Journal of Loss Prevention in the Process Industries 24 (2011) 208e213

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

journal homepage: www.elsevier.com/locate/jlp

History of Dutch process equipment failure frequencies and the Purple Book
Hans J. Pasman*
Emeritus Chemical Risk Management, Delft University of Technology, Delft, The Netherlands

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

Article history: In the middle of 70s concerns of the public in the Netherlands about fire, explosion and toxic risks due to
Received 23 May 2010 mishaps in the expanding process industry, causing political pressure led to the embracement of
Accepted 30 August 2010 quantitative risk analysis as tool for licensing and land-use planning. Probabilistic treatment of risk had
been exercised before to design flood defense. A ‘test’ on six different plants, the COVO study, favored the
Keywords: idea. Failure rate values were in immediate need. For storage vessels AKZO’s chlorine vessel data and
Equipment failure rates
British steam boiler data have been the first. Risk criteria to make decisions on were also developed and
Quantitative risk assessment
in 1985 embodied in legislation. As licensing and land-use planning are tasks of provincial authorities,
Process safety
Risk acceptance
under the auspices of the Inter-Provincial Consultation (IPO), further details such as failure frequency
values have been worked out. In the late 90s the Purple Book consolidated the information as a guideline
for Dutch quantitative risk assessment of process installations.
The paper will give a condensed historical overview, guidance to published papers; it will further make
comments, explain policy backgrounds, present comparison with other data and will briefly indicate in
which direction developments should go to improve QRA.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
The effect of the very small equipment failure rate values on
final event risk result is relatively large. Other probability values
in risk analysis such as the probability of wind in a given direc-
tion at given speed or a damage probit are for the major part
much closer to unity. In the early days of process installations
risk assessment not many failure data were available, although at
the Loss Prevention Symposium in Newcastle in 1971 (where the
European Loss Prevention Working Party was born) already an
excellent and very clear paper has been presented by Buffham,
Freshwater and Lees (1971) which explained the method as
well as the complexities in determining failure frequencies. In the
second half of the 70s in the Netherlands the formal Committee
for the Prevention of Disasters (CPR), founded in the 60s because
of several industrial accidents, had formed working groups on
various aspects with industry experts and scientists to draw up
guidelines and discuss safety improvements. Then, for the first
time quantitative risk analysis was performed and data had to be
* Present address: Mary Kay O’Connor Process Safety Center, Texas A&M
University, College Station, TX 88843-3122, USA. Tel.: þ1 356 21378271.
E-mail address: hjpasman@gmail.com.
collected which form still the basis of the present collection. In
the following the history of the data in the later (Purple Book,
1999) is briefly described.
2. History of failure rate values in the Netherlands
Various papers have been published on this history already.
A paper at the 10th Int’l Symposium on Loss Prevention in Stock-
holm in 2001 introducing the first edition of the Purple Book by Uijt
de Haag, Ale, & Post (2001) from RIVM, the National Institute for
Public Health and the Environment, discussed the selection of the
frequency values. RIVM is tasked to shape the use of quantitative
risk analysis (QRA) for Dutch users seeking a permit to operate
a plant under the Seveso II directive, implemented in the
Netherlands in the BRZO 1999. In the more recent Bevi regulation,
the Decree on External Safety of Establishments (Besluit externe
veiligheid inrichtingen) from 2004, relevant for licensing and spatial
planning, the sphere of application with respect to installations
covered, widened extensively compared to the BRZO. Two, more
recent papers by the same Dutch authors (Beerens, Uijt de Haag,
Post, & Ale, 2003; Beerens, Post, & Uijt de Haag, 2006) treated
more explicitly the history of the data. In summary and using their

0950-4230/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jlp.2010.08.012
 H.J. Pasman / Journal of Loss Prevention in the Process Industries 24 (2011) 208e213
Purple Book (1999)
RE-95-1 (version 2/2/1996)
IPO (1994)
COVO study (1981)
AKZO Chlorine study Appendix 9: General failure data
Smith & Warwick (1974) Bush (1975, Review failure data)
Fig. 1. Flow scheme of information on process equipment failure frequency data in use in the Netherlands for QRA as given by Beerens et al. (2003) and Beerens et al. (2006) of
RIVM.
scheme as shown in Fig. 1 the story can be told in just a few
sentences.
