Quantitative Risk Analysis of Oil and Gas Fires and Explosions for FPSO Systems in China

Abstract

The LH11-1 FPSO is an 80,000 t cylindrical structure that is responsible for the processing, storage, and offloading of process oil from existing and newly developed oilfields. In this paper, a full probabilistic analysis was developed based on very detailed CFD simulation results to evaluate ventilation, gas dispersion, explosion, and fire scenarios. A detailed fire and explosion risk analysis of LH11-1 FPSO was performed based on NORSOK Z-013 and FABIG Technical Note 11. The risk-based calculations were performed applying FLACS, KFX, and DNV GLEXPRESS Fire. Finally, the oil and gas dispersion, fire, and explosion consequence risks were calculated under the credible combination of leak frequency and leak location. By this probabilistic risk analysis, it was found that the west wind could generate optimal ventilation conditions for the topside process area of FPSO compared to other wind directions, while the hull region was poorly ventilated for all wind directions. The explosion risk analysis showed that the FPSO system would take no action in terms of explosion risk according to the acceptance criteria of 10⁻⁴ occ/year. Meanwhile, the fire risk analysis demonstrated that the PR1 first and second deck, PR2 first and second deck, process deck, offloading on starboard, and the base of the flare tower would have different impacts from the fire, and the PR1 first deck and PR2 first deck at the topside deck would be severely impacted regions.

1. Introduction

Floating Production Storage and Offloading (FPSO) systems are offshore oil and gas production and exploration facilities, which are capable of floating production, storage, and offloading. Due to the advantage of the huge storage capacity, the adaptability for water depth, the mobility, and the low cost and high efficiency for construction, a large number of FPSOs have been applied for the exploration and production of deep-water oil and gas, especially the unconventional oil and gas reserves under the marine region over 5000 feet [1,2]. Such deep-water oil and gas exploration would contribute to the increasingly severe global petrochemical energy supply situation [3,4]. However, FPSO is naturally a complex system integrating all kinds of processing units and existing in a high-temperature and high-pressure processing environment, and so exists a relatively high accident likelihood in terms of oil and gas leakage [5]. As we all know, the dispersing of oil and gas would form a flammable cloud and finally pose a potential fire and explosion risk. In February 2015, an FPSO explosion accident in the Brazilian offshore sector resulted in nine fatalities, and this accident consequence was close to that of the Macondo disaster in 2010 [6,7].

In order to mitigate fire and explosion disasters for FPSO systems, quantitative risk assessment (QRA) has become a promising tool for determining the Design Accidental Loads (DALs) for critical regions and structures to improve the platforms’ safety levels. Different from the traditional qualitative risk analysis approach, QRA approaches could integrate the probabilistic modeling of accident frequencies and CFD-based consequence modeling of accident scenarios, which can provide an accurate and efficient probabilistic analysis result for the safety design and decision [8,9]. Until now, a large number of fire and explosion quantitative risk analyses have been conducted for all kinds of offshore platforms [9,10,11,12,13,14,15,16,17]. Therein, J. Li et al. [10] conducted a gas dispersion and explosion risk analysis for FLNG platforms with and without blast walls and determined the optimal blast wall layout with lowest the explosion overpressure exceedance frequencies. A. Rajendram et al. [13] applied a CFD-based Fire Dynamic Simulator for the fire risk modeling of offshore platforms and compared and analyzed the accuracy of the CFD fire simulation results. S. Park et al. [14] conducted a risk analysis of vapor cloud explosions for FLNG topside liquefaction modules and assessed the effect on adjacent modules under the corresponding explosion based on the finite element approach. Nowadays, most literature considers and assesses the one risk of fire or explosion accidents. However, in the FPSO accident scenario. There may be a cascading effect of oil and gas dispersion, fire, or explosion, and this accident result can often be more severe than their individual impacts [18,19]. Meanwhile, risk analysis targeting FPSO has mostly focused on the traditional ship-type FPSO, and the risk investigation on circular FPSO is very limited [20,21]. It can be seen that the structural complexity of the cylindrical FPSO is higher, and the gas leakage in the vertical direction needs to be considered due to the facility’s layout. Therefore, the gas cloud distribution is more complicated, increasing the factors considered in probabilistic analysis. Thus, conducting the full probabilistic risk analysis of fires and explosions will have important reference value for the evaluation of the risk to specified individuals and overall potential loss of life per annum, as well as the overall safety and ventilation design of FPSO.

