RR402 Natural ventilation of offshore modules [5.6MB]

HSEHealth & Safety

Executive

Natural ventilation of offshore modules

Prepared by the Health and Safety Laboratoryfor the Health and Safety Executive 2005

RESEARCH REPORT 402

HSEHealth & Safety

Executive

Natural ventilation of offshore modules

C J Saunders BSc & M J Ivings BSc PhD Health and Safety Laboratory

Broad Lane Sheffield

S3 7HQ

Natural ventilation is a common method for mitigating the hazard posed by gas and vapour leaks on offshore platforms. Openings in wind walls and doors allow the wind to blow through a module and hence the ventilation is not generally dependent on the operation of any devices such as mechanical fans.

Information on how to optimise natural ventilation of offshore platforms is available in a number of documents. Guidance aimed specifically at the offshore industry can be found in the International Standard IS 15138 and is also discussed as a topic, usually within a dedicated chapter, in most other ‘Area classification’ documents. The one most relevant to the UK offshore industry is the ‘Area classification for Petroleum Industries. Part 15’ (IP 15) published by the Institute of Petroleum. This guidance is based on the European Standard BS EN60079-10. IP 15 describes the assessment of areas based upon different levels of ventilation, which range from ‘open’, ‘sheltered’ to ‘enclosed’. The level of ventilation within an area, which is described by either air velocities or air change rates, greatly affects area assessment.

This report investigates the effectiveness of natural ventilation of offshore platforms, focusing on the non-uniformity of the ventilation within a module and its dependence on the wind conditions. This has been achieved by making detailed experimental measurements on three typical offshore platforms and applying Computational Fluid Dynamics (CFD) to one of these.

This report and the work it describes were funded by the Health and Safety Executive (HSE). Its contents, including any opinions and/or conclusions expressed, are those of the authors alone and do not necessarily reflect HSE policy.

HSE BOOKS

© Crown copyright 2005

First published 2005

All rights reserved. No part of this publication may bereproduced, stored in a retrieval system, or transmitted inany form or by any means (electronic, mechanical,photocopying, recording or otherwise) without the priorwritten permission of the copyright owner.

Applications for reproduction should be made in writing to: Licensing Division, Her Majesty's Stationery Office, St Clements House, 2-16 Colegate, Norwich NR3 1BQ or by e-mail to [email protected]

ii

ACKNOWLEDGMENTS

The authors would like to express their thanks to BP Exploration, Conoco (UK) Ltd. and Phillips Production Company Ltd. for their co-operation during this project. Also thanks to HSL staff, Mr A E Johnson and Mr B Fletcher who helped to carry out the offshore experimental measurements.

�iii

CONTENTS

1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 AREA CLASSIFICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.1 IP 15 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.2 API RP500 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

2.3 OTHER RELEVANT STANDARDS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 EXPERIMENTAL MEASUREMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

3.2 INSTALLATION C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3.2.1 Velocity measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

3.2.2 Tracer gas tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

4 EXPERIMENTAL RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4.1 VELOCITY MEASUREMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

4.2 TRACER GAS TESTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

5 COMPUTATIONAL FLUID DYNAMICS MODELLING . . . . . . 14

5.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

5.2 VALIDATION CALCULATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

5.3 GENERAL WIND CONDITIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

5.4 GAS RELEASE CALCULATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

6 CFD MODEL DETAILS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

6.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

6.2 EXTERNAL FLOW CALCULATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

6.3 INTERNAL FLOW CALCULATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

6.3.1 High pressure jet releases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

7 CFD RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

7.1 EXTERNAL FLOW CALCULATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

7.2 CFD MODEL VALIDATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

7.2.1 Validation Case V1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22



7.2.4 Sensitivity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

7.3 GENERAL WIND CONDITIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

v

7.3.1 Ventilation field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

7.3.2 Air change rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

7.3.3 Tracer gas calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

7.4 JET RELEASE PREDICTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

8 PURGE TIME CALCULATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

9 REMEDIAL MEASURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

9.1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

9.2 CFD MODELLING APPROACH . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

9.3 CFD RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

9.3.1 Tracer gas simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

9.3.2 Purge time calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

9.3.3 Gas release simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

10 CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

10.1 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

10.2 VENTILATION ASSESSMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

10.3 FURTHER WORK . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

11 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

A1 FIGURES: CFD RESULTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

A1.1 EXTERNAL FLOW CALCULATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

A1.2 CFD MODEL VALIDATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

A1.2.1 Validation case V1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40



A1.2.4 Sensitivity tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

A1.3 GENERAL WIND CONDITIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

A1.3.1 Low wind speed cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

A1.3.2 High wind speed cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

A1.3.3 Tracer gas calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

A1.4 JET RELEASE PREDICTIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

A2 FIGURES: REMEDIAL MEASURES . . . . . . . . . . . . . . . . . . . . . . . . 80

vi

EXECUTIVE SUMMARY

Background

Natural ventilation is a common method for mitigating the hazard posed by gas and vapour leaks on offshore platforms. Openings in wind walls and doors allow the wind to blow through a module and hence the ventilation is not generally dependent on the operation of any devices such as mechanical fans.

Information on how to optimise natural ventilation of offshore platforms is available in a number of documents. Guidance aimed specifically at the offshore industry can be found in the International Standard IS 15138 and is also discussed as a topic, usually within a dedicated chapter, in most other ‘Area classification’ documents. The one most relevant to the UK offshore industry is the ‘Area classification for Petroleum Industries. Part 15’ (IP 15) published by the Institute of Petroleum. This guidance is based on the European Standard BS EN60079-10. IP 15 describes the assessment of areas based upon different levels of ventilation, which range from ‘open’, ‘sheltered’ to ‘enclosed’. The level of ventilation within an area, which is described by either air velocities or air change rates, greatly affects area assessment.

This report investigates the effectiveness of natural ventilation of offshore platforms, focusing on the non-uniformity of the ventilation within a module and its dependence on the wind conditions. This has been achieved by making detailed experimental measurements on three typical offshore platforms and applying Computational Fluid Dynamics (CFD) to one of these.

The completion of this project has led to the production of two additional reports that describe the experimental measurements in detail on the two platforms where CFD was not used.

Objectives

The main objectives of this project are:

(1) To gain an improved understanding of the extent of the problem of poorly ventilated regions on naturally ventilated offshore modules.

(2) To correlate localised velocity / ventilation rate measurements and CFD predictions against wind conditions.

(3) To assess the feasibility of using tracer gas techniques to measure ventilation rates in naturally ventilated offshore modules.

(4) To assess and validate the use of CFD as a technique for predicting the effectiveness of natural ventilation by comparison against detailed on-site measurement.

(5) To use CFD to investigate the effectiveness of ventilation under different weather conditions for diluting and dispersing flammable gases produced from realistic gas releases.

(6) To evaluate the effectiveness of remedial measures to mitigate local poorly ventilated areas.

(7) To assess the practicality of complying with guidance for natural ventilation and assessing the likelihood of this leading to effective ventilation. en

vii

Main Findings

(1) Velocity measurements on three separate offshore platforms and a detailed CFD study on one platform have shown that the effectiveness of natural ventilation is highly dependent on the external wind conditions and is unlikely to be uniform throughout the module.

(2) The flow paths through a naturally ventilated module have been found to be very complex and often non-intuitive. It is therefore imperative that ventilation assessments are based on a detailed programme of local velocity measurements carried out over a range of wind conditions, preferably in conjunction with CFD predictions, to provide a clear picture of the effectiveness of the ventilation.

(3) Measurements made on the three offshore platforms have shown that natural ventilation of a deck is affected by the arrangement and amount of equipment on a deck and also by the number of openings and their position at the perimeter of the deck. A further influence on the natural ventilation within a module is the sheltering offered by adjoining platforms.

(4) The air movement on the three platforms studied displayed different flow regimes:

a) Under certain wind conditions, air flow patterns on the cellar deck of Installation A, which was the most open of the three platforms, could loosely be described as ‘plug flow’ . This was where the air entered the module via one fully open face and was swept across the deck before leaving via the opposite open face. This type of ventilation would be described as ‘desirable’ . When the wind direction was parallel to the open sides the air flow patterns on the deck were complex and highly three dimensional, illustrating the effect of wind direction on flow patterns within modules. Even though this is a relatively open platform, all measured air velocities were found to be less than the prevailing wind speed.

c) Measurements on the cellar deck of Installation B, which had partial wind walls fitted to all four sides, highlighted a degree of short circuiting. Also the sheltering effect of an adjoining platform was thought likely to reduce the ventilation rate through the deck.

d) Installation C, which was the most enclosed of the three platforms studied, was surrounded on three sides by adjoining platforms. This platform displayed complex flow patterns which were highly variable and dependent on the wind speed and direction. This was the platform where CFD was applied.

(5) CFD has been shown to be a powerful tool for the prediction of the effectiveness of natural ventilation on offshore platforms. The present CFD model has been validated against a detailed set of offshore measurements and the predictions have been shown to be in good agreement for a range of wind conditions. We can expect that, with the increases in computer power and availability of CAD data for platforms, the use of CFD modelling in ventilation assessments will increase.

(6) CFD predictions have shown that the rate of gas build-up from realistic gas leaks on an offshore platform is highly dependent on the wind conditions and hence the effectiveness of the ventilation. The rate at which gas leaks are diluted and dispersed following a leak is also strongly dependent on the wind conditions.

viii

(7) It has been found that, even for a platform that was well ventilated, large gas leak rates will lead to the formation of large explosive gas clouds. However, once the leak has stopped, good ventilation can remove the flammable gas from the platform efficiently. For smaller releases poor ventilation will allow significant gas clouds to form above the lower flammability limit. Therefore effective ventilation can be expected to remove large volumes of flammable gas after a large leak or limit the build-up of flammable gas clouds for smaller leaks.

(8) It should be possible to measure ventilation rates on offshore platforms, for certain platform configurations/geometries and under certain weather conditions using either velocity measurements or tracer gas techniques. However, where the ventilation openings are positioned so as to create complex flow patterns inside modules, it is unlikely that accurate measurements would be possible.

(9) Whilst tracer gas measurements are difficult to carry out offshore, CFD modelling of tracer gas tests has proved to be useful and has helped to identify the least well ventilated areas. These predictions have been used to calculate purge times. These have proved to be useful for comparing the level of ventilation in different areas under arbitrary, or all, wind conditions.

(10) If local poorly ventilated areas are identified on a module it should be possible to improve the ventilation by a number of means: introduction of fans, nozzles, air movers, removal of wind walls or replacement of floor areas with open grating. However, these measures need to be designed and assessed in each case to ensure that they will actually lead to an improvement in the ventilation.

(11) Local areas of poor ventilation can often be found on a platform even where the remainder of the platform is well ventilated. This therefore suggests that determining an air change rate for a module is not a clear indication of whether or not the ventilation within a module is effective.

(12) Complying with guidance on natural ventilation that requires a minimum air change rate is not straightforward due to the difficulties in measuring the ventilation rate in modules. It has also become apparent that even if these conditions are adhered to, that this does not necessarily lead to effective ventilation.

(13) The combination of offshore velocity measurements and CFD modelling has proven to be an invaluable and practical tool for evaluating the effectiveness of ventilation on offshore modules.

Main Recommendations

(1) The current guidance is difficult to implement and its usefulness is questioned. HSE should consider influencing improvements to current guidance. Possible improvements could include the replacement of air change rates with volumetric flow rates that are dependent on the size of the potential leak, not the size of the deck in which it may occur. Alternatively ventilation rates could be completely removed from guidance and replaced with acceptable geometries/open areas which give acceptable ventilation. A further possibility is to evaluate the effectiveness of the ventilation by its ability to dilute and disperse a given range of gas leak rates.

ix

(2) One question in particular needs to be addressed before any guidance is amended. That is, “For what range of gas leak sizes can natural ventilation be expected to provide mitigation against gas build up?” This is not clear in the current guidance.

