Weight-Estimation Method of FPSO weight-estimation process is proposed in this study.

Using the
WBS of the FPSO topsides, the corresponding presentation of the
Topsides Considering the Work weight-estimation process makes the process usable in the field
work regarding the WBS-item estimations. Accordingly, estimates
Breakdown Structure of the detailed units (disciplines, modules, and areas) inside the
topside that were previously not possible were performed. In addi-
tion, a prototype program was developed using the proposed
Ki-Su Kim method, and the applicability of the proposed method was eval-
Department of Naval Architecture and Ocean Engineering, uated through the application of three projects.
Seoul National University, [DOI: 10.1115/1.4037828]
1 Gwanak-ro, Gwanak-gu,
Keywords: weight estimation, FPSO topsides, work breakdown

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022

Seoul 08826, South Korea
structure, statistical method, optimization method
e-mail: kisu2511@snu.ac.kr

Myung-Il Roh1
Department of Naval Architecture and Ocean Engineering, 1 Introduction
Research Institute of Marine Systems Engineering, 1.1 Research Background. Normally, floating, production,
Seoul National University, storage, and offloading (FPSO) projects starts with exploration
1 Gwanak-ro, Gwanak-gu, and feasibility studies for reservoirs. If the investigation results
are deemed as valid, an owner (generally an oil company) per-
Seoul 08826, South Korea
forms a front-end engineering design (FEED) considering the var-
e-mail: miroh@snu.ac.kr ious characteristics of the reservoir and the operating conditions
and philosophies of the FPSO. After performing the FEED, the
Sung-Min Lee owner customarily devises a contract with an engineering, pro-
Department of Naval Architecture and Ocean Engineering, curement, and construction (EPC) contractor (generally a ship-
Seoul National University, building company) to proceed with the design details and the
1 Gwanak-ro, Gwanak-gu, FPSO construction. From the standpoint of the EPC contractor,
the FPSO project starts from the contractual agreement with the
Seoul 08826, South Korea
FPSO owner. Before the contract is devised, the EPC contractor
e-mail: heaven603@snu.ac.kr receives the FEED results such as the specification documents,
drawings, model tests, analysis reports, and main equipment list,
Han-Sung Kim among others, from the owner. The EPC contractor conducts a
Technology Strategy Department, detailed review of the documents for the cost estimation during a
Daewoo Shipbuilding and Marine Engineering, period from one month to three months. After the review process,
125 Namdaemun-ro, Jung-gu, the EPC contractor decides whether to contract the FPSO project.
After the contractual agreement, the EPC contractor proceeds
Seoul 04521, South Korea
with the project. In this process, the detail design, production
e-mail: kzephyrk@dsme.co.kr design, procurement, and construction are included. When per-
forming the cost estimation, the EPC contractor considers the
Hyunsik Ahn material amount and the labor costs. The material cost is mainly
Technology Strategy Department, determined by the FPSO weight in consideration of the other phi-
Daewoo Shipbuilding and Marine Engineering, losophies of the owner, and the labor costs are largely affected by
125 Namdaemun-ro, Jung-gu, the type, size, and complexity of the FPSO. In this situation, the
weight of the FPSO is the main driver of the cost estimation; fur-
Seoul 04521, South Korea
thermore, the weight is also very important in the controlling of
e-mail: ahnmotors@dsme.co.kr the progress of the project. Accordingly, the accuracy of the
weight estimation is a crucial factor, not only in the cost-
estimation stage, but also over the entire duration of the project.
With the recent international economic downturn, most engineer- During the cost-estimation stage, the weight estimation is
ing, procurement, and construction (EPC) contractors are incur- mainly performed using the FEED documents and data of the pre-
ring deficits in their floating, production, storage, and offloading vious projects. The weight that is calculated by the weight estima-
unit (FPSO) projects. Numerous reasons underpin these situa- tion, however, is commonly not matched with the actual weight
tions. One of the most important reasons is the cost-estimation that is calculated using the production-design drawings and the
failure. The cost estimation is the key contractual point and three-dimensional computer-aided design data. The weight mainly
mainly depends on a weight estimation of the FPSO topsides. consists of the topside weight and the hull weight. Between the
Because the topsides contain a lot of equipment and complex two, the topside weight mainly causes the uncertainty and the
structures, it is very difficult to make an estimation at the contrac- latent growth of the FPSO weight. In general, the topsides include
tual stage. To overcome this problem, many methods have been various subcomponents such as the process modules, utility mod-
proposed to estimate the weight of offshore topsides; however, ules, pipe rack, accommodations, and flare tower. The hull
most of the methods involve the top–down approach, making it includes the machinery space, cargo oil tank, and produced water
difficult to obtain a sufficiently accurate prediction for field-work tank. For the various modules and equipment, a lot of the equip-
usage in terms of the weight estimation. Therefore, a ment pipes, pipe racks, and supports are installed in the topsides,
work breakdown structure (WBS) for the performance of the and it causes numerous uncertainties and weight growth as a con-
sequence. On the basis of the analysis and observations of the
projects in the independent project analysis database, the weights
1
Corresponding author. of the one-half topsides grow by more than 10% from the contrac-
Contributed by the Ocean, Offshore, and Arctic Engineering Division of ASME
for publication in the JOURNAL OF OFFSHORE MECHANICS AND ARCTIC ENGINEERING.
tual stage to the construction stage [1]. To overcome these issues,
Manuscript received March 27, 2017; final manuscript received August 16, 2017; a systematic method for the weight estimation of the FPSO top-
published online October 4, 2017. Assoc. Editor: Jonas W. Ringsberg. sides, like the suggestion of this study, should be provided.

