DNV Assumptions.pdf

DET NORSKE VERITAS

TM

APPENDIX A - ASSUMPTIONS REGISTER

SKANGASS AS

R EPORT EPORT NO./DNV R EG EG NO.: 2013-4091 / 17TLT29-4 EV 1, 11.06.2013 R EV

DET NORSKE VERITAS Report for Skangass AS Appendix A - Assumptions Register

MANAGING RISK

T able able of of C onte ontents Assumption no.

Subject Description and Background Data 1-A Manning Level 2-A Meteorological Data 3-A Meteorological Parameters 4-A Ignition sources – sources – Equipment Equipment 5-A Ignition sources – sources – Traffic Traffic 6-A Ignition Sources – Sources – People People 7-A Ignition sources – sources – Hot Hot work 8-A Bunkering installation – installation – Base Base case design and inventory 9-A Escape and Evacuation of Passengers and Personnel LNG accidents Representative Scenario Assumptions 1-C Release Location / Height 2-C Release Sizes Frequency Analysis Assumptions 3-C Leak frequencies Event Tree Modelling Assumptions 4-C Detection and Isolation Times 5-C Isolation Failure 6-C Immediate Ignition Probability 7-C Event Tree Framework 8-C Event Tree Probabilities Consequence Modelling Assumptions 9-C Dispersion Parameters 10-C Consequence Modelling Parameters Storage & loading – loading – Specific Specific 1-D Bunkering Frequency Impact Criteria 1-H End Point (Impact) and Vulnerability (Fatality) Criteria

DNV Rep. No.: 2013-4091 Revision No.: 1 Date : 11.06.2013

Page 1 of 1


MANAGING RISK

LNG Bunkering Terminal, Risavika Harbour

Date: 20.05.2013

Category:

1-A Description and Background Data

Subject:

Manning level and distribution

Assumption No.:

Revision: 1

Specifications: Manning levels are defined for the foll owing time periods: ‘Day’: Morning shift (08:00 to 16:00) and Afternoon shift (16:00 to 00:00). ‘Night’: 00:00 to 08:00.  

The risk analysis is based on the t he onsite population and off-site population : -

1st party: Personnel in the plant; Fjordline and Skangass personnel involved in the bunkering operation; Skangass personnel.

-

2nd party: Risavika harbour personnel, Fjordline personnel not involved in t he bunkering operation, Ferry terminal workers.

-

3rd party: Population (workers and public) working/evolving in the Container ara, Ernegiveien + Risavika, Rest Companies areas. Fjordline’s Passengers, public evolving around the plant, Tananger residents.

The LNG bunkering operation of Fjordline which are not to be present 24 hours a day, all days throughout the year, are taken into i nto account. The bunkering operations are planned to last 1.5 hours every day. 1 hour for LNG cool down and transfer and ½ an hour for connection and disconnection of loading arm to ship manifold. Hence, possible leakage, from downstream ESD valve, may occur only during 1 hour per day. In order to obtain a good representation of the risk picture towards the ferry passengers and to take a conservative approach, it has been decided that the same population is present for 30 % of the totality of the LNG bunkering operations. A total number of 4 workers (1st party) will be present at the bunkering terminal. 1 to 2 Fjordline personnel at the ship manifold manifold and 1 to 2 Skangass personnel personnel at the jetty. This figure is used to st calculate the average individual risk for 1 party population. The distribution of on-site (inside the ISPS area of Risavika jetty 38) and off-site off -site people is summarised in Table 1 and Table 2. Indoor and outdoor factors used in the analysis for the different parties are presented in Table 3 and Table Table 4. Implication of assumption:

Societal risks are directly influenced by t he numbers of personnel exposed to hazardous events and hence the results are sensitive to the manning assumptions. Key influence on societal risk risk / FAR. Reference: /1/ LNG Bunkering of Fjordline ferries Project Design Basis 17.02.2012, Draft Version Sign: J-B, Berthomieu Date: 21.05.213 Prepared by: Internal Verification:

Sign:

Date:

Sign:

Date:

Comment from Skangass AS: Approved by Skangass AS: DNV Reg. No.: 17TLT29-4 Revision No.: 1 Date :11.06.2013

Page 1 of 32


MANAGING RISK

Attachment to Assumption 1-A

st

Table 1 Base case manning within each area for 1 party Area/Category

During Bunkering, i.e. afternoon

Fjordline personnel involved in bunkering activity Skangass Bunkering personnel

2 2

Table 2 Base case manning and population ‘Day’ Area/Category

Peninsula Hiking Track Skangass LNG plant* Ferry Terminal – Office Workers Ferry Terminal – Industry Workers Ferry Terminal – Passengers incl. ferry crew Energiveien+Risavika – Office Workers Energiveien+Risavika – Industry Workers Container Area – Office Workers Container Area – Industry Workers Rest Companies – Office Workers Rest Companies – Industry Workers Living Quarters Tananger*

‘Morning’

‘Afternoon’

16*1 8* 9 100 10 0 400 559 10 50 1 139 715 60 5 964

16*1 8* 4 10*3 10 1500 40*3 56* 10 50 114* 72* 60 5964

‘Night’

24 hrs average

0 0 4 10*3 14 0 5 0 1 0 10 0 60 5 964

2 1 6 40 11 500 148 205 7 33 421 262 60 5 964

*1 In a non-working day *2 24 hours average will be used when no bunkering is on-going. ‘Afternoon’ will be used when bunkering is on-going. *3 Assuming that 10% are working overtime. *4 Population updated at the 1st of January 2011, source: Statistisk sentralbyrå.