To explore the integrated impact of oil and gas dispersion, fires, and explosions, relevant research has proposed the full probabilistic analysis approach. The main research basis for the probabilistic risk analysis comes from databases, such as the statistical data of the leakage probability of DNV.GL, CNOOC, etc. The full-probability risk analysis based on real statistical data can accurately and intuitively calculate the individual and overall risks of the corresponding modules [22,23,24]. J.K. Paik et al. [25,26] proposed a fire and explosion risk quantitative assessment procedure, and performed the occurrence frequency and design loads of fires and explosions for the topside process region of FPSO. T. Baalisampang et al. [18] proposed a risk assessment methodology considering an evolving accident scenario from gas dispersion to fire and explosion. This result showed that pool fires can transit to an explosion under meeting certain conditions, and so the fire and explosion risk consequence results should be comprehensively considered. Moreover, for the specific FPSO system, the topside region contains several processing installations and units, and so there is a high probability of oil and gas leakage [27,28]. Therefore, most quantitative risk assessments mainly focus on the effect of the fire and explosion on these regions, which results in the lack of completeness of corresponding risk assessments for FPSO systems [29].

Under the defined acceptance criteria, this paper aims to conduct fully probabilistic fire and explosion analyses and determine the DALs of fire and explosion accidents for the process region and adjacent region of LH11-1 FPSO, which is one of the main offshore production facilities in the LH11-1 oilfield. Different from the traditional ship-type FPSO, the LH11-1 FPSO is a cylindrical structure, which leads to cases where some reference data cannot obtain specific values. Therefore, this paper evaluates the applicability of the industry data from the Hydrocarbons Release Database (HRDB 2006–2015), and, when necessary, factor corrections are conducted based on the engineering judgment and the risk assessment experience of DNV GL in the field of marine engineering. This study considers the oil and gas leak scenario under the combination of leak frequency and leak location. Then, based on these credible accident scenarios, CFD-based consequence modeling was conducted for the assessment of ventilation, dispersion, fires, and explosions, and consequences with overpressure and drags of explosion, heat dose, and fire duration were obtained. Finally, based on the EXPRESS and EXPRESS fire, a probabilistic assessment for the explosion and fire results was conducted for determining the fire and explosion exceedance curves and DALs according to acceptance criteria for the LH11-1 FPSO system.

2. Full Probabilistic Risk Analysis of Fires and Explosions

2.1. Main Steps of Risk Analysis

Due to presenting different effect phenomena and damage mechanisms between fires and explosions, accident consequence modeling was divided into explosion modeling and fire modeling. The full probabilistic risk assessment procedure included geometry modeling, leak frequency, ignition assessment, ventilation and dispersion modeling, explosion modeling, fire modeling, and risk assessment. Each part is described in detail as follows:

• Geometry modeling. We imported the provided geometry model into the FLACS/KFXs model and made the necessary changes and complements (since the 3D model is not completed yet). The small pieces (i.e., small-scale geometry important to the pressure build-up) are added based on the DNV GL internal database, which contains the average amount of small pieces of similar installations, to represent the expected congestion level.
• Leak frequency. A detailed parts counting of all hydrocarbon-carrying equipment (valves, flanges, vessels, pumps, etc.) is performed based on the engineering diagrams. The applied documentation and respective versions were detailed in the FERA Study Basis. The parts counting is performed segment-by-segment (section between SDV valves). The results from the parts counting are inserted into the software DNV Leak, which applies the Hydrocarbons Release Database (HRDB 2006–2015), to calculate the leak frequency based on the amount of equipment. Pipe lengths are estimated considering the approximate distance between the main equipment and modules (applying a correction factor) in order to calculate inventories and frequency estimations.
• Ignition assessment. In the present work, the ignition sources and the ignition mitigation systems have been obtained in detail and modeled in EXPRESS and EXPRESS FIRE. The methodology applied is developed during a Joint Industry Project (JIP). The ignition probability model included in the DNVGL EXPRESS tool (TDIIM—Time-Dependent Internal Ignition Model) is applied for jet/gas releases. The current assessment considers UKOOA look up correlations to estimate the ignition probability of liquid releases.
• Ventilation and dispersion modeling. Eight wind directions are applied for the ventilation simulations. Representative segments in the installation are those that have the most critical combination of inventory and leak frequency. Additionally, the position of the leak location is also considered for the representative leak selection. Therefore, the leak positions selected are in the representative segments and the dispersion scenarios are the combination of leak rates and wind speeds that can create larger and the most critical clouds.
• Explosion modeling. Scenarios are selected with distribution in cloud sizes, cloud location, and ignition location. Results in terms of pressure on large surface sub-panels (4 × 4 m) and drag and pressures at selected monitor points in each area are obtained for each scenario. Loads on the most critical elements are further selected and a full probabilistic analysis is performed on these loads.
• Fire modeling. Several fire scenarios are simulated with KFX (jet and pool fires) considering different jet fire directions and leak rates. The wind direction and speed selected are given by the most common weather conditions for the jet fire scenarios; for pool fire scenarios, the worst-case condition considering calm weather conditions is considered. These different fire scenarios, with constant leak rates, are chosen to mimic the development of a real transient HC leak. Constant leak rates are simulated instead of a fully transient leak to reduce the computational times. The fires out of the constant leak rates are then considered as snapshots of real transient HC leaks. The fire frequency is calculated based on the leak frequency and the ignition probability (fire frequency = leak frequency × ignition probability) for each process segment using the DNV GL EXPRESS Fire approach.
• Risk assessment. DNV GL’s EXPRESS and EXPRESS Fire are applied in the assessment. Inputs are functional descriptions of the cloud size, explosion pressure obtained from the results of the ventilation, dispersion, and explosion simulations. Additionally, radiation loads derived from fire simulations are applied to the calculated probabilistic heat doses associated to fire scenarios. These are explicit functions (named “response surfaces”) that output the cloud size, explosion pressure, and radiation loads as mathematical functions of the dependent variable, such as the wind speed, wind direction, leak rate, leak direction, and leak location. For each area/target, new response surfaces are fitted to the CFD results, reflecting the unique characteristics of the geometry. Further input to EXPRESS is the probability distributions of the dependent variables. The wind probability is obtained from the wind rose.