(3) Assessing an area as ‘open’ based on the measurement of air velocities is not ideal. However, this is the approach adopted in IP 15. Offshore velocity measurements have shown that if guidance is followed, areas with no wind walls present on Installation A should be classified as ‘sheltered’ . As any equipment will cause an obstruction to the flow, it is possible that there are few areas on offshore modules that could be classified as ‘open’.

(4) CFD modelling has been shown to be a powerful tool that can be used in ventilation assessments. However, it is important that CFD results are interpreted with caution and the models should be validated against offshore velocity measurements to establish confidence in the predictions. Ventilation assessments based on CFD predictions alone must therefore be treated with caution.

(5) It has been shown that due to complex geometries and the influence of adjoining platforms that flow paths through modules and thus ventilation effectiveness is difficult to anticipate and can be counter-intuitive. Offshore measurements, ideally supported by CFD modelling, should be undertaken in order to demonstrate the adequacy of the ventilation.

(6) Ventilation assessments could usefully be used to optimise the positioning of gas detectors.

(7) Further work is needed to assess the relative effectiveness of different methods for improving the ventilation in local poorly ventilated areas. Possible methods include the installation of air movers, fans and nozzles, removal of wind walls and replacing sections of floor areas with open grating. A cost-benefit analysis is needed.

x

1 INTRODUCTION

One of the most significant hazards faced by offshore workers is the risk of an explosion resulting from the ignition of a cloud of flammable gas or vapour following a leak. Ventilation can help to maintain safe working conditions on offshore modules by diluting and dispersing the gases/vapours from such leaks. At best the ventilation will dilute the gas down to a level such that its concentration everywhere is below the lower explosive limit. Under other conditions, e.g. for large gas leaks, the ventilation can help to reduce the size of the flammable gas cloud and also remove it quickly from the module once the gas leak has stopped.

There are two main methods for ventilating a space: mechanical (or forced) and natural. Mechanical ventilation is typically used in enclosed spaces. Fans are used to supply and/or extract air from the space and are usually connected to ductwork to form the ventilation system. The inlet(s) and outlet(s) can be positioned strategically to produce effective ventilation throughout the space, ideally eliminating areas which are poorly ventilated. As fans are used to move the air, mechanical ventilation is largely independent of weather conditions.

Natural ventilation is achieved by making openings in the walls, floors and ceiling of the structure. The size and position of any opening should be planned at the design stage, so as to strike a balance between the required airflow rate through the platform and the protection of the workers and equipment from the, often harsh, environment. Natural ventilation is widely used offshore because of its simplicity and it is also considerably less expensive than mechanical ventilation. The International Standard IS 15138 (2000), provides guidance for design, testing and commissioning of Heating and Ventilation and Air Conditioning (HVAC) and pressurised systems and equipment on all offshore production installations. It states that natural ventilation is preferred over mechanical (or artificial) ventilation where practical. This is mainly because it is available throughout any emergency and does not rely on the operation of equipment. The main disadvantage of natural ventilation is that, by its nature, its effectiveness is variable and strongly dependent on weather conditions.

During periods of low wind speeds or unfavourable wind direction, the flow of air through certain areas, particularly in congested regions, may not be as high as expected. This can lead to the creation of poorly ventilated areas that may pose a potential explosion hazard in the event of a gas leak.

To understand and quantify the extent of the problem on naturally ventilated modules, an approach combining both experimental measurements, which were carried out on three platforms, and Computational Fluid Dynamics(CFD), applied to one platform, has been undertaken.

Velocity measurements were made on all three platforms. The instruments used measured all three components of the flow, so that at each position the air speed and direction was measured. On two of the platforms, tracer gas measurements were carried out with the aim of measuring the ventilation rates. CFD modelling has been used to make predictions of the ventilation flow for one of the three platforms and the results have been validated against the experimental measurements. The validated CFD model then enabled predictions to be made of the ventilation flow for a set of general wind conditions. Realistic gas releases on the platform have then been modelled to study the interaction of a high pressure gas leak with the ventilation flow under different weather conditions. The CFD model has also been used to make some initial investigations into the

1

effectiveness of remedial measures for improving the ventilation locally in areas that may be poorly ventilated.

The remainder of this report is structured as follows: Section 2 outlines the guidance and standards relevant to the ventilation of offshore modules. A description of the experimental measurements that have been carried out, and the methodology used, are described in Section 3 and the results are given in Section 4. The CFD modelling carried out is described in Section 5. The technical details of the modelling is described in Section 6 and the results are presented and discussed in Section 7. Section 8 describes a method for assessing the effectiveness of the ventilation locally based on the calculation of the ‘purge time’ , which is defined in the report. Methods for improving the ventilation are discussed in Section 9 and one method is investigated using CFD modelling. The conclusions of the work are summarised in Section 10. The report also includes two appendices which include more detail on both the CFD results and remedial measures.

2

2 AREA CLASSIFICATION

Natural ventilation on offshore modules is of primary importance when determining hazardous zones, and its role is considered by most International Codes and Standards. For this reason guidance on natural ventilation for offshore platforms is found in area classification codes. The areas where measurements were made on two of the three platforms visited were classified according to the Institute of Petroleum, Area Classification for Petroleum Installations, Part 15 (IP 15), (1990). The third platform visited was classified in accordance with the American Petroleum Institute (API) RP500 (1997).

2.1 IP 15

IP 15 provides a guide for the classification of the areas around equipment handling or storing flammable fluids in order to provide a basis for the correct selection and location of fixed electrical equipment in those areas. These areas can either be classified as hazardous or non-hazardous. Hazardous area are the subdivided into ‘zones’ . Ventilation plays an important role when determining the extent of these zones and is covered in chapter 6 of the code. Chapter 6 is based upon the framework set out in BS EN 600079-10 (1996) and describes ventilation as being of different types and levels. It even has a ‘finding tree’ to assist in the assessment. The different types are defined as follows:

(i) Open area; “an area in an open air situation without stagnant areas where vapours are readily dispersed by wind and natural convection. Typically air velocities should rarely be less than 0.5 ms-1 and should frequently be above 2 ms-1”.

(ii) Sheltered area; “an area within, or adjoining, an open area (which may include a partially open building or structure) where, owing to obstruction, natural ventilation may be less than in a true open area, and this may enlarge the extent of the hazard zone”.

(iii) Enclosed area; “an enclosed area is any building, room or enclosed space within which, in the absence of artificial ventilation, the ventilation will be limited and any flammable atmosphere will not disperse naturally”.

Sheltered and enclosed areas can have either adequate or inadequate ventilation. IP 15 defines adequate ventilation as the achievement of a uniform ventilation rate of at least 12 air changes an hour with no stagnant areas.

The code gives examples of the different types of sheltered area. It also states that, if a walled building or offshore module which contains a source of release is naturally ventilated by permanent openings or louvres, a judgement needs to made to ascertain whether the openings are sufficient to give adequate ventilation (uniform rate of 12 air changes per hour). The code goes on to say that a specialist should confirm this and also confirm that there are no stagnant areas.

2.2 API RP500

API RP500 provides guidelines for classifying locations at petroleum facilities for the selection and installation of electrical equipment. A class I location is defined as a location which flammable gases or vapours are, or may be, present in the air in quantities sufficient to produce explosive or ignitable mixtures. Class I locations include the following:

3

Class I, division 1 location: A location in which ignitable concentrations of flammable gases or vapours are expected to exist under normal operating conditions or in which faulty operation of equipment or processes might simultaneously release flammable gases or vapours and also cause failure of electrical equipment.

Class I, division 2 location : (a) A location in which flammable gases or vapours may be present, but normally are confined within closed systems; (b) are prevented from accumulating by adequate ventilation; (c) or the location is adjacent to a Division 1 location from which ignitable concentrations might occasionally be communicated.

The decision to classify a location as Division 1, Division 2, or unclassified (non-hazardous location), depends in part on the degree of ventilation of the location.

API RP500 states that “ In general, a naturally ventilated location (building, room, or space) should be substantially open and free from obstruction to the natural passage of air through it, vertically and horizontally. Such locations may be roofed or partially closed on the sides, or both.” An enclosed area is defined as “ A three-dimensional space enclosed by more that two-thirds of the possible projected plane surface area and of sufficient size to allow the entry of personnel. For a typical building, this would require that more that two-thirds of the walls, ceiling, and/or floors be present.”

The document goes on to define adequate ventilation as “ ventilation that is sufficient to prevent the accumulation of significant quantities of vapour-air or gas-air mixtures in concentrations above 25 % of their lower flammable (explosive) limit” . The document describes a number of possible strategies for achieving adequate ventilation (but states that the list is not intended to be all-inclusive).

One strategy for achieving adequate ventilation in enclosed areas prescribes a volume flow rate of air throught the structure based on floor area but states the minimum number of air changes per hour should be 6. The ventilation in this case may be natural or mechanical. If the space is ventilated naturally and the ventilation adequacy is to be determined by a mathematical analysis, which is described in the document, then a safety factor of two should be used. i.e. the minimum number of air changes is increased from 6 to 12. Another method determines the adequacy of ventilation based on the percentage open area in the walls and whether or not a floor or roof is present. No mention of air speeds is made.

2.3 OTHER RELEVANT STANDARDS

There are a number of other relevant standards which cover the subject of natural ventilation and are worthy of mention. These are:

� ISO 15138 � BS EN60079-10 (upon which IP 15 is based)

The International Standard ISO 15138 provides guidance for design, testing and commissioning of HVAC and pressurised systems and equipment on all offshore production installations. It also covers, ventilation of production areas and states that natural ventilation is preferred over mechanical ventilation where practical. It also states that displacement ventilation, where the air

4

moves like a piston through a space and is spread uniformly over the whole cross section, is desirable for removing pollutants generated within the space.

The Standard defines open areas as “ an open air situation where vapours are readily dispersed by wind” and that typically velocities in such areas should be rarely less than 0.5 ms-1 and frequently be above 2 ms-1. This is similar the definition in IP 15.

For natural ventilation, the Standard notes that the distribution of air within the area/module is considered to be at least as important as the quantity of air supplied, consequently minimum ventilation rates throughout an area are required. It should be noted that the Standard is not prescriptive with respect to natural ventilation and refers to ventilation rates rather than ‘air change rates’ .

The European Standard BS EN60079-10 is concerned with the classification of hazardous areas where flammable gas or vapour risks may arise, in order to permit the proper selection and installation of apparatus for use in such areas. The Standard is not exclusive to offshore applications. It notes the two types of ventilation (natural or mechanical) but does not say which is preferred. It does however state that the most important factor is that the degree or amount of ventilation is directly related to the types of source of release and their release rates, irrespective of the type of ventilation.

For outdoor areas, it states “ the evaluation of ventilation should normally be based on an assumed minimum wind speed of 0.5 ms-1, which will be present virtually continuously. The wind speed will frequently be above 2 ms-1.” (Note, BSI have confirmed that the reference to wind speed that would be found within the open structure i.e. at the point of release of flammables; not the prevailing wind speed). Again, this is consistent with IP 15.

