SHIP REPAIRING TIME AND LABOUR

Arun Kr Dev * and Makaraksha Saha**

* Associate Professor, Newcastle University Singapore

**Technical Advisor, MK Marine Pte Ltd, Singapore

Abstract

Ship repairing time, duration of stay of a ship in a shipyard (also a slipway or a floating dock), is a part of
the routine maintenance schedule of a ship, mainly required by the classification societies and the flag
states. With an interval of approximately 24 to 30 months, the regulations of both the flag state and the
classification society call for a ship to carry out docking survey, intermediate survey or special survey (once
in 5 years), depending on the age of the ship, in a shipyard/floating dock/slipway. It is mandatory to fulfil
the requirements and comply with the rules and regulations for maintaining statutory certification. Ship
owners and shipyards always try to reduce the repairing time to lessen the loss of income (for an owner)
and maximise the annual turnover through handling more ships (for a shipyard).

Labour cost is an important and sensitive issue in labour intensive industry. Ship repairing work is, by
nature, labour intensive and not prone to automation. In regular ship repairing or routine maintenance of
a ship, labour cost contributes the highest amount in the final invoice. This figure may go up to 70% of the
total cost. This cost is directly provided by labour (man-days) utilised for the ship repairing works. Ship
owners and shipyards are always very keen for lowering the man-days value. Lesser man-days can directly
be translated into the lower final invoice (for the ship owner) and higher productivity (for a shipyard),
which can help the shipyard to stay in a competitive market.

Ship repairing time (days) related information for 600 cargo ships and ship repairing labour (man-days)
related information for 50 cargo ships of various ages, sizes and types were collected from a single
shipyard. A multiple linear regression model was developed and analysed using these primary data of time
and labour each. Ship repairing time was then expressed as a function of a ship’s age, deadweight,
repairing works of mainly hull coating, piping, structural steel and tank coating. Similarly, ship repairing
labour was then expressed as a function of a ship’s age, deadweight, type and repairing works of mainly
hull coating, piping, structural steel and tank coating. “Method of least squares” was applied to estimate
the regression coefficients.

In this paper, the authors have made an attempt to identify the number of those independent variables
that influence ship repairing time and labour (the dependent variables) and their inter-relationship. A
mathematical model has been developed for both time and labour and proposed, as a guiding tool, for
the decision maker to estimate a more realistic ship repairing time and labour for ships to be under repair.

Keywords: Ship repairing time, Ship repairing labour, Age, Deadweight, Types of ships, Hull coating,
Piping, Structural steel, Tank coating, Regression coefficient, F-statistic, Co-relation coefficient,
Coefficient of multiple determination.

1
1. Introduction

It is a fact that reduction in ship operating expenditures and increase in earnings (from cargo freight) are
the two vital issues of survival in a shipping business. Earning from freight is beyond the control of a
shipping company and heavily dependent on the external factors including global economic and trade
environment. The operating expenditures are mostly under the supervision of the shipping companies.
Therefore, it is very critical for the management to gather the required information about the operating
costs so that a proper financial planning can be drawn to monitor and improve the overall performance
of the company. Besides the bunker cost, a significant part of the operating expenditures is the ship
repairing cost. This value is very much dependent on the repairing time. So, reduction in repairing time
can directly be translated into a reduction in operating expenditures.

In routine ship repairing/maintenance activities, 60% to 70% of ship repairing cost is contributed by labour
(Ozgur, U.S. 2008) and the rest is from materials, services and utilities. Therefore, it is very critical for the
shipyard management to gather the required information about the repairing expenditures, more
specifically the labour cost. Lesser labour cost (means fewer man-days utilisation to get the repairing
works done resulting indirectly higher productivity) is a strong foundation to stay competitive. This
competitiveness is the key factor to survive in any business. For ship repairing activities, the ship repairing
cost is very much dependent on ship repairing labour (man-days). So, reduction in repairing man-days can
directly be translated into a reduction in repairing expenditures. This decrease will enhance the shipyard’s
chances to survive in any adverse market conditions.

There are various ways to reduce the repairing time and labour, such as to lessen the idle man-power, to
reduce the wastage of man-power, more specifically, the optimum utilisation of available man-power and
other resources. To achieve all or any of these, a proper manpower planning, by the repair works need to
follow. To prepare such plan, one should have the information about the inter-relationship between
repairing time and labour (dependent variables) and the repairing works and other factors (independent
variables) that influence the time (days) and labour (man-days).

For the ship repairing time, data analysis of 600 cargo ships of different ages, deadweights and types
suggests that all these items (variables) have a positive influence on the total duration of repairing time.
Data analysis of 101 crude oil tankers of different ages, deadweights, their particular quantity of repairing
works like hull coating, piping, structural steel and tank coating suggest that all these items (variables)
have a positive influence on the total duration of repairing time. For ship repairing labour, data analysis
of 50 cargo ships of various ages, deadweights, types and repair activities suggest that all these items
(variables) have a positive influence on the entire ship repairing labour (man-days).

First, a mathematical model was developed for both ship repairing time and ship repairing labour, where
repairing time and labour are treated as the dependent variable and age, deadweight and type are treated
as the independent variables. The multiple linear regression methods are used to estimate the
interrelationship between the ship repairing time and labour and its independent variables like age,
deadweight and type using collected sample data (600 cargo ships for ship repairing time and 50 cargo
ships for ship repairing labour). It revealed a stronger positive correlation among repairing time and age,

2
type but comparatively a weaker positive relationship with deadweight. For the ship repairing time,
though the multiple regression equations pass the F-statistic test (f > f α where α = 0.05), yet the coefficient
of multiple determination is comparatively small (R2 = 0.2201). For ship repairing labour, though the
multiple linear regression equations pass the F-statistic test (f > f α where α = 0.05), yet the coefficient of
multiple determination is comparatively small (R2 = 0.338). It prompted the authors to rethink about the
various independent variables used in the mathematical model.

The ship repairing works are now considered as independent variables, and the mathematical models for
ship repairing time and ship repairing labour were revised and re-developed. In this mathematical model,
ship repairing time is treated as the dependent variable and age, deadweight, hull coating repair, piping
repair, structural steel repair and tank coating repair are handled as the independent variables for crude
oil tankers only (101 ships). The multiple linear regression methods are used to estimate the
interrelationship between ship repairing time and its independent variables like age, deadweight and
repairing works using collected sample data. This revised mathematical model revealed a positive
relationship between ship repairing time and age, deadweight, piping repair, structural steel repair, tank
coating repair and a negative correlation with hull coating. For ship repairing labour, the improved
mathematical model revealed a positive relationship between repairing labour and age, deadweight, type,
structural steel repair and a negative correlation with hull coating (blasting and painting) and piping repair.

In this paper, ship repairing time is considered as the total time for which a ship stays in the shipyard (in
the dock and at the quayside). This time is counted from the date of arrival in the ship and departure from
the shipyard. Ship repairing labour (man-days) is considered as the total number of men used to complete
the repairing of a ship. Man-days are counted for hourly rated men and men by jobs (contracted by volume
of work) such as blasting, painting, staging, etc. between the date of arrival in the shipyard and the time
of departure from the shipyard. This research covers only the ships with normal docking repair, not the
emergency repair, damage repair or afloat repair to get more uniform, reliable and realistic relationship
among the variables mentioned above.

The aims of this analysis are (1) to establish the inter-relationship among dependent and independent
variables and (2) to develop a mathematical model to estimate ship repairing time and repairing labour.
In doing so, data collection and analysis (both analytical and graphical) are carried out.

The remainder of the paper is arranged as follows. A review of the related literature with a starting
paragraph of a brief background is highlighted in Section 2. In Section 3, the problem formulation and
associated assumptions are discussed. Collected sample, initial findings, methodology and analysis, are
presented in Section 4. In section 5, an elaborate discussion on analysis, results, the empirical formula, its
application and validation are performed. Finally, Section 6 concludes the paper and proposes
recommendations for future research.

3
2. Literature review

Though sufficient literature is available on ship operating expenditures, ship maintenance expenses and
related issues, very few research on ship repairing related subjects were found. Not much research was
found particularly on ship repairing time and labour. The probable reason for the nonexistence of
investigation on ship repairing labour seems to be the non-availability of such commercial data and
information that are strictly considered as the trade secret of the concerned shipyard, ship owner and
ship manager. This lack of ship repairing research area motivated the authors, who have more than three
decades of direct experiences in marine and offshore industries particularly in repairing and new building
of various types and sizes of ships, to take the task of investigating ship repairing time (days) and labour
(man-days).

