries to nuclear sector.
Since HAZID is applied to all aspects and operations and a complete evaluation of all hazards is performed, an extended
list of potential hazards and recommendations for avoidance is produced. To this respect
a well-defined system or
activity is required in order to minimize time needed for the analysis [38,
39].
HAZID’s in depth and time consuming analysis can counterbalance any omissions that can come of analysts’ lack of
experience as in the case of FMECA. Also, since FMECA covers mainly equipment failure modes and effects, HAZID
comes to fill the gap in safety related studies.
BBN
Reasoning with uncertainty is common in all aspects of everyday life, so dealing with it has forced scientists even from
sixteenth century to develop several approaches such as the frequentist or the subjective Bayesian that was widely
adopted in more recent years. Bayesian Belief Networks (also known as Belief Networks, Causal Probabilistic Networks,
Causal Nets, Graphical Probability Networks and Probabilistic Cause-Effect Models) were first developed in 1980s and
they are based in statistics and artificial intelligence and they provide a simple way of building a "picture" of a decision
5
problem. They make a framework for decision support that takes into consideration variables with unknown state and
influence on outcomes. So it is basically a tool for modeling under uncertainty by using conditional probabilistic
calculations and graphical representation of the logical relationships between variables. This method is a more flexible
tool than other methods (FTA, DFTA ETA etc.) as it can combine objective and subjective data like expert judgment and
can resolve some disadvantages that more traditional methods have as intrinsic limitations such as system and component
interconnectivities in multiple layers so that they simulate real state conditions, fault detection and system degradation.
[40, 41, 42].
The approach is based on conceptualizing a model domain of interest as a graph of connected nodes and linkages. In the
graph, nodes represent variables (XX
,…,X
,…,X
) and arcs represent direct connections between them (X
→X
).
Probabilistic relations rather than deterministic are used to describe the dependency relations [42].
The construction of a BBN is simple but as the variables increase in number, the complexity rises. Figure 1 shows the
logical relationship between five hypothetical variables A through E. D is called parent or predecessor of B. Equivalently
B is called a child or descendant of D. The arrow illustrates that the parent node has a direct influence on the child node.
D and E do not have any parents and are called root nodes. A does not have any descendants and it is called leaf. The
conditional probabilities are specified for each node to represent the influence of the parent nodes on its value using the
chain rule from probability theory:
P
A, B, C, D, E
PA/B,C,D,E∙PB/C,D,E∙PC/E∙PE∙PD
Figure 1a. Schematic BBN diagram.
Thus, the BBN structure represents the independence between the variables by using conditional probabilities that
represent the degree of belief in these relationships [40]. Main advantage of this method is the comprehensible graphic
display of interrelation of examined system and failure modes.