DNV Reg. No.: 17TLT29-4 Revision No.: 1 Date :11.06.2013

Page 2 of 32


MANAGING RISK Table 3 Indoor/outdoor fraction – 1st party Area/Category Administration Building Operator/Maintenance Truck Loading (1 person per truck per 1.2h) Ship Loading (Jetty- only during connection and disconnection) Ship Deck (during loading only) Ship Bridge (during loading only)

Table 4 Indoor/outdoor fraction - 2 nd and 3rd party Location/Category Peninsula Hiking Track Ferry Terminal – Office Workers Ferry Terminal – Industry Workers Ferry Terminal – Passengers Ship passengers Energiveien+Risavika – Office Workers Energiveien+Risavika – Industry Workers Container Area – Office Workers Container Area – Industry Workers Rest Companies – Office Workers Rest Companies – Industry Workers Living Quarters Tananger Population

Indoor fraction 0 0 1 0 0.75 0.75 1 0 1 0 1 0 0.75 0.75


Page 3 of 32

Indoor fraction 1 0.8 0 0 0 0.75

Outdoor fraction 0 0.2 1 1 1 0.25

Outdoor fraction 1 1 0 1 0.25 0.25 0 1 0 1 0 1 0.25 0.25


MANAGING RISK

Skangass LNG anlegg

Figure 1 Risavika harbour and population location


Page 4 of 32


MANAGING RISK

LNG Bunkering Terminal, Risavika Harbour Assumption No.: Category: Subject:

Date: 20.05.2013

2-A Description and Background Data Meteorological Data

Revision: 1

Specifications:

Data on the wind direction, wind speed and atmospheric stability are combined to form a set of representative weather categories in the surroundings of Sola, which are taken from the QRA for Lyse LNG Base Load Plant (ref A). Table 5 shows the wind data and Figure 2 shows the wind rose for Sola. Implication of assumption:

The weather conditions have a key infl uence on flammable cloud dispersion and fore heat loads, hence the consequences associated with any release. The influence of any specific weather category and direction will vary for each and every release, where on balance the resulting influence of any changes in the meteorological assumptions will have a negli gible influence on the risk results. Relevant to specific consequences – risk is not sensitive to individual meteorological assumptions. Reference: QRA for Lyse LNG Base Load Plant – Train 1 (R100-LE-S-RS0003) Sign: J-B, Berthomieu Prepared by: Internal Verification:

Sign:

Date: 21.05.213 Date:

Sign:

Date:

Comment from Skangass AS: Approved by Skangass AS:


Page 5 of 32


MANAGING RISK

Attachment to Assumption 2-A Table 5: Rationalised representative weather categories for Sola o

Wind Direction ( )

292.5 – 337.5 337.5 – 22.5 22.5 – 67.5 67.5 – 112.5 112.5 – 157.5 157.5 – 202.5 202.5 – 247.5 247.5 – 292.5 All

% Occurrence of Weather Classes (Pasquill Stability, Wind Speed) D1.5 (day) / F1.5 (Night) D6 D12 Total 1.5m/s 6 m/s 12 m/s 1.99 14.71 2.79 19.49 0.961 7.09 1.346 9.397 1.012 7.47 1.417 9.899 1.633 12.04 2.293 15.966 1.335 9.89 1.878 13.103 0.501 3.69 0.702 4.893 0.807 5.96 1.13 7.897 1.977 14.57 2.76 19.307 10.216 75.42 14.316 100

Figure 2 Wind rose data for Sola, Rogaland


Page 6 of 32


MANAGING RISK

LNG Bunkering Terminal, Risavika Harbour Assumption No.: Category: Subject:

Date: 20.05.2013

3-A Description and Background Data Meteorological Parameters

Revision: 1

Specifications:

In addition to the weather categories, certain meteorological constants are defined as inputs to the consequence modelling. These values are summarised below: Parameter

Atmospheric temperature Atmospheric pressure Relative humidity Surface temperature Surface roughness parameter

Solar flux

Wind speed reference height

Value

10ºC 101 325 N/m2 68% 10ºC 0.3 for land 0.05 for water 100 W/m2

10 m

Notes and References

Range is 5º to 45ºC. Average sea level pressure. Range is 50% to 85%. Taken to be the same as atmospheric temperature. Land value appropriate for terrain with varying geometry, water value for coastal waters. The maximum solar flux (i.e. midday midsummer) is about 1320 W/m2. However, the solar flux varies diurnally, annually and with cloud amount. Hence the annual mean value will be less than half the maximum. 100 W/m2 is a representative value. Standard for meteorological measurements.