Despite the unavailability of sufficient literature on ship repairing time (days) and labour (man-days) in
particular, there are already some works done on ship repairing activities and related fields.

Dev, A.K. and Saha, M. 2015, reviewed the ship repairing time (total days counting from the arrival at the
shipyard to the departure from the shipyard). The research shows that the ship repairing time (day) is
linearly related to the age, size (deadweight) and repairing works, namely, external hull coating, structural
steel, tank coating and piping. A mathematical model was developed and proposed a multiple linear
regression equation to estimate the expected ship repairing time using age, size and quantity of repairing
works.

Naffisah, M. S., Surjandari, I., Rachman, A. and Palupi, R. (2014) examined and analysed the real life dry
docking maintenance period (days) using Artificial Neural Network technique with back propagation
algorithm. They used 29 types of works in dry dock maintenance activities as input and developed and
proposed a mathematical model of dry docking maintenance duration estimation.

Surjandari, I. and Novita, R. (2013) examined and emphasised on dry docking period using data mining
technique. This paper explores and identifies the relationship between the dry docking time and other
variables responsible for dry docking works. The authors then propose a mathematical model for
estimation of the dry docking duration using CART (Classification and Regression Tree) method of
estimation. This dry docking period referred to the time that a ship stays in the dock for routine
maintenance works.

Dev, A.K. and Saha, M. (2016), reviewed the ship repairing labour (total man-days counting from the
arrival at the shipyard to the departure from the shipyard).The research shows that the ship repairing
labour (man-day) is linearly related to the age, size (deadweight) and repairing works, namely, external
hull coating, structural steel, tank coating and piping. A mathematical model was developed and proposed
a multiple linear regression equation to estimate the expected ship repairing labour using age, size and
quantity of repairing works.

Emblemsvag, J. (2014) demonstrated the Lean Project Planning (LPP) technique in shipbuilding for project
planning. The LPP is a combination of the Earned Value Method (EVM) and the Last Planner System (LPS),

4
which deals with planning, execution, monitoring and correction. This paper presented a case study of a
platform supply vessel using the LPP technique.

Jose, R.S.C. (2009) investigated and studied the dry docking time and cost and used multi-criteria decision-
making methods called the goal programming model to minimise the dry docking time and cost. This
paper demonstrates the technique of the goal programming model to balance the time and cost of dry-
docking of a ship. Dry-docking is sometimes referred as ship repairing.

Meland, K. and Spoulding, R. (2003) investigated and studied the workload and labour resource planning
in Northrop Grumman Newport News and provides an overview of the approach the shipyard uses to
develop and manage its workload and labour supply forecasts.

Victoria, D. et al. (2010) investigated and has shown insight into the key findings of the project entitled
“Customization of Web-Based Planning and Production Engineering Technologies to Support Integrated
Shipyard Work Flow” initiated by NSRP, with an objective to enable shipyards to achieve reduction in
project costs and cycle time through project standardization and the ability to perform re-planning. This
paper also provides a quantification and appreciation of the resulting cost benefits experienced by each
participating shipyard.

Apostolidis, A., Kokarakis, J. and Merikas, A. (2012) investigated and highlighted on the dry docking cost
for tankers. This paper explores and identifies the relationship between the dry docking cost and other
variables responsible for dry docking cost. The authors then propose a mathematical model for estimation
of the dry docking cost using Generalised Method of Moment (GMM). This dry docking cost referred to
the routine repairing cost, not the damage repairing cost.

Turan, O., Olcer, A.I., Lazakis, I., Rigo, P. and Caprace, J.D. (2009) researched the elements of life cycle cost
and earning and the effect of the change in structural weight due to optimisation experiments on life cycle
cost and earning. They developed the life cycle cost/earning model for structural optimisation during the
conceptual design stage.

3. Problem formulation and related assumptions

3.1 Problem formulation
After successful completion of repairing of a ship with complete satisfaction of the owner, the game is not
over yet. The bitter part is still to come. That is the final invoice. This invoice is the hot topic on the
negotiation table between the shipyard and the owner. The owner always complains “Yard has dumped
all the idle and unskilled labours to our ship resulting in longer repairing time”. The shipyard always replies
“Yard put maximum efforts and best and skilled workers to your ship to complete with quality and in
shortest possible time.” Still, the question is “Is there any scope and venue to reduce the time and labour
utilised?” The answer is not straightforward and easy. Past data of ship repairing time and labour can
provide the valuable information to get an appropriate answer to this question. Given the above difficulty,
this paper focuses on the inter-relationship between ship repairing time and labour and various variables
like age, deadweight, type and routine maintenance/repairing works. Using this above relationship, the
necessary ship repairing time (days) and labour (man-days) could be estimated, which may then be used
to solve the above problem, if not entirely, at least partly.

5
3.2 Related assumptions
3.2.1. Age
The age of a ship at the time of repairing is the time (nos. of years) counted from the date of delivery of
the ship from its builder to the owner. It is the time of birth of a ship, and it is used in all
documents/certificates issued for the first time by the underwriter, the flag state and the classification
society. It is thus expected that older ships having old machinery and equipment will experience higher
wear and tear depending on the maintenance policy of the ship owner. Also, the flag state rules and the
class rules demand higher standards/criteria of testing, inspection or survey for older ships. In the end,
older ships need more extensive repair and maintenance than the newer ones. Therefore, the first
assumption is that the age of a ship has a positive impact on ship repairing time and labour independently
and the relation is assumed to be linearly dependent. Ceteris paribus, the age of a ship has a positive
impact on ship repairing time (days) and labour (man-days). Mathematically, both ship repairing time and
labour are independently a function of the age of a ship and considered to be linearly associated.

3.2.2. Deadweight
The size of a ship can be defined by its dimensions (length, breadth and depth) or by its capacity (gross
tonnage, net tonnage or deadweight). In this paper, the deadweight is considered as the size that also
comprises the cargo carrying capacity. A big size ship means greater dimensions with larger machinery
and equipment. Logically, the bigger ships need longer repairing time for even regular
repairing/maintenance works. This analogy can be translated into the fact that the bigger ships need more
man-days for routine repairing/maintenance works than, the smaller ones. Therefore, the second
assumption is that deadweight of a ship has a positive impact also on the repairing time and labour
independently and the relation is assumed to be linearly dependent. Ceteris paribus, the deadweight of a
ship has thus a positive impact on the ship repairing time (days) and labour (man-days). Mathematically,
both ship repairing time and labour are independently a function of the deadweight of a ship and
considered to be linearly associated.

3.2.3. Type
The type of ship mainly refers to the type or nature of cargo it carries. Because of the type of cargo and
nature of the cargo, the configuration of ship varies widely including machinery and equipment. Also,
there are inherent differences between types of ships on machinery and equipment, piping arrangement,
tank arrangement, geometrical configurations, cargo handling facilities, etc. Due to this fact, it is also
logical that the different types of ships may, thus, require different repairing time. Therefore, the third
assumption is that the type of ship has a positive impact on the repairing time and labour independently
and the relation is assumed to be linearly dependent. Ceteris paribus, type of ship has a positive influence
on the ship repairing time (days) and labour (man-days). Mathematically, both ship repairing time and
labour are independently a function of the type of a ship and considered to be linearly associated.

