Professional Documents
Culture Documents
1–8
http://dx.doi.org/10.5957/JSPD.29.1.120045
1. Introduction block sequence generation. They also mention the different strat-
egies for erection used in practice such as starting from the
BLOCK ERECTION is the process in which the hull of a ship is built middle, fore, or aft part of the ship; erecting the ship slice by slice;
by joining blocks together. It is also called hull erection. Blocks or using a pyramidal strategy. Another recent article in which
are large welded steel structures that are typically manufactured as different erection constraints are explored is by Meijer et al. (2009).
single pieces. When a block is ready, it is typically arranged near They focus on risk management and their aim is to find the most
the launching place, lifted to the launching place, and welded to critical blocks in the schedule. Both articles lack further analysis
other blocks there. Block erection in shipbuilding is a unique phase of the effects of the erection parameters on the erection schedule,
when shipbuilding is compared with the normal type of mechanical which is dealt with in this article.
engineering production. It is also typical that block erection is Block erection scheduling is generally considered a hard prob-
a bottleneck for the whole process (Yoon et al. 2006). The block lem and several articles give heuristic solutions to the problem
erection sequence typically defines the production order for the (Okumoto 2002; Varghese & Yoon 2005, 2006; Yoon et al. 2006;
previous stages as well. This all makes the scheduling of a block Chung et al. 2009; Jinsong et al. 2009). In the problem setting,
erection order important. the blocks form a hierarchical succession network, which means,
In the literature, articles about block erection exist. In many e.g., that a block on the second level is a successor of the cor-
articles such as in an article by Lee et al. (Lee et al. 1997), block responding block on the first level. This kind of property with
erection is considered together with the whole production process. the different processing times can make the problem hard to solve
In most of those articles, the erection is typically just mentioned optimally. Because of this, different heuristics such as genetic
briefly as one part of a larger problem. There are some exceptions algorithms are used to solve the problem. An article by Okumoto
to this. In an article by Caprace et al. discrete-event simulation (2002) proposes a solution for positioning a block with dimen-
is used to solve the erection process (Caprace et al. 2011). Their sional errors during the erection. A solution in an article by Yoon
approach gives a list of constraints to be considered during the et al. (2006) optimizes the makespan in the situation with flexi-
ble lifting dates of the blocks. An article by Varghese and Yoon
Manuscript received by JSPD Committee August 21, 2012; accepted (2005) deals with the spatial block arrangement problem, which
November 19, 2012. is the step before the erection. An article by Jinsong et al. (2009)
studies the block erection with multiple cranes and shows how pre-erection area to their place on the hull of the ship. After lifting
the problem can be defined as an identical parallel machines the block, the crane becomes available for other blocks, but the
scheduling problem. There are a few other kinds of solutions to lifted block is positioned in its place and joined to the other
the erection problem. Another article by Varghese and Yoon blocks by welding. To keep the partial construction stable and
(2006) shows that the problem can be formulated as a traveling support the joining work, temporary supporting structures are
salesman problem. The theory of constraints approach is briefly needed. Lifting typically takes several hours, whereas the join-
considered in an article by Chung et al. (2009). All of these ing work can take days. Figure 2 shows an overview of the block
articles typically focus on how a specific method can be used erection process.
to model the problem and the focus of those articles is on the The block erection schedule determines the lifting times for
details of the specific methods rather than the effects of the the blocks. For the schedule to be made, capacity and block pre-
parameters on the whole production, which are explored in this cedence constraints have to be considered. Later in this section,
article. In addition, this article also explores how and when the the constraints of erection scheduling are discussed and a mathe-
problem can be solved analytically. matical notation of block erection scheduling is formed. Using
This article constructs a mathematical model of erection sched- the constructed model, the effects of lifting and joining times are
uling, analyzes the special cases of the model, and shows how studied later in the article, in Section 3.
in many cases the model can be solved analytically using well-
known algorithms. The analysis and algorithms are used to gain 2.1. Constraints of block erection scheduling
important insights into how the lifting time and block joining
time after erection affect the scheduling of the erection. To the Erection scheduling orders the block erection sequence so that
best of our knowledge, this is the first time that a detailed model the makespan, i.e., the length of the whole schedule, is minimized.