Journal of Offshore Mechanics and Arctic Engineering FEBRUARY 2018, Vol. 140 / 014501-1
C 2018 by ASME
Copyright V
 Table 1 Summary of the related works and a comparison with this study

Studies Application target Used data Category of estimation method Estimation approach

Bolding [2] FPSO topsides and modules 60 offshore topside data Particular-based method Top–down
(volumetric-density method)
Rasmussen [3] Offshore topside and equipment 1 offshore topside data Statistical method Top–down
Aasen and Bjorhovde [4] Lightweight of ship 17 ship data Statistical method Top–down
Ha et al. [6] FPSO topsides 30 FPSO topside data Statistical method Top–down
Ha et al. [10] FPSO topsides 37 FPSO topside data Optimization method Top–down
This study FPSO topsides, disciplines, areas, Three FPSO topside data Statistical and optimization Semi bottom–up
modules, and WBS items methods

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022

1.2 Related Works. Numerous approaches exist for the substantial method for the weight estimation in the contract phase
prediction of the FPSO-topside weight. Generally, many of the is lacking, which is a risk for both the EPC contractor and the
methods are based on the systemic particulars like the ship and owner.
the offshore platform. Bolding [2] used the volumetric-density To overcome these difficulties, a systematic method for the
method that belongs to a particular-based method to estimate the weight estimation of the FPSO topside is proposed in this study.
weight of the FPSO topsides, where the data of 60 offshore proj- Generally, the statistical and optimization methods are more
ects were used to make the bulk-factor weight of each system and powerful than the particular-based methods regarding the identifi-
discipline that consist of the FPSO topside. Using the bulk cation of the highly correlated elements and a regression equation
factors, the weights of the other FPSO topsides were estimated. of a system in terms of the accuracy of the equation. To use these
Rasmussen [3] estimated the weight of offshore topsides using an methods to find the appropriate equations for the FPSO topsides, a
equation that was made by a regression analysis of 10  106 data preceding analysis of the work breakdown structure (WBS) for
samples, which had been cloned from one offshore topside using the FPSO topside is required. Therefore, the WBS for the weight
the Monte Carlo method and the main equipment list factors. He estimation of the FPSO topside for which a synthesis analysis of
first made the equation for the equipment weight, and then, he the weight data of the FPSO topsides has been proposed, and
estimated the topside weight by multiplying several factors with the weight equation for each WBS item was proposed using the
the equipment weight. Aasen and Bjorhovde [4] also used a statis- statistical and optimization methods. Through the proposed pro-
tical method to estimate the lightweights of ships. They first cess for the FPSO-topside weight estimation, a systematic weight-
defined the breakdown structure for the ship subcomponents, estimation method named semi-bottom–up approach is presented
followed by a calculation of the coefficient (similar to the factor) here.
for each component using the Watson and Gilfillan method [5]. The reminder of this paper is organized as follows: Sec. 2
Last, the equation for the lightweight of a ship was determined describes the WBS configuration of the FPSO topside; in Sec. 3,
using a coefficient regression analysis. Ha et al. [6] suggested two the estimation method for each WBS item is presented; in Sec. 4,
methods for the FPSO-topside weight estimation. Based on a com- the proposed estimation method is applied to an FPSO topside,
bination of the correlation analysis and the multiple regression and a discussion is presented for which the application results are
analysis, an equation that contains several parameters such as the used, and the final section presents the overall results of this study
main dimensions and specifications for the FPSO-topside weight and a brief discussion of the remaining future works.
was proposed. In addition, an optimization method, genetic pro-
gramming (GP) [7–9], for the construction of the equation for the 2 Hierarchical Work Breakdown Structure for the
FPSO-topside weight was proposed. Ha et al. [10] compared the Floating, Production, Storage, and Offloading Topsides
accuracy of these two methods with the same FPSO records from
the past. Table 1 shows a summary of the related works and a The main challenge for the floating offshore structures is the
comparison of the works for the weight estimations of ships and development of a WBS that defines the logical relationship
offshore platforms. between all of the project components and the cost elements of
As mentioned earlier, there are many methods including the the project, while its natural extension with the design and con-
particular-based, statistical, and optimization methods for the esti- struction requirements is maintained [16]. In this section, the pro-
mation of the weight of FPSO topsides. The particular-based posed hierarchical WBS for the FPSO topsides is in consideration
method uses an empirical formula based on the systemic particu- of the efficiency of the weight-estimation methods. In addition,
lars such as the length, height, volume, and density [6]. The statis- the concept of the weight-estimation process for the FPSO top-
tical method is based on the previous actual reference. It looks sides is presented using the WBS. To use the proposed method of
similar with the particular-based method, but it is strictly based on this study effectively, it should resemble the field works; there-
a statistic technique such as regression analysis [11]. The optimi- fore, a thorough understanding of these works is preferred.
zation method uses an optimization technique when it makes an A WBS means a hierarchical structure in which an object is
equation for the weight estimation [10,12–15]. The accuracy and divided into detailed components, based on the specific criteria.
applicability of the proposed methods can differ depending on the Since the criteria depend on the needs, most of the WBSs are dif-
systemic characteristics and resource data. However, the main ferent from each other according to their purposes and applica-
focus of most of the existing research studies is the estimation of tions. Therefore, a WBS for the weight estimation of
the total systemic weight (so-called top–down approach), and a FPSO topsides might be different according to shipbuilding or
detailed weight estimation regarding the FPSO topside or the sub- engineering companies.
systemic discipline (so-called bottom–up approach) is nonexis-
tent. The bottom–up approach has not been properly researched, 2.1 Concept of the Weight-Estimation Process for Floating,
either because most of the ship and offshore-platform information Production, Storage, and Offloading Topsides. Conventionally,
is not publicly disclosed, or it is difficult to analyze the discipline most of the cost-estimation processes are performed by each disci-
and subsystem of the FPSO topside. In addition, it is very difficult pline in the field despite whether the discipline categorization is
to reassemble the weight data in terms of the possible analytical different from each company. In this process, the weight estima-
area for not only owners but also EPC contractors. So far, a tion is also performed for each discipline. The total weight of the