It all started with the COVO study (COVO, 1982)1 concerned with
the safety of employees and residents in the Rotterdam Rijnmond
area. It was a pilot study to evaluate methods of process risk
assessment, get experience and develop a policy. It started in 1978
on six plants, mostly storage installations (largest quantities, most
visible): acrylonitrile, ammonia, chlorine, LNG, propylene, and
a hydrodesulphuriser plant (using diethanol amine). Data on failure
frequencies of equipment were collected as inputs for the analysis.
Failure data on chlorine equipment originated from AKZO,
a company which had considerable experience with that chemical.
However the data on catastrophic failure of chlorine pressure
vessels have been calculated by means of fault tree analysis because
of lack of historical data, while only leakage failures could be based
on observations (Beerens et al, 2003; Beerens et al., 2006).
Appendix IX of the COVO study: ‘Failure rate data’ (In Fig. 1
described as ‘Appendix IX: general failure data’) contained failure
information on many different items. Event probability was built
through Fault Tree Analysis (including Human Error) and historical
data.
The various, later publications by Uijt de Haag et al. (2001) and
Beerens et al. (2003) and Beerens et al. (2006) focused on failure of
pressure vessels because these data are very determining overall
risk, and this paper will do the same. The data on the vessels
selected for the COVO study by Cremer & Warner originated from
survey reports of the UK Atomic Energy Authority by Phillips and
Warwick (1969) and by Smith and Warwick (1974), and from
a publication by Bush (1975). Lees (1996) in his compendium on
Loss Prevention gives some background information while covering
a longer time span of literature on the subject and summarizing
also later studies comprising process vessels. The interest of
mentioned authors was not process vessels but ‘nuclear primary
circuit envelopes’, hence the data refer mostly to steam vessels and
1
COVO is a Dutch acronym for Contact Group for Safety of Residents. This group
reported to the competent authority, the then existing regional Rijnmond Public
Authority, which later was absorbed by the Province of South-Holland. The COVO
steering committee chaired by a representative of the regional Rijnmond Central
Environ-mental Control Agency (DCMR) consisted of (expert) representatives of
two, later three ministries, ten companies and TNO. It guided the work carried out
by Cremer & Warner Ltd, London with counter expertise from Battelle Institut e.V.,
Frankfurt and Science Applications Inc., Los Angeles. Some references to the older
studies mentioned will not be repeated here as these can be easily found in the
recent articles (Beerens et al., 2003; Beerens et al., 2006). However crucial original
sources such as the COVO study have been consulted again for this paper.
209
Philips & Warwick (1969)
few to process vessels. The vessels considered were built to the
highest standard. The first study contained a considerable number
of failures observed in construction and 100,000 vessel-years in
service. The 1974 follow-on study increased significantly the latter
sample size. Two failure modes were distinguished: ‘catastrophic
failure’ and ‘potentially dangerous failure’; the former are disrup-
tions of the vessel requiring major repair or scrapping, the latter are
defects which might deteriorate under working conditions and
which require remedial action. Bush reviewed data, applied
a slightly different failure classification: ‘non-critical’ and ‘(poten-
tially) disruptive failure’, and published in the Journal of Pressure
Vessel Technology.
For the COVO storage vessels ‘base failure rate’ data were
required of static vessels, free of vibration, corrosion and thermal
cycling and absence of human operator error. (The COVO report
states that for vessels in a process environment, where conditions
are not free of the above influences, the base failure rate is normally
multiplied by a factor 3e10 depending on conditions.) Two modes
of failure were defined: ‘serious leakage’ and ‘catastrophic rupture’.
‘Serious leakage’ occurs if a hole arises equivalent to 50 mm
diameter. Because this hole was considered larger than most
‘potentially dangerous failure’ cases collected in the data, for the
COVO analysis a reduction factor on these data was applied of 3.
Appendix IX of COVO states that because catastrophic failure
was only about 10e25% of the failures, for the probability value of
catastrophic failure ‘the range quoted in the above references has
been divided by a factor of ten’ without further explanation. In
Fig. 2, borrowed from Beerens et al. (2003) and Beerens et al.