3.2.4 Scope of repairing works

Repairing works refer to the category and volume of works carried out during the repairing period. In this
paper, mainly, four areas (category) are considered such as hull coating repair, piping repair, structural
steel repair and tank coating repair. Without any doubt, it can be concluded that higher volume of

6
repairing works requires more repairing time and labour. Therefore, the fourth assumption is that the
volume of repairing works has a positive impact on the repairing time and labour independently and the
relation is assumed to be linearly dependent. Ceteris paribus, the volume of ship repairing works has a
positive impact on the ship repairing time (days) and labour (man-days). Mathematically, both ship
repairing time and labour are independently a function of the quantity of ship repairing works of a ship
and considered to be linearly associated.
So far, it has been highlighted theoretically that ships’ age, deadweight, type and scope/quantity of ship
repairing works (hull coating, structural steel, piping and tank coating) are directly, positively and linearly
associated with the corresponding ship repairing time (days) and labour (man-days). In other words, ship
repairing time and labour (dependent variable) is a linear function of age, deadweight, type, the scope of
repairing works (independent variables). Mathematically, the above assumptions can be expressed in the
form of equations 1-6 for ship repairing time and equations 7-13 for ship repairing labour respectively as
follows,

𝑆𝑆𝑆𝑆TIME = f(𝑆𝑆A ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑆𝑆A (1)

𝑆𝑆𝑆𝑆TIME = f(𝑆𝑆D ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑆𝑆D (2)

𝑆𝑆𝑆𝑆TIME = f(𝑅𝑅hc ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅hc (3)

𝑆𝑆𝑆𝑆TIME = f(𝑅𝑅tc ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅tc (4)

𝑆𝑆𝑆𝑆TIME = f(𝑅𝑅s ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅s (5)

𝑆𝑆𝑆𝑆TIME = f�𝑅𝑅p � = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅p (6)

𝑆𝑆𝑆𝑆LABOUR = f(𝑆𝑆A ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑆𝑆A (7)

𝑆𝑆𝑆𝑆LABOUR = f(𝑆𝑆D ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑆𝑆D (8)

𝑆𝑆𝑆𝑆LABOUR = f(𝑆𝑆T ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑆𝑆T (9)

𝑆𝑆𝑆𝑆LABOUR = f(𝑅𝑅hb ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅hb (10)

𝑆𝑆𝑆𝑆LABOUR = f�𝑅𝑅hp � = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅hp (11)

𝑆𝑆𝑆𝑆LABOUR = f(𝑅𝑅s ) = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅s (12)

𝑆𝑆𝑆𝑆LABOUR = f�𝑅𝑅p � = 𝑎𝑎 + 𝑏𝑏 ∗ 𝑅𝑅p (13)

Where 𝑆𝑆𝑆𝑆TIME : Ship repairing time (days)

𝑆𝑆𝑆𝑆LABOUR : Ship repairing labour (man-days)

𝑆𝑆A : Age of a ship (years)

𝑆𝑆D : Deadweight of a ship (tonnes)

7
 𝑆𝑆T : Type of a ship

R hc : Hull coating repair (m2)

R tc : Tank coating repair (m2)

R hb : Hull blasting repair (m2)

R hp : Hull painting repair (m2)

R s : Structural steel repair (kg)

R p : Piping repair (m)

a, b : Regression coefficients

4. Sample data and methodology

4.1 Data collection
In order to develop the repairing time function, ship repairing labour function and verify the assumptions,
a total of 600 cargo ships for ship repairing time and 50 cargo ships for ship repairing labour were selected.
These ships were docked and repaired in a shipyard during the period 1999 to 2011. Their repairing time,
repairing labour, age, deadweight, type and repairing scopes were collected. Data were gathered from a
single shipyard to eliminate the geographical influences. Geographical influences refer to the difference
in the skill of the workforce, material handling and processing and overall facilities and capabilities of
shipyards in various shipyards in various regions, even in the same region. All these activities have a direct
impact on ship repairing time (days) and labour (man-days).
A general picture of sample data for ship repairing time is presented in Table 1 and Figure 1 through Figure
3. Table 1 shows that container carriers and crude oil tankers cover 30% and 26% respectively of the
sample data, followed by chemical/product tankers and bulk carriers of 15% each. The rest comprises car
carriers, general cargos, L.P.G. carriers, L.N.G. carriers and dredgers.

Table 1 Sample ships for repairing time at a glance

Types of No. of Average Average Average

ships ships age deadweight repairing time
General cargo 19 8.53 12,394 11.79
Container carrier 180 10.48 37,150 12.31
Car carrier 15 13.80 15,502 12.47
Chemical/Product Tanker 91 7.44 38,766 13.73
Bulk carrier 91 8.95 75,653 13.73
Dredger 6 15.33 8,274 16.17
Crude oil tanker 154 9.79 171,535 19.42
L.P.G. carrier 40 12.93 47,196 21.1
L.N.G. carrier 4 6.50 72,776 23.25
Total 600 9.82 77,020 15.25

8
Figure 1 shows the distribution of age of sample data with a mean and standard deviation of about 9.95
years and 6.81 years respectively and suggests that the majority of samples (90%) fall within 20 years of
age and hence, the result will not be affected by the samples beyond that age. Figure 2 shows the
distribution of deadweight of sample data with a mean and standard deviation of 76,455 tonnes and
79,676 tonnes respectively and suggests that the majority of samples (91%) fall within 180,000 tonnes
deadweight range. In this case, also the result will not be affected by the sample data beyond that
deadweight. Figure 3 shows the distribution of repairing time of sample data with a mean and standard
deviation of about 15.25 days and 10.82 days respectively. It also suggests that the majority of samples
(90%) fall within 25 days of repairing time and hence; the result will not be affected by the sample beyond
that repairing time.

200

160
NO. OF SHIPS

120

80

40

0
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29
SHIPS' AGE (YEARS)

Figure 1 Distribution of age of sample ships

100
NO. OF SHIPS

80
60
40
20
0
100,000 ~…
110,000 ~…
120,000 ~…
130,000 ~…
140,000 ~…
150,000 ~…
160,000 ~…
170,000 ~…
90,000 ~…
~ 10,000
10,000 ~ 20,000
20,000 ~ 30,000
30,000 ~ 40,000
40,000 ~ 50,000
50,000 ~ 60,000
60,000 ~ 70,000
70,000 ~ 80,000
80,000 ~ 90,000

SHIPS' DEADWIGHT (TONNES)

Figure 2 Distribution of deadweight of sample ships

9
 100

80
NO. OF SHIPS

60

40

20

0
4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34
SHIP REPAIRING TIME (DAYS)

Figure 3 Distribution of ship repairing time of sample ships

A general picture of sample data for ship repairing labour is presented in a tabular form in Table 2 and a
graphical form in Figures 4-10 respectively. Table 2 shows the average values of various independent
variables on different types of sample ships. Figures 4-10 show the distributions of different independent
variables. It also shows the limiting value of the variable. It can be seen in Figure 4 that the most of the
sample ships are within the 14 years of age limit. Figure 5 shows that the most of the sample ships are
within the 125,000 tonnes deadweight limit. Similarly, other figures also depict the limiting values of other
variables accordingly.

Table 2 Sample ships for repairing labour at a glance

Types of No. of Average

Ships Ships SA SD R hb R hp Rs R p SR LABOUR
Crude oil tanker 13 7 188,005 2,845 42,367 2,795 41 1,925
Container carrier 12 11 34,620 3,821 29,838 11,603 6 3,065
Chemical tanker 9 7 46,136 2,073 23,766 1,694 23 2,481
L.P.G. carrier 8 14 41,866 2,402 27,101 14,916 37 4,773
Bulk carrier 5 9 85,409 1,681 26,848 5,560 27 1,489
L.N.G. carrier 1 5 75,248 3,172 54,309 597 0 3,681
General cargo 1 20 9,594 316 5,214 2,255 5 2,857
Car carrier 1 2 15,154 300 10,287 7,064 0 1,216
Total 50 10 82,733 2,658 30,872 6,957 25 2,750

10
 NO. OF SHIPS NO. OF SHIPS NO. OF SHIPS

0
5
10
15
20
10

0
2
4
6
8

10
15
20

0
5
~ 500
501 ~ 1,000 -25,000
1,001 ~ 1,500
1,501 ~ 2,000 25,001-50,000
2,001 ~ 2,500
2,501 ~ 3,000 50,001-75,000
3,001 ~ 3,500
3,501 ~ 4,000 75,001-100,000
4,001 ~ 4,500
4,501 ~ 5,000 100,001-125,000
5,001 ~ 5,500
5,501 ~ 6,000 125,001-150,000
6,001 ~ 6,500

11
6,501 ~ 7,000 150,001-175,000
7,001 ~ 7,500
7,501 ~ 8,000 175,001-200,000
8,001 ~ 8,500
SHIPS' AGE (YEARS)

8,501 ~ 9,000
200,001-225,000
9,001 ~ 9,500

SHIPS' DEADWEIGHT (TONNES)

HULL BLASTING REPAIR, Rhb (M2)

9,501 ~ 10,000
225,001-250,000
Figure 4 Distribution of age of sample ships

10,001 ~ 10,500
10,501 ~ 11,000

Figure 5 Distribution of deadweight of sample ships

250,001-275,000
11,001 ~ 11,500

Figure 6 Distribution of hull blasting repair of sample ships

11,501 ~ 12,000
275,001-300,000
12,001 ~ 12,500
12,501 ~ 13,000
300,001-325,000
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