014501-2 / Vol. 140, FEBRUARY 2018 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022
Fig. 1 Concept of the weight-estimation process for the FPSO topsides

FPSO topside is estimated by integrating the results of each disci- the purpose of estimation. From this perspective, an appropriate
pline. The weight estimation of the field works, however, is not WBS for the weight estimation of the FPSO topsides should be
subsequently performed correctly according to that which is men- made to achieve a precise estimation. The works division for each
tioned previously; therefore, the proposed FPSO-topside weight- company, however, is different, and this also applies to the FPSO
estimation process is one of the effective methods also applicable philosophies, each of which varies from project to project.
to the field works. Consequently, it is difficult to derive a suitable WBS for the
First, the as-built data that will be used to formulate the weight weight estimation. In this study, the proposed WBS for the weight
equation are reassembled to correspond to the WBS. In the estimation describes the general and diverse FPSO characteristics.
reassembling task, the data of the weight and the parameters are Functionally, the FPSO topside consists of areas. The areas’
reorganized following the implementation of the WBS. Using the roles cover the FPSO functions such as oil processing, electricity
reassembled database that contains the as-built records, the weight generation, and pipe support. By dividing the topsides into the
equation for each WBS item can be made based on the statistical areas, the weight and parameters can be effectively organized fol-
and optimization methods. After formulating all of the equations lowing each area function. Depending on the company circum-
for each WBS item, the WBS can be used for the estimation of stances, a various combination of areas need to be considered. In
the new project. In addition, the estimated weight can be catego- this study, six areas were proposed to divide the function of the
rized according to the discipline and used to analyze the accuracy topside. The details of each area are listed in Table 2.
of the field works. Figure 1 shows the concept of the weight- The areas can also be divided into modules for specific func-
estimation process that is proposed in this study. tions. Typically, the construction unit has been integrated into the
modules, and the module’s roles are for specific FPSO functions
such as oil separation, water injection, and power generation.
2.2 Work Breakdown Structure for the Weight Estimation These functionalities mean the modules comprise several equip-
of the Floating, Production, Storage, and Offloading Topsides. ment and facilities so that each module can function effectively.
The WBS is a vehicle to deconstruct an engineering project into Overall, regarding the functional hierarchy of the FPSO topside,
the relevant components like the subprojects, tasks, subtasks, and the topside consists of areas and the areas contain several mod-
work packages [17]. For the fine estimation of some systemic val- ules. In Fig. 2, categories with dashed lines (topside, area, and
ues, it is essential to divide the system into proper categories for module) show the functional hierarchy of the FPSO topsides.

Table 2 Areax list of the FPSO topsides

Area Description

Power-utility area Set of power-generation modules

Utility area Set of utility modules including the water-injection and electricity-building modules
Process area Set of oil-processing modules including the separation and compression modules
Pipe-rack area Set of pipe-rack modules that is placed along the middle of the FPSO topsides
Flare-tower area Area of the flare tower
ETC area Set of extra modules that is not included in the areas described above

Journal of Offshore Mechanics and Arctic Engineering FEBRUARY 2018, Vol. 140 / 014501-3
 module; therefore, the WBS items in each module should be esti-
mated using the appropriate dataset. In this study, the modules in
each area share similar functions and characteristics so that the
WBS-item weights are differently estimated according to the area.
By summarizing the weight-estimation process of this study,
the weights of the WBS items are estimated using the module data
in the same area. By adding the WBS items in each category that
is shown in the WBS hierarchy, the modules, areas, and discipline
weights can be estimated. Finally, by adding all of the discipline
weights, the FPSO-topside weight can be calculated.