(2006), the assumed pressure storage vessel data are shown at
the top line (pressure reactor vessels are an order of magnitude
worse). In Table 9.1 ‘Base failure rate data’ of Appendix IX of COVO
catastrophic failure of pressure vessel was quoted as 106/yr with
a range of 6.3∙107e4.6∙105/yr and for serious leakage 105 and
range of 6  106e2.6  103/yr.2 In the COVO study itself evalu-
ating comments were given. On pages 5e44 and 5e45 critical
remarks by DCMR (chair) are collected referring to a much more
2
At the pressure vessel data in the table beside the references mentioned above,
also two others are mentioned: Green, A.E. and Bourne A.J., Safety Assessment with
Reference to Automatic Protective Systems for Nuclear reactors-Part 3, UKAEA
AMSB(S) R117 e 1976 and Marsall W. et al., An Assessment of the Integrity of PWR
Pressure Vessels, UKAEA Report (October 1976).
210 H.J. Pasman / Journal of Loss Prevention in the Process Industries 24 (2011) 208e213
Fig. 2. History of pressure vessel failure rate data as given in the subsequent Dutch documents. The COVO failure modes were split in the IPO document in instantaneous release,
‘continuous’ release of largest connection, emptying of the vessel in 10 min and leaks from openings of 50 and 10 mm, according to Beerens et al. (2003) and Beerens et al. (2006).
extensive study of Boesebeck3 leading to the conclusion that cata-
strophic base failure rate should have been a factor three higher
than assumed in COVO noting that in the Canvey study4 even
a value of 105/yr is used and for a NH3-sphere due to stress
corrosion 104/yr.
Beside the application of steam equipment data for process
plant in the COVO study Beerens et al. (2003) and Beerens et al.
(2006), question rightfully how ‘potentially dangerous failure’ is
interpreted in the COVO study and translated into failure causing
‘serious leakage’. For the ‘catastrophic’ failure data derived a similar
restriction holds. The literature failure data were divided by a factor
to obtain the desired data for the COVO study without solid ground.
Lees (1996) gives a summary table of Phillips and Warwick (1968)
data distinguishing failure in construction (12700 vessels) and
failure in service (100300 vessel-years, 132 failures, 89% by cracks).
Failure in service shows a rate of potentially dangerous failures of
1.25  103/yr and catastrophic failure of 0.7  104/yr. From the 7
catastrophic ones 4 were by mal-operation, 2 by fatigue and 1 by
manufacture. By further reduction for their nuclear assessment
they arrive at 62 applicable failures of which 2 catastrophic, and
rates become 6  104 and 2  105/yr. Lees quotes many additional
data which in no way indicate these rates are too high. The total
lack of data at the time of the COVO study and the pressure to get
the study done may have had an influence on results.
3
Boesebeck, K., Schadenswahrscheinlichkeiten für Reaktordruckbehälter abge-
leitet aus Schadensstatistiken für Druckbehälter aus dem konventionellen Bereich,
TÜ 16 (1975) Nr. 10 p. 281e286.
4
Canvey, An investigation of potential hazards form operations in the Canvey
Island/Thurrock area, Health and Safety Executive, London, 1978, Her Majesty’s
Stationary Office ISBN 011883200X.
As licensing of plant and land-use planning for ‘Seveso’ instal-
lations are very much tasks of provincial authorities in the
following years when QRA became standard practice the provincial
umbrella organization, the Inter-Provincial Consultation (IPO),
became involved in further details such as failure frequency values
of the whole variety of plant equipment components. Since the
safety reports containing a QRA had in principle to be drafted by
industry and submitted to the competent authority, in consultation
with industry recommended data were provided. The document
with data became known as IPO 1994. In this, the COVO data on
pressure vessel failure were modified with respect to failure mode.
A further differentiation was made of release scenarios. Cata-
strophic failure became split in three modes: instantaneous failure,
failure of largest connection to vessel or complete release of vessel
contents in 10 min, and serious leakage in large or small continuous
leak, while the overall failure rates remained roughly the same as
shown in Fig. 2. (Industry could take own data if they could prove
them to be better than the public data.) Amendment in the RE-95-1
document introduced again simplifications as shown in Fig. 2, as
e.g. leakage from a 50 mm hole was already covered by failure of
pipe work.