2.2. EXPRESS/EXPRESS Fire

EXPRESS is a model for the calculation of explosion/fire risk in gas process modules/areas. The EXPRESS program includes consequence models for a transient general gas leak, including the depressurization of the ESD segment, leak rate calculations, gas dispersion modeling, gas detection, ignition, and explosion pressure models. The mathematical models are developed as explicit formulae, also called response surfaces, and are well-suited for running thousands of Monte Carlo simulations quickly. These response surfaces are applied in the EXPRESS program, where the pressure exceedance curves are found to be applied to the ProExp Monte Carlo engine. EXPRESS Fire takes into account the heat radiation history curves, leak frequencies, inventory, ignition probability, characteristics of the segment feeding the fire (pressure, temperature, composition, time to isolation, and blowdown), and runs a series of Monte Carlo simulations to generate the heat dose exceedance curve. The combination of the exceedance curve with the dimensioning criterion (general criteria, probabilistic DAL assessment) gives the Design Accidental Load for the process area considered.

An essential input to the response surfaces is the results from the CFD ventilation, dispersion, explosion, and fire simulations. By using CFD results from the actual installation, generic formulations are avoided, and the need to make conservative assumptions is reduced. Due to its general consequence model and the inclusion of explosion mitigation efforts in the models, EXPRESS is ideal to determine the most effective fire and explosion mitigation strategy. The following effects can be investigated using this methodology:

• Air ventilation. The effect of opening wind walls, louvers, etc. may be obtained through changes in ventilation and cloud size, and give the total effect on the pressure exceedance curves;
• Confinement. The effect of plated or grated decks, wall configurations, etc. may be obtained through changes in the explosion pressure and partly in cloud size, and give the total effect on the pressure exceedance curves;
• Protection systems. Gas detection settings, shutdown, blowdown, and shutdown of ignition sources may be modeled in detail with the transient model. The effect of changed philosophy or settings may be found by simple parameter changes in EXPRESS;
• Ignition sources. Different ignition sources are easily implemented in EXPRESS and the influence on the pressure exceedance curves may be obtained.

3. Probabilistic Explosion and Fire Simulation of LH11-1 FPSO System

3.1. Modeling of LH11-1 FPSO System

3.1.1. Geometry Model

The LH11-1 FPSO is mainly responsible for the processing, storage, and offloading of process oil from existing and newly developed oilfields. This system includes a series of oil and gas processing units, which are shown in

Table 1. Process and utility systems of LH11-1 FPSO.