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022

3 Weight-Estimation Method for the Floating,
Fig. 2 WBS hierarchy of the FPSO topsides for the weight
Production, Storage, and Offloading Topsides
estimation As mentioned in Sec. 2, the WBS items represent a weight-
estimation unit; that is, equations were formulated for the weight
Different from the functional hierarchy, the cost-estimation pro- estimation of each WBS item. Regarding the equation formula-
cess does not follow the hierarchy, as the EPC contractors apply tions, appropriate methods that are also suitable for the WBS-item
their own discipline configurations. In addition, the work scope of dataset are also needed. For the weight estimation of this study,
the cost estimation is performed alongside the discipline. As a the proposed methods of previous studies including statistical [6]
similar task is adopted by each discipline, the functional-system and optimization methods [10] were used.
characteristics can be well categorized in correspondence with the A statistical method allows for the derivation of a simplified
disciplines. The weight category is also combined alongside the nonlinear weight-estimation model; here, the correlation and mul-
discipline, and the proposed discipline of this study is listed in tiple regression analyses were performed. Before the correlation
Table 3. The disciplines are divided into WBS items that handle analysis, a variable transformation can be used to choose a nonlin-
the smallest concept of the weight items and are proposed in this ear form of each variable through a rough regression analysis. A
study. In Fig. 2, categories with lines (discipline, and WBS item) weight-estimation model can also be made through optimization
show the work hierarchy of the FPSO topsides. methods. In this study, the weight-estimation model was made
The WBS items represent the lowest weight-item level of this using the GP. To improve the calculation time, a correlation anal-
study. The WBS items were set to the weight-estimation unit. By ysis was performed prior to the GP. Figure 3 shows the process
adding the WBS items in the disciplines, the discipline weight can for generation of weight estimation models using two methods.
be estimated. To estimate the weight of the FPSO topsides, the
efficient way for the FPSO-topside WBS construction is the 4 Application of the Weight-Estimation Method for the
combination of two hierarchies; that is, the two hierarchies are Floating, Production, Storage, and Offloading Topsides
appropriately mixed in a single hierarchy. Figure 2 shows the
WBS hierarchy that is proposed in this study. 4.1 Prototype Program for the Weight Estimation of the
First, regarding the weight estimation, the FPSO topside is the Floating, Production, Storage, and Offloading Topsides. In this
ultimate purpose and highest object level. The topside is divided study, a prototype program was developed based on the proposed
by the disciplines in consideration of the work scope. Each disci- method. A screenshot of the program is shown in Fig. 4. The proto-
pline handles the entire topside area, and it can be divided type program is written in the C# language, and the graphic user inter-
along the purpose of the area. Inside of each area, several mod- face was developed using the Windows Presentation Foundation.
ules, each of which assumes a role for a specific FPSO function,
are installed. Each module consists of the WBS items that repre- 4.2 Data Categorization for the Weight Estimation of the
sent the bundle of specific equipment or a system. In this study, Floating, Production, Storage, and Offloading Topsides. For
each WBS item is the target of the weight equation; that is, the the weight estimation of the FPSO topsides, statistical and optimi-
WBS item is the minimum size of the weight estimation. By add- zation methods were used in this study, and this required the com-
ing all of the WBS items in each category, the weights of the pletion of a pre-analysis data categorization. The data structure
higher categories can be estimated. and the categorization of each ship-building company are differ-
Fundamentally, as with the work hierarchy, the discipline ent, and all of the ship-building companies do not share their pro-
weight can be made from the WBS items in the discipline. In ject data, making it difficult to obtain publicly available detailed
addition, the addition of the related WBS items in all of the disci- specification and weight data. The as-built data of three FPSO top-
plines can also be used to estimate the module weight. In each sides that were gathered from Daewoo Shipbuilding & Marine
module, several WBS items are included according to the module Engineering were used for this study.
function. For example, various pipes, structural members, separa- Only three of the FPSO-topside data were available, so if the
tors, pumps, and related items are included in separation module; data are divided on a per-module basis, it can be used for statisti-
accordingly, each module can be the set of WBS items. However, cal methods, but it cannot be used for optimization methods; this
the same WBS items in the other modules can be varied for each is because the weight-estimation model was devised for each

Table 3 Discipline list of the FPSO topsides

Discipline Description

Structure Structural-support items, e.g., plate, web, support, etc.

Machinery Contained topside equipment, e.g., compressor, separator, pump, etc.
Piping Pipes and associated accessories, e.g., pipe, flange, fitting, valve, etc.
Outfitting e.g., material-handling system, lifesaving equipment, firefighting equipment, etc.
HVAC Heating, ventilation, air-conditioning equipment, e.g., duct, discharger, etc.
Accommodation Equipment for accommodation, e.g., door, deck, furniture, etc.
Electric Electrical material and accessories, e.g., lighting, cable, cable tray, etc.
Instruments and telecoms e.g., ICSS (integrated control and safety system), telecommunication system, etc.