In the late 90s Seveso II directive gave new momentum. RIVM
got tasked by the Ministry VROM. As a result the Committee for the
Prevention of Disasters published Part I of the Purple Book (1999) as
a guideline for Dutch quantitative risk assessment of process
installations. It summarized again the information to be used as
default in Dutch QRA and stated in a commentary that no
management influence was accounted for and that the pressure
vessel data hold for static, non-vibrating, not corroding vessels
which are not subjected to thermal cycling. The Purple Book was
presented at the Stockholm Loss Prevention symposium by Uijt de
Haag et al. (2001). The presentation mentioned a number of then
recent review studies on failure rates, such as the one by
 H.J. Pasman / Journal of Loss Prevention in the Process Industries 24 (2011) 208e213 211

Logtenberg (1998) of TNO and by Taylor (1998) ‘which show Table 1

a tendency towards higher failure frequencies’ than the ones Comparison by Nussey (2006) of failure rate values of pressure vessels from the
Purple Book (PB99) with HSE adopted data and a ‘correction’ of Purple Book for
collected in the Purple Book, as illustrated by the TNO figures in human factor and external impacts. The data are expressed as cpm or chance of
Fig. 3. Also the AMINAL data of the Flemish, Belgian authorities failure per million years.
were mentioned in this context. An investigation and update of the
Type of failure PB99 default PB99 ‘complete’ HSE
data were announced.
Catastrophic 0.5 5.5 2e6
More or less in parallel, experts at Health and Safety Executive
Large hole 0.5 5.5 5
(HSE) in UK being involved in Land-Use Planning advice to local Small hole 10 10 55
authorities and having developed Risk Assessment Tool RISKAT All types 11 21 62e66
including risk criteria, built their own failure frequency database.
Hurst, Hankin, Wilkinson, Nussey, and Williams (1992) presented
a paper on their generic failure rates collection. Historical data of
The HSE data are close to the median of the logarithms of failure
vessel failure were tried to back by structured, fault tree analysis;
frequency values.
further, causes of failure were investigated in depth so that
a statistical breakdown could be given. HSE was aware of the
influence of management on the rates and intended to investigate 3. Qbjectives to perform a QRA
the effect on final risk values. They also tried to get clarity and
uniformity in the terminology of failure modes. Based on 230 Risk assessments are conducted for four possible reasons:
reported vessel failures a cold catastrophic vessel failure rate was
derived of 6  106/yr. The theoretical rate for vessels producing 1. For making plant safer
a BLEVE was two orders of magnitude lower than the mean of seven 2. For licensing of plant
historical surveys, which showed that fault tree analysis does not 3. For land-use planning
provide a complete picture. Spread in the observed rates for pipes 4. For improving emergency planning
was between þ and e one order of magnitude.
Recently Nussey (2006) on behalf of HSE, reviewed the then Dutch QRAs have been mainly applied for reasons nos. 2 and 3.
available information on failure frequency high pressure storage As already mentioned the analyses are in general performed by
vessels to investigate why HSE’s applied values were higher than a consultant on behalf of a plant owner. Both licensing and land-use
the Dutch ones of the Purple Book. Nussey’s overview is a very planning have legal implications and where there are conflicting
extensive one; Appendix 4 reviewed the information in the Purple interests between parties fuzziness and ambiguity are not accept-
Book. Apart from new information e.g. on the use by HSE of 1981 able. The QRA results should be reliable and beyond doubt. Once
data of Smith and Warwick, he also comments on sources already a decision is made and a license is granted it should not be possible
mentioned here, e.g. noting that Logtenberg (1998) reported that to arrive on the basis of the same data and input information to
the IPO data are in essence expert judgments. Nussey produced other results and challenge the license. It is known that given
much evidence that the Dutch values are rather optimistic but that a certain installation QRA outcomes of individual and societal risk
including estimated effects of human factor and external impacts values produced by different analysts scatter over various orders of
e.g. corrosion etc., increasing the rates by a factor ten would bring magnitude. The EU sponsored two projects comparing calculation
them close to the HSE values not even considering the confidence results of QRAs by different teams: the first in the early 90s; the
interval of one to two orders of magnitude. This is illustrated by second one ten years later known as project ASSURANCE reported
Table 1 borrowed from his report. by Lauridsen et al. (2002). In this project also scenario definition
In fact, in the light of other data sources even the HSE data are was included and although experience was gained over the years,
still optimistic. Fig. 3 presenting data from Logtenberg (1998) gives spread in risk figures remained to be orders of magnitude. The
an idea of how wide the actual range of data for pressure vessels is. project investigated the various factors playing a role in causing the
deviations.