Process Systems	Utility Systems
Crude oil-treatment system Fuel oil-treatment system Gas compressor system Gas desulfurization system Fuel gas system Produced water-treatment system Flare system Open drain system Chemical injection system	Diesel system Aviation fuel system Utility air and instrument air system Nitrogen system Inert gas system Sea water system Fresh water system Heating medium system Gas turbine generator system and crude oil separators

. Moreover, this geometry model for LH11-1 FPSO was constructed according to the information regarding the conversion from a 3D model and layout arrangements. The general layout of LH11-1 FPSO is presented in Figure 1.

Referring to the above general layout, the processing region and equipment layouts were initially determined. In order to fully consider the congestion level of the FPSO system, some small-scale elements and pipes need to be added to the geometry model, which can accurately calculate the overpressures and accurately reflect the final as-built geometry of the installation. According to DNV GL’s experience of detailed geometry modeling of many previous facilities, the average congestion level was set to 2.9 m/m³ [30]. Based on the DNV GL internal database, this 3D model applied many small-scale elements for representing piping, fittings cable trays, railings, etc., which are shown in Figure 2.

3.1.2. Weather Statistics

The wind distribution is applied in the ventilation and dispersion assessment for LH11-1 FPSO in order to calculate the expected occurrence frequency of the evaluated conditions. Figure 3 and Figure 4 summarize the FPSO orientation and wind speed distribution statistics according to True North. Figure 5 presents the wind occurrence frequency related to project North and considering only the eight main wind directions at 45°.

3.1.3. Target Definition

Due to the geometrical characteristics, the FPSO system was divided into different monitored areas as detailed below: (1) Hull Deck, (2) Process Module 2-1st Deck (PR 2_1st), (3) Process Module 2-2nd Deck (PR 2_2nd), (4) PD-Process Deck (Desulfurization Module and Fuel Oil), (5) Process Module 1-1st Deck (PR 1_1st), (6) Process Module 1–2nd Deck (PR 1_2nd), (7) Offloading Starboard, (8) Offloading Portside, (9) Firewall, (10) Power-Generation Module, (11) Electrical Module 1, (12) Electrical Module 2, (13) Utility Module, (14) Living Quarter, (15) Helideck, (16) Base of Flare Tower, (17) Closed Drain Area, (18) Flare KO Drum Area, (19) Risers, (20) Topside Deck, (21) Crane Portside, and (22) Crane Starboard. Figure 6 shows the regions defined for the probabilistic explosion and fire analysis.

3.2. Determination of Credible Leak Scenario

3.2.1. Determination of Leak Frequency

According to the engineering diagrams of the FPSO system, a detailed part counting of all oil and gas carrying equipment (valves, flanges, vessels, pumps, etc.) was performed. Referring to the above-defined target regions, the calculation of the leak frequency was conducted segment-by-segment and sub-segment-by-sub-segment. The results from the parts counting were processed to calculate the leak frequency based on the amount of equipment for each sub-segment by using the software DNV GL LEAK, which applied the Hydrocarbons Release Database (HRDB) 2006–2015 version created based on the leak statistics of offshore installations. By the leak frequency calculation, the main leak accident was selected and determined, which is shown in Figure 7.

3.2.2. Determine of Leak Location

Probabilistic FERA analysis using CFD combined with EXPRESS/EXPRESS fire methodology considers a set of representative leak locations to model the expected consequences in terms of gas dispersion, radiation, and overpressure that can reach the target areas to calibrate the models. The aim of this section is to present the release scenarios that were evaluated using CFD for the FERA of LH11-1 FPSO.

According to the leak frequency calculation, a total of six representative leak locations were selected, which are shown in Figure 8. For each segment (sub-segment), this includes the evaluation of different release rates. It should be noted that the CFD consequences (gas dispersion, jet, or pool fire consequences) evaluated for the representative segment were used to evaluate the risk contribution of all of the remaining segments not explicitly modeled with CFD. In this way, the overall fire and explosion risk figure will be accurately obtained.

3.3. CFD-Based Consequence Simulation

3.3.1. Ventilation Modeling

A total of eight simulations were performed to assess the ventilation on LH11-1 FPSO based on the wind directions N, NE, E, SE, S, SW, W, and NW (related to project north) as detailed in

Table 2. CFD ventilation simulations performed.

Case#	Ventilation Scenario	Wind Direction (Project North)	Wind Speed (m/s)
01	660,001	N	6.0
02	660,002	NE	6.0
03	660,003	E	6.0
04	660,004	SE	6.0
05	660,005	S	6.0
06	660,006	SW	6.0
07	660,007	W	6.0
08	660,008	NW	6.0

. Plots of the wind absolute velocity (UVW) are presented in Table 2 as 2D slices (XY Plane) in three elevations, Z = 34 m, Z = 40 m, and Z = 45 m. Areas colored as dark red present wind speeds above 6.0 m/s, while the white areas present wind speeds below 0.2 m/s and thus are regions of potential gas accumulation (stagnant zones).