Implication of assumption:

The dispersion and consequences associated with LNG and other dense gas releases are relatively sensitive to assumptions affecting the heat transfer to the cloud. Hence, the above values are relatively conservative representative conditions, but will not necessarily correspond to the worst-case dispersion conditions that may occur. Representative conditions used – relevant to consequences, with relatively minor influence on subsequent risks. Reference: QRA for Lyse LNG Base Load Plant – Train 1 (R100-LE-S-RS0003) Sign: J-B, Berthomieu Date: 21.05.213 Prepared by: Sign: Date: Internal Verification: Comment from Skangass AS: Approved by Skangass AS:

Sign:

Date:


Page 7 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Ignition Sources – Equipment

Assumption No.:

Revision: 1

Specifications:

The basis for defining ignition source probabilities for equipment items within the plant is taken as the JIP Ignition Modelling study /1/. The values given in the JIP Ignition Modelling study are summarised below, together with modification factors proposed within the study and the value a dopted for this analysis. It is considered that with respect to ignition sources associated with electrical and rotating equi pment, the Skangass LNG Plant and the LNG bunkering facility is consistent with a modern, best-practice offshore facility. Hence the recommended modification factors are adopted for all except the ‘Other’ category, which is considered to be less practical to control for land based facilities.

Equipment Type

Electrical Other Pump Compressor Generator / Turbine

Base Ignition Probability, per second of exposure 2.7E-08 per m2 2.1E-09 per m 2.1E-07 per item 5.1E-06 per item 6.2E-06 per item

JIP Modification Factor 0.49 0.65 0.61 0.61 0.61

Modification factor Used

Ignition probability Flare 54-FC-101 0.25 ) Fired Heater for Hot Oil 52-FA-101 0.12) H2S Converter 0.1 ) Electrical Substation1) 0.13) Ferry terminal and surroundings (e.g. lighting poles) 0.7 Ignition Source

0.5 1 0.6 0.6 0.6

Ignition Probability Used, per second of exposure 1.4E-08 per m2 2.1E-09 per m 1.3E-08 per item 3.1E-06 per item 3.7E-06 per item

Time of exposure (s) 10 60 60 600 60

1) The flare is located 70m above the ground level, and only large leaks from segment 1 and 2 are assessed to reach the flare. For other scenarios the ignition probability will be 0. 2) 60 seconds of exposure with an operating probability of 0.5 3) It is assumed that if the air intake of a substation or building is exposed to gas for 600 seconds, which then enters the building, ignition will occur (i.e. ignition probability of 1). It is, however, assumed that the gas detection and automatic closure of the HVAC intake dampers is effective in isolating the sources within the building i n 90% of such cases, i.e. the activation frequency for each substation / building ignition source is taken as 0.1. In case of gas detection, the bunkering activity will be automatically stopped Implication of assumption:

Key influence in determining the likelih ood of flash fire, pool fire and explosion hazards and the extent of each (i.e. time of ignition relative to size of cloud). Note, however, that the overall effect is that there are a significant number of very low ignition probabilities. Overall effect is a key influence on the risks, but not sensitive to any particular ignition source.


Page 8 of 32


MANAGING RISK

Reference: /1/ JIP, 1998. Ignition Modelling, Time Dependent Ignition Probability Model , Joint Industry Project – DNV, Scandpower, et al. DNV Report No. 96-3629, Revision 4, February 1998 /2/ QRA for Lyse LNG Base Load Plant – Train 1 (R100-LE-S-RS0003) Sign: J-B, Berthomieu Date: 21.05.213 Prepared by: Internal Verification:

Sign:

Date:

Sign:

Date:



Page 9 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Ignition sources – Traffic

Assumption No.:

Revision: 1

Specifications: 







Bunkering will only take place when southbound ferry lay at port. Boarding will take place either by passenger tube or cars/trucks by the ro-ro ramp. Consequently, vehicles do not board through ISPS area, where bunkering station is to be located. Vehicles board from the parking area south/southwest to the ISPS area, at a minimum distance of 40 m from the bunkering station. Bunkering is planned to take place at the same time as cars are boarding. However, no passengers are allowed in the tube during this operation. Unloading of vehicles only takes place when northbound ferry lay at port. Vehicles are then routed through the ISPS area, i.e. directly past the bunkering station. However, bunkering is not taking place when unloading vehicles. The jetty is assumed closed for traffic except for when boarding/unloading takes place. There are 150 parked cars and trucks waiting to board, with the possibility to have their engines running.



150 cars and trucks are assumed to be boarding the ferry by the south gate, planned LNG bunkering operations is allowed during this step.