For licensing large differences in results of two analyses are
a rather bad situation. The Dutch ministry VROM confronted with
the scatter in QRA outcomes by different consultants analyzing the
1.E-03
same plant, decided some years ago to standardize the whole
procedure and to remove uncertainty where possible. It means the
use is restricted to only one particular model, in one particular
1.E-04
version with a particular set of model options (SAFETI-NL), (see Uijt
Failure frequency (/ yr)

Inst. Hole de Haag, 2007). The Purple Book was replaced by the Reference
Manual Bevi Risk Assessments, (see Uijt de Haag et al., 2009), with
1.E-05 all data given (for vessel failure rates no change in values). Users of
SAFETI-NL get training. Loss of containment scenarios are stan-
dardized and kept simple; many details in a scenario that require
1.E-06
a decision from an analyst are prescribed. Effects of management
quality and human factors are excluded. The user influence on the
results is in this way minimized. It may seem pragmatic and from
a juridical aspect making a lot of sense, but it is from a scientific
1.E-07
point of view unsatisfactory.
Purple Book For application of QRA to make plants safer, which should be the
first reason to do it, reality content of a QRA for an actual installa-
Fig. 3. Pressure vessel failure frequencies (Inst. ¼ instantaneous/catastrophic;
Hole ¼ 50 mm hole/vessel drainage in 10 min) in the Purple Book (1999) compared to
tion should be as high as possible. It holds too for emergency
various literature data collected by Logtenberg (1998) as shown by Uijt de Haag et al. planning which is a growing need and which adds another
(2001). dimension to QRA because time functions in a scenario now
212 H.J. Pasman / Journal of Loss Prevention in the Process Industries 24 (2011) 208e213
become essential to model as well for evaluating self-rescue and
evacuation possibilities.
4. Brief history of risk acceptance criteria
After the flooding disaster in 1953 with thousands of fatalities
and the effort to design improved dike and coastal defense (Delta
plan) political debate on risk in terms of probability and effect
versus benefit or at least damage reduction had been practiced. So,
in the 70s when chemical risk became an issue the debate went also
that direction. In the middle of 80s an individual risk contour
threshold was introduced of 106 probability of a fatality/year to
stay outside residential area and a societal risk one such that an
accident with 10 or more fatalities at the side of residents in the
plant’s vicinity shall be less than once in 100,000 years, 100 fatal-
ities less than once in 10 million and 1000 in a billion years (NMP,
1988). In particular the latter criterion became over the years more
relaxed and nuanced. The rationale behind the criteria has been
explained by Ale (1991). The risk by an industrial activity should
remain very low with respect to the natural risks present. The
reason to mention the development of these criteria here is that it is
quite reasonable to assume it had an influence on the attitude with
respect to the failure frequencies.
Before the time of the actual introduction of the criteria in the
law, consensus seeking meetings between industry and govern-
mental representatives under the umbrella of the Committee of
Prevention of Disasters shall have had its shadows already cast
forward as the low chance risk acceptance levels met resistance by
industry circles. Stringent risk requirements will have been
compensated by low, hence optimistic failure rate values. This is of
course just psychological explanation and speculation by the
present author. The system is now in operation for more than 20
years. A change in probability values will cause protests in any
direction. If the probabilities and hence the risk will become lower
industry may complain that they have been forced to higher cost
levels unnecessary; if it is the other way complaints will be stronger
because it will require additional safeguarding measures.
5. How shall we advance
As mentioned the first use of risk assessment shall be making
plants and transportation activities of hazardous materials safer,
and safe enough for workers in the plant and population outside.
There are quite a few tools to assist in such effort, but risk assess-
ment and certainly if it is as comprehensive as the ARAMIS meth-
odology developed at the start of this century as an EU project (see
Salvi & Debray, 2006) will help as a tool of last resort in complex
cases in which one is likely to lose overview. Reality content shall be
as high as possible. Human factor and management effectiveness
shall be difficult but essential to take into account. A prescribed,
standardized approach will however discourage incentives to
improve. Instead use should be made of better knowledge,
improved scenario generation, better models and data, progress in
IT and computer technology, Bayesian belief networks, economic
aspects and decision theory to name a few. Nobody will in the end
be served well with unrealistic results obtained with the aid of
unreliable data and non-transparent models.