Figure 9 displays the 2D slice distribution of the simulated wind velocity of ventilation case 660,001 under different XY slice elevations. It can be seen that the ambient wind velocity of the FPSO system was about 6 m/s, and the velocity of the inlet was relatively low. In the interior region, the region near the wind inlet had good ventilation conditions, epically the region of XY slice elevation Z = 34 m and Z = 45 m. When the XY slice elevation was set to 34 m, the FPSO region near the wind inlet wind velocity ranged from 1.5–2.0 m/s; when the XY slice elevation was set to 40 m, that of the region near the wind inlet was 2.0–3.0 m/s; when the elevation was 45 m, the wind velocity was 3.0–5.5 m/s.

3.3.2. Dispersion Modeling

The dispersion simulations performed in this study aimed to identify the maximum size of the flammable gas cloud that can be formed from a gas leak in the investigated area. Further, simulations were performed to evaluate how the size of this flammable cloud varies with leak direction, leak rate, wind speed, and wind direction. The understanding of the dependence of these variables on the cloud’s behavior is essential to model the probabilistic assessment.

A set of twenty-four dispersion simulations were performed to determine the behavior of explosive gas cloud build-up originating from each leak location.

Table 3. Dispersion cases set up and Q9 explosive gas cloud result for Leak1.

Case#	Leak Direction (to Project North)	Leak Rate (kg/s)	Wind Direction (to Project North)	Wind Speed (m/s)	Q9 Gas Cloud Size (m3)
661,101	North	25	North	6	210
661,102	South	25	North	6	277
661,103	East	25	North	6	120
661,104	Down	25	North	6	237
661,105	North	50	North	6	378
661,106	South	50	North	6	361
661,107	East	50	North	6	158
661,108	Down	50	North	6	372
661,109	North	25	West	6	128
661,110	South	25	West	6	132
661,111	East	25	West	6	187
661,112	Down	25	West	6	226
661,113	North	50	West	6	199
661,114	South	50	West	6	138
661,115	East	50	West	6	180
661,116	Down	50	West	6	312
661,117	North	25	East	6	206
661,118	South	25	East	6	351
661,119	East	25	East	6	271
661,120	Down	25	East	6	285
661,121	North	50	East	6	238
661,122	South	50	East	6	496
661,123	East	50	East	6	303
661,124	Down	50	East	6	601

presents the cases’ definitions and the results obtained for the simulated cases. From the results, it can be seen that the largest explosive gas cloud was found to be 601 m³ and the average gas cloud size was found to be 265 m³. Figure 10 and Figure 11 show examples of gas clouds obtained during CFD dispersion simulations.

Figure 10 displays the dispersion simulation result of a leakage gas cloud under dispersion case 661124. This result demonstrates that the gas cloud is mainly distributed in PR1_1st deck, PR1_2nd deck, and the firewall, especially in the process module region, where there is a large number of Q9 gas clouds. This gas cloud would gradually extend to offloading portside and the utility module due to the influence of the east wind. Simultaneously, Table 3 demonstrates that this case would produce the maximum Q9 gas cloud volume of 601 m³, which was caused by the combined effect of downwards leakage and wind east direction. Looking at the leakage gas cloud under case 661,117, from Figure 11, it can be seen that this leakage gas cloud was mainly distributed in PR1_1st deck and PR1_2nd deck. Table 3 demonstrates that the Q9 gas cloud volume was 206 m³.

3.3.3. Explosion Modeling

Thirty explosion simulations were performed to determine the overpressures associated with different explosive gas cloud sizes, shapes, positions, and ignition locations.

Table 4. Main explosion simulation results read in all targets for Leak 1.