Ignition Source

Trucks/cars loading Maintenance traffic Parking area traffic Ferry

Ignition probability, per vehicle/vessel 0.4 0.4 0.4 0.5

Time of exposure (s) 60 60 60 60

Traffic density, per day

150 1 20 1


Probability of ignition in case of release. Influences societal risk result. Reference: /1/ LNG Bunkering of Fjordline ferries Project Design Basis 17.02.2012, Draft Version /2/ Unloading Loading routes Layout Draft /3/ PGS 3, Guideline for quantitative risk assessment (“Purple book”), Ministry of VROM, 2005 Sign: J-B, Berthomieu Date: 21.05.213 Prepared by: Sign: Date: Internal Verification: Comment from Skangass AS: Approved by Skangass AS:

Sign:

Date:


Page 10 of 32


MANAGING RISK

Figure 3 Loading/unloading routes; tube and bunkering station (‘manifold’) location; parking areas


Page 11 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Ignition Sources – People

Assumption No.:

Revision: 1

Specifications:

The default value assigned within PhastRisk for the ignition source associated with people corresponds to 1.68E-4 per person per second of cloud exposure. This value has been derived to account for the probability of ignition associated with people in general, and includes an allowance for smoking and general human behaviour associated with residential areas. The value assigned to personnel working offshore by the JIP Ignition Study (Reference /2/) is significantly lower, being almost 4 orders of magnitude lower even f or dense populations. The value applicable to Skangass personnel at the bunkering station (i.e. a trained workforce, with no smoking and with traffic and hot work ignition sources accounted for separately) would be much closer to that recommended by the JIP study i n practice; however the default PhastRisk value will be conservatively applied in the analysis. Implication of assumption:

Key influence in determining the likelihood of flash fire, pool fire and explosion hazards and the extent of each (i.e. time of ignition relative to size of cloud). Note, however, that the overall effect is that there are a significant number of very low ignition probabilities. Overall effect is a key influence on the risks, but not sensitive to any particular ignition source. Reference: /1/ PhastRisk version 6.7., 2011 /2/ JIP, 1998. Ignition Modelling, Time Dependent Ignition Probability Model, Joint Industry Project – DNV, Scandpower, et al. DNV Report No. 96-3629, Revision 4, February 1998. Sign: J-B, Berthomieu Date: 21.05.213 Prepared by: Internal Verification:

Sign:

Date:

Sign:

Date:



Page 12 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Ignition sources – Hot work

Assumption No.:

Revision: 1

Specifications:

The analysis is based on no hot work taking place within the bunkering area.


Key influence in determining the likelihood of flash fire, pool fire and explosion hazards and the extent of each (i.e. time of ignition relative to size of cloud). Note, however, that the overall effect is that there are a significant number of very low ignition probabilities. Overall effect is a key influence on the risks, but not sensitive to any particular ignition source. Reference: /1/ LNG Bunkering of Fjordline ferries Project Design Basis, 17.02.2012, Draft Version Prepared by: Internal Verification:

Sign: J-B, Berthomieu Sign:

Date: 21.05.213 Date:

Sign:

Date:



Page 13 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Bunkering station – Base case design and inventory

Assumption No.:

Revision: 1

Specifications: The base case design is described by Skangass in Design Basis document /1/ and drawings /2/, /3/ and /4/.

The ferry bunkering system relevant for the analysis corresponds to /2/ :

-

1 LNG booster pump

-

1 4” Vapour Return Line

-

1 flowmeter package

-

1 LNG loading arm

-

Small bore fittings, manual and actuated valves.

1 8” LNG pipeline

The details of equipment are available in the appendix D of this study. The LNG booster pump, the fiscal metering package and the first part of the pipeline is located inside the plant area. The pipeline within the plant will be routed in the existing pipe rack /1/. Outside the plant, the pipeline will be routed in an underground tunnel up to jetty 38 following the route indicated in reference /3/. Part of this section will be alongside water pipes. The LNG pipeline will run underground between the ferry terminal building and the jett y (together with power cables) before going vertically up. The pipe will then go above ground the last 7 meters before the bunkering station. The design of the piping has the following characteristics:

-

Double-wall (pipe-in-pipe) stainless steel pipe

-

Full containment

-

Routed in an underground tunnel

-

Carried on pipe supports

-

Leak detection between the double walls

The loading arm is assumed to be equipped with a break-away coupling according to SiGGTO standard. The inventory of a given section is defined as the isolatable mass within that section under normal operating conditions and in addition to t hat, the inventory released prior to t he segment isolation. The total inventory is according to Skangass 21,8 m3. /4/ Pressure used in the analysis (leak frequency and r elease rate), when bunkering is on-going, is 10 barg downstream the booster pump /4/. This is according to Skangass based on pump vendor’s specifications. Pressure used between bunkering operations is the settle out pressure of 7 barg, ref. /1/. Note that all the pipes and equipment downstream the ESD valve will be filled with LNG during the standby mode. Implication of assumption: DNV Reg. No.: 17TLT29-4 Revision No.: 1 Date :11.06.2013