13,001 ~ 13,500
 NO. OF SHIPS NO. OF SHIPS NO. OF SHIPS

0
10
20
30
40
0
5
10
15
20

0
5
10
15
20
~ 10 ~ 500 ~ 5000

501 ~ 1,000 5,001 ~ 10,000

11 ~ 20
1,001 ~ 1,500 10,001 ~ 15,000
21 ~ 30
1,501 ~ 2,000 15,001 ~ 20,000

31 ~ 40 2,001 ~ 2,500 20,001 ~ 25,000

2,501 ~ 3,000 25,001 ~ 30,000

41 ~ 50
3,001 ~ 3,500 30,001 ~ 35,000
51 ~ 60
3,501 ~ 4,000 35,001 ~ 40,000
61 ~ 70 40,001 ~ 45,000

12
4,001 ~ 4,500

71 ~ 80 4,501 ~ 5,000 45,001 ~ 50,000

5,001 ~ 5,500 50,001 ~ 55,000

81 ~ 90

PIPING REPAIR, Rp (M)

5,501 ~ 6,000 55,001 ~ 60,000
91 ~ 100
HULL PAINTING REPAIR, Rhp (M2)

6,001 ~ 6,500 60,001 ~ 65,000

STRUCTURAL STEEL REPAIR, RS (KG)

101 ~ 110 6,501 ~ 7,000 65,001 ~ 70,000

7,001 ~ 7,500 70,001 ~ 75,000

Figure 9 Distribution of piping repair of sample ships

111 ~ 120
Figure 7 Distribution of hull painting repair of sample ships

7,501 ~ 8,000 75,001 ~ 80,000

Figure 8 Distribution of structural steel repair of sample ships

121 ~ 130
8,001 ~ 8,500 80,001 ~ 85,000
131 ~ 140 8,501 ~ 9,000 85,001 ~ 90,000
 20
NO. OF SHIPS 15
10
5
0

1,001 ~ 1,500
1,501 ~ 2,000
2,001 ~ 2,500
2,501 ~ 3,000
3,001 ~ 3,500
3,501 ~ 4,000
4,001 ~ 4,500
4,501 ~ 5,000
5,001 ~ 5,500
5,501 ~ 6,000
6,001 ~ 6,500
6,501 ~ 7,000
7,001 ~ 7,500
7,501 ~ 8,000
8,001 ~ 8,500
8,501 ~ 9,000
9,001 ~ 9,500
~ 500

9,501 ~ 10,000
501 ~ 1,000

SHIP REPAIRING LABOR, SRLABOUR (MAN-DAYS)

Figure 10 Distribution of ship repairing labour of sample ships

4.2 Methodology

To establish the relationship between the ship repairing time (days) and labour (man-days) and its
independent variables, the following function of the linear equations 14 and 15 are chosen because
individually the independent variables are linearly associated with the dependent variable as per initial
assumptions.

𝑆𝑆𝑆𝑆TIME = f �𝑆𝑆A , 𝑆𝑆D , 𝑅𝑅hc , 𝑅𝑅p , 𝑅𝑅s , 𝑅𝑅tc � (14)

𝑆𝑆𝑆𝑆LABOUR = f �𝑆𝑆A , 𝑆𝑆D , 𝑆𝑆T , 𝑅𝑅hb , 𝑅𝑅hp , 𝑅𝑅s , 𝑅𝑅p � (15)

Where 𝑆𝑆𝑆𝑆TIME : Ship repairing time (days). It is the time for which a ship stays in the shipyard. It is
calculated from the date of arrival into the shipyard and the date of departure from the shipyard.

𝑆𝑆𝑆𝑆LABOUR : Ship repairing labour (man-days). These are the total no. of men utilised for repairing
a ship. It is calculated by adding the no. of men engaged each day of ship repairing.

𝑆𝑆A : Age of a ship (years) at the time of docking. The age of a ship is calculated from the date of
delivery of the ship by the builder to its owner.

𝑆𝑆D : Deadweight (tonnes) of a ship. This deadweight is mainly the cargo carrying capacity plus the
consumables, crews, and fixed stores.

𝑆𝑆T : Type of ships like the crude oil tanker, container carrier, chemical tanker, L.P.G. carrier, bulk
carrier.

R hc : Quantity of hull coating repair (m2) works. It is measured in terms of area (m2) irrespective
of the standard of blasting such as SA 2.5, SA 2.0 and so on.

13
 R tc : Quantity of tank coating repair works (m2). It is measured in terms of area (m2) irrespective
of type of surface preparation (grit blasting or hydro blasting) and their standard (SA 2.5, SA 2.0,
etc. or 25,000 psi, 30,000 psi, etc.).

R hb : Quantity of hull blasting work. It is measured in terms of area (m2) irrespective of the
standard of blasting such as SA 2.5, SA 2.0 and so on.

R hp : Quantity of hull painting work. It is measured in terms of area (m2) irrespective of types of
paints such as anticorrosive, antifouling, silicon and so on.

R p : Quantity of piping repairing works. It is measured in terms of running length (m) irrespective
of location (in the tank or on deck), the system (steam line, cargo line, fire line, fuel line, etc.) and
their sizes (diameter).

R s : Quantity of structural steel repairing work. It is measured in terms of weight (kg) irrespective
of their thickness, quality (mild steel or high tensile steel) and locations (bottom plate, deck plate,
shell plate, longitudinals, transverses, bulkheads, brackets, collar plates, etc.).

4.2.1 Ship repairing time

Initial investigations on pairs of variables of interests are shown in Figure 11 through 17. Figure 11 shows
the ship repairing time against the age groups with a strong linear relationship. Figure 12 shows the ship
repairing time against the deadweight with a strong linear relationship. Figure 13 shows the average ship
repairing time against the type of ship with a strong linear relationship. Figure 14 shows the ship repairing
time against the hull coating repair works with a weak linear relationship. Figure 15 shows the ship
repairing time against piping repair works with a significant linear relationship. Figure 16 shows the ship
repairing time against structural steel repair works with a significant linear relationship. Figure 17 shows
the ship repairing time against tank coating repair work with a significant linear relationship. However, all
these findings confirm the initial assumptions that ship repairing time is a function of age, deadweight,
type and repairing works.

Repairing time Linear (Repairing time)

100
80
SRTIME (DAYS)

60
40
20
0
0 4 8 12 16 20 24 28 32 36 40
SHIPS' AGE (YEARS)

Figure 11 Ship repairing time versus age of sample ships

14
 Repairing time Linear (Repairing time)

100
80
SRTIME (DAYS)

60
40
20
0
0 50,000 100,000 150,000 200,000 250,000 300,000 350,000 400,000 450,000
SHIPS' DEADWEIGHT (TONNES)

Figure 12 Ship repairing time versus deadweight of sample ships

Average by type Linear (Average by type)

30
SRTIME (DAYS)

20
10
0
Chem/Pro

Bulk carrier
Car carrier

L.N.G
General

Container

Dredger

L.P.G
Crude oil
cargo

tanker
carrier

Tanker

TYPES' OF SHIPS

Figure 13 Average ship repairing time versus type of sample ships

Repairing time Linear (Repairing time)

100
80
SRTIME (DAYS)

60
40
20
0
0 4,000 8,000 12,000 16,000 20,000 24,000
2
HULL COATING REPAIR, Rhc (M ) 

Figure 14 Ship repairing time versus hull coating repair of crude oil tankers

15
 Repairing time Linear (Repairing time)

100
80
SRTIME (DAYS)

60
40
20
0
0 100 200 300 400 500 600 700 800
PIPING REPAIR, Rp (M)

Figure 15 Ship repairing time versus piping repair of crude oil tankers

Repairing time Linear (Repairing time)

100
80
SRTIME (DAYS)

60
40
20
0
0 5,000 10,000 15,000 20,000 25,000
STRUCTURAL STEEL REPAIR, Rs (KG)

Figure 16 Ship repairing time versus structural steel repair of crude oil tankers

Repairing time Linear (Repairing time)

100
80
SRTIME (DAYS)

60
40
20
0
0 5,000 10,000 15,000 20,000 25,000
2
TANK COATING REPAIR, Rtc (M )

Figure 17 Ship repairing time versus tank coating repair of crude oil tankers

16
4.2.2 Ship repairing labour

Initial investigations on pairs of variables of interests are shown in Figures 18-24. Figure 18 shows ship
repairing labour against the age with a positive and strong linear relationship. Figure 19 shows ship
repairing labour against the deadweight with a linear relationship. Figure 20 shows the average ship
repairing labour against the type of ship with a strong relationship. Figure 21 shows ship repairing labour
against the hull blasting repairing work with a positive and linear relationship. Figure 22 shows ship
repairing labour against the hull painting repairing work with a positive and linear relationship. Figures 18
and 19 also reveal very critical phenomena that need to be addressed. These figures show the ship
repairing labour against the ships’ age and ships’ deadweight respectively. In both cases, the ship repairing
labour against same ships’ age (Figure 18) and same ships’ deadweight (Figure 19) respectively, are shown
to vary widely. The apparent reason is mainly the scope of repairing works. It is very usual in routine ship
repairing activities that the scope of ship repairing works are significantly different from ship to ship even
though their ages and deadweights are same. It happens due to many reasons, one of which is the in-
service maintenance schedule and the subject is beyond the scope of this paper.