Case#	Cloud Volume (m3)	Effect Type
		MP (Barg)	Drag (Barg)	PP (Barg)
662,101	2316.6	0.42	0.13	0.40
662,102	2316.6	0.12	0.03	0.09
662,103	2316.6	0.16	0.05	0.12
662,104	2316.6	0.28	0.09	0.18
662,105	2316.6	0.07	0.02	0.06
662,106	1782.0	0.34	0.11	0.23
662,107	1782.0	0.07	0.02	0.04
662,108	1782.0	0.13	0.06	0.11
662,109	1782.0	0.12	0.03	0.07
662,110	1782.0	0.06	0.02	0.04
662,111	1275.0	0.29	0.08	0.22
662,112	1275.0	0.05	0.02	0.04
662,113	1275.0	0.07	0.02	0.06
662,114	1275.0	0.19	0.06	0.11
662,115	1275.0	0.06	0.02	0.06
662,116	960.0	0.26	0.06	0.06
662,117	960.0	0.05	0.01	0.04
662,118	960.0	0.19	0.04	0.15
662,119	960.0	0.24	0.05	0.15
662,120	960.0	0.04	0.01	0.04
662,121	612.0	0.12	0.04	0.04
662,122	612.0	0.02	0.00	0.02
662,123	612.0	0.05	0.02	0.03
662,124	612.0	0.08	0.02	0.07
662,125	612.0	0.07	0.02	0.06
662,126	450.0	0.13	0.04	0.06
662,127	450.0	0.03	0.00	0.02
662,128	450.0	0.06	0.01	0.04
662,129	450.0	0.09	0.02	0.06
662,130	450.0	0.01	0.00	0.01

presents the main results, including pressure on monitor points (MP), drag, and pressure on Panels (PP) considering the peak value registered in all targets defined in this study.

Figure 12 displays the simulation dynamic development of the gas cloud and overpressure wave versus time under explosion case 662,101. It can be seen that the gas cloud expanded slightly outwards due to the influence of the explosion wave and was consumed and finally used up during the process of the gas cloud explosion. As for the overpressure wave, the range of the wave would be extended versus time so that the wave would impact process module 1, process module 2, offloading portside, and the utility module. Table 4 demonstrates that the corresponding pressure value on MP, drag, and PP were 0.42, 0.13, and 0.40, respectively. Among these explosion cases, this overpressure under case 662,101 was the most serious, which was mainly caused by the maximum cloud volume. Moreover, the ignition location had a certain impact on the explosion result. For example, compared with case 2, the ignition position in case 662,101 was closer to the leakage position, so the explosion consequences were more serious.

3.3.4. Fire Modeling

Twenty-five jet fire scenarios and four pool fire scenarios were simulated for Leak 1 to evaluate the radiation loads in the case of the ignition of hydrocarbon oil and gas releases. The jet fire scenarios were performed by varying the leak direction and leak rate, while the pool scenarios were evaluated varying the pool size to represent four different pool sizes for the assessment. The radiation results for all fire simulations performed were post-processed to evaluate the incident radiation loads on the defined target. The simulations performed are detailed in

Table 5. Fire scenarios simulated for Leak 1.

Fire Type	Leak Direction	Wind Direction	Wind Speed (m/s)	Fire Sizes (Leak Rate (kg/s))
Jet	East	North	6.0	2.5	5.0	10.0	25.0	50.0
Jet	West	North	6.0	2.5	5.0	10.0	25.0	50.0
Jet	North	North	6.0	2.5	5.0	10.0	25.0	50.0
Jet	South	North	6.0	2.5	5.0	10.0	25.0	50.0
Jet	Down	North	6.0	2.5	5.0	10.0	25.0	50.0
Pool	N/A	N/A	Calm weather	2.5	5.0	10.0	18.0

.

Figure 13, Figure 14 and Figure 15 display the radiation flux of jet fire of process deck first floor, jet fire of process deck second floor, and pool fire of process deck second floor under different flammable gas leak rates. Therein, the larger leak rate would induce more severe fire heat radiation flux, especially under the leakage rate of 50 kg/s, when the ambient radiation flux of the jet could reach the range of 5 × 10⁴ to 2 × 10⁵ W/m², and that of the pool fire could reach the range of 3 × 10³ to 1.5 × 10⁴ W/m². Moreover, comparing Figure 13 and Figure 14, we could find that the process deck second floor would suffer more heat radiation than the process deck first floor. At the same time, the radiation flux of pool fires is much smaller than that of jet fires.

4. Results and Discussion

The probabilistic analyses for the LH11-1 FPSO were carried out with six different EXPRESS scenarios, corresponding to one for each leak location. All leak scenarios that occurred inside these explosion/fire areas were contributing to the pressure/heat dose on the several analyzed structures. Verified by numerical simulation, leaks smaller than 0.1 kg/s did not contribute to the explosion pressure and, hence, were not included in the analysis. The six leak locations selected as representative were used to assess the contribution of different segments/equipment on the overpressure DAL for each of the regions analyzed.