5

3 EXPERIMENTAL MEASUREMENTS

3.1 INTRODUCTION

Experimental measurements were made on three disparate offshore installations. All three installations afforded differing levels of weather protection to workers and equipment. The main features of the areas where measurements were made are detailed below:

� Installation A: This is an integrated platform consisting of four main areas; process, well bay, utilities and living quarters. The platform is arranged on three main levels (cellar, intermediate and weather deck levels) and afforded the lowest level of weather protection of the three platforms investigated. During the two visits to the platform, velocity measurements were made in the process and well bay areas on the cellar deck level. The process area was 24.5 m by 34.5 m by 7.8 m high and was open on three sides; the fourth side consisted of a solid fire and blast wall separating the area from the well bay. Both the floor and ceiling were solid. The well bay (19.3 m by 30.3 m by 7.8 m high) was located centrally, having blast walls on either side, with the remaining two sides open. The floor comprised of solid areas with open mesh areas around the well head valve control systems (Christmas Trees). The ceiling was solid. The classification of the areas around equipment handling or storage of flammable fluids followed the scheme specified in IP 15. At the time of this report, the area assessment, i.e. Open, sheltered or enclosed, was not known.

� Installation B: The complex consisted of three platforms, connected via bridges. Velocity measurements were made on the cellar deck of the production platform. The cellar deck was approximately 36.5 m by 18.3 m by 3.6 m high and had solid wind walls fitted to all four sides. The walls extended from the floor to a height of 2.45 m. The majority of the space above the wind walls was open. There were also a number of floor to ceiling gaps. The total open area in the four sides is approximately 40%. The deck had a solid roof and an open mesh floor. The classification of the areas around equipment handling or storage of flammable fluids followed the scheme specified in API RP500.

� Installation C: The installation consisted of five platforms, connected via bridges. Measurements were made on the cellar deck of the production platform, which was 30.25 m by 36.7 m by 9.8 m high. The cellar deck had a mezzanine level at approximately mid-height (velocity measurements were also made on this level). Solid wind walls were fitted to three sides, with gaps of 1 m and 0.35 m at the top and bottom respectively; the fourth side consisted of a solid wall. This resulted in a wall open area of approximately 20%. The deck had both a solid ceiling and floor. The classification of the areas around equipment handling or storage of flammable fluids followed the scheme specified in IP 15. At the time of this report, the area assessment, i.e. open, sheltered or enclosed, was not known.

Installation C was chosen as the main focus of the project due to its enclosed nature and the possible sheltering effects created by the adjacent platforms on three sides. This was the only platform where CFD was applied.

The measurements made on Installations A and B have been described in detail in two separate reports, IR/ECO/02/10 (2002) and IR/ECO/00/12 (2001) respectively. This report will focus on Installation C, with references made to both Installation A and B.

6

3.2 INSTALLATION C

A plan view of Installation C is shown in Figure 3.1.

N

Production platform

Platform

Accomodation platform

Figure 3.1 Plan view of Installation C showing the layout of the five interconnecting platforms

3.2.1 Velocity measurements

Between 22 and 28 measurement locations were selected on the production platform cellar deck and 16 locations on the cellar deck mezzanine level. The criteria for selection were to ensure:

(i) measurements were made in potential areas of poor ventilation,(ii) a number were selected in relatively ‘open areas’ and(iii) a relatively uniform coverage of the cellar deck level was achieved.

At each location, velocity measurements were made at 3 heights; 1 m, 2 m, and 2.75 m or 2.82 m (depending of test number). A total of 21 sets of measurements were completed, although two of the data sets were later split due to a change in wind speed and direction during the tests. The velocity measurements were made using ultrasonic anemometers. These instruments measure the time of flight of pulsed sound waves across a 14.6 cm path, from which the three orthogonal air velocity components are calculated. The resolution of the instrument is 0.01 ms-1 and it has an accuracy of 1.5 % over the range 0 to 20 ms-1, decreasing to 3 % between 20 and 60 ms-1. At each measurement position data were collected every 0.5 s for a period of 2 minutes and stored on a laptop computer. The software calculated average air velocites for each measurement position. Figure 3.2 shows typical velocity data.

7

0

1

-7

-6

-5

-4

-3

-2

-1

Air

vel

oci

ty :

ms -1

U component V component W component

0 20 40 60 80 100 120

Time : seconds

Figure 3.2 Typical velocity data showing all three components of the air flow

In order to correlate measured data with meteorological conditions, the wind speed and wind direction, measured at a height of 5m above sea level, were recorded every 30 minutes by equipment located on the standby vessel. The platform did have a working cup anemometer and a direction vane, however they were sited close to obstructions and we were informed that their readings were not accurate.

3.2.2 Tracer gas tests

The use of tracer gas techniques to measure ventilation rates is well documented (Etheridge and Sandberg (1996)). Two tracer gas techniques were used to measure (i) the air change rates using the step down method (ii) ventilation rate by the principle of dilution. Both techniques were carried out on the cellar deck of the production platform.

(i) Measurement of air change ratesA commonly used methodology for measuring the air change rate of an enclosed space is to use a tracer gas technique. A tracer gas, often sulphur hexafluoride (SF6), is released into the space and a mixing fan used to achieve a uniform gas concentration throughout. Once the release is terminated the tracer gas concentration will diminish with time as clean air enters the space. The decay rate is then measured using a calibrated gas analyser. For accurate results the ventilation rate should be uniform throughout the test. A plot of the log of the tracer gas concentration with time will then yield the air change rate of the space. This method was attempted on Installation C, however, it became apparent after a trial test that a uniform gas/air mixture would not be

achievable without first temporarily covering all large openings. This would allow the tracer to mix thoroughly with the air inside the module. Once mixed, the temporary coverings would be removed and the test started. This, however, was considered to be not practical for Installation C, but may be feasible on a platform which has adjustable louvres in the wall.

8

An alternative method of release was chosen. The tracer gas was released through a network of tubing to 4 release points. The network of tubes was designed so that the length of tubing to each release position was the same. The end of each tube was connected to the supply of an air mover. Air movers use the energy from a small volume of compressed gas to induce an air movement many times that of the supply air. These were used to help dilute the SF6 tracer gas, which has a density relative to air of 5.11. The tracer was released for 6 minutes and the gas concentration measured at two positions.

(ii) Measurement of ventilation rate (dilution measurements)The principle of dilution was used to attempt to determine the rate of airflow through the cellar deck during a period when the wind was from platform Southwest. A tracer gas (SF6) of known concentration (Cfeed) was released at a continuous rate (Qfeed) of 18 l.min-1 through a network of tubes to 4 release positions. The positions were spaced equidistantly across the South wind wall at an approximate height of 2 m. The end of each tube was connected to the supply of an air mover. Again, these were used to help dilute the tracer gas, and reduce the effect of gas density.

The concentration of SF6 (Cnorth) was measured at two positions using an infrared gas analyser at the North side of the deck.

To use this method it is assumed that:

i) the majority of the air passes through the bulk of the deck, ii) there is no transfer of air between the cellar deck and the mezzanine deck, and iii) the tracer gas is fully mixed with the air entering the well bay by the time it reaches the sample positions at the North side of the deck.

The mass flow rate of the tracer gas at the North sample positions equals the rate of the tracer gas released upstream. i.e.

�Qfeed � Cfeed = Qnorth Cnorth (3.1)

Cfeed Qnorth = Qfeed �

Cnorth (3.2)

Two tracer gas tests were carried out on the cellar deck. Test durations were approximately 28 minutes and 40 minutes.

9

4 EXPERIMENTAL RESULTS

4.1 VELOCITY MEASUREMENTS

In this section the velocity measurements are presented and discussed. In all of the 21 tests, velocity measurements were made at three heights using the ultrasonic anemometers as described in Section 3.2.1. For all tests, including the tests carried out on the mezzanine level, the lowest air velocities were measured at the highest positions (2.75 m). The ceiling height of the lower cellar deck and the mezzanine level is approximately 5m and therefore the measurements were some distance from the ceiling. The reason low velocities were measured at this height may be due to pipework etc., positioned well above head level, creating congestion.

An example of the test results are shown in Figure 4.1. The figure shows the velocity vectors at a height of 2 m above the cellar deck floor. The length of the arrow indicates the magnitude of the velocity component, whilst the angle indicates its direction. The base of the arrow locates the measurement position. The true North and the designated platform North are also shown in the figure. The average wind speed during this test, measured on the standby vessel, was 28 knots (14.4 ms-1) from a platform SSW direction. The highest measured air velocity for this height was 5.8 ms-1 and was taken in the doorway on the South East corner of the platform, which opened onto an exposed walkway. When all three measurement heights are considered for this test, 61 % of the measurements are found to be less than 2 ms-1 and 8 % less than 0.5 ms-1 (the wind speed values specified in IP 15).

o

Reference vector : 5 ms -1

True North

Platform North

45

Wind direction

Figure 4.1 Measured velocity field at a height of 2 m, for a wind speed of 28 knots (14.4 ms-1) from platform SWS

10

Table 4.1 shows the percentage of measured velocities below 0.5 ms-1 and 2 ms-1 for all 21 data sets. All tests were carried out on the cellar deck, except tests 18 to 24, which were made on the mezzanine level. It can be seen that there are a large number of measurements below 2 ms-1

occurring at relatively high wind speeds. However, wind direction plays an important role, as is illustrated by comparison of tests 10 and 17. The results show that for a wind direction of 1370

just 14% of all measurements were recorded below 0.5ms-1, but for a wind direction of 800, 74% of measurements were below 0.5ms-1.

Table 4.1 Percentage of measured air velocities below 0.5 ms-1 and 2 ms-1 for all data sets

Test number Average wind speed

Average wind direction, with respect to

platform North (degrees)

Number of measurements

Percentage less than 0.5 ms-1

(%)

Percentage less than

2 ms-1

(%)knots ms-1

17 16 15 14 13 12 11 10 9

8 (part test) 8 (part test)

7 6

5 (part test) 5 (part test)

4 3 2 1

22 18 15 19 30 28 28 23 20 11 8 8 13 6 9

18 18 17 15

11.3 9.3 7.7 9.8

15.4 14.4 14.4 11.8 10.3 5.7 4.1 4.1 6.7 3.1 4.6 9.3 9.3 8.7 7.7

80 127 205 207 141 199 193 137 133 145 77 92 295 288 137 339 336 325 349

66 66 66 66 66 66 66 66 66 18 30 66 66 24 24 84 82 82 82

74 53 12 11 8 8 5

14 17 56 100 95 15 63 38 21 11 16 20

95 95 79 71 61 61 64 82 86 83 100 100 89 96 88 92 85 90 91

18 25 12.9 187 48 2 40 19 15 7.7 211 48 15 69 20 14 7.2 175 48 21 79

21 31 16.0 81 48 38 100

When the wind was from platform East the production platform is sheltered by the accommodation platform (see Figure 3.1). This sheltering effect significantly reduced the air speed on the deck and had an effect on the air flow paths. This is illustrated in Figure 4.2, which shows the data from Test 7 at a height of 2 m above the floor of the cellar deck. The average wind speed during this test was 8 knots from approximately platform East. Excluding the 2 part tests, this test produced the highest percentage of measurements below 0.5 ms-1, with all measurements falling below 2 ms-1. Intuitively the movement of air through the deck would be expected to follow the wind direction, however, the measurements showed that the general flow direction through the platform actually opposed the wind direction. i.e. moving from platform West to East.

11

o

Reference vector : 1 ms -1

True North

Platform North

45

Wind direction

Figure 4.2: Measured velocity field at a height of 2 m, for a wind speed of 8 knots (4.1 ms-1) from platform East

Velocity data such as this gives a good impression of the overall air movement through the cellar deck for the correlated wind speed and direction and can give a good indication of areas that may be poorly ventilated, e.g. identification of recirculation areas.

From this example it should be noted that the presence of relatively high air speeds does not provide conclusive proof that an area is well ventilated, and vice versa.It should also be noted that measured air velocities of this nature do not give any information on the ventilation rate or the global air change rate.