014501-4 / Vol. 140, FEBRUARY 2018 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022
Fig. 3 Process for creation of the weight estimation models

Fig. 4 Screenshot of the prototype program for the weight estimation of the FPSO topsides

WBS item, and therefore, the available data for each WBS item is Table 4 Examples of the WBS-item categorized data
insufficient from a statistical point of view. As mentioned in
Sec. 2, the modular characteristics of the same area are similar; Independent variables
accordingly, the modules for each area were categorized into the
same attribute data. By categorizing the data, more than three Projects (module in
pieces of data can be used for each WBS item. Examples of the the same area) Parameter 1 Parameter 2 … Weight
categorized data are listed in Table 4, wherein “module AP1” is
A (module AP1) 5512.4 1095.1 … 109.5
the first module of the process area in the A project. The number
⯗ ⯗ ⯗ … ⯗
of independent variables can include general parameters (main B (module BP1) 2879.8 1222.6 … 60.2
dimension of the vessel, topside total volume, etc.) and specific ⯗ ⯗ ⯗ … ⯗
parameters (projected area, number of nodes, weight of other C (module CP1) 4372.7 472.7 … 117.9
WBS items, etc.), and they can vary according to the WBS item. ⯗ ⯗ ⯗ … ⯗
By using the categorized data according to the above process, the

Journal of Offshore Mechanics and Arctic Engineering FEBRUARY 2018, Vol. 140 / 014501-5
Table 5 Data portion for one WBS item (plate) regarding the Table 6 Correlation-analysis results for a WBS item (plate)
structural discipline regarding the structural discipline

Module Volume (m3) Area (m2) MA þ PI þ EL (ton) … Weight (ton) Parameter Correlation coefficient

AP1 8302.1 1273.5 1091.8 … 47.0 Yield strength 0.44

AP2 3943.9 497.8 123.2 … 96.8 Environmental condition 0.09
AP3 11,462.8 8.9 289.1 … 80.6 Section modulus 0.15
AP4 1505.0 86.4 1104.9 … 196.1 Vessel breadth 0.85
BP1 1553.0 1678.5 1172.4 … 228.9 Vessel depth 0.85
BP2 3683.1 231.3 76.0 … 294.6 Vessel length 0.28
BP3 6777.2 1007.7 164.4 … 159.7 Projected area 0.85
BP4 13,344.1 2478.3 1064.7 … 19.4 Number of nodes 0.73

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022

CP1 6225.9 3039.3 1021.2 … 92.2 Number of decks 0.44
CP2 971.1 2249.2 666.8 … 35.2 Design life 0.85
CP3 3368.9 1101.0 462.9 … 152.6 Volume 0.75
CP4 4208.8 691.4 332.4 … 33.3 Area 0.97
⯗ ⯗ ⯗ ⯗ ⯗ ⯗ MA þ PI þ EL 0.79

weight-estimation model (weight-estimation equation) can be two candidates (vessel breadth and vessel depth) were excluded.
made using either a statistical or an optimization method, meaning As a result, six parameters (projected area, number of nodes,
that a weight equation can be made using the parameters in each design life, volume, area, and MA þ PI þ EL) were selected for
WBS item. In this study, the “3” project data were divided by “6” the regression analysis.
areas, and then each area was also divided by a total of “63” After the parameter selection, a multiple nonlinear-regression
modules. The categorized data were then used to make “558” analysis was conducted. By performing the regression process
weight-estimation models for “93” WBS items in “6” areas three times, two parameters (number of nodes and volume) were
(93  6 ¼ 558). In addition, the “93” WBS items can also be cate- excluded according to the multicollinearity of the independent
gorized by “8” disciplines. variables (parameters). The final result of the regression analysis
One of important points when selecting parameters is to include is presented in Eq. (1), where “PA” is the projected area, “DL” is
the design basis and the philosophy. A FPSO has large changes in design life, and “A” is area. Consequently, the plate weight of one
design according to its basis and concept. They are difficult to be module in the process area can be estimated by using Eq. (1). The
quantified as parameters to estimate the weight of FPSO topsides. adjusted R2 of the weight-estimation model is 0.96, the average of
Instead, these might be reflected partly by environmental condi- error is 7.32%, and the standard deviation of error is 0.05. Figure
tions because the conditions are a big part of the design basis and 5 shows the regression-graph results, showing that the regression
philosophy. Therefore, they were included as formations of para- equation fits soundly with the actual data. The dark grey-color
meterized environmental conditions in this study. Especially, region in the graph represents the 1-sigma (approximately 68% of
among the environmental conditions, oil-processing capacity, gas- the average area) area of the weight-estimation model and the
processing capacity, water-injection capacity, etc., were used as light grey-color region in the graph represents the 2-sigma
parameters. (approximately 95% of the average area) area

4.3 Example of Weight-Estimation Model for Each Work weightplate ¼ 155:03 þ 0:22PA þ 4:28DL þ 0:11A
Breakdown Structure Item. In this section, representative
weight-estimation models from among “558” cases are presented. þ 0:02ðMA þ PI þ ELÞ (1)

