Searching for the source of a hose leak frequency
DNV GL’s internal guideline on LNG QRA  mentions a leak frequency for hoses used in loading/unloading of ships. The frequency is 4 x 10-5 per hour for leaks and 4 x 10-6 per hour for ruptures, giving a total frequency of 4.4 x 10-5 per hour. What is the basis for these figures? Here is my attempt to find out.
Bevi Manual (2009)
The referenced source is a risk assessment manual for external safety (known as the Bevi manual) published in 2009 by the National Institute of Public Health and the Environment (RIVM) in the Netherlands . This specified that the figures were for loading of liquids from a storage tank to a transport unit (road tanker, tank wagon or ship) or vice versa. The source for the Bevi manual was an earlier document known as the Purple Book.
Purple Book (1999)
The Purple Book was published in 1999 and updated in 2005 by the Ministry of Housing Spatial Planning and the Environment (VROM) in the Netherlands . This quoted the same values above for loading of road tankers and tank wagons (as another source was used for ships). It attributed the rupture frequency to a value for “lightly-stressed” hoses in an earlier COVO study, and assumed that the frequency of leaks was 10x the value for ruptures.
COVO Study (1981)
The COVO study  was a risk analysis of major hazard installations in the Netherlands, published in 1981. This still sits on an old library shelf in our London office. It gave two values for hose ruptures: 4 x 10-6 per hour for “lightly-stressed” hoses, and 4 x 10-5 per hour for “heavily-stressed” hoses. It attributed them to two earlier studies by Welker (1976) and Jacobs (1971).
The study by Welker  was a detailed risk analysis of LNG ship transfer, but it made use of generic data due to lack of experience with LNG at that time. It emerged from our archives on a microfiche, a format that now requires me to visit the local history library to read. It gave a median rupture frequency of 2 x 10-5 per hour for hoses (no stress level specified), with a range from 4 x 10-6 to 4 x 10-5 per hour. It also stated that these were generic values appropriate for land installations, but could be applied to ships by multiplying by a service factor of 2.
It attributed the frequencies to two sources, Green & Bourne (1972) and Siccama (1971), and to engineering judgement based on failure modes of related equipment. It also mentioned a USCG report on flammable liquid spills (unidentified at present), in which the frequency of spills over 1000 gal due to cargo transfer, mainly hose failures, was 8 x 10-6 per hour, and cited a further report by Jacobs (1971) as an independent check.
A similar report by the same author , which is available on the internet, quotes the same value of 2 x 10-5 per hour for hoses, combined with a service factor of 2 for ships, attributed to Green & Bourne (1972).
Green & Bourne (1972)
The earliest definite source of the figures is a textbook on reliability, published in 1972 by Green & Bourne of the UK Atomic Energy Authority . The British Library holds a copy of this. Listed under average component failure rates for pneumatic and hydraulic components are failure rates of 40 per million hours for heavily-stressed hoses and 4 per million hours for lightly-stressed hoses. The book states that they are derived from past experience, and suggests they may have been corrected to apply to a general-purpose ground-based environment, but does not reveal the ultimate source. There is no confirmation that the hose data referred to ruptures, and pneumatic and hydraulic components in general were allocated 80% to leaks and 20% to blockages, which leaves the rupture contribution uncertain. The table of component failure rates is reproduced in Table A14.3 of the widely-used textbook by Lees .
The paper by Siccama  is a red herring. Although cited by Welker, it does not contain any hose leak frequencies, and states that only land storage tanks had information available to quantify risks. Based on all types of tanks in the Rotterdam industrial area during 1945 to 1970, it estimates a serious incident frequency of 1 per 3000 tank-years. This is equivalent to 3.8 x 10-8 per hour. The “engineering judgement” that Welker applied was presumably that the hose leak frequency would be higher than this tank incident frequency.
A possible source of the Green & Bourne value, and certainly a source referenced by Welker, was a paper published by Jacobs in 1971 by the American Society for Quality Control . As part of an illustrative reliability analysis of a domestic washing machine, this gave a failure rate for “hoses & joints” of 4.12 per million hours. There was no information on the source, or on whether the data referred to leaks or ruptures, although “failure” of a washing machine hose could reasonably be assumed to refer to ruptures.
Having followed the trail back over 40 years, I have still found no actual data underlying the hose leak frequencies. In other words, the original hose population and number of failures are unknown. It appears the rupture frequency comes from either washing machines or an unknown nuclear industry source, both from the 1960s or perhaps earlier. The frequency of smaller leaks is entirely based on a leak-to-rupture ratio assumed by the Netherlands government.
It seems disappointing that such an old and uncertain source should be repeated and recycled, over a period during which much better frequencies could have been collected. Perhaps the lack of transparency about its origins allows users to believe it is more robust than it actually is. Whatever the explanation, it is time to do better.
 DNV, “LNG QRA Guideline”, Confidential Internal Guideline G16, 2012.
 “Reference Manual BEVI Risk Assessment”, National Institute of Public Health and the Environment (RIVM), the Netherlands, 2009.
 “Guidelines for quantitative risk assessment”, Ministry of Housing Spatial Planning and the Environment (VROM), the Netherlands, 2005.
 “Risk analysis of six potentially hazardous industrial objects in the Rijnmond area, a pilot study”, COVO Commission report to the Rijnmond public authority, 1981.
 Welker, J.R. et al, “Fire safety aboard LNG vessels”, NTIS AD/A 030 619, January 1976.
 Green A.E. & Bourne, A.J., “Reliability Technology”, Wiley 1972.
 Lees, F.P., “Loss Prevention in the Process Industries: Hazard Identification, Assessment and Control”, Butterworth-Heinemann, 1996.
 Siccama, E.H. “The environmental risk arising from the bulk storage of dangerous chemicals”, Conference on Hazard Evaluation and Risk Analysis, Houston, August 1971, proceedings pp118-139.
 Jacobs, R.M., “Minimising hazards in design”, Quality Progress, October 1971.