4.2 TRACER GAS TESTS

(i) Air change rate measurements.These tests were carried out whilst the wind direction was from approximately platform East at a speed of between 8 and 9 knots; conditions similar to those during test 7. When the wind is from this direction the accommodation platform offers significant sheltering, as described in Section 4.1. It was thought that these were the best conditions during the visit to the platform to carryout the air change rate measurements. However, it was still not possible to create a uniform concentration of tracer throughout the cellar deck. Whilst data was obtained and a decay rate measured, the data was deemed to be unreliable.

(ii) Dilution measurementsDilution measurements were carried out whilst the wind was from platform Southwest at a speed of 25 knots. Under these conditions the predominant direction of air movement on the cellar deck is from the south wind wall to the north wall. It was thought releasing the tracer gas at a known

12

rate at the South wind wall and detecting at the north side would allow a dilution calculation to be carried out (as described in Section 3.2.2). However, whilst data was collected, variations in concentration between the two sample points indicated there would be appreciable errors in any calculations. For this reason the data has not been included.

Dilution measurements were also carried out on Installation A, and whilst data was obtained, the reliability was again questioned. However, the geometry of the well bay on Installation A lends itself well to dilution measurements for certain wind directions and a experimental method is described for such measurements in the separate report IR/ECO/02/10.

These tracer gas tests have led us to conclude that this method of ventilation assessment is not particularly appropriate for naturally ventilated platforms where the area is relatively ‘open’.

13

5 COMPUTATIONAL FLUID DYNAMICS MODELLING

5.1 INTRODUCTION

Computational Fluid Dynamics (CFD) can be used in conjunction with experimental measurements to investigate the effectiveness of natural ventilation in offshore modules under arbitrary wind conditions. Clearly, measurements made on platforms are limited by the weather conditions experienced on that day and therefore may not provide a full picture of the adequacy of the natural ventilation over a given period. Application of CFD modelling to arbitrary weather conditions can therefore be used in conjunction with wind rose data to provide a complete picture of the natural ventilation.

A further advantage of CFD modelling is that it can be used to investigate practical measures to improve the ventilation, before implementation, if poorly ventilated areas have been identified. Such methods could include the installation of fans and/or removal of wind walls.

A detailed ventilation study of the Installation C production platform was made using CFD as it was felt that this would provide more useful information than a less detailed study of two or three platforms. Indeed, the large amount of CFD results generated require careful analysis and interpretation and have allowed many interesting conclusions to be drawn.

The following three Sections describe the three main phases of the CFD modelling carried out to investigate the ventilation on the Installation C platform. Before CFD results can be used and interpreted with confidence, the model must first be validated against reliable experimental data. How this was done is described in Section 5.2. The application of the CFD model to arbitrary wind conditions is described in Section 5.3 and the programme of high pressure gas leak calculations carried out on the platform is described in Section 5.4. The details of the CFD modelling approach are described in Section 6.

5.2 VALIDATION CALCULATIONS

The series of velocity measurements made on the production platform of Installation C, described in Section 4.1 above, were used to validate the CFD model. Three tests were chosen to apply the CFD model to, which allowed the CFD model to be tested in a variety of conditions. The first test (Test 12) had a high wind speed and the wind direction was approximately platform Southwest. The ventilation would be expected to be good in this case. The second test chosen was one where the wind speed was low and was from approximately platform East (Test 7). In this case, one would expect that the geometry of the platform and the low wind speed would lead to a low ventilation flow through the platform. The final case that was modelled was an approximation to Tests 15 and 19. These two tests were carried out under very similar wind conditions. Test 19 was one of just a few tests where measurements were made on the mezzanine deck. This allowed a fuller validation of the CFD model to be made. Table 5.1 below summarise the three validation calculations undertaken. The results of the validation exercise are described in Section 7.2.

14

Table 5.1 Summary of test data used for validation exercise Experimental Test data CFD model parameters

Validation test number

Test number Average wind speed (knots)

Average wind direction, with

respect to platform North

Wind speed (knots)

Wind direction, with respect to platform North

V1 12 28 199 28 199

V2 7 8 92 8 92

V3 15 15 205

15 208 19 15 211

5.3 GENERAL WIND CONDITIONS

Having validated the CFD model it can then be systematically applied to give predictions for the platform ventilation under a range of weather conditions. This can therefore provide statistical information on the adequacy of the natural ventilation in combination with the wind rose data for the platform. Although it is important to note that this is not what we seek to achieve here.

The aim of this exercise is to understand under what conditions a typical platform may be poorly ventilated and how the natural ventilation is effected by the wind conditions. The CFD model was applied to give ventilation predictions for a set of 16 weather conditions defined by eight wind directions and two wind speeds. The wind speeds were chosen with reference to the Meteorological Office data for the installation. Wind speeds up to the lower wind value, 6.5 knots (3.3 ms-1), occur approximately 15% of the time. The higher wind speed, 14 knots (7.2 ms-1), was chosen as it is the average wind speed for the platform. The eight equally spaced wind directions were measured relative to the platform (N, NE, E, SE, etc.).

The results of these calculations were used to predict flow velocities throughout the platform and a global air change rate for the platform. The latter was calculated by predicting the total air flow through the openings in the platform (doors, gaps in wind walls etc.) and dividing by the internal volume of the platform (allowing for the volume occupied by equipment). The results of the CFD calculations for the general wind conditions are presented in Section 7.3.

‘Tracer gas’ calculations were also used to provide more detailed information about the local air change rates. In a similar way to the experimental technique, the platform was filled with an imaginary tracer gas and the resulting dispersion of the gas by the natural ventilation was monitored. This data was then used to calculate local ‘purge times’ . This concept is described in detail in Section 8.

5.4 GAS RELEASE CALCULATIONS

The CFD calculations described above provide velocity predictions, global air change rates and local purge times. However, it is still not completely clear how this data will relate to the key issue of the prevention of the build-up of flammable gas clouds. Therefore to obtain a correlation between the ventilation and the build-up / dispersion of a flammable gas cloud a number of gas releases where modelled on the platform. The details of the dispersion calculations are summarised in Table 5.2 below. The releases have been classified in terms of the categories given

15

in the Hydrocarbons Release Database (HCR) (OTO, 1999 079). The tests that are on the borderline between two classes are marked as such.

Table 5.2 Summary of dispersion CFD calculations of jet release on CONOCO platform including Hydrocarbons Release Database (HCR) Classification

Dispersion test number

Wind direction Release Rate

(kgs-1) Wind speed

(knots)

Release duration (mins)

HCR Classification

D1 W

1.0

6.5

5 Major / Sig.

D2

2

Significant

D3 SE Significant

D4 N Significant

D5 W 14.0

Significant

D6 SE Significant

D7 SE 0.1

6.5

Minor / Sig.

D8 W Minor / Sig.

D9 SE 5.0

Major

D10 N Major

The position and direction of the jet release is the same for all of the tests. The release is assumed to occur in the export riser (85 bar) and is directed towards the east side of the platform. This choice of location and direction has been identified as a realistic worst case scenario. The gas release is assumed to continue at a constant rate for two or five minutes and then stop instantaneously. A further two minutes of the gas dispersion has then been modelled.

In general, gas releases on offshore platforms consist of high pressure releases resulting in under-expanded jets. The method used to represent the jet in the CFD model is fairly basic but should give a qualitatively correct representation of the high momentum jet and will capture the key feature of a ‘point’ release of gas not represented in the tracer gas calculations. The calculations will also allow a comparison to be made of the effects of the wind conditions on the build up of a gas cloud and therefore the effectiveness of the ventilation. The air velocity predictions, measurements, global air change rates and purge times can then be compared against the results of the gas release calculations, to give feedback on their relative importance.

16

6 CFD MODEL DETAILS

6.1 INTRODUCTION

The CFD modelling was undertaken in two stages. For each simulation, the wind flow over the exterior of the installation was modelled to predict the pressure gradients over the production platform. These were then used as boundary conditions for a model of the interior of the production platform.

All of the CFD calculations have been carried out using the unstructured mesh codes CFX5.3 and CFX5.4. The following two Sections describe the external and internal flow calculations respectively.

6.2 EXTERNAL FLOW CALCULATION

An unstructured mesh was generated using approximately 59,000 nodes representing the domain surrounding the installation. An inflated mesh was used on the bottom of the domain to facilitate the modelling of the boundary layer (see boundary conditions below). Mesh refinement studies were carried out using three progressively finer meshes. The three meshes had approximately 30,000, 59,000 and 96,000 nodes respectively. The results showed that the solution was largely independent of the mesh. Figure 6.1 shows the surface grid over the platforms. The mesh was finest around the production platform to capture the detail of the flow there. The details of the other platform geometries were not included in the model as it is unlikely to have a large effect on the pressure distribution over the production platform.

The size of the domain was chosen to be 720 m long by 360 m high by 360 m wide. Sensitivity tests were carried out by making predictions using smaller (600 � 300 � 300 m) and larger (864 � 432 � 432 m) domains. The results showed that the predictions were not sensitive to the size of the domain and it was therefore concluded that the boundary conditions did not affect the flow over the installation.

17

XY Z

XY Z

CFXFigure 6.1. Surface mesh over Installation C

The wind flow over the installation was assumed to have a simple logarithmic profile and neutralatmospheric conditions were imposed. The wind speed for the validation cases is specified at 5 mabove sea level which is where the measurements were made. For the remaining CFD calculationsthe wind speed is specified 10 m above sea level to agree with the wind rose data. An inletboundary condition was used to model the wind flow and the installation was rotated so that theinflow boundary was always perpendicular to the wind direction. The imposed logarithmic windprofile on the inlet boundary was found to be stable by running the model without the installation,i.e. the logarithmic wind profile was maintained downstream of the inlet boundary. The boundarycondition representing the sea was defined as a rough wall with a roughness length of 0.5 mm.Symmetry boundary conditions and an outlet boundary condition were applied to the sides andend of the domain respectively.

The flow over the installation was modelled as an incompressible isothermal flow and steady statesolutions were sought. The standard k- � turbulence model was used. The CPU time for runningthese calculations was typically about an hour.

6.3 INTERNAL FLOW CALCULATION

Modelling the interior of the platform is very challenging due to the highly complex geometry ofthe platform structure and equipment. Ideally the computational mesh would be generated byimporting CAD data for the platform directly into the CFD mesh-generating package. However,

18

CAD data was not available for this platform, therefore the geometry was read from the plans for the platform. The plans used to build up the CFD geometry were not up to date and so a good deal of information had to be gathered from site visits. This clearly leads to uncertainty in the accuracy of the CFD geometry. As an example, on some site visits temporary scaffolding and tarpaulin had been erected on the mezzanine deck, which could lead to significant changes in the ventilation flow.

The larger features on the cellar and mezzanine decks were represented explicitly within the computational mesh. Smaller objects, such as areas of congested pipe work, were modelled by assuming that they acted as momentum sinks. A full porosity model is not currently available in CFX5. The linear and quadratic momentum sink terms were expressed as a function of the estimated volume porosity. The volume porosity for each area of pipe-work was estimated to be one of three values: 0.4, 0.6 or 0.8, where 0.4 represents the most congested region. Sensitivity tests were carried out to the choice of these values and it was found that the ventilation flow did not change much using different values for the volume porosity.

The calculations were carried out on an unstructured mesh using approximately 214,000 nodes. Note that a computational cell is based around each node in the computational mesh, computational cells are not equivalent to the mesh elements. There are over a million mesh elements used in the CFD calculations. The smallest dimension of an object that was resolved by the grid was 50 cm, (that is not to say that all objects bigger than 50 cm were resolved by the grid). A brief mesh dependence study was undertaken by using a finer mesh of 328,000 nodes. The solution did not change appreciably on the finer mesh, see Section 7 for a description of the results. Figure 6.2 shows a view of the surface mesh over the platform equipment and areas representing the congested regions.