4.3.1 Results of Correlation and Regression Analyses for the

Work Breakdown Structure Items (Structural Discipline). The 4.3.2 Results of Correlation and Regression Analyses for One
weight-estimation model for one of the structural WBS items was Work Breakdown Structure Item (Piping Discipline). The weight-
made and analyzed in this section. Among several of the WBS estimation model for one of the piping WBS items was made
items in the structural discipline, the weight-estimation model for and is analyzed in this section. Among several of the WBS
module plate in the process area is presented. For the analysis, 18 items in the piping discipline, the weight-estimation model for
module data and 13 parameter data were collected. Some of the the pipe weights under 6 in of the modular process area is pre-
collected data are shown in Table 5. In Table 5, “MA þ PI þ EL” sented. For the analysis, 18 module data and 12 parameter data
means the sum of the discipline weights (machinery, piping, and were collected. Some of the collected data are shown in Table
electric) of the module. The last column of Table 5 shows the 7, wherein “MA” is the weight sum for the modular machinery
weight of the plate (target of the weight-estimation model) in each discipline and PI > 6 in is the pipe-weight sum over 6 in
module. As shown in Table 5, the weights of the other disciplines (because pipe over 6 in weight is highly related to pipe under 6
can affect the structural disciplines. in). The last column of Table 7 shows the pipe weights of each
As mentioned previously, a total of 13 parameters were used to module that are less than 6 in (target of the weight-estimation
make the weight-estimation model for the plate. The correlation model). As shown in Table 7, the weight of the machinery disci-
coefficients of each of the parameters are listed in Table 6, where pline can affect the piping disciplines.
the larger the correlation coefficient, the higher the correlation The correlation coefficient of each parameter is listed in Table 8.
between the WBS-item weight and the parameters. Generally, if Nine of the parameters (water-injection capacity, produced water
absolute correlation-coefficient value is larger than 0.5, the corre- capacity, oil-processing capacity, gas-processing capacity, total
lation is considered as high enough; therefore, eight parameters liquid capacity, oil-storage capacity, area, MA, and PI > 6 in) can
(vessel breadth, vessel depth, projected area, number of nodes, be the candidates for the multiple regression analysis; however,
design life, volume, area, and MA þ PI þ EL) can be the candi- the collected parameters for the weight estimation are generally
dates for the multiple regression analysis. A positive correlation, considered a positive correlation with respect to the weight, simi-
however, is generally considered between the parameters that are lar to the structural-discipline case. For this reason, six candidates
collected for the weight estimation and the weight; for this reason, (water-injection capacity, produced-water capacity, oil-processing

014501-6 / Vol. 140, FEBRUARY 2018 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022
Fig. 5 Regression results of the weight-estimation model for the plate using a statistical method

Table 7 Data portion for one WBS item (pipes under 6 in) machinery weight, and “P” indicates the pipe weights over 6 in.
regarding the piping discipline Consequently, the pipe weight of one module in the process
area can be estimated using Eq. (2). The adjusted R2 of the
Module No. of equipment PI > 6 in MA … Weight (ton) weight-estimation model is 0.93, the error average is 57.63%,
and the standard error deviation is 0.77. Figure 6 shows the
AP1 17 68.6 216.1 … 28.5 regression-graph results, where the regression equation mostly
AP2 0 5.2 0.0 … 2.1
AP3 44 11.1 218.2 … 18.2
fits with the actual data; however, some of the data are off the
AP4 19 9.6 118.7 … 5.6 weight-estimation model, increasing the error average and the
BP1 24 11.0 488.4 … 22.1 standard error deviation compared with the structural-discipline
BP2 47 1.9 35.4 … 13.1 case
BP3 44 146.4 69.6 … 20.1
BP4 72 82.2 624.6 … 112.5 weightpipes under 6 in ¼ 13:36 þ 0:004A þ 0:002MA þ 0:36P (2)
CP1 45 8.3 715.7 … 99.8
CP2 14 0.9 154.2 … 12.7
CP3 111 7.4 270.8 … 3.6
The adjusted R2 is close to “1,” which is a satisfactory result,
CP4 84 2.8 334.9 … 13.5 but the optimization method was performed for a higher accuracy
regarding the weight-estimation model; here, the same is applica-
⯗ ⯗ ⯗ ⯗ ⯗ ⯗
ble for the statistical method, where three parameters (area, MA,
and pipe weights over 6 in) were selected for the estimation
through the correlation analysis. The weight-estimation model of
Table 8 Correlation-analysis results for one WBS item (plate) the optimization method is presented in Eq. (3). The adjusted R2
regarding the piping discipline of the weight-estimation model is 0.98, the error average is
36.32%, and the standard error deviation is 0.23. Figure 7 is the
Parameter Correlation coefficient graph of the regression results, wherein the regression equation
fits soundly with the actual data
Power-generation capacity 0.42
Water-injection capacity 20.97 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Produced-water capacity 20.60 ð1  PÞP2
weightpipes under 6 in ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (3)
Oil-processing capacity 20.82 900  2P
Gas-processing capacity 0.68
Total liquid capacity 20.98
Oil-storage capacity 20.80
Comparing the results between the statistical and optimization
No. of equipment 20.08 methods, the result for which the optimization method was used is
Volume 0.43 more accurate; however, the result cannot be analyzed because
Area 0.81 of the complexity of the optimization-method-derived weight-
MA (machinery weight) 0.64 estimation model. Although the statistical method is slightly inac-
PI > 6 in (pipe weights over 6 in) 0.92 curate, it is possible for the user to grasp the parameter influence
relatively easily, whereas the optimization method is more accu-
rate than the statistical method, but the influence of the parameters
is difficult to grasp.
capacity, gas-processing capacity, total liquid capacity, and oil-
storage capacity) were excluded. As a result, three parameters 4.4 Verification of the Weight-Estimation Model. As
(area, MA, and PI > 6 in) were selected for the regression described in Sec. 4.3, a total of 558 WBS-item weight-estimation
analysis. models were generated in this study. Because of the ease of the
After the parameter selection, the multiple nonlinear-regression analysis, most of the weight-estimation models were generated
analysis was conducted. All of the parameters are the satisfied cri- using a statistical method, but a number of them did not fit
teria (p-value and VIF). The final result of the regression analysis soundly with their models; therefore, the optimization method
is presented using Eq. (2), where “A” is the area, “MA” is the was applied in these cases, as shown in Sec. 4.3.2.