Page 14 of 32


MANAGING RISK

Key influence in determining the likelihood of flash fire, pool fire and explosion hazards and the extent of each (i.e. time of ignition relative to size of cloud). Note, however, that the overall effect is that there are a significant number of very low ignition probabilities. Overall effect is a key influence on the risks, but not sensitive to any particular ignition source. Reference: /1/ LNG Bunkering of Fjordline ferries Project Design Basis, 17.02.2012, Draft Version /2/ Cryonorm Project BV documents, P&IDs number 1301-1100-100, sheets TA01, TB01, TC01, received 06.05.2013 /3/ Suggested pipeline_trase_24.2.2012.pdf /4/ Eivind Anfindsen, Skangass, 06.05.2013 Sign: J-B, Berthomieu Date: 21.05.2013 Prepared by: Internal Verification:

Sign:

Date:

Sign:

Date:



Page 15 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Escape and Evacuation of Passengers and Personnel

Assumption No.:

Revision: 1

Specifications:

It is assumed that the escape and evacuation of passengers and personnel are following the LNG plan evacuation of Fjordline /1/


The above assumptions each have a qualitative influence on risks to 1st, 2nd and 3rd parties. Reference: /1/ LNG Bunkering of Fjordline ferries Project Design Basis 17.02.2012, Draft Version Sign: J-B, Berthomieu Date: 21.05.2013 Prepared by: Internal Verification:

Sign:

Date:

Sign:

Date:



Page 16 of 32


MANAGING RISK


Date: 20.05.2013

Category:

1-C Representative Scenario Assumptions

Subject:

Release Location / Height

Assumption No.:

Revision: 1

Specifications: Release location for each section is derived from the plot plan/drawings of the respective area. The location is generally selected as that of the vessel containing the main inventory of the section or, where a number of vessels apply, as the centre of the section.

The representative release height for the plant is 1.5 meter. The representative release height for the underground pipelines is 0 meter. The representative release height for the bunkering station is 1 meter.

Implication of assumption: Dispersion is based on those inputs, and location and height contributes significantly for the consequence modelling. Reference: Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 17 of 32


MANAGING RISK


Date: 20.05.2013

Category:

2-C Representative Scenario Assumptions

Subject:

Release Sizes

Assumption No.:

Revision: 1

Specifications:

To define the hazardous release events applying to each Process Accident release scenario (QRA section), representative hole sizes are modelled. The selection of the hole sizes is made based on the need of defining different leak categories, and for each leak size a leak rate is associated. It depends on how refined the assessment can be, meaning that a s more leak categories chosen, a better di stribution of the frequency per leak size is obtained. Based on previous experience, Skangass proposes the selection of three different leak cat egories ranging from the following sizes: Range of Leak Sizes: Small 1 to 10 mm Medium 10 to 50 mm Large greater than 50 mm

Representative Hole Size / Equivalent Diameter: To be calculated by LEAK To be calculated by LEAK To be calculated by LEAK

Skangass assessed that the maximum volume LNG released due to leak from the loading arm was estimated to be 3 m3 /1/. Cryonorm assessed that the inventory in the system is 21.8 m3. /2/ Skangass assessed that the main LNG line up to the ESV valve at the ferry terminal jetty is 15 m3. /3/ Implication of assumption: The release size taken as representative is a key factor in the release parameters and subsequent consequences in each case. However, the use of representative releases is inherent in QRA and the frequencies are assigned according to each of the defined leak size ranges, such that the overallrisks should not be sensitive to the specific values selected. Nevertheless, the representative nature of each release size should be recognised. Reference: /1/ Assumption 4-C /2/ Email received from Eivind Anfindsen 08.05.2013 /3/ Email from Eivind Anfindsen received 06.05.2013 Sign: J-B, Berthomieu Prepared by: Internal Verification:

Sign:

Date: 21.05.2013 Date:

Sign:

Date:



Page 18 of 32


MANAGING RISK


Date: 20.05.2013

Category:

3-C Frequency Analysis Assumptions

Subject:

Leak frequencies

Assumption No.:

Revision: 1

Specifications: The generic failure data used as the basis of the frequency analysis of valves, flanges and pipes is the UK HSE’s Hydr ocarbon Release Database from 2010, or HCRD 2010 /1/, /2/. Although the lea k frequency data refer to offshore leaks, it is commonly applied for onshore installations. The leak frequencies for LNG piping are considered as having a fail ure frequency of 10% of regular process pipes.

Equipment count is based on P&ID (Cryonorm Project BV documents, P&IDs number 1301-1100100, sheets TA01, TB01, TC01), received 06.05.2013. Equipment downstream ESV valve for ferry bunkering is assumed to only be in use during bunkering. Loading arm leak frequency is calculated based on ACDS data covering both connection failures and ranging failures (leading to disconnection) /2/. The ACDS data is considered to be the most representative data for liquefied gas loading arms but can be considered a conservative estimate for LNG. The table below gives a breakdown of contributors to the loading arm failure frequency from ACDS data, predicted by DNV /2/: Cause / Type of failure

Connection Failures

Ranging Failures

Failure Frequency (per visit)

Failure of arm Failure of quick release connection Failure of ship's pipework Operator error Mooring fault

5.7E-05 5.7E-06 6.1E-06 6.1E-06 6.7E-07

Passing ships

2.3E-07

All

7.6E-05

Leak frequency per visit in the ACDS data /2/ is for filling of LNG tankers, which typically lasts for 18-24 hours, whereas the ferry bunkering duration is significantly shorter, see Assumption no. 1-A and 1-D. The total generic frequency above is this thus reduced accordingly. It should also be noted that the generic frequency data is not modified to account for dropped objects, this should be considered. The generic data i ncludes leaks from all causes, including dropped objects, such that additional dropped object risks should only be included where identified as a particular hazard or potential leak cause. (The passengers have access to the sun deck, however due to the design of the ship it is assumed that it will not be possible to drop any objects from the deck and onto the bunkering station). Implication of assumption: Key influence on the risks (i.e. risk is directly proportional to frequency).