Tracer gas calculations were carried out for the full set of sixteen weather conditions. The initial conditions for these runs were taken from the steady state solutions. A time dependent scalar equation was then solved using the fixed velocity field. The results were saved at regular time intervals so that the decay of the tracer gas could be closely monitored.

19

Figure 6.2. Surface mesh on interior of Installation C production platform

A total of 27 pressure boundaries conditions were used to represent the wind interaction with the platform. The boundary conditions were applied in the external doorways, gaps in the wind walls etc.. The pressure values were taken from the results of the external flow field calculation. The flow direction on the boundary conditions was taken to be the wind direction except in cases where the wind direction was parallel (or near to parallel) to the boundary surface. In this case the inflow direction was taken to be normal to the boundary surface. It is hard to predict exactly the angle that the wind will enter the platform due to the complex flow patterns around the module. The results of the calculations were found to be fairly sensitive to the direction used. The approach used was therefore a best guess approach and was carefully validated by comparison with the experimental measurements made in the doorways.

For each of the given wind conditions steady state solutions were sought to the incompressible isothermal Navier-Stokes equations. To aid convergence a first order differencing scheme was used to generate a converged solution and this was then used as an initial guess when running with the higher order differencing scheme. In some cases a laminar flow calculation also had to be performed first to help the first order solution start converging. Each of these simulations took approximately 15 to 45 hours to reach a converged solution depending on the speed of the computer. Turbulence was modelled using the standard k- � turbulence model. For the tracer gas calculations an additional scalar transport equation was solved.

20

6.3.1 High pressure jet releases

A high pressure gas release usually consists of an under-expanded jet. For the purpose of the current research, the detail of the near source region was not modelled as it was not required. This approach will allow a qualitative view of the interaction between the jet release and the local ventilation flow. The CFD model of the jet was started downstream of the release point where it is assumed that the pressure has returned to ambient. It was therefore a reasonable approximation to continue to use an incompressible approximation to the ventilation flow.

The inlet conditions for the jet were described as Gaussian profiles for velocity and gas concentration over a circular face within the CFD geometry. The Gaussian functions (Ewan and Moodie (1986)) for velocity, V, and gas concentration, Y, are given by the following expressions

rV = Vc exp(−( bV )2 ) (6.1)

rY = Yc exp(−( bY )2 ) (6.2)

,where r is the radial distance, bY = 0.126Z bV = 0.107Z (List, 1982) and Z is the axial distance from the source. The centreline velocity, Vc, and centreline gas concentration, Yc, are given by

� −1 ambient

(6.3) Vc = 1 − exp 0.7 − Kv

Z � Re eVe

� −1 ambient

(6.4) Yc = 1 − exp 0.7 − 0.104 Z � Re e

respectively. For sonic jets, Kv = 0.0672 and Ve, Re and � e are the velocity, radius and density of the jet at the exit respectively. The axial distance Z was chosen such that the width of the jet at this point would approximately fit inside the inlet in the computational domain.

21

7 CFD RESULTS

Due to the large amount of data generated by the CFD calculations, the figures showing the CFD results are presented in Appendix A1 of this report. A description of the calculations carried out is included below.

7.1 EXTERNAL FLOW CALCULATION

Figure A1.1 shows the predicted flow field over Installation C where the wind direction is 92 degrees and the wind speed is 8 knots. This corresponds to validation case V3. The wind field can be seen to be dominated by flow separation around the platforms. The results of the external flow calculations are examined to determine the pressure at the locations on the production platform corresponding to the gaps in the wind walls and doorways.

7.2 CFD MODEL VALIDATION

The aim of this exercise is not to try and obtain a one to one correspondence between the experimental data and the CFD predictions. This is not a feasible objective as the platform geometry is complicated and the ventilation flow in the platform is highly complex and time dependent. The aim of this exercise therefore, is to try and capture the general properties of the ventilation flow. For example, the main flow paths through the platform and the relative ventilation velocities. Section 6.3 describes the assumptions made by the CFD model for the internal flow calculation.

The validation exercise consisted of modelling the three tests listed in Table 5.1. The results showed that the CFD predictions were in good agreement with the experimental measurements, i.e. the general flow patterns for the cases considered were all captured well by the CFD model. The three test cases are examined in detail below.

For validation cases V1 and V2 the following CFD predictions are shown, where the height is measured relative to the cellar deck floor:

� Velocity vectors at experimental measuring positions 1.00 m above floor � Velocity vectors at experimental measuring positions 2...01 m above floor � Velocity vectors at experimental measuring positions 2...82 m above floor � More detailed velocity vectors 2.01 m above floor � Air speed above and below 0.5 and 2.0 ms-1, 2.01 m above floor

For validation case V3 the above results are shown for both the cellar deck and the mezzanine deck.

7.2.1 Validation Case V1

Summary: Wind speed = 28 knots (14.4 ms-1) Wind direction = 199 degrees

The results of applying the CFD model to the first validation case and the comparison with the experimental measurements are shown in Figures A1.2.1-A1.2.9. In this case the wind is coming from a fairly favourable direction, i.e. the platform is not sheltered by the other platforms, and the

22

wind speed is quite high. The ventilation on the platform is therefore relatively good in this case. Comparison of the experimental results with the CFD predictions shows a good correlation. The CFD model predicts that the general flow direction through the platform is from south to north. It appears that the bulk of the air moves towards the large gap in the north wind wall towards the west side of the platform. This is highlighted in Figure A1.2.7 which shows the velocity field 2.01 m above the floor in more detail. The predictions for the ventilation velocities are in reasonable agreement with the experimental measurements. The region of lowest flow is in the sheltered area on the east side of the platform. This is shown by the experimental measurements and CFD predictions in Figures A1.2.8 and 7.2.9 respectively. The ventilation velocities are all generally above 2.0 ms-1. However, note that the CFD results showing the velocity magnitude in Figure A1.2.9, the plane on which the speed is shown intersects with a number of regions where a momentum sink has been applied (see Section 6.3). In these small localised areas, where there are generally low velocities, we are not confident of the velocity predictions.



A comparison between the CFD predictions and the experimental results for the second validation case are shown in Figures A1.2.10-A1.2.18. The wind direction in this case is from the East. This means that the accommodation module is sheltering the production platform from the wind. The wind speed is also quite low in this case. These two factors lead to a much reduced ventilation rate. Contrary to what one might expect the general flow through the platform is from West to East. This unexpected phenomena is captured by the CFD model. The ventilation velocities are significantly lower in this case compared to the previous validation case. Note that the vector scale in these Figures is five times larger than the previous set of results, i.e. a vector of length 5 m equals a velocity of 1 ms-1. The areas of lowest flow in this case are on the west side of the platform in the middle and, again, in the sheltered area on the east side of the platform. Figures A1.2.17 and A1.2.18 show that the ventilation velocities are generally below 0.5 ms-1.



The results of applying the CFD model to the final validation case and the comparison with the experimental measurements are shown in Figures A1.2.19-A1.2.36. In this case experimental measurements were made on both the cellar and mezzanine decks which allows a fuller comparison to be made against the CFD predictions. The wind direction is very similar to the first test case and this is reflected in both the experimental measurements and the CFD predictions. The general direction of air movement is from South to North on both decks. It appears, again, that the bulk of the air exits the platform through the large gap in the north wall on the west side. The areas of lowest flow on the cellar deck appears to be in the sheltered area on the east side and also in the south east corner of the platform. The experiments indicate that there may be a recirculating flow in this area although the CFD results just show a low flow area. This area is sheltered by the wind wall on the south side. The mezzanine level appears to be generally better ventilated than the cellar deck. This is most likely due to it being less congested with equipment than the cellar deck. Also, it does not have a semi-enclosed area on the east side like the cellar deck, although the East wall is still solid. The magnitude of the velocity predictions on

23

the east side of the mezzanine deck are not in agreement with the experimental measurements. This is probably due to the inaccuracies in the platform geometry in that area, or the CFD predictions are shown too close to a solid object and are therefore sheltered. On both decks the ventilation velocities are rarely below 0.5 ms-1.

7.2.4 Sensitivity tests

A number of sensitivity tests were carried out on the CFD model to assess the effect of a number of parameters on the solution. The results of all the sensitivity tests cannot be shown here due to the limitation on space. The grid dependence of the solution was assessed by repeating Validation Case 1 on a finer grid employing approximately 328,000 computational nodes, an example of the results are shown in Figure A1.2.37. Comparison with the regular mesh (Figure A1.2.7.) shows that the CFD solution does not change appreciably under mesh refinement.

The sensitivity of the solution to the choice of differencing scheme and values used in the momentum sink terms was also assessed. CFD simulations were carried out with a first and second order differencing scheme and the results showed the expected differences. The general flow pattern was the same but the first order results failed to pick up smaller flow features, such as small areas of recirculation. The choice of values in the momentum sink terms also did not change the general flow pattern.

7.3 GENERAL WIND CONDITIONS

7.3.1 Ventilation field

Figures A1.3.1-A1.3.16 show the velocity field 2 m above the cellar deck floor for the 16 different wind conditions considered. Clearly the flow through the platform is strongly three dimensional, however these figures give a good idea of the general flow through the platform under these wind conditions. The first eight figures show the velocity field at the lower wind speed, 6.5 knots, and the remaining eight figure, A1.3.9-A1.3.16, show the results at the higher wind speed, 14 knots. The general flow pattern appears to stay the same as the wind speed is varied but the magnitude of the ventilation velocities varies accordingly.

The flow velocities through the platform are at their lowest when the wind is coming from the East and the West. We would therefore expect that the effectiveness of the ventilation is at it’ s lowest when the wind is coming from these directions. For all other wind directions, where the wind speed is 6.5 knots the ventilation velocities are typically around 1 ms-1 except near to the doors and gaps in the wind walls etc. where they are around 2 to 3 ms-1. At the higher wind speed, again excepting the East and West wind directions, the ventilation velocities vary from 1 to 3 ms-1

through the platform and are much higher near the doorways. At the lower wind speed when the wind is coming from the East or the West, the ventilation velocities are almost all below 1 ms-1. At the higher wind speed they are around 1 to 2 ms-1

The reason for the lower flow velocities when the wind is coming from the East and West is predominately due to the sheltering effect of the adjoining modules, see Figure 3.1. The solid walls on the east side of the platform will also have a significant effect when the wind is coming from these directions.

It is interesting to note that when the wind is coming from the East the general predicted flow direction through the platform is not from East to West as expected but is actually from West to

24

East. These predications agree with the experimental measurements that were made on the platform when the wind was coming from an easterly direction (see Section 7.2.2). This example shows that it is not possible to guess the general flow patterns through a module. A quick look at the results presented in this report shows that the general flow patterns are not uniform across the module. For example, there are many areas of recirculation and counter flow. The flow pattern through the module also varies strongly with height (not shown in this Section). The flow patterns through the mezzanine deck for example are very different from those on the cellar deck.

7.3.2 Air change rates

The global air change rate through the platform (both cellar and mezzanine decks) can be estimated using the CFD model. This is achieved by measuring the total volume flow through the openings in the platform and dividing this by the platform volume. An allowance is made for the volume occupied by the equipment. This is only a rough estimate of the volume of the equipment on the platform. The results of these calculations are shown in Figure 7.1 for each of the wind directions and speeds. These measurements show how the ventilation rate changes with wind conditions and clearly show that the air chnage rate is at its lowest under Eastery and Westerly winds. However, they do not provide any information on the effectiveness of the ventilation as they do not account for the presence of either recirculation zones or ventilation ‘short circuiting’ that leads to stagnant regions. The shortcoming of this approach are addressed in the next Section.