A way out of the observed dilemma will be to continue for the
time being using the frozen model and data for licensing and LUP in
the Netherlands until a major overhaul is ready and proven.
Moreover before a change can be made a solution shall be found
how to protect old rights and how to deal with residual uncertainty.
Further the question shall be answered whether the regime is
already stringent enough, which is a political matter. Meanwhile it
will be necessary to strive for a broad international cooperation to
further develop risk assessment methodology which is in the
interest of all parties in a situation of ever decreasing available
space. Failure rate data will be a crucial part of it. It means in
a practical way collecting data, describing the equipment contour
limits (stumps, nozzles etc.), categorizing failure modes and leak
size, determining external influences and loading history, effect of
inspection intensity and maintenance, if needed describing the
failure rate as time function (ageing), producing the data not only as
point values but defining confidence intervals and where useful to
back empirical data by theoretical analysis and making use of
Bayesian statistics. The nuclear industry did this and also the
offshore sector had a reasonable success with their data. The
needed inputs can only come from process industry. Companies
may however be reluctant to share the information afraid of dis-
turbing a status quo and getting more severe restrictions for their
activities.
In fact, in Europe the Commission installed in 2003 the Euro-
pean Working Group on Land Use Planning (EWGLUP, 2003) with
Terms of Reference covering exactly the objectives described. There
are of course a few fundamental differences in thinking to over-
come between European countries with regard to risk assessment
than just the models and data. But a pursuance of getting the
technical aspects right will help. The meetings EWGLUP organized
such as the one on failure frequencies in March, 2005 might be
repeated with more input from industry and backed by research
incentives and a program.
The Quality of Risk Assessment for Process Plant or QRAQ
project by Taylor Associates (Taylor, 2009) together with the
RELBASE data shows what is possible with decisiveness and
endurance. Given an installation results of risk calculation shall be
repeatable (by the same group), reproducible (by another group)
and hence the approach robust, while model and data are trans-
parent and verified, where necessary validated experimentally and
the software is reliable. Quality QRA methodology shall be a longer
term goal to be achieved in an open international cooperation.
6. Conclusions
 The history of the Dutch failure rate data for QRA started in the
‘dark Middle Ages’ of risk analysis for process plant.
 The data contain considerable expert judgment.
 The data are also the result of consensus between industry and
authorities and have to be viewed in the light of the stringent
risk acceptance criteria formulated in the middle 80s.
 To improve the reality content of risk analysis improvements
are much desirable.
 Acquiring good data is no sinecure. Many factors play a role.
Beside a point value a confidence interval is needed. The EU
with EWGLUP set an objective for improvement with regard to
more uniform data in view of land-use planning.
 Quality QRA methodology shall be a longer term goal to be
achieved in an open international cooperation.
Acknowledgement
The support and comments of Paul Uijt de Haag are gratefully
acknowledged.
References
Ale, B. J. M. (Jan 1991). Risk analysis and risk policy in the Netherlands and the EEC.
Journal of Loss Prevention in the Process Industries, 4, 58e64.
Beerens, H. I., Post, J. G., & Uijt de Haag, P. A. M. (2006). The use of generic failure
frequencies in QRA: the quality and use of failure frequencies and how to bring
them up-to-date. Journal of Hazardous Materials, 130, 265e270.
 H.J. Pasman / Journal of Loss Prevention in the Process Industries 24 (2011) 208e213
Beerens, H. I., Uijt de Haag, P. A. M., Post, J. G., & Ale, B. J. M. (Nov 2003). Validity of
the use of generic failure frequencies in a QRA, IMECE 2003-43444, 2003 ASME
international Mechanical Engineering Congress & Exposition. Washington, D.C.:
pp. 16e21.
Buffham, B. A., Freshwater, D. C., & Lees, F. P. (1971). Reliability engineering-a rational
technique for minimising loss, major loss prevention in the process industries. In
IChemE symp series, 34. 106th Event EFCE. pp. 87e98.
Bush, S. H. (February 1975). Pressure vessel reliability. Transactions of ASME - Journal
of Pressure Vessel Technology, 97(1), 5470.
COVO. (1982). COVO Commission, risk analysis of six potentially hazardous industrial
objects in the rijnmond area, a pilot study. A Report to the Rijnmond Public
Authority, Central Environmental Control Agency, Schiedam, The Netherlands,
1981. Dordrecht, The Netherlands: D. Reidel Publishing Co, ISBN 90 277
1393 6.