Page 19 of 32


MANAGING RISK

Reference: /1/ HSE, 2010. Offshore Hydrocarbon Release Statistics, 2010 (until march 2010) /2/ DNV Guideline 16, LNG QRA Guideline, 09.11.2011 Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 20 of 32


MANAGING RISK


Date: 20.05.2013

Category:

4-C Event Tree Modelling Assumptions

Subject:

Detection and Isolation Times

Assumption No.:

Revision: 1

Specifications: The times required to detect a release and then to initiate isolation are summarised in the tables below, which give the representative times assumed for Process Barge and ‘Other’ process events, respectively.

F&G Detection

The fire and gas detection depends on the location and magnitude of the event, the number, location of detectors and their PFD (probability of failure on demand). However, the basic design of the LNG bunkering terminal is considered to have enough gas detectors. The F&G system is automatic activated upon gas detection.  

Automatic shutdown of the ESV valve at the bottom of the loading arm: Manual activation of the emergency shutdown and isolation push-buttons by t he operator in the CCR. The F&G detection system is the basis of ESD duration time.

ESD system – release duration

The initial release rate [in kg/s] is calculated within the PHAST RISK discharge model and set constant during the representative release duration. In reality, the internal pressure is reduced and this reduces the release rate (please refer to assumption 11-C), meaning that the release r ate should drop in time. Larger release rates have shorter duration than smaller scenarios for the same segment, as the inventory after the ESD closure is the same. At the jetty, DNV recommends to use ESD total time of 90 seconds when the operation is continuously supervised by operators (60 seconds for detection and initiati on, 30seconds for isolation). The main study will be based on this time at the jetty. For the other areas, the detection and response time are in compliance with the table provided by Skangass below. The line will constantly have a flow of 330 m3/h during bunkering operation and no fl ow during no bunkering operation. The line will then be liquid filled with stagnant LNG.


Page 21 of 32


MANAGING RISK

Representative detection and response times

Based on the experience of Skangass and the use of fast responsive gas detectors, ensuring automatic closure of ESV upon detection, the following response times have been e stimated: Response Time (min) Leak Size Small (<10 mm) Medium (10-50 mm) Large (>50 mm)

Detection 2 0.2 0.2

Isolation 2 0.3 0.3

Cumulative Time to Initiation (min) Isolation 4 0.5 0.5

The ESV valve is based on Skangass input, able to be closed in 6 seconds. /1/ Note: Skangass will use assumptions for response times as requirements for selection of designer/vendors. DNV recommends that equipment is qualified for compliance with these requirements. Implication of assumption:

The detection and isolation assumptions are key influences on t he release duration and impact on the selection of representative release rates. On balance, any specific inventory assumption will have a limited influence on the overall risks, although the inventory is a key parameter with respect to the detailed modelling of each scenario. Reference: /1/ Eivind Anfindsen email received on the 06.05.2013 Sign: J-B, Berthomieu Date: 21.05.2013 Prepared by: Sign: Date: Internal Verification: Comment from Skangass AS: Approved by Skangass AS:

Sign:

Date:


Page 22 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Isolation Failure

Assumption No.:

Revision: 1

Specifications: To account for the possibility of failure to isolate occurring either due to failure of the relevant ESDs (PESD) or due to human error (Phuman), the probability of isolation failure is determined as:

P isolation failure = 1 – (1-P human )*(1-P ESD ) Where: N

P ESD = 1 – (1-PFD ESD ) And:

PFDESD is the probability of failure on demand of the ESD(s)., and is t he sum of technical failure rate of logic (0.15%) , valve (2% to 5%) and actuator (0.5%), ref. A and B. N is the number of ESDs required for isolation, and On average 2 valves are assumed to be required to isolate a section, hence N = 2. Phuman = Probability of human failure, and is set to 10%

The general rule-set adopted is that two ESD valves are required for isolation of a section. For liquid and gas sections, a probability of failure on demand of 2% is assumed. As a result, the probability of isolation failure applied within the study is calculated as follows: 

Gas QRA Sections: Pisolation failure = 0.15



Liquid QRA Sections:

Pisolation failure = 0.15

Reference: A: TD0096, Akseptkriterier og beskrivelser av svikt for utvalgte sikkerhetskomponenter (DRAFT) B: Anbefalte sviktrater – risikoanalyser, Skangass AS 29.03.05. Implication of assumption:

The probability of isolation failure has a key influence on the frequency of release events that have sufficient duration to lead to escalation. Reference: Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 23 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Immediate Ignition Probability