0

N E S W

i i i

i i i

50

100

150

200

250

NE SE SW NW

Air

ch

ang

e ra

te p

er h

ou

r

Pred cted a r change rate for a w nd speed of 14 knots

Pred cted a r change rate for a w nd speed of 6.5 knots

Figure 7.1. CFD predictions of global air change rates over a range of wind conditions

If we assume that the global air change rate varies linearly with the wind speed it is possible to determine the wind speed at which the global air change rate would fall below 12. Use of the wind rose data would then show what percentage of the time the ventilation would be below this value. However, although the CFD predictions show a fairly strong linear relationship between the air change rate and wind speed, it may not be a reasonable assumption to make at lower wind speeds. For instance, at these lower wind speeds thermal effects will have a more significant impact, and are likely to increase the ventilation rate. Thermal effects were not included in the CFD as model as it was assumed their effect would be small at the wind speeds considered. For information this is shown in table 7.1.

25

Table 7.1 CFD prediction of wind speeds at which the global air change rate would fall below 12.

Wind direction

Air change rate at 6.5 knot wind

speed (h-1)

Air change rate at 14 knot wind

speed (h-1)

Wind speed required to produce an air change rate of 12

(knots)

Frequency of wind speed (from wind

rose) (%)

N 93 200 0.8 -

NE 48 105 1.6 <0.6

E 27 57 2.9 0.4

SE 51 109 1.5 <0.4

S 73 157 1.1 <0.5

SW 65 140 1.2 <0.4

W 23 50 3.3 0.5

NW 86 185 0.9 0

7.3.3 Tracer gas calculations

Tracer gas calculations have been carried out for the 16 wind conditions considered. An example of one of the results of the calculations is shown in Figure A1.3.17. In this case the wind is coming from the South East. The red areas show where the tracer gas has not been removed and the blue regions show the areas that have been purged of the tracer gas. Comparison of this Figure with the velocity field shown in Figure A1.3.4 helps to explain the result. The areas where the tracer gas is slow to disappear correspond to the areas of lowest ventilation velocities. Also note that the areas nearest to the air inlets are quickest to be purged. This is an obvious statement, but it is important to note that the areas that are slowest to clear do not necessarily imply the lowest ventilation rate. For example, the area near to the door in the North East corner has a reasonable flow through it. However, because it is a long way from the nearest air inlet the tracer gas is slow to clear there. This point is discussed further in Section 8, where the results of these tracer gas calculations are analysed further.

7.4 JET RELEASE PREDICTIONS

It is not easy to visualise the results of the jet release predictions calculations as the shape of the gas cloud is very complex and time dependent. The best method for visualising the results appears to be showing the gas concentrations on a slice through the platform at regular time intervals. Although this does not give much of a feel for the three dimensional characteristics of the gas cloud it does represent fairly well the extent to which the gas cloud is being dispersed by the ventilation. This is therefore the method that has been used and the results for all the jet release calculations are shown in Figures A1.4.1 - A1.4.10.

The results of these tests give a good indication of the effect of natural ventilation on the build up and dispersion of flammable gas on an offshore platform. However, due to the assumptions made in modelling the source of the jet, the results should be interpreted in a qualitative fashion; i.e. the predicted gas clouds sizes can be used as a good initial estimate. The results do, however, give a very good indication of the relative effects of the wind speed and direction etc. on the dispersion of the gas cloud. Note that all of the results shown use the same scale, shown in Figure A1.4.1,

26

and areas that are shown in red indicate gas concentrations at or above the lower explosive limit (i.e. 100% LEL).

Figure A1.4.1 shows the results of the first test (D1) where the gas builds up over a five minute period. After these five minutes the dispersion still hasn’ t reached a steady state although the flow is not changing a great deal. The remainder of the tests are for a two minute release. This test, therefore, can help to indicate the expected behaviour in the remaining tests if the release where for a longer duration.

The results of tests D2-D4 are shown in Figures A1.4.2-A1.4.4. These tests show the effect of three different wind directions on the build up and dispersion of the flammable gas. All three tests are carried out using the lower wind speed, 6.5 knots, and a 1 kgs-1 release rate. The tests clearly show the strong effect that the wind direction has on the gas dispersion. For the case where the wind is coming from the North, the gas is much slower to build up and is dispersed very quickly after the release has stopped. The other two cases are fairly similar to each other. Where the wind is coming from the Southeast the gas appears to build up more quickly than when the wind is coming from the West. However, after the gas release has stopped the gas in the southeast case appears to be dispersed slightly more quickly. Note the corresponding ventilation flow predictions in Figures A1.3.1, A1.3.4 and A1.3.7 which help to explain these results. In the southeast case the east side of the platform appears very sheltered although the rest of the platform is better ventilated than the west case.

Tests D5 and D6 show the effects of the higher wind speed on the gas dispersion (see Figures A1.4.5 and A1.4.6). Here, as expected, the gas is dispersed more quickly than the corresponding low wind speed case. However, the gas is not dispersed as quickly as in the low wind speed case where the wind is coming from a more favourable direction (i.e. from the North, as in Figure A1.4.4). In both the high wind speed cases the extent of the flammable region of gas is fairly small and is dispersed quickly after the gas release is finished.

The next two test cases, D7 and D8, show the results of a much smaller release rate. Figures A1.4.7 and A1.4.8 show that even with a low wind speed and an unfavourable direction the gas is unlikely to build up into a significant flammable cloud. The classification of these releases as ‘Minor’ (see Section 5.4) therefore seems appropriate.

Test cases D9 and D10 show the results of the dispersion of a 5 kgs-1 release (see Figure A1.4.9 and A1.4.10). These much larger releases lead to the build up of significant clouds of flammable gas at or above the LEL. The difference between the two wind directions in this case is clear. Where the wind is coming from the Southeast the cloud of gas at or above the LEL builds up quickly and remains for over two minutes after the release has finished. However, in the case where the wind is coming from the north, the gas cloud builds up much more slowly and is rapidly dispersed after the gas leak has stopped. Note that this is the case even though this Test problem is still at the lower wind speed. These two test problems also highlight the sensitivity of the gas cloud volume above LEL to the gas leak rate. The five-fold increase in the mass release rate has lead to significantly higher quantities of gas on the platform at or above LEL. This indicates the importance of modelling the source accurately.

27

8 PURGE TIME CALCULATIONS

Whilst a global air change rate can be used to calculate the volume of air moving through a platform, it does not give any information on the distribution of the air and therefore the effectiveness of the ventilation. In order to predict the locations of the least well ventilated areas, a tracer gas release was modelled, see Section 7.3.3. This method allowed areas where the tracer decayed more slowly to be identified. In order to quantify, and allow comparison of the ventilation of areas for different wind speeds and directions, the concentration/time data was integrated from the point at which the tracer concentration started to decrease. This, therefore, accounted for the time taken for clean air to reach the area of interest. The integral was defined as the ‘purge time’ . The larger the calculated purge time, the less well the area was likely to be efficiently ventilated. It should be noted that the values, whilst having units of time, do not provide any information on the time taken for tracer removal.

Figure 8.1 shows the calculated purge times at a number of positions on the cellar deck at a height of 2 m for a wind speed of 6.5 knots from platform North. For comparison Figure 8.2 and Figure 8.3 shows purge times for the same wind speed but from platform East and platform West respectively. These are considerably higher, particularly in the recess at the East side of the deck and also an area at the West side of the deck. This technique allows a valuable insight to be gained of the relative effectiveness of the ventilation at various locations within the platform. However, uncertainties are introduced into these calculations as air arriving from other parts of the deck is likely to contain the tracer and will, therefore, not dilute and remove the tracer from the area of interest as quickly as clean uncontaminated air.

The purge time data can be used as part of a technique for assessing the local ventilation effectiveness as a function of the likelihood of a given set of wind conditions being encountered for that platform. This is done by multiplying the local purge time by the frequency at which that wind condition occurs and then averaging over all wind directions. This will then give an averaged purge time for each chosen location on the platform and an indication of the effectiveness of the ventilation at that point over all the wind conditions experienced specific for that platform.

28

0.14

0.25

0.45

0.18

0.15

0.12

0.52

0.81

1.78

0.24

0.58

1.37

0.57

0.72

0.91

0.67

1.08

0.9

0.47

1.66

0.35

0.58

0.87

Figure 8.1 Calculated purge times for a wind speed of 6.5 knots from platform North

2.85

1.32

5.16

0.16

2.22

0.91

1.87

2.77

1.83

2.32

1.06

1.98 2.93

4.52

Figure 8.2 Calculated purge times for a wind speed of 6.5 knots from platform East

29

3

4.01

0.61

1.4

1.14

2.22

0.88

1.74

2.02

4.64

2.74 5.14

2.47

Figure 8.3 Calculated purge times for a wind speed of 6.5 knots from platform West

30

9 REMEDIAL MEASURES

9.1 INTRODUCTION

Ventilation assessments may show that some local areas may be inadequately ventilated. In this section we look at a method for improving the local ventilation, assuming that such an area has been identified. Sections 7 and 8 of this report have shown that the region on the east side of the platform is the least well ventilated. Therefore, this area has been chosen as an example to see if the ventilation can be improved here.

There are many different ways in which this can be achieved, the two obvious ones are improvement of the natural ventilation by removal, or part removal, of wind walls/floor areas or the installation of mechanical ventilation. It is difficult to know which method is most appropriate, as both have advantages and disadvantages. For example, the quality of the working environment could be seriously compromised by opening up large wall/floor areas and may introduce other risks. Additionally, equipment may be subjected to harsher weather conditions, including salt spray, which may increase corrosion. The removal of wind walls or floor areas is likely to be a large and expensive operation, which would also carry a degree of risk. The main advantage of improving the ventilation by natural means is the area of interest will be continually ventilated throughout an emergency. Improving the ventilation using mechanical methods ensures the ventilation rate in the area varies little with weather conditions, however in the event of a power failure or shutdown, the ventilation will revert back to being ‘inadequate’.

An alternative to improving the ventilation, is to install additional gas detectors in the area deemed to be poorly ventilated. This method links the localised efficiency to the gas detection strategy.

From conversations with the platform operators, they strongly suspected that if any remedial measures were taken, they would be in the form of ducted mechanical fans. These have been used in the past to ventilate the cellar deck of Installation C and ventilation equipment, including fans, were still evident during the platform visits. By installing ducted mechanical fans the intention would not be to significantly increase the overall ventilation rate on the cellar deck, but to improve the ventilation effectiveness in the identified region at the east side of the platform by the introduction of clean air.

To actually install a fan on the cellar deck of the platform would have taken considerable time to organise, therefore CFD has been applied in order to explore this scenario further. The outlet of a fan was added to the existing CFD model at the east side of the platform. The outlet velocity was fixed at 3 ms-1, as it was thought that this should not unduly affect the comfort of personnel working in this area.

9.2 CFD MODELLING APPROACH

The ventilation fan was modelled by specifying an inlet boundary condition on a solid face 3.1 m above the cellar deck floor at an angle of 15 degrees below horizontal. The fan was pointed in a southerly direction relative to the platform. A uniform velocity of 3 ms-1 was specified across the 30 cm square inlet face giving rise to a volume flow rate of 972 m3h-1. Note that this is roughly equivalent to an air change rate of 0.1 per hour for the whole platform. For a Westerly wind at

31

6.5 knots, this represents an increase in the global air change rate of around 0.4%, see Figure 7.1. The air entering the platform through this inlet is assumed to be clean air.