Other probable reasons for different labours may be the level of skill of the workforce allocated to various
ships, locations of repairing items (such as piping in tanks and on deck) and structural steel in tanks and
at the shipside (outside).

Also, high level of differences may be the result of weather factor, materials, and spare parts availability.
However, these are the areas of further research.

Figure 23 shows ship repairing labour against structural steel repairing work with a positive and strong
linear relationship. Figure 24 shows ship repairing labour against piping repairing work with a positive and
weak linear relationship. All these findings are summarised in Table 3. However, all these findings confirm
the initial assumptions that ship repairing labour (man-days) is a function of age, deadweight, type and
scope of repairing works and associated to a linear relationship. Therefore, it is justified theoretically, as
well as from sample data processing, to adopt a linear equation.

Observed Linear (Observed)

8,000
SRLABOUR (MAN-DAYS)

6,000

4,000

2,000

0
0 2 4 6 8 10 12 14 16 18 20 22
SHIPS' AGE (YEARS)

Figure 18 Ship repairing labour versus age of sample ships

17
 Observed Linear (Observed)

15,000
SRLABOUR (MAN-DAYS)

10,000

5,000

0
0 50,000 100,000 150,000 200,000 250,000 300,000 350,000
SHIPS' DEADWEIGHT (TONNES)

Figure 19 Ship repairing labour versus deadweight of sample ships

Repairing man-days Linear (Repairing man-days)

5,000
SRLABOUR (MAN-DAYS)

4,000
3,000
2,000
1,000
0
Bulk Carriers Chemical Crude Oil Container L.P.G Carriers
Tankers Tankers Carriers
TYPES OF SHIPS

Figure 20 Average ship repairing labour versus type of sample ships

Observed Linear (Observed)

8,000
SRLABOUR (MAN DAYS)

6,000

4,000

2,000

0
0 1,000 2,000 3,000 4,000 5,000 6,000 7,000
HULL BLASTING REPAIR, Rhb (M²)

Figure 21 Ship repairing labour versus hull blasting repair of sample ships

18
 Observed Linear (Observed)

8,000
SRLABOUR (MAN DAYS)

6,000

4,000

2,000

0
0 10,000 20,000 30,000 40,000 50,000
HULL PAINTING REPAIR, Rhp (M²)

Figure 22 Ship repairing labour versus hull painting repair of sample ships

Observed Linear (Observed)

8,000
SRLABOUR (MAN-DAYS)

6,000

4,000

2,000

0
0 4,000 8,000 12,000 16,000 20,000
STRUCTURAL STEEL REPAIR, RS (KG)

Figure 23 Ship repairing labour versus structural steel repair of sample ships

Observed Linear (Observed)

8,000
SRLABOUR (MAN DAYS)

6,000

4,000

2,000

0
0 25 50 75 100 125 150 175
PIPING REPAIR, RP (M)

Figure 24 Ship repairing labour versus piping repair of sample ships

19
 Table 3 Summary of findings

Figures Variables Co-relation coefficient ( r ) Relationship

18 SR LABOUR vs S A 0.4791 Exist & Positive
19 SR LABOUR vs S D 0.0959 Exist & Positive
20 SR LABOUR vs S T NA Exist
SR LABOUR vs
21 0.3655 Exist & Positive
R hb
SR LABOUR vs
22 0.1609 Exist & Positive
R hp
23 SR LABOUR vs R s 0.7353 Exist & Positive
24 SR LABOUR vs R p 0.0436 Exist & Positive

4.2.3 Multiple linear regression analysis

The multiple linear regression analysis is a mathematical procedure to determine the mathematical
relationship involving more than one independent variable, unlike a single independent variable in simple
linear regression analysis. This method uses the past data of both dependent and independent variables
to establish the relationship to predict the dependent variable against a set of independent variables.

A general idea of the multiple linear regression analysis can be explained as follows. Let us consider an
equation of the form: y i = b 0 + b 1 *x 1i + b 2 *x 2i + b 3 *x 3i ...........+b k *x ki , where b 0 , b 1 , b 2 , b 3 , ......b k are
the regression coefficients, k is the size of the independent variables and each set of data point {(x 1i , x 2i
, x 3i ,............ x ki , y i ); i = 1, 2, 3, 4,.........n, where n is the sample size and n >2 } satisfy the equation.
Applying the method of least squares (Walpole, R.E. and Myers, R.H. 1978), the following normal
equations can be obtained.

𝑛𝑛 ∗ 𝑏𝑏0 + 𝑏𝑏1 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 + 𝑏𝑏2 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥2𝑖𝑖 + ⋯ … … … … … … . . … … + 𝑏𝑏𝑘𝑘𝑘𝑘 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑘𝑘𝑘𝑘 = ∑𝑛𝑛𝑖𝑖=1 𝑦𝑦𝑖𝑖 (16)

𝑏𝑏0 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 + 𝑏𝑏1 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 2 + 𝑏𝑏2 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 . 𝑥𝑥2𝑖𝑖 + ⋯ … … + 𝑏𝑏𝑘𝑘𝑘𝑘 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 . 𝑥𝑥𝑘𝑘𝑘𝑘 = ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 . 𝑦𝑦𝑖𝑖 (17)

𝑏𝑏0 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥2𝑖𝑖 + 𝑏𝑏1 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥1𝑖𝑖 . 𝑥𝑥2𝑖𝑖 + 𝑏𝑏2 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥2𝑖𝑖 2 + ⋯ … … + 𝑏𝑏𝑘𝑘𝑘𝑘 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥2𝑖𝑖 . 𝑥𝑥𝑘𝑘𝑘𝑘 = ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥2𝑖𝑖 . 𝑦𝑦𝑖𝑖 (18)

. . . . . . . . . . . . .

. . . . . . . . . . . . .

𝑏𝑏0 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑘𝑘𝑘𝑘 + 𝑏𝑏1 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑘𝑘𝑘𝑘 . 𝑥𝑥1𝑖𝑖 + 𝑏𝑏2 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑘𝑘𝑘𝑘 . 𝑥𝑥2𝑖𝑖 + ⋯ … … + 𝑏𝑏𝑘𝑘𝑘𝑘 ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑘𝑘𝑘𝑘 2 = ∑𝑛𝑛𝑖𝑖=1 𝑥𝑥𝑘𝑘𝑘𝑘 . 𝑦𝑦𝑖𝑖 (19)

Using the observed sets of data points, values of the above statistical notations can be calculated. By
inserting those values in the above normal equations (Equations 16 - 19), these can be converted into
simultaneous equations with unknown regression coefficients. By solving those simultaneous equations
by any standard procedure, the values of regression coefficients can be estimated, and the required
regression equation can be formed. Various statistical parameters are used as references to demonstrate

20
the adequacy of the mathematical model considered. In this modelling, these are mainly (i) standard
deviation, (ii) coefficient of multiple determination and (iii) F-statistic (the calculated value and the critical
value at 5% significance level).

𝑆𝑆𝑆𝑆𝑆𝑆
Standard deviation, 𝑠𝑠 = �
𝑛𝑛−𝑘𝑘−1

𝑆𝑆𝑆𝑆𝑆𝑆
Coefficient of multiple determination, 𝑅𝑅 2 =
𝑆𝑆𝑆𝑆𝑆𝑆

𝑆𝑆𝑆𝑆𝑆𝑆 ⁄𝑘𝑘 𝑆𝑆𝑆𝑆𝑆𝑆 ⁄𝑘𝑘

F-statistic, 𝑓𝑓 = =
𝑆𝑆𝑆𝑆𝑆𝑆 ⁄(𝑛𝑛−𝑘𝑘−1) 𝑠𝑠2

𝑓𝑓𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = 𝑓𝑓0.05 (𝑘𝑘, 𝑛𝑛 − 𝑘𝑘 − 1), from standard statistical tables at 5% significance level.