Assumption No.:

Revision: 1

Specifications:

The probability of immediate ignition is derived as a function of the release rate and release phase using the framework set out below. This immediate ignition probability model is the same as derived for the corresponding Kårstø QRA study conducted by DNV /1/. It should be noted that the basis of the derived ignition probabilities is the energy associated with each release. As such it is important to note that the “liquid” category, in this context, corresponds to the phase of the material at standard pressure and temperature. Hence, in this study, only the condensate release scenarios are treated as liquid in terms of ignition probability (i.e. all other liquid releases are actually gas under cryogenic, or pressurised, conditions). Leak Size Category Small Medium Large

Size Interval, Release Rate (kg/s) Gas Liquid <1 < 1.2 1 – 10 1.2 - 25 > 10 > 25

Immediate Ignition Probability 0.01% 0.1% 1%


The immediate ignition probability has a direct influence on the risks associated with jet and pool fire risks to personnel (and to assets), which contribute around a quarter of the overall risks. Reference: /1/ “Kårstø Plant”. DNV Report No. 98-3090, Rev. 01. Sign: J-B, Berthomieu Date: 21.05.2013 Prepared by: Sign: Date: Internal Verification: Comment from Skangass AS: Approved by Skangass AS:

Sign:

Date:


Page 24 of 32


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Event Tree Framework

Assumption No.:

Revision: 1

Specifications: Figure 4 shows the framework for the modelling of a release within the PHAST RISK risk model, as summarised below:  







Immediate ignition has a defined probability for each release, detailed in Assumption 6-C. Given that immediate ignition occurs, the majority of release scenarios will be modelled as a jet fire, for gas releases. Where rainout occurs (i.e. where some liquid is present in the release) a similar event tree applies where the equivalent outcome will be a pool fire (liquid only), or both pool and jet fires (where liquid rains out from the initial discharge). However, the event tree structure enables a proportion of short duration releases (defined as less than 20 seconds, in this study) to be modelled separately. The event tree enables the user to define the proportion of these short duration events that are fireballs, flash fires or explosions (conditional probabilities A, B and C in Figure 4, which are defined in Assumption 8-C). As above, where liquid is present, the event tree enables pool fires to be either neglected, modelled as the only outcome, or modelled in addition to the gas impacts (where the latter option is applied within this study). Delayed ignition is calculated within the risk model for each release, as described in the Appendix A assumptions. Where delayed ignition occurs, the outcome is split into flash fire and explosion scenarios (conditional probabilities D and E in Figure 4, which are defined in Assumption 8-C). This applies equally to vapour clouds arising from gas releases or clouds flashed from liquid releases, where delayed ignition of liquid releases will have an additional (“late”) pool fire outcome.


The event tree framework is a key aspect of the QRA model, although the main influence on the risk results is the probabilities applied within the framework, as described elsewhere within this appendix. Reference: Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 25 of 32


MANAGING RISK

Attachment to Assumption 7-C Source

Immediate Ignition ?

Evaluate short release ?

Short release ?

Y

Y

Y

Defined for each release

Releas e

User defined (p=1)

Consequence (model used) ?

y p=F Determined by PHAST RISK according to release properties

p=A

Firebal l

n y p=F

p=B

Flash Fire

n y p=F

p=C N

Residua l Pool Fire?

Explosi on

n

N Jet Fire

Delayed Ignition ?

N

Consequence (model used) ?

Y

y p=G Determined by PHAST RISK according to dispersion, duration, ignition sources

p= D

Flash Fire

y p=G p=E

Explosi on

N No ignitio n Figure 4 Example risk model event tree structure


Page 26 of 32

N

N


MANAGING RISK


Date: 20.05.2013

Category:


Subject:

Event Tree Probabilities

Assumption No.:

Revision: 1

Specifications:

The development of a release is largely defined by the stage at which ignition occurs, where the immediate and delayed ignition parameters are described elsewhere within this a ppendix. Immediate ignited releases with pool formation are likely to develop pool fires. Thus in case of a release with liquid drop which ignites immediately, the pool fire is considered as one of the consequences. 

Process areas: % fireball, % flash fires, % explosion, % pool fires (i.e. A=, B=, C=, F= 1)

Delayed ignition events are split between flash fire and explosion outcomes (probabilities D and E, respectively, in Figure 4) which may result or not in pool fires as follows: 

Process areas: % flash fire, % explosion, % pool fires (i.e. D=, E=, G= 0.15)

Those probabilities will come from the explosion assessment. Implication of assumption:

Short duration events (in the context of PHAST RISK, i.e. less than 20 s) are very limited, and the difference in risks to personnel associated with flash fire and explosion events are not major. Hence, the above values do not have a major influence on the overall risks, although the influence will accumulate for all release scenarios for which each parameter set is applied. Reference: Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 27 of 32


MANAGING RISK


Date: 20.05.2013

Category:

9-C Consequence Modelling Assumptions

Subject:

Dispersion Parameters

Assumption No.:

Revision: 1

Specifications: The key inputs into the dispersion modelling are the release / discharge parameters. Additional assumptions that influence the dispersion are: 





Weather. The wind speed, direction and stability have a key influence on the downwind dispersion distance of vapour clouds (and to a lesser extent the radiation contours associated with fires). These parameters are defined within Assumption 2-A. Ambient conditions. The air and surface temperature, together with other local parameters such as atmospheric pressure and relative humidity, will also have an influence on dispersion. These parameters are defined within Assumption 3-C. Congestion / impingement. The dispersion parameters are derived for an idealised release, with no consideration of potential obstructions. It is li kely that a release will impinge on equipment; therefore the release is treated as impinged releases.