To assess the effect of the fan on the ventilation of the platform, and specifically on the local area, the tracer gas calculations, as described in Section 7.3.2, are repeated here. These calculations allow the decay of the tracer gas to be monitored at any location within the module and therefore the adequacy of the ventilation can then be inferred. As we are interested in the cases where the ventilation is poorest, the weather conditions chosen for this test are those highlighted earlier in this report as leading to the poorest ventilation; i.e. Westerly wind at 6.5 knots. The results of these calculations, along with purge times, are described in the next Section.

The effect of the introduction of a fan on a realistic gas release has also been assessed by repeating the gas dispersion test D2 (see Table 5.2). The wind conditions for this test are the same as those described above. The results of this simulation are presented and discussed in the next Section.

9.3 CFD RESULTS AND DISCUSSION

9.3.1 Tracer gas simulations

A comparison of the tracer gas simulations, with and without the remedial measure in place, is shown in Figure A2.3.1. The figure shows a slice through the platform 2 m above the cellar deck floor indicating the local tracer gas concentration. The position of the fan can be seen in the pictures on the right, near to the East wall. After one minute (i.e. looking at the two top pictures) the fan can be seen to introduce clean air into the sheltered area on the east side of the platform. Without the fan this area is one of the last to be purged of the tracer gas at its highest concentration, this is shown by the red area in the picture on the left at 2 minutes. However, with the fan this area is diluted to a concentration below 0.4 (see Figure). Further dilution effects can be seen at three minutes.

These results, however, do not tell the full story as they only show the results on a single plane through the solution. A post-processing utility allows the volume of the tracer gas to be calculated between given concentrations. This, therefore, allows the total volume of tracer gas to be calculated and compared for the two corresponding cases. These results are given in Figure A2.3.2 and show the volume of tracer gas above three concentrations plotted against time. The most important point here is that the volume of tracer gas is not always lower when the fan is running. This is most notable at the higher gas concentration at around 2-3 minutes where the volume of tracer gas is slightly higher with the remedial measure in place. There are a number of reasons why this may occur. One mechanism that could lead to this phenomenon is that the fan dilutes small volumes of high concentrations of the tracer gas, thereby leading to larger volumes of low concentration gas. Another possibility is that the air movement induced by the fan is opposing the natural ventilation flow. In this case the ventilation could actually be reduced in small areas. Finally, these volume calculations are themselves subject to a degree of uncertainty and therefore the small differences in volume shown in Figure A2.3.1 are not that significant. Further work would certainly be required to provide more understanding here.

32

9.3.2 Purge time calculations

A comparison of the purge times with and without the remedial measures in place is shown in Figure 9.1. The purge times in the recess on the east side of the platform are not significantly lower, in fact one position has increased whilst the other has decreased. Overall it can be concluded that the introduction of the ventilation fan has not had the desired effect of significantly reducing the purge times.

The fan appears to have increased the mixing within the area but has failed to purge the area with clean air. The additional air from the fan may also effect the efficiency of the ventilation elsewhere on the deck, as can be noted from the increase in the purge time at one of the positions close to the recess area. It should also be noted that most of the purge times have changed elsewhere on the cellar deck, if only slightly. This illustrates how even a localised change to the flow field can have wide ranging effects. As the purge times were only calculated at a height of 2m, it is not possible to say whether purge times at other heights have increased or decreased.

3

4.01

0.61

1.4

1.14

2.22

0.88

1.74

2.02

4.64

2.74 5.14

2.47 (3.20) (4.95)

(1.06) (1.65)

(2.18)

(2.86) (2.05)

(1.63)

(4.21)

(0.61)

(2.73)

(4.44)

(0.72)

Figure 9.1 Calculated purge times for a wind speed of 6.5 knots from platform West with (bracketed) and without remedial measure

9.3.3 Gas release simulations

The results of carrying out dispersion test D2 with the remedial measure in place are shown in Figure A2.3.3. These results can be compared to the corresponding test, without the remedial measure, shown in Figure A1.4.2. The results appear to be very similar, especially at the earlier times during the release where the gas release dominates the flow. After the release has finished the fan clearly has an effect on where the gas accumulates but it is not possible to say from these pictures whether the gas is being effectively diluted or just moved around to a different position in the platform.

33

As above, the volume of gas at different concentrations can be calculated at different times to assess the impact of the fan. These results are shown in Table 9.1 below. As the release is not particularly large, the volume of gas above the lower explosive limit (LEL) is small. It is therefore not possible to draw any conclusion from those figures. However, it is interesting to note that, counter-intuitively, the volume of gas is slightly higher when the fan is used. The differences in these volumes are of the same order as the volume of a single computational cell and therefore, cannot be viewed as significant.

The volume of gas above 50% LEL also provide some surprising results. During the release and for at least the next 30 s the volume of gas above 50% LEL is lower with the fan. However, after this time the volume of gas above 50% LEL is actually higher with the fan running. Further work is needed here to provide more understanding on the mechanisms that lead to these results. This would include looking at different release rates and wind conditions in conjunction with different fan flow rates.

Table 9.1 Gas cloud sizes. Wind direction: West, wind speed 6.5 knots

Time (mins:secs)

Volume of gas cloud (m3)

With ducted fan (1000 m3h-1) Without fan

>100% LEL >50% LEL >100% LEL >50% LEL

0:30 0.9 43.1 0.8 47.9

1:00 1.0 135.3 0.9 180.5

1:30 1.1 607.2 1.0 647.6

2:00 1.4 1,289.5 n/a n/a

2:30 0.0 764.3 0.0 780.9

3:00 0.0 519.9 0.0 501.8

3:30 0.0 300.6 0.0 243.1

n/a - data not recorded

34

10 CONCLUSIONS

10.1 SUMMARY

A combination of experimental measurements and CFD have been used to study natural ventilation on offshore platforms. The measurements have shown that the effectiveness of natural ventilation is highly dependent on the external wind conditions and is unlikely to be uniform throughout the module, with areas of low air velocity and areas of recirculation identified. Also, the work has demonstrated that, due to complex geometries and the influence of adjoining platforms, flow paths through modules is difficult to anticipate and can be counter-intuitive.

The combination of the experimental measurements and CFD have proved to be an invaluable tool for assessing the effectiveness of the ventilation. One of the key advantages of this approach is that it allowed an assessment to be made of the effectiveness of the ventilation for diluting and dispersing gas/vapours following a credible gas leak. As a result a useful tool has now been developed to assess ventilation on offshore modules.

Tracer gas tests, which have been carried out on two platforms, have demonstrated the difficulty in gathering reliable data. It should be possible to measure ventilation rates on offshore platforms for certain platform configurations/geometries and under certain weather conditions. However, where the ventilation openings are positioned so as to create complex flow patterns inside modules, it is unlikely that accurate tracer gas measurements would be possible.

The CFD model has been used to calculate a ventilation rate for a module. However, this value gives no information on the effectiveness of the ventilation or the flow paths through the module. Measurements have shown that different flow regimes exists for different geometries; from plug flows on some platforms to areas of short circuiting on others. It is therefore unlikely that the ventilation will be uniform with time or with position.

In order to compare the effectiveness of the ventilation at different points on a module a ‘purge time’ has been defined. Realistically, this parameter can only be calculated using data generated from a CFD simulation and, whilst it allows the least well ventilated areas to be identified, it does not give information on the adequacy of the ventilation. Nevertheless, purge times could be used to optimise the positioning of gas detectors. Further work in this area is required.

If an area was deemed to be inadequately ventilated, there are a number of remedial measures which could be applied. These range from improving the natural ventilation by removal of wall, floor and/or roof sections to complimenting the natural ventilation with mechanical ventilation via the installation of fans, air movers etc. This report has considered the use of a single fan placed in the area to the East of the cellar deck on Installation C. The results have shown that the purge times have not significantly improved. The additional air appears to have increased the mixing within the region but has failed to purge the area with clean air. Further work is required to ensure that remedial measures are designed and assessed to ensure that they will actually lead to an improvement in the ventilation.

The CFD simulation of gas releases have shown that for moderate wind speeds and favourable wind directions natural ventilation is capable of preventing the build up of flammable gas clouds for small leaks and dispersing the gas clouds that form following larger leaks once they have finished. However, these simulations have also shown that even where the wind speed is of a

35

moderate strength, gas clouds from fairly small leaks can build up and form potentially explosive gas clouds. This raises the question as to what natural ventilation should be expected to be able to achieve in terms of preventing the build up of flammable gas clouds and dispersing them once they have been created. A possible way forward would be to define a maximum leak size that the ventilation should be able to control. The CFD simulations of the gas releases have also shown that large gas releases dominate the ventilation flow through the module. It is therefore very difficult to predict how a gas cloud may build up within a module and how it will interact with the natural ventilation.

10.2 VENTILATION ASSESSMENTS

According to IP 15, if an area is classed as open it should not have stagnant areas and “ Typically air velocities should rarely be less than 0.5 ms-1 and should frequently be above 2 ms-1” . From the considerable number of velocity measurements made on all three installations (one with no wind walls fitted), measurements made within the structure are always less than the prevailing wind speed, often a significant number were below 0.5 ms-1 and many below 2 ms-1. However, this is not unexpected; no matter how ‘open’ the platform structure is, it will still present a significant blockage to the air flow and therefore the air will move around the structure, reducing the air velocity within it. If this criteria is to be used for assessment of areas, i.e. ‘open’ or ‘sheltered’ , many areas which are currently classed as ‘open’ may need to be redefined as ‘sheltered’.

If an area is assessed as ‘sheltered’ the ventilation rate needs to be quantified. The measurement of ventilation rates on offshore modules provides a challenging task. If accurate measurements cannot be made, it becomes difficult for operators to demonstrate that they are complying with current Standards and Codes.

Cleaver (1994) discusses the assessment of natural ventilation and comments that ventilation rates can be calculated by measuring air speeds at the perimeter of platforms and multiplying these by the appropriate open areas to give a volume flow rate. This technique could be applied to certain module designs and would be applicable to the well bay of Installation A. However, for Installation B, the design of the wind wall around the perimeter of the cellar deck of the production platform was such that ‘short circuiting’ could occur. This is when air entering the platform resides for only a short period of time without passing through the whole space. On this platform the primary route for air entering the deck was through the gaps above the wind walls. The air entered in a downward direction towards the open mesh floor (probably due to the external flow pattern around the platform). This suggests that a significant quantity of air ‘short circuits’ the majority of the deck. Measurements made at openings which acted as air inlets to the deck would have overestimated the volume flow rate through the majority of the deck.

To determine whether or not a sheltered area is adequately ventilated, IP 15 states that there should be a uniform ventilation rate of at least 12 air changes an hour, with no stagnant regions. ‘Uniform ventilation rate’ raises a number of questions. If the ventilation was spatially uniform then there would be no stagnant zones. Clearly with natural ventilation the ventilation rate cannot be uniform with respect to time and a finite minimum rate cannot be guaranteed. Chapter 6 of IP 15 is based upon the framework of BS EN 60070-10, which does not refer to air change rates in the context of adequate ventilation, but states “ the most important factor is that the degree or amount of ventilation is directly related to the types of source of release and their release rates".

36

Expressing a ventilation rate in terms of air changes per hour for places such as offshore modules may not be the ideal method. The required ventilation should depend on the size of the potential leak, not the size of the deck in which it may occur. Both Dagested and Holdφ (1990) and Gale (1985) have raised this question in the past. It is also known that an offshore operator has an internal standard for predicting the hydrocarbon fugitive emission for a given plant design/layout. This enables them to calculate the quantity of dilution ventilation required to maintain a non-flammable atmosphere i.e. the ventilation rate is calculated based on the predicted fugitive emissions. Whilst the standard implies that it should only be applied to enclosed areas with mechanical ventilation, it is a departure from IP 15. It should be noted that the standard is only applicable to refurbishment/upgrade work to existing installations; new installations are based on IP 15, i.e. a minimum of 12 air changes an hour with no stagnant regions.