𝑆𝑆𝑆𝑆𝑆𝑆 = 𝑆𝑆𝑆𝑆𝑆𝑆 − 𝑆𝑆𝑆𝑆𝑆𝑆

2
�∑𝑛𝑛
𝑖𝑖=1 𝑦𝑦𝑖𝑖 �
𝑆𝑆𝑆𝑆𝑆𝑆 = ∑𝑛𝑛𝑖𝑖=1(𝑦𝑦𝑖𝑖 − 𝑦𝑦�)2 = ∑𝑛𝑛𝑖𝑖=1 𝑦𝑦𝑖𝑖 2 −
𝑛𝑛

2
�∑𝑛𝑛
𝑖𝑖=1 𝑦𝑦𝑖𝑖 �
𝑆𝑆𝑆𝑆𝑆𝑆 = ∑𝑘𝑘𝑗𝑗=0 𝑏𝑏𝑗𝑗 𝑔𝑔𝑗𝑗 − 𝑛𝑛

For this research, the following multiple linear regression model is chosen to represent the relationship
expressed in equations 14 and 15.

𝑆𝑆𝑆𝑆TIME 𝑖𝑖 = 𝑏𝑏0 + 𝑏𝑏1 ∗ 𝑆𝑆A 𝑖𝑖 + 𝑏𝑏2 ∗ 𝑆𝑆D 𝑖𝑖 + 𝑏𝑏3 ∗ 𝑅𝑅hc 𝑖𝑖 + 𝑏𝑏4 ∗ 𝑅𝑅p + 𝑏𝑏5 ∗ 𝑅𝑅s 𝑖𝑖 + 𝑏𝑏6 ∗ 𝑅𝑅tc 𝑖𝑖 (20)
𝑖𝑖

𝑆𝑆𝑆𝑆LABOUR 𝑖𝑖 = 𝑏𝑏0 + 𝑏𝑏1 ∗ 𝑆𝑆A 𝑖𝑖 + 𝑏𝑏2 ∗ 𝑆𝑆D 𝑖𝑖 + 𝑏𝑏3 ∗ 𝑆𝑆T𝑖𝑖 + 𝑏𝑏4 ∗ 𝑅𝑅hb 𝑖𝑖 + 𝑏𝑏5 ∗ 𝑅𝑅hp𝑖𝑖 + 𝑏𝑏6 ∗ 𝑅𝑅s 𝑖𝑖 + 𝑏𝑏7 ∗ 𝑅𝑅p (21)
𝑖𝑖

Where 𝑆𝑆𝑆𝑆TIME : Ship repairing time (days)

𝑆𝑆𝑆𝑆LABOUR : Ship repairing labour (man-days)

𝑆𝑆A : Age of the ship at the time of docking (years).

𝑆𝑆D : Deadweight (tonnes)

S T : Type of ships

R hc : Hull coating repair (m2)

R tc : Tank coating repair (m2)

R hb : Hull blasting repair (m2)

R hp : Hull painting repair (m2)

21
 R s : Structural steel repair (kg)

R p : Piping repair (m)

b 0 , b 1 , b 2 , b 3 , b 4 , b 5 , b 6 and b 7 : Regression coefficients

i : 1,2,3,4.............n

n : Sample size

Using the method of least squares (Walpole, R.E. and Myers, R.H. 1978), the simultaneous equations are
obtained (Dev, A. and Saha, M. 2015, 2016).

Using the observed data of 𝑆𝑆𝑆𝑆TIME, 𝑆𝑆A , 𝑆𝑆D, R hc , R p , R s , and R tc , the values of the statistical notations,
mentioned in the above equations, are calculated. They are inserted in the equation 7 through equation
13 (Dev, A. and Saha, M. 2015) and obtained seven simultaneous equations in terms of regression
coefficients. The solution of these simultaneous equations yields the estimate of regression coefficients.
Subsequently, the statistical testing parameters are calculated to demonstrate the adequacy of the model.

Similarly, using the observed data of 𝑆𝑆𝑆𝑆LABOUR , 𝑆𝑆A , 𝑆𝑆D, S T , R hb , R hp , R s , R p , the values of the statistical
notations, mentioned in the above equations, are calculated. Those are inserted in the equation 14
through equation 21 (Dev, A. and Saha, M. 2016) and obtained eight simultaneous equations in terms of
regression coefficients. The solution of these simultaneous equations yields the estimate of regression
coefficients. Subsequently, the statistical testing parameters are calculated to demonstrate the adequacy
of the model.

In developing the regression equation, the “stepwise regression” technique is applied using the “forward
selection” method (Walpole, R.E. and Myers, R.H. 1978). This method ensures the selection of the most
effective variable from a set of variables in each step by comparing the contribution of each variable (R
(β j |β 1 , β 2 ,....β j-1 ) and F-statistic, at every step. The final regression estimation passes the statistical quality
test by F-statistic (calculated value and critical value at 5% significance level) and coefficient of multiple
determination (R2).

5. Discussion

5.1 Ship repairing time

The initial analysis was based on the assumption that ship repairing time is directly proportional to age,
deadweight and type of ships with a sample size of 600 cargo ships (Figure 11 through Figure 13). The
resulting regression equation (𝑺𝑺𝑺𝑺𝐓𝐓𝐓𝐓𝐓𝐓𝐓𝐓=-1.9921+0.55*𝑺𝑺𝐀𝐀 +2.304x10-5*𝑺𝑺𝐃𝐃 +7.78*𝑺𝑺𝐓𝐓 ) passes the F-statistic
test and gives a value of 0.2201 for the coefficient of multiple determination (R2).

22
At this stage, the authors realised three points. Firstly, an important factor responsible for ship repairing
time is the scope of repairing works (category and quantity). The scope of repairing works mainly refers
to hull coating, piping, structural steel and tank coating. Secondly, use of numerical values for types of
ships in the regression equation made the equation biased for a set of sample data. The reason is that the
numerical values, used for the types (Table 4 ), is assigned based on the average repairing time of various
types of ships (Figure 13) which is very much dependent on the sample data size. Thirdly, inherent
differences among types of ships have an impact on the normal repairing time even though the scope of
repairing works remains same.

Table 4 List of values assigned to the types of ships

Type General Container Car Chemical/ Bulk Dredger Crude Oil L.P.G. L.N.G.
of Ship Cargo Carrier Carrier Product Tanker Carrier Tanker Carrier Carrier
Values 1.000 1.044 1.058 1.165 1.165 1.372 1.647 1.790 1.972

All the above points led to include the scope of repairing works also as independent variables. Accordingly,
it was decided to use the data of crude oil tankers only, which has the second highest data population
(sample size of 101 ships). This restriction has increased the uniformity and confidence of the result of the
mathematical model and data analysis and various conclusions drawn.

Unquantifiable repairing works refer to works related to the rudder, propeller, tail shaft, sea valve and
chest, bow thruster, hull anodes, deck fittings, main engine and auxiliary engine, etc.

Using the mathematical model under the estimation method of least squares and collected data, the final
regression equation for the ship repairing time for crude oil tankers is as follows:

𝑺𝑺𝑺𝑺𝐓𝐓𝐓𝐓𝐌𝐌𝐄𝐄 = 8.4557+0.57x𝑺𝑺𝐀𝐀 +1.65x10-5 𝑺𝑺𝐃𝐃 -1.37x10-4R hc +2.035x10-2R p +1.86x10-4R s +5.09x10-4R tc (22)

The findings are in line with the assumptions made earlier. In the regression equation, all the variables,
except the hull coating works (R hc ), have the similar sign as assumed. Age, deadweight, piping work,
structural steel work and tank coating work have the positive and significant impact on the repairing time.

Negative (equivalent to zero) impact of hull coating work is, probably, due to the fact that the hull coating
work is totally related to the docking time (duration of stay inside the dock), not the repairing time. More
specifically, hull coating repairing works are completed during docking time. Hull coating repair works for
routine maintenance too does not influence repairing time (Figure 14). Therefore, it’s inclusion in the
regression equation does not contribute significantly. The situation is explained by the same value of the
variance for the regression equations, without hull coating work and with hull coating work, about 104
and 105 respectively.

Table 5 displays the estimates of regression coefficients and other statistical parameters of the regression
equation during the process of adding new variables. From the table, one can easily comprehend that the
successive inclusion of variables of scope of works in the model contributed higher R2 value, which is the
main measurement of the adequacy of the model.