In terms of explosion DAL, for each of the targets, the pressure on panels (for deck floors, ceilings, or firewalls, when applicable) and monitoring points (for equipment and large pipes) were used in the exceedance curves, produced by DNV GL’s probabilistic approach. Pressure on other targets and regions was provided through correlations of drags, obtained directly from CFD simulations and the probabilistic results obtained for the targets next to them. These correlations would be provided only if the DAL value read at the acceptance criteria in the produced exceedance curves was found to be greater than 0 barg. If the frequency of occurrence was found to be below the acceptance criteria, the drags would be considered to be 0.0 barg.

For the fire DAL calculation, the probabilistic heat dose was obtained based on the monitoring of the radiation heat flux over the targets during the fire scenarios evaluated, also combining leak frequency and ignition probabilities; DNV GL’s EXPRESS Fire probabilistic was also applied to generate the fire exceedance curves.

4.1. Exceedance Frequency Curves

Dimensioning Accidental Loads (DAL) applying a frequency of 10⁻⁴ (1/year) for all monitored targets are presented in this section in terms of the fire and explosion loads. For each region, two exceedance curves are provided: one for overpressure on panels or monitoring points, and another for the probabilistic fire heat dose. The heat dose fire durations were calculated based on 150 kW/m² and 250 kW/m² reference radiation values for pool fires and jet fires, respectively. For each module/area, the event that contributed most to the probabilistic DAL was used as a basis for the probabilistic fire load duration.

Taking process module 1 first deck as an example, the probabilistic analyses results are as follows:

(1)

Explosion DAL

Figure 16 displays the explosion exceedance frequency curve for the PR1_1st under the above-mentioned six leak locations. From it, we could find that the exceedance frequency would decline versus the explosion pressure increases. This frequency was below 1.0 × 10⁻⁵, which could meet the designed accidental load requirement under acceptance criteria. Therefore, the drags will be considered as 0.0 barg.

(2)

Fire Heat Dose DAL

Figure 17 displays the exceedance frequency of the jet fire heat dose and pool fire heat dose under six different leak locations. From it, we could find that, when the leak location was set to leak 3 on PR2_1st deck, the exceedance frequency would be generally greater than the acceptance criteria, and that of the other leak locations was below the acceptance criteria. In order to meet the acceptance criteria for jet and pool fires, the fire DAL could be determined to be 180 MJ/m² for achieving an exceedance frequency below 10⁻⁴ (1/year).

4.2. Summary of Probabilistic Loads

The Dimensioning Accidental Loads (DAL) for the targets identified in the previous sections are presented here. For each target, based on the exceedance curves presented above, the fire and explosion DALs are provided for the acceptance criterion of 1.0 × 10⁻⁴ accidents per year.

4.2.1. Probabilistic Explosion Loads

This section addresses the probabilistic explosion loads obtained in the analysis. The DNV GL program EXPRESS was run for each target to obtain the probabilistic pressure in the respective targets.

Table 6. Summary of results in terms of the Dimensioning Accidental Loads (DAL), pressure on monitor points, and drag in each target.

Load on	Applying 1.0 × 10−4 acc./Year
	DAL (Barg)	Duration (ms)	MP Pressure (Barg)	Duration (ms)	Drag (Barg)
Hull Deck	0.0	-	0.0	-	0.0
PR2_1st deck	0.0	-	0.0	-	0.0
PR2_2nd deck	0.0	-	0.0	-	0.0
PR1_1st deck	0.0	-	0.0	-	0.0
PR1_2nd deck	0.0	-	0.0	-	0.0
Process deck (Desulfurization Module and Fuel Oil)	0.0	-	0.0	-	0.0
Offloading starboard	0.0	-	0.0	-	0.0
Offloading starboard	0.0	-	0.0	-	0.0
Firewall	0.0	-	0.0	-	0.0
Power Gen. Mod.	0.0	-	0.0	-	0.0
Electrical Module 1	0.0	-	0.0	-	0.0
Electrical Module 2	0.0	-	0.0	-	0.0
Utility Module	0.0	-	0.0	-	0.0
Living Quarter	0.0	-	0.0	-	0.0
Helideck	0.0	-	0.0	-	0.0
Base of Flare Tower	0.0	-	0.0	-	0.0
Closed Drain Area	0.0	-	0.0	-	0.0
Flare KO Drum Area	0.0	-	0.0	-	0.0
Risers	0.0	-	0.0	-	0.0
Topside deck	0.0	-	0.0	-	0.0
Portside Crane	0.0	-	0.0	-	0.0
Starboard Crane	0.0	-	0.0	-	0.0