Journal of Offshore Mechanics and Arctic Engineering FEBRUARY 2018, Vol. 140 / 014501-7
 Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022
Fig. 6 Regression results of the weight-estimation model for pipes under 6 in using a statistical method

Fig. 7 Regression results of the weight-estimation model for the pipes under 6 in using an optimization method

Table 9 Estimation results of Project A

Actual weight Weight ratio Estimated weight Weight ratio Ratio

Discipline (ton) (A) (%) (A/C) (ton) (B) (%) (A/D) (%) (B/A)

Structure 12,631.1 38.2 12,955.7 38.3 102.5

Machinery 5973.9 18.1 5858.6 17.3 98.0
Piping 7298.1 22.1 7459.4 22.1 102.2
Outfitting 1985.2 6.0 2124.6 6.3 107.0
HVAC 720.4 2.2 749.5 2.2 104.0
Accommodation 1061.1 3.2 1235.3 3.7 116.4
Electric 2458.4 7.4 2581.4 7.6 105.0
Instruments and telecoms 935.8 2.8 883.9 2.6 94.4
Topside (C) 33,064.0 100.0 (D) 33,784.8 100.0 102.1

Obviously, for verification, the data that were used to generate As a result of the Project A estimation, the average discipline
the weight-estimation model should not be used; however, a error is 5.59%, and the total-weight error of the topside is 2.18%.
dearth of well-categorized data regarding FPSO projects exists. The detailed results of each discipline of Project A are listed
The models were applied to three projects that were used to make in Table 9 and presented in Fig. 8 in graph form. As shown in
the models for the verification of the weight-estimation model. Table 9, the errors of most of the disciplines are less than 10%.
First, the weight-estimation model for each WBS item was cate- Especially, the accommodation error of 16.42% is the largest,
gorized for each of the modules, because every module does not meaning that the accommodation weight-estimation models are
contain all of the WBS items depending on the modular not sufficiently accurate for an estimation; that is, the accommo-
characteristics. After the categorization, the input data (value of dation characteristics in the three projects that were used as data
parameters) for each WBS item were applied to each weight- are far different. Therefore, the other parameters, or the other ana-
estimation model. lytical methods, are needed to make the accommodation weight

014501-8 / Vol. 140, FEBRUARY 2018 Transactions of the ASME

 Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022
Fig. 8 Graph of the estimation results of Project A

Fig. 9 Graph of the estimation results of the three projects

estimation. Similar to the results of Project A, the weights of Proj- in the related works [6,10] were 19.0% and 18.3%, respectively.
ects B and C were also estimated, and the results are shown in From these results, it can be seen that the proposed method could
Fig. 9. From the results of all of the projects, the average disci- effectively represent the weight characteristics of three example
pline error is 6.16%, and the average error of the total topside projects.
weights is 1.69%.
The comparative test between the proposed method and the
existing methods is important and will help to show the superior- 5 Conclusion and Future Works
ity of the proposed method. However, a study that classifies the In this study, a method for the generation of a weight-
weight of FPSO topsides like in this study is very rare. That is, estimation model for which the work breakdown structure was
from the literature survey, we could not find any study that is considered during its design is presented through an expansion of
based on the semi-bottom up approach like this study to estimate our previous study [18]. By considering the WBS, the weight of
the weight of FPSO topsides. Therefore, we compared the accu- each WBS item can be estimated or analyzed. In addition, it is
racy of the proposed method with that of the related works [6,10] possible to easily estimate the weight of each area or discipline, or
(only studies that confirmed the accuracy) in terms of the total even the total topsides, through a well-organized WBS. Most of
weight of FPSO topsides. The errors of the estimated total weight the pre-existent weight-estimation methods are top–down and do

Journal of Offshore Mechanics and Arctic Engineering FEBRUARY 2018, Vol. 140 / 014501-9
not accurately estimate the actual weight. In addition, it is not pos- [3] Rasmussen, K. R., 2015, “Optimisation of Central Offshore Process Equipment
sible to estimate the details of the topside weights such as the dis- and Weight Estimations,” B.Sc. thesis, Aalborg University, Aalborg, Denmark,
pp. 22–53.
ciplines, modules, or areas; therefore, the authors conceived of [4] Aasen, R., and Bjorhovde, S., 2010, “Early Stage Weight and Cost
“semi-bottom up” methods for the estimation of the weight of the Estimation Using Parametric Formulas and Regression on Historical Data,”
FPSO topsides that have not been tried until now for the proposal 69th Annual Conference of the Society of Allied Weight Engineers (SAWE),
of this paper. To evaluate the applicability of the proposed method Virginia Beach, VA, May 23–26, Paper No. 3515-S.
[5] Watson, D. G. M., and Gilfillan, A. W., 1976, “Some Ship Design Methods,”
and the prototype program, the proposed methods were applied to Tran. R. Inst. Naval Archit., 119, pp. 279–324.
a three-project estimation problem. The results showed that the [6] Ha, S., Seo, S. H., Roh, M. I., and Shin, H. K., 2016, “Simplified Nonlinear
proposed method can be used as a new method for the FPSO- Model for the Weight Estimation of FPSO Plant Topside Using the Statistical
topside weight estimation. Method,” Ship Offshore Struct., 11(6), pp. 603–619.
[7] Koza, J. R., 1992, Genetic Programming: On the Programming of
In the future, the proposed method will be applied to additional Computers by Means of Natural Selection, MIT Press, Cambridge, MA, pp.
projects. Because only three projects were used in this study, the