Air changes per hour are more applicable to describing ventilation rates in buildings, where the air entering the building will mix thoroughly with the existing air, effectively diluting any contaminants. In areas such as the well bay on Installation A and to some extent the cellar deck on Installation B, the degree of mixing within the space was low; the flow regime was closer to piston flow or plug flow, where the air moved in a definite direction across the deck. The Industrial Ventilation manual (2001) maintains that air changes per hour are a poor basis for ventilation criteria where environmental controls of hazards are required. Guidance based on volume flow rates may be more applicable in this scenario. Alternatively, ventilation rates could be completely removed from guidance and replaced with acceptable geometries/open areas which give acceptable ventilation. A further possibility is to evaluate the effectiveness of the ventilation by its ability to dilute and disperse a given range of gas leak rates.

What may need to be addressed is whether or not guidance on ventilation on offshore modules should be issued as a separate document to the area classification codes.

10.3 FURTHER WORK

Whilst this project has addressed the main aims and objectives of the work, further questions have been raised. These could be addressed through further work and include:

(i) If an area on a module has been classified as ‘inadequate’ there are a number of methods available to improve the ventilation. The work in this report on remedial methods has highlighted the need to assess any measures before implementation. Further work on the best ways of improving the ventilation need to be explored further.

(ii) The need to address current guidance.

Before further work is undertaken it may be worth considering publishing this work in peer review press and/or industry newsletters to benefit from comments and suggestions from those operating platforms and other interested parties.

37

11 REFERENCES

ISO 15138, (2000), Petroleum and natural gas industries – Offshore production installations – Heating, ventilation and air-conditioning, International Standard ISO 15138:2000(E).

API Recommended Practice 500. American Petroleum Institute. November 1997.

Area Classification Code for Petroleum Installations. Part 15 of Institute of Petroleum ModelCode of Safe Practice in the Petroleum Industry. John Wiley and Sons, March 1990.

BS EN 60079-10 : 1996. Electrical apparatus for explosive gas atmospheres, Part 10,classification of hazardous areas.

Cleaver, R. P., Carol, E. H. and Robinson, C. R., ‘Accidental generation of gas clouds onoffshore process installations’. J. Loss. Prev. Process Ind., Vol 7, No 4. pp 273-280. 1994.

Dagested, S. and Holdφ, A. E. ‘Natural ventilation in offshore modules of rectangular shape’.Trans. Inst. Marine Eng, Vol 103, pp 85-96. 1990.

Etheridge, D and Sandberg, M, ‘Building ventilation - Theory and Measurement’ (1996).Chapter 12, pp591-625. Wiley. ISBN 0 471 96087 X.

Ewan B. C. R. and Moodie K. ‘Structure and velocity measurements in under-expanded jets’Combustion Science and Technology 45 pp275-288. (1986)

Gale, W. E. ‘Module ventilation rates quantified’. Oil and Gas Journal. Dec. 1985.

Hydrocarbons Release Database (HCR) (OTO, 1999 079).

Industrial Ventilation: A manual of recommended practice. (2001). Chapter 7. pp7-16. ACGIH.24th Edition.

IR/ECO/02/10 (2002) and IR/ECO/00/12 (2001)

List E. J. ‘Mechanics of turbulent buoyant jets and plumes; in turbulent and buoyant plumes’ Ed.Rodi W.; Pergamon Press Ltd. (1982)

38

A1 FIGURES: CFD RESULTS

EXTERNAL FLOW CALCULATIONA1.1

Figure A1.1.1 CFD predictions of flow field over Installation C

A1.2 CFD MODEL VALIDATION

39

A1.2.1 Validation case V1

Figure A1.2.1 Experimental velocity vectors 1.0 m above cellar deck floor. 1 m vector = 1 ms-1

Figure A1.2.2 CFD Velocity vectors at experimental measuring positions 1.00 m above cellar deck floor. 1 m vector = 1 ms -1

40


Figure A1.2.4 CFD velocity vectors at experimental measuring positions 2.01 m above cellar deck floor; 1 m vector = 1 ms -1

41

Figure A1.2.5 Experimental velocity vectors 2.82 m above cellar deck floor; 1 m vector = 1 ms-1

Figure A1.2.6 CFD velocity vectors at experimental measuring positions 2.82 m above cellar deck floor. 1 m vector = 1 ms -1

42

Figure A1.2.7 CFD velocity vectors 2.01 m above cellar deck floor. 1 m vector = 5 ms-1

43

Figure A1.2.8 Experimental measurements showing air speeds above and below 0.5 and 2.0 ms-1, 2.01 m above cellar deck floor

Figure A1.2.9 CFD predictions showing air speed above and below 0.5 and 2.0 ms -1, 2.01 m above cellar deck floor

44


Figure A1.2.10 Experimental velocity vectors 1.00 m above cellar deck floor. 1 m vector = 0.2 ms-1

Figure A1.2.11 CFD velocity vectors at experimental measuring positions 1.00 m above cellar deck floor. 1 m vector = 0.2 ms-1

45



46



47

Figure A1.2.16 CFD velocity vectors 2.01 m above cellar deck floor; 1 m vector = 1 ms-1

48



49



Figure A1.2.20 CFD velocity vectors at experimental measuring positions 1.00 m above cellar deck floor; 1 m vector = 1 ms-1

50

Figure A1.2.21 Experimental velocity vectors 2.01 m above cellar deck floor . 1 m vector = 1 ms-1


51

Figure A1.2.23 Experimental velocity vectors 2.82 m above cellar deck floor . 1 m vector = 1 ms-1


52

Figure A1.2.25 CFD velocity predictions 2.01 m above cellar deck floor; 1 m vector = 5 ms-1

53



54

Figure A1.2.28 Experimental velocity vectors 1.00 m above mezz deck floor; 1 m vector = 1 ms-1

Figure A1.2.29 CFD velocity vectors at experimental measuring positions 1.00 m above mezz deck floor; 1 m vector = 1 ms-1

55

Figure A1.2.30 Experimental velocity vectors 2.01 m above mezz deck floor . 1 m vector = 1 ms-1

Figure A1.2.31 CFD Velocity vectors at experimental measuring positions 2.01 m above mezz deck floor; 1 m vector = 1 ms-1

56

Figure A1.2.32 Experimental velocity vectors 2.82 m above mezz deck floor; 1 m vector = 1 ms-1

Figure A1.2.33 CFD Velocity vectors at experimental measuring positions 2.82 m above mezz deck floor; 1 m vector = 1 ms-1

57

Figure A1.2.34 CFD velocity predictions 2.01 m above mezz deck floor; 1 m vector = 5 ms-1

58

Figure A1.2.35 Experimental measurements showing air speeds above and below 0.5 and 2.0 ms-1, 2.01 m above mezz. deck floor


59

A1.2.4 Sensitivity tests

Figure A1.2.37 CFD Velocity vectors on fine mesh at experimental measuring positions 2.01 m above cellar deck floor; 1 m vector = 1 ms-1

60

A1.3 GENERAL WIND CONDITIONS

A1.3.1 Low wind speed cases

Figure A1.3.1 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: north, wind speed 6.5 knots

Figure A1.3.2 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: northeast, wind speed 6.5 knots

61

Figure A1.3.3 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: east, wind speed 6.5 knots

Figure A1.3.4 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: southeast, wind speed 6.5 knots

62

Figure A1.3.5 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: south, wind speed 6.5 knots

Figure A1.3.6 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: southwest, wind speed 6.5 knots

63

Figure A1.3.7 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: west, wind speed 6.5 knots

Figure A1.3.8 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: northwest, wind speed 6.5 knots

64

A1.3.2 High wind speed cases

Figure A1.3.9 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: north, wind speed 14 knots

Figure A1.3.10 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: northeast, wind speed 14 knots

65

Figure A1.3.11 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: east, wind speed 14 knots

Figure A1.3.12 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: southeast, wind speed 14 knots

66

Figure A1.3.13 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: south, wind speed 14 knots

Figure A1.3.14 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: southwest, wind speed 14 knots

67

Figure A1.3.15 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: west, wind speed knots

Figure A1.3.16 CFD predictions of ventilation flow 2 m above cellar deck floor. Wind direction: northwest, wind speed 14 knots

68

A1.3.3 Tracer gas calculations

Figure A1.3.17 Tracer gas calculation 90 s after release of gas. Wind direction: southeast, wind speed 6.5 knots

A1.4 JET RELEASE PREDICTIONS

69

Figure A1.4.1 Test D1: Predicted gas concentration (%LEL) on cellar deckWind: westerly at 6.5 knots; Release: 1 kgs -1 for 5 mins;

(subsequent Figures use same scale)

70

Figure A1.4.2 Test D2: Predicted gas concentration (%LEL) on cellar deckWind: westerly at 6.5 knots; Release: 1 kgs -1 for 2 mins;

(see Figure A1.4.1 for scale)

71

Figure A1.4.3 Test D3: Predicted gas concentration (%LEL) on cellar deckWind: south-easterly at 6.5 knots; Release: 1 kgs -1 for 2 mins.


72

Figure A1.4.4 Test D4: Predicted gas concentration (%LEL) on cellar deckWind: northerly at 6.5 knots; Release: 1 kgs -1 for 2 mins.


73

Figure A1.4.5 Test D5: Predicted gas concentration (%LEL) on cellar deckWind: westerly at 14 knots; Release: 1 kgs -1 for 2 mins.


74

Figure A1.4.6 Test D6: Predicted gas concentration (%LEL) on cellar deckWind: south-easterly at 14 knots; Release: 1 kgs -1 for 2 mins.


75

Figure A1.4.7 Test D7: Predicted gas concent ration (%LEL) on cellar deck Wind: south-easterly at 6.5 knots; Release: 0.1 kgs -1 for 2 mins.


76

Figure A1.4.8 Test D8: Predicted gas concent ration (%LEL) on cellar deckWind: westerly at 6.5 knots; Release: 0.1 kgs -1 for 2 mins.


77

Figure A1.4.9 Test D9: Predicted gas concent ration (%LEL) on cellar deckWind: south-easterly at 6.5 knots; Release: 5 kgs -1 for 2 mins.


78

Figure A1.4.10 Test D10: Predicted gas concent ration (%LEL) on cellar deck Wind: northerly at 6.5 knots; Release: 5 kgs -1 for 2 mins.


79

A2 FIGURES: REMEDIAL MEASURES

One minute

Two minutes

Three minutes

Without Fan With Fan

Figure A2.3.1 Comparison of tracer gas calculations with and without ventilation fan. The scale is the same as that used in Figure A1.3.17

80

0 1 2 3 4 5 6

Time (mins)

0

2000

4000

6000

8000

10000

12000 l

>0.75 (remedial) >0.50 (remedial) >0.25 (remedial) >0.75 >0.50 >0.25

Vo

ume

(m^3

)

Figure A2.3.2 Volume of tracer gas above a given concentration with and without remedial measure in place

81

Figure A2.3.3 Test D2: Predicted gas concent ration (%LEL) on cellar deck With air vent; Wind: westerly at 6.5 knots; Release: 1 kgs -1 for 2 mins;


Published by the Health and Safety Executive 10/05

RR 402

RR402 Natural ventilation of offshore modules [5.6MB]

Documents

Transcript of RR402 Natural ventilation of offshore modules [5.6MB]