. presents the explosion DAL for the several targets analyzed in LH11-1 FPSO, which are the input data for the design of structure strength and equipment foundation. The DALs in terms of monitoring pressure and drag in Table 6 were valid for all equipment and piping inside the analyzed deck. As presented in Table 6, all DALS were found to be 0.0 barg, since the explosion frequency was below the acceptance criteria, which means that the explosion accident load will not be considered. Table 6 presents the explosion DAL, explosion duration, monitoring pressure, and drag for the above-mentioned FPSO regions. From this table, we can find that the DAL of the FPSO system was 0 barg, which was caused by the explosion frequency being below the acceptance criteria. Therefore, the FPSO explosion risk was within the acceptable range under the acceptance criteria. Looking at the monitoring pressure and drag, the corresponding value was also 0.

4.2.2. Probabilistic Fire Loads

Table 7. Summary of results in terms of the Dimensioning Accidental Loads (DAL) and heat dose (MJ/m²) obtained applying the acceptance criterion of 1.0 × 10⁻⁴ accidents per year.

Load on	Applying 1.0 × 10−4 acc./Year
	DAL (Barg)	Duration (s)
Hull Deck	0	-
PR2_1st deck	180	1200
PR2_2nd deck	20	133
PR1_1st deck	180	1200
PR1_2nd deck	85	567
Process deck (Desulfurization Module and Fuel Oil)	30	200
Offloading starboard	5	33
Offloading starboard	0	-
Firewall	0	-
Power Gen. Mod.	0	-
Electrical Module 1	0	-
Electrical Module 2	0	-
Utility Module	0	-
Living Quarter	0	-
Helideck	0	-
Base of Flare Tower	1.5	10
Closed Drain Area	0	-
Flare KO Drum Area	0	-
Risers	0	-
Topside deck	180	1200
Portside Crane	0	-
Starboard Crane	0	-

presents the DAL and duration of the fire heat dose for the FPSO system under the acceptance criterion. Therein, the heat dose fire durations were calculated based on 150 kW/m² and 250 kW/m² reference radiation values for pool fires and jet fires, respectively. From this table, apart from the process module and topside module, other modules were less affected by the fire; therefore, the corresponding DAL was 0 and the duration was also 0. As for the process module and topside module, the PR2 first deck, PR1 first deck, and topside deck were vulnerable to the fire; therefore, the DAL were determined to be 180 barg and the duration was 1200 s.

5. Conclusions

A detailed fire and explosion risk analysis of the LH11-1 FPSO was performed based on NORSOK Z-013 and FABIG Technical Note 11. The risk-based calculations were performed applying FLACS, KFX, and DNV EXPRESS Fire. The acceptance criterion defined for the current analysis was equal to 1.0 × 10⁻⁴/year for fire/explosion scenarios. This risk analysis mainly assessed the dimensioning accidental loads, including overpressure, drags of explosion, heat dose, and duration of fire. The related conclusions are as follows:

• The ventilation CFD simulation results showed that winds coming from the west (project North) generated good ventilation in the topside region, while winds coming from the east (from project North) generated low ventilation profiles in the topside region, where hydrocarbons can be released. Low-ventilation regions are more favorable for explosive gas formation under low gas release rates, which was the case for most of the gas releases. Additionally, from the ventilation results, it can be seen that the hull region was poorly ventilated for all wind directions considered, since it was almost fully enclosed.
• Probabilistic explosion analysis showed that all of the targets defined in Section 3.3 presented DAL values of 0.0 barg, meaning that there were no actions to be taken regarding explosion events. This was because of the low gas leak frequency, as mentioned previously, and the relatively small gas clouds formed due to the low number of congested/confined FPSO process areas. Additionally, most of the flammable gas is lighter than air, which makes the buoyancy effect take place in the early stages of dispersion and lead the gas outside the more confined/congested regions.
• Probabilistic fire analysis showed that equipment, structures, and/or pipes located at the PR1 first and second decks, PR2 first and second decks, process deck (desulfurization module and fuel oil), offloading on starboard, and the base of the flare tower were impacted on different levels due to fires and presented a fire DAL value different from zero. Among them, the PR1 first deck and PR2 first deck at the topside deck were the most impacted targets, with a DAL value of 180 MJ/m² and a fire duration of 1200 s under a dominating pool fire scenario of 150 kW/m². In the remaining analyzed areas, fire DAL were not observed and presented a DAL value of 0.0 MJ/m².