Downloaded from http://asmedigitalcollection.asme.org/offshoremechanics/article-pdf/140/1/014501/6247676/omae_140_01_014501.pdf by Newcastle University user on 16 February 2022

73–120.
accuracy and the generality of the weight-estimation model were [8] Koza, J. R., 1994, Genetic Programming II: Automatic Discovery of Reusable
not verified correctly. A collection of more of the data of the Programs, MIT Press, Cambridge, MA, Chap. 2.
[9] Banzhaf, W., Nordin, P., Keller, R. E., and Francone, F. D., 1997,
FPSO topsides is consequently necessary using the big data plat- Genetic Programming: An Introduction on the Automatic Evolution of
form [19], and they will be used to generate more accurate Computer Programs and Its Applications, Morgan Kaufmann, San Francisco,
weight-estimation models. Among the characteristics of the FPSO CA, pp. 105–236.
topsides, the owner philosophy and the installation-site flag are [10] Ha, S., Um, T. S., Roh, M. I., and Shin, H. K., 2017, “A Structural Weight Esti-
mation Model of FPSO Topsides Using an Improved Genetic Programming
very important, and these characteristics will be reflected in the Method,” Ship Offshore Struct., 12(1), pp. 43–55.
estimation process. [11] Cohen, J., Cohen, P., West, S. G., and Aiken, L. S., 2002, Applied Multiple
Regression/Correlation Analysis for the Behavioral Sciences, 3rd ed., Lawrence
Erlbaum Associates, Mahwah, NJ, pp. 64–100.
Funding Data [12] Xin, P., Khan, F., and Ahmed, S., 2016, “Layout Optimization of a Floating
Liquefied Natural Gas Facility Using Inherent Safety Principles,” ASME J. Off-
shore Mech. Arct. Eng., 138(4), p. 041602.
 Daewoo Shipbuilding and Marine Engineering. [13] Yazdani, M., and Jolai, F., 2016, “Lion Optimization Algorithm (LOA): A
Nature-Inspired Metaheuristic Algorithm,” J. Comput. Des. Eng., 3(1), pp.
 MSIP (Ministry of Science, ICT and Future Planning), 24–36.
Republic of Korea, under “Development of Ship Design [14] Tejani, G. G., Savsani, V. J., and Patel, V. K., 2016, “Adaptive Symbiotic
Standard PLM Platform based on Big Data” (Grant No. Organisms Search (SOS) Algorithm for Structural Design Optimization,”
NIPA-2016-S1106-16-1025) supervised by the NIPA J. Comput. Des. Eng., 3(3), pp. 226–249.
[15] Kim, S. K., Roh, M. I., and Kim, K. S., 2017, “Arrangement Method of
(National IT Industry Promotion Agency). Offshore Topside Based on an Expert System and Optimization Technique,”
 BK21 Plus Program (Education and Research Center for ASME J. Offshore Mech. Arct. Eng., 139(2), p. 021302.
Creative Offshore Plant Engineers) funded by the Ministry [16] Cocodia, E. O., 2008, “Creating Better Cost Estimates for Floating Offshore
of Education, Republic of Korea. Structures by Assessing Cost Correlation and Understanding Risk,” Eighteenth
International Offshore and Polar Engineering Conference, Vancouver, BC,
 Research Institute of Marine Systems Engineering of Seoul Canada, July 6–11, SPE Paper No. ISOPE-I-08-192.
National University. [17] Tausworthe, R. C., 1980, “The Work Breakdown Structure in Software Project
Management,” J. Syst. Software, 1, pp. 181–186.
[18] Roh, M. I., Seo, S. H., Shin, H. K., Ku, N. K., Ha, S., and Kim, K. S., 2014,
References “Simplified Model for the Weight Estimation of Floating Offshore Plant Using
[1] Rui, R., and Walker, J., 2015, “Upstream Offshore-Facility Weight-Growth the Statistical Method,” ASME Paper No. OMAE2014-24379.
Study,” Oil Gas Facil., 4(2), pp. 107–112. [19] Kim, S. H., Roh, M. I., Kim, K. S., and Oh, M. J., 2017, “Big Data Platform
[2] Bolding, A., 2001, “Bulk Factor Method Estimates FPSO: Topsides, Weight,” Based on Hadoop and Application to Weight Estimation of FPSO Topside,”
Oil Gas J., 99(10), pp. 49–53. J. Adv. Res. Ocean Eng., 3(1), pp. 32–40.

014501-10 / Vol. 140, FEBRUARY 2018 Transactions of the ASME