23
 Table 5 Different mathematical model values

Mathematical Model b0 SA SD R hb Rp Rs R tc s R2 f f 0.05

SR TIME = f(R p ) 15.86 NA NA NA 3.4x10-1 NA NA 10.57 0.1640 19.380 3.950

SR TIME = f(R p ,S A ) 10.99 0.023 NA NA 6.4x10-1 NA NA 10.30 0.2139 13.340 3.100

SR TIME = f(R p ,S A ,S D ) 8.20 0.070 1.5x10-5 NA 2.4x10-2 NA NA 10.27 0.2268 9.490 2.710
SR TIME =
1.5x10-5 2.2x10-2 NA 4.8x10-4 10.20 7.794 2.482
f(R p ,S A ,S D ,R tc ) 8.32 0.610 NA 0.2451
SR TIME =
1.6x10-5 2.1x10-2 1.8x10-4 5.0x10-4 10.22 6.320 2.323
f(R p ,S A ,S D ,R tc ,R s ) 8.33 0.558 NA 0.2497
SR TIME =
1.7x10-5 2.0x10-2 1.9x10-4 5.0x10-4 10.27 5.256 2.205
f(R p ,S A ,S D ,R tc ,R s ,R hb ) 8.46 0.570 -1.4x10-4 0.2512

Focusing on the final regression equation, which passes the F-statistic test, the value of the coefficient of
multiple determination is low (0.2512). Mathematically, it means that 25% of the variation in the
dependent variable is contributed for the variation in independent variables and remaining is called the
error of estimation. This error of estimation is due to the absence of one or more influential independent
variables (unquantifiable variables) responsible for the change in the dependent variable.

In this case, two reasons were identified as responsible for small value R2. Firstly, all the ships did not go
through all types of repair works. For example, some ships have piping repair but no structural steel repair
and tank coating repair; some ships have tank coating repair but no piping repair and structural steel
repair. So, there are various combinations of repair categories and quantities including zero repairs for
some categories. Mathematically, it will result in lower cell values in matrix A (coefficients in simultaneous
equations which are derived from the independent variables) and matrix g (right-hand side of the
simultaneous equations which are derived from independent variables and the dependent variable
jointly) and yield lower regression coefficients and R2.

Secondly, some repair works are a combination of more than one step by nature with different quantities
but used only one quantity to represent the quantity of the whole work. For example, hull coating and
tank coating repair work consists of blasting (grit blasting or hydro blasting) and painting (touch up and
full with multiple coats), but in the calculation of the regression equation, the blasting area only was used
for the whole works ignoring the quantity of painting work. As a result, the effect of painting works is not
accounted for ship repairing time in the regression equation which also resulted in lower regression
coefficients and R2.

However, despite the limitations mentioned above, various findings and the mathematical model can be
useful to the ship managers/ship superintendents and the shipyard as a guide to determine the expected
ship repairing time against a set of the vessel’s particulars and expected the scope of repairing works for
crude oil tankers. The shipyard can make use of this model to estimate the repairing time which can be
translated into total man-days and finally into labour cost for the budgetary purpose. On the other hand,

24
ship owners also can make use of the model to estimate the cost of repairing based on the repairing
specification and the loss of income and to reflect this in the company budget.

5.2 Ship repairing labour

Similarly, the initial analysis was based on the assumption that ship repairing labour (man-days) is directly
proportional to age, deadweight and type of ships with a sample size of 50 cargo ships (Table 2 and Figures
18-20). The resulting regression equation (𝑺𝑺𝑺𝑺𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋 = 224*𝑺𝑺𝐀𝐀 + 0.0083*𝑺𝑺𝐃𝐃 + 1,138 *𝑺𝑺𝐓𝐓 – 2,024) passes
the F-statistic test (𝑓𝑓 > 𝑓𝑓α where 𝛼𝛼 = 0.05) and gives a value of 0.338 for the coefficient of multiple
determination. Assigned values of S T are given in Table 6.

Table 6 List of values assigned to the type of ships

Type Bulk Chemical Crude Oil Container L.P.G.

of Ship Carrier Tanker Tanker Carrier Carrier
ST 1.00 1.35 1.59 1.90 2.90

There are no definite criteria or limiting values for correlation coefficient (r2), multiple correlation
coefficients (R) and coefficient of multiple determination (R2) that reject or accept the adequacy of a fitted
regression model. Similarly, there are also no limiting values to decide strong or weak relationships among
the variables. Mathematically, co-relation coefficient (r2) or coefficient of multiple determination (R2) are
the ratios of the explained variation to the total variation in dependent variable accounted for the change
in independent variables and values lies between -1 to +1. Signs + and - indicate the nature of the linear
co-relationship in the form of positive slope and negative slope respectively. The values 0 and 1 indicate
no linear relationship and perfect linear relationship respectively. The computed value of r or R, other
than 0 and 1, measures the degree of relationship of variables of the assumed model.

The values of r in Table 3 suggest that there is a positive correlation between the variables considered. It
also indicates that considering a simple linear regression, structural steel repair has the highest degree of
co-relation (0.74) followed by ships’ age (0.48), hull blasting repair (0.36), and hull painting repair (0.16)
respectively. On the other hand, the values of r2 of above indicate that 54% of the variation in man-days
is accounted for the difference in structural steel repair followed by 23%, 13%, 2.6% for ships’ age, hull
blasting repair and hull painting repair respectively. The value R=0.5814 (R2=0.338), indicates the degree
of the relationship between the independent variables like ships’ age, deadweight, type and the
dependent variable, man-days in the multiple linear regression models assumed. On the other hand, R2 =
0.338 indicates that about 34% of the variation in man-days is accounted for the changes in ships’ age,
deadweight and type. Remaining 66% is the error of estimation. It means that the unexplained variation
is almost double the explained variation. Unexplained errors occurred for many reasons. One of them is
the absence of one or more independent variables that have a strong relationship with the dependent
variable.

At this stage, the scope of repairing works (category and quantity); mainly, hull blasting, hull painting,
structural steel and piping were also considered as the independent variables.

25
Using a mathematical model under the estimation method of least squares and collected data, the final
regression equation for the ship repairing labour (man-days) is as follows:

𝑺𝑺𝑺𝑺𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋𝐋 = 178*S A + 0.0021*S D + 534*S T - 0.0091*R hb - 0.0026*R hp + 0.078*R S -3.78*R p - 417 (23)

In the regression equation, all the variables, except the hull blasting (R hb ), hull painting (R hp ) and piping
(R p ), have the similar sign as assumed. Age, deadweight, type and structural steel have the positive and
significant impact on the ship repairing labour (man-days). The probable reason for the negative sign for
hull blasting, hull painting and piping could be the quantity of works. In the case of piping works, the
quantity for most of the ship is within 10m (Figure 9) irrespective of locations and systems. This amount
of piping works, invariably, can be completed within much less time than the ship repairing time. As such,
it does not influence the ship repairing time and hence, the ship repairing labour. However, the findings
are in line with the assumptions made earlier except the hull blasting, hull painting, and piping.

Table 7 displays the estimates of regression coefficients and other statistical parameters of the regression
equation during the process of adding new variables. From the table, one can easily recognise that the
successive inclusion of variables of the scope of works in the model contributed higher R2 value, which is
the primary measurement of the adequacy of the model.