The above assumptions each have key influences on the consequence results. Reference: Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 28 of 32


MANAGING RISK


Date: 20.05.2013

Category:

10-C Consequence Modelling Assumptions

Subject:

Consequence Modelling Parameters

Assumption No.:

Revision: 1

Specifications: The key inputs into the consequence modelling are taken directly from the discharge and dispersion modelling inputs and results. A wide range of additional parameters are applied within the models, where in general the widely accepted PHAST Risk default values are applied. The key parameters that are specific to the above consequence models are summarised below.  

Jet fire – maximum surface emissive power (SEP): 250 kW/m2 Jet fire – rate modification factor (the mass of vapour that remains in cloud calculated by PHAST is multiplied by this factor – determines the proportion of the liquid fraction that contributes to the jet fire for 2-phase jets): 3



Pool fire – minimum duration – 10 seconds



Fireball / BLEVE – maximum SEP: 300 kW/m2





Fireball / BLEVE – mass modification factor (the mass of vapour that remains in cloud calculated by PHAST is multiplied by this factor – determines the proportion of the liquid fraction that contributes to the fireball/BLEVE): 3 Flash fire – The size is calculated based on mass between LFL and UFL (for ignition probabilities, the 50% LFL is used)



Explosion – minimum explosion energy: 5 x 106 kJ



Explosion – explosion efficiency: 10%


The above assumptions each have key influences on the consequence results. Reference: Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 29 of 32


MANAGING RISK

LNG Bunkering Terminal, Risavika Harbour Category:

1-D Frequency Assumptions

Subject:

Bunkering Frequency

Assumption No.:

Date: 20.05.2013 Revision: 1

Specifications:

Planned frequency of LNG bunkering is 1 per day. Bunkering duration is 1 hour for LNG cool down and transfer and ½ an hour for connection and disconnection of loading arm to ship manifold. The ferry is scheduled to arrive at 8 in the afternoon. Fjordline /2/ has estimated the following ferry arrival delays: 

Delay less than 3 hours: 4 %



Delay more than 3 hours: 2 %

According to Fjordline /2/ 1 % of scheduled ferry travels are cancelled.


The above assumptions each have key influences on t he frequency estimates. Reference: /1/ LNG Bunkering of Fjordline ferries Project Design Basis 17.02.2012, Draft Version /2/ Email from Larsen/Fjordline to Gautestad/Skangass, 24.02.2012, subject: “SV: Forsink elser av anløp pga vær etc.” Sign: J-B, Berthomieu Date: 21.05.2013 Prepared by: Internal Verification:

Sign:

Date:

Sign:

Date:



Page 30 of 32


MANAGING RISK


Date: 20.05.2013

Category:

1-H Assumptions

Subject:

End Point (Impact) and Vulnerability (Fatality) Criteria

Assumption No.:

Revision: 1

Specifications:

Human Impact

(End Point) Impact Criteria

Event Location

Explosion (Heavy Blast)

Heavy blast damage – 350 mbarg

Explosion (Light Blast)

Vulnerability Parameters

Notes

Outdoor

Indoor

Other areas

0.3

1

Default – building collapse is main impact potential

Light blast damage – 100 mbarg

Other areas

0.1

0.3

Default – building damage is main impact potential

Flash fire

50% LFL

All

1

0.1

Fatality assumed if outdoors; shielded if indoors

Fireball / BLEVE

250 kJ/m2 thermal dose

All

0.7

0.1

High fraction killed if outdoors; shielded if indoors

Storage & Loading

0.5

0.1

Jet fire

12.5 kW/m2 radiation level

LP – spray rather than jet fires; open area - good escape prospect

Process Areas

0.7

0.1

High fraction killed if outdoors, but some escape possible

Storage & Loading

0.5

0.1

Open area - good escape prospects for pool fires

Process Areas

0.5

0.1

Open area - good escape prospects for pool fires

12.5 kW/m2 radiation level

Pool fire


The risks are directly influenced by the impact and fatality assumptions, which quantify the severity of the consequences. The above assumptions include some allowance for differing escape characteristics in different areas of the facility, but remain consistent with established, conservative best-practice. Reference:

-

DNV expert judgement – using PHAST Risk defaults and DNV Technical data


Page 31 of 32


MANAGING RISK

Prepared by: Internal Verification:


Date: 21.05.2013 Date:

Sign:

Date:



Page 32 of 32

DNV Assumptions.pdf

Recommend Documents