EWGLUP. (2003). European working group on land use planning. http://landuseplanning.
jrc.it/index.html.
Hurst, N. W., Hankin, R. K. S., Wilkinson, J. A., Nussey, C. & Williams, J. C. (May 1992).
Failure Rate and Incident Databases for Major Hazards. In 7th International
Symposium on Loss Prevention and safety Promotion in the Process Industries,
454th Event EFCE, Vol. 3, 143, pp. 1e22, Taormina, Italy; 4e8.
Lauridsen, K., Kozine, I., Markert, F., Amendola, A., Christou, M., & Fiori, M. (May
2002). Assessment of Uncertainties in Risk Analysis of Chemical Establishments.
Risø National Laboratory. The ASSURANCE project, Final summary report. www.
risoe.dk/rispubl/sys/syspdf/ris-r-1344.pdf pp. 52.
Lees, F. P. (1996)Loss prevention in the process industrieseHazard identification,
assessment, and control, Vol. 1e3. Oxford: Butterworth-Heinemann. 12/94e98.
Logtenberg, M. Th (1998). Derivation of failure frequencies for LOC cases. TNO report
MEP-R98/501, Apeldoorn.
NMP. (1988). Dutch National Environmental Policy Plan “Premises for Risk
management”, Second Chamber of the State General 1988e89, session 21137,
nos 1e2, (Nationaal Milieu Beleidsplan, NMP, “Omgaan met risico’s, de ris-
icobenadering in het milieubeleid”, Handelingen Tweede kamer, vergaderjaar
1988e1989, 21 137, nr. 5, p. 7).
213
Nussey, C. (Dec 2006). Failure frequencies for major failures of high pressure storage
vessels at COMAH sites: A comparison of data used by HSE and the Netherlands.
http://www.hse.gov.uk/search/googlecoop.htm?q¼failureþfrequencyþpressureþ
vesselþTaylor&cof¼FORID%3A11&cx¼015848178315289032903%3Akous-jano
68#1235 pp. 1e36.
Phillips, C. A. G., & Warwick, R. G. (1969 as cited in COVO, 1982; and 1968 as cited by
Lees, 1996). A survey of defects in pressure vessels built to high standards of
construction and its relevance to nuclear primary circuits, UKAEA AHSB(S) R162.
Purple Book. (1999). Committee for the Prevention of Disasters, Guideline for
Quantitative Risk Assessment, CPR 18E, SDU The Hague, renumbered to PGS 3
in 2005, http://www.vrom.nl/pagina.html?id¼20725.
Salvi, O., & Debray, B. (2006). A global view on ARAMIS, a risk assessment meth-
odology for industries in the framework of the SEVESO II directive. Journal of
Hazardous Materials, 130, 187e199.
Smith, T. A., & Warwick, R. G. (1974). The second survey of defects in pressure
vessels built to high standards of construction and its relevance to nuclear
primary circuits. Safety and Reliability Directorate, SRD R30.
Taylor, J. R. (2009)The QRAQ project, quality of risk assessment for process plant, Vol.
1e5. Taylor Associates ApS. 2005e2009. http://itsa.dk/default.htm.
Taylor, J. R. (1998). Review of failure data for risk analyses. Version 1 Issue 1.
Glumsoe: Neste and Taylor Associates ApS.
Uijt de Haag, P. A. M., Ale, B. J. M., & Post, J. G. (June 2001). Guideline for quantitative
risk assessment in the Netherlands. 19e21, Stockholm, Sweden. In H. J. Pasman,
O. Fredholm, & Jacobsson. (Eds.), Proceedings 10th international symposium on
loss prevention and safety promotion in the process industries, Vol. 2 (pp.
1429e1438), ISBN 0 444 50699 3.
Uijt de Haag, P. A. M., et al. (2009). Reference manual bevi risk assessments. Bilthoven,
The Netherlands: National Institute of Public Health and the Environment
(RIVM). http://www.rivm.nl/milieuportaal/bibliotheek/veelgesteldevragen/-
waar-vind-ik-de-handleiding-risicoberekeningen-bevi.jsp.
Uijt de Haag P. (2007). QRAs for Dutch installations, 12th symp loss prevention and
safety promotion in the process industries, Edinburgh, U.K.: pp. 22e24 May, on
CD-ROM.