Table 7 Different mathematical model values

Regression coefficients Statistical parameter

Mathematical models b0 b1 b2 b3 b4 b5 b6 b7 s R2 F f 0.05
SR LABOUR = f(R s ) 1,951 0.12 NA NA NA NA NA NA 1,565 0.54 65.40 4.05
SR LABOUR = f(R s ,S A ) 626 0.08 162 NA NA NA NA NA 1,312 0.68 50.09 3.20
SR LABOUR = f(R s ,S A ,S T ) -241 0.08 161 501 NA NA NA NA 1,285 0.72 36.73 2.83
SR LABOUR =
-398 0.08 162 512 0.004 NA NA NA 1,298 0.72 27.02 2.60
f(R s ,S A ,S T ,R hp )
SR LABOUR =
-444 0.08 166 552 -0.005 0.002 NA NA 1,308 0.73 21.37 2.45
f(R s ,S A ,S T ,R hp ,S D )
SR LABOUR =
-454 0.08 167 548 -0.004 0.002 -0.005 NA 1,324 0.73 17.37 2.34
f(R s ,S A ,S T ,R hp ,S D ,R hb )
SR LABOUR =
-417 0.08 178 534 -0.003 0.002 -0.009 -3.78 1,318 0.73 15.20 2.26
f(R s ,S A ,S T ,R hp ,S D ,R hb ,R p )
NA : Not Applicable

Focusing on the final regression equation, which passes the F-statistic test, the value of the coefficient of
multiple determination is 0.73 which can be considered safe. Mathematically, it means that 73% of the
variation in the dependent variable is contributed to the change in the independent variables, and
remaining (27%) is called the error of estimation. Statistically, it is referred to as an error of the sum of
squares (SSE) or unexplained variation. The variation behaves randomly or unpredictably (Murray, R.S.
1992). It also reflects the variation about the regression line. However, this error of estimation is due to
the absence of one or more important independent variables, (unquantifiable variables) responsible for
the change in the dependent variable. The coefficient of multiple determination and the error of

26
estimation are inter-related, meaning that for the higher value of the coefficient of multiple
determination, the error of estimation will be little and vice versa. The lower value of the coefficient may
also occur if the collected data suffers inconsistency. It means that when some of the independent
variables have zero values. Received data sheet indicates that some ships do not have any piping works,
and structural steel works and this situation lead to lower coefficient value (Dev, A.K. and Saha, M. 2015).

Validation of a postulated mathematical model is an alternative way to demonstrate the adequacy of

fitness of the model to the system in question. To carry out the validation of the mathematical model
(Equation 23), a total of 11 ships were selected randomly from the database of crude oil tankers, container
carriers, chemical tankers, L.P.G. carriers and bulk carriers to get a fair judgement. Their age, deadweight,
type and scope of repairing works (hull blasting, hull painting, structural steel and piping) were used in
the model to estimate SR LABOUR for each ship and compared with actual data of man-days regarding
deviation. Details descriptions, explanations and discussions are given in Dev, A. and Saha, M. 2016. It is
concluded that the proposed mathematical model cannot capture the effects of uncertainty of weather
condition, the supply of materials, spare parts and so on, which has a significant impact on ship repairing
time and labour.

Normally, when the values of independent variables are closer to the average value of the sample, the
model will yield a good estimate of negligible deviation. It can be demonstrated in Figure 25 (μ refers to
the hypothetically equivalent mean values of various independent variables of sample data). The figure
shows that the values of the independent variables higher than the average value of the sample data
resulted in an increase in the estimated ship repairing labour giving a greater positive deviation. Inversely,
values of the independent variables lesser than the mean value of the sample data resulted in a decrease
in the estimated ship repairing labour giving a higher negative deviation. However, someone may choose
to use the special relationship of the independent variables with the dependent variable provided the
independent variable in question is a major contributor to the total man-days compared to others
independent variables. In other words, the variable must be the one that takes the longest time.

10
DEVIATION IN SR LABOUR (%)

8
6
4
2
0
-2
-4
-6
-8
-10
0.7μ 0.75μ 0.8μ 0.85μ 0.9μ 0.95μ μ 1.05μ 1.1μ 1.15μ 1.2μ 1.25μ 1.3μ
CHANGE IN INDEPENDENT VARIABLES

Fig.25 Change in independent variable versus change in SR LABOUR

27
Finally, the proposed mathematical model may be used to estimate ship repairing labour for overall
planning of man-days for a ship to undergo routine maintenance program. Using the above model as a
guide, shipyards may be able to estimate the expected ship repairing labour (man-days) against an
expected scope of repairing works of a ship. The shipyard can make use of this model to predict the ship
repairing man-days that can be translated into total no. of men and finally into labour cost for commercial
purpose and control of labour cost. While using the model to estimate the expected ship repairing labour,
one may be aware of potential demerit in predicting ship repairing labour for a set of the scope of work.
If the expected independent variables are close to mean value (Table 2), then the model will provide a
good estimate of labour. However, for low and high value, the estimate will be low and high accordingly.
Also, one may consider allowing some allowance on top of the model value to accommodate the effect
of location of repair and skill of workforce. Shipyards may think of predicting the ship repairing labour
using the model for a ship, then collect actual labour data from that ship and compare with the predicted
value. By doing this exercise for few ships, shipyards may be able to adjust the predicted value in
percentage with the actual value.

6. Concluding remarks.

This paper explores and identifies the possible independent variables responsible for ship repairing time
(day) and labour (man-days). This paper also suggests a possible relationship between various variables
in the form of a mathematical equation using multiple linear regressions and verified with statistical
testing parameters to demonstrate the adequacy of the model for the system. The condition f > f α
suggests the rejection of the null hypothesis. As such, it is logical to conclude that there is a significant
amount of variation in their response (the dependent variable) due to the differences in independent
variables in the postulated model.

However, despite the limitations mentioned above in the discussion, various findings, and the
mathematical model can be useful to shipyards. Shipyards may use the model as a guide and may be able
to estimate the expected ship repairing time (days) and labour (man-days) against an expected scope of
repairing works of a ship. The shipyard can make use of this model to estimate the ship repairing time
(days) and labour (man-days) that can be translated into total no. of men and finally into labour cost for
budgetary purpose and control of labour cost.

This research may be considered as a first step in uncovering the relationships that exist among the
possible variables in ship repairing activities and entire ship repairing time (days) and labour (man-days).
This research can further be fine-tuned with larger sample data size with all types of
repairing/maintenance scopes. Also, it can be considered to use partly of sample data (say 80%) for the
model formulation and the remaining for the validation of the developed model. Furthermore, Artificial
Neural Network may be considered as an alternative method to develop the mathematical model for the
estimation of ship repairing time and labour.

In connection with ship repairing time, future studies could be for other types of ships. Another area is to
investigate to find ways to incorporate different types of ships in a single equation. Similarly, for ship
repairing labour, the further investigation may consider addressing (i) the reasons that result in the

28
 difference between actual value and model value of ship repairing labour and (ii) to incorporate the effect
of skill of workforce, materials, and spare parts availability, and location of repairing items.

References

Apostolidis, A., Kokarakis, J. and Merikas, A. 2012 Modeling the Dry-Docking Cost - The Case of Tankers,
Journal of Ship Production and Design, Vol.28, No.3, pp. 134-143.

Dev, A.K. and Saha, M. 2015 Modeling and Analysis of Ship Repairing Time, Journal of Ship Production and
Design, Vol.31, No.2, pp. 129-136.

Dev, A.K. and Saha, M. 2015 Modeling and Analysis of Ship Repairing Labor, Journal of Ship Production
and Design, Vol.32, No.4, pp. 258-271.

Emblemsvag, J. 2014 Lean Project Planning in Shipbuilding, Journal of Ship Production and Design, Vol. 30,
No. 2, pp. 79-88.

Jose, R.S.C. 2009 A Goal Programming Model for Vessel Dry Docking, Journal of Ship Production, Vol. 25.
No. 2, pp. 95-98.

Meland, K. and Spoulding, R. 2003 Workload and Labor Resource Planning in a Large Shipyard, Journal of
Ship Production, Vol. 19, No. 1, pp. 38-43

Murray, R.S. 1992 Theory and Problems on Statistics, McGraw-Hill International, UK.

Naffisah, M. S., Surjandari, I., Rachman, A. and Palupi, R. 2014 Estimation of Dry Docking Maintenance
Duration using Artificial Neural Network, International Journal of Computing, Communications &
Instrumentation Engineering (IJCCIE), Vol.1, Issue.1, pp. 113-115

Ozgur, U.S. 2008 Interaction between the Ship Repair, Ship Conversion and Shipbuilding Industries, report
prepared by Organization for Economic Co-operation and Development, Government of Turkey.

Surjandari, I. and Novita, R. 2013 Estimation Model of Dry Docking Duration Using Data Mining, World
Academy of Science, Engineering and Technology, Vol.7, No.7, pp 1718-1721.

Turan, O., Olcer, A.I., Lazakis, I., Rigo, P. and Caprace, J.D. 2009 Maintenance/repair and production-
oriented life cycle cost/earning model for ship structural optimization during conceptual design stage,
Ships and Offshore Structures, Vol.4, Issue 2, pp 107-128.

Victoria, D.; Dennis, F.; Lisa, H. and David, T. 2010 Transforming the Ship Building and Ship Repair Project
Environment, Journal of Ship Production and Design, Vol. 26, No. 4, pp. 265-272.

Walpole, R. E. and Myers, R. H. 1978 Probability and Statistics for Engineers and Scientists, Macmillan
Publishing Co., Inc., New York.

29

View publication stats