for the block erection has been used to study the impact of the Here the block sequence has the following constraints:
lifting time and block joining time, which includes, e.g., the weld- 1. Lifting capacity: only one block can be lifted at a time.
ing time and creation of any support structures for other blocks. Because of this, there has to be a certain time lapse between
The rest of the article is organized as follows. Section 2 two successive blocks.
describes the block erection model. Section 3 shows how joining 2. Joining time of adjacent blocks: there is typically a mini-
times affect the total production time, or makespan, in three dif- mum amount of time between the erection of two adjacent
ferent cases. Section 4 outlines the insights from the earlier results blocks. This includes time for joining the first block to the
of the article and, finally, Section 5 concludes the article. hull and construction of the support structures for the next
block. This article studies a version of this constraint in which
horizontal and vertical joining times are different.
2. Block erection schedule 3. Structural stability: the construction should be stable,
e.g., the construction should have more material horizon-
The shipbuilding process typically consists of the manufactur-
tally than vertically.
ing of blocks, block outfitting, block erection, and final outfitting.
4. No-skipping-blocks property: because of the tolerances,
At first, blocks are typically constructed in a factory-like envi-
blocks should not be inserted between others.
ronment. Then basic outfitting is done for them in the block
outfitting phase. After this the blocks are erected to build the The constraints used here are similar to the ones that were
ship and finally the remaining outfitting is done. Figure 1 shows described in the article by Meijer et al. (2009). However, in that
a simplified overview of the shipbuilding process. article, the descriptions of the constraints were quite brief and the
Block erection is the focus of this study. In it, the blocks are effects were not studied, which is done in this article. This study
typically erected onto the ship hull by a crane. In block erection, concentrates on erecting the blocks as fast as possible, which might
the blocks are first attached to the crane and then lifted from the not be optimal for the whole shipbuilding process in some cases,
Nomenclature
H ¼ height of hull p ¼ minimum delay between consequent DH ¼ joining delay in the direction of the height
L ¼ length of hull erections (lifting time) of the hull, i.e., vertically
A ¼ lifting time matrix; Ahl is the lifting time DL ¼ joining delay in the direction of whl ¼ weight for the lifting time of the block at
of the block at level h and column l the length of the hull, i.e., horizontally level h and column l
Fig. 3 Example of ship block erection schedule, i.e., the lifting time matrix A when H is 4 and L is 9
where, e.g., closing the deck would be useful after most of the Figure 3 shows an example schedule and describes how these
blocks had been erected. The closing deck constraint appears in variables and block structure are related. The ship is built starting
another article by Meijer et al. (2009). In another article by Meijer from the middle at the time 1 and there is a delay of 1 between
et al. (2010), there is a yard-dependent preference as well, which consequent blocks. Because there are no additional delays, the
is not considered here. last block is erected at the time HL ¼ 36.
in each direction to finish the ship, each of which has to wait If the joining times are zero, then there are no precedence
DL time units. Thus, it takes at least (L-1) DL /2 time units to delays in the problem. This case is also applicable if the joining
finish the corner part of the ship. This is the same as with the times are short compared with the lifting times (i.e., DL << p,
previous case in which DH DL. From this part, it is easy to DH << p). The problem can be solved efficiently, because all
see that starting from the middle of the ship minimizes the feasible solutions have a minimum makespan. Because there is
makespan, because it would take more time to fill the larger corner no waiting after lifting, the makespan, m, is clearly the time that
than two equal, but smaller, corners. Therefore, the makespan it takes to lift all blocks, i.e.
for filling the corners of the ship when work is started from the m ¼ HLp ð13Þ
middle of the ship is
L1 This approach can be adapted for block completion times with
m2 ¼ DL ð11Þ a nondecreasing cost function C(l, h, t) for the time t and a block
2
in the position l, h. To do that, the first block on each level has
Now the makespan can be calculated from these equations. to be fixed, because it defines the relative precedence order
By summing up the previous equations, (10) and (11), for m1 and of the blocks on a single level. Now, when there are only fixed
m2, we get a formula for the makespan m, which is precedence constraints between the blocks, the problem can be
L1 arranged as a single machine problem with the precedence con-
m ¼ ðH1Þ maxðDL ; DH Þ þ DL ð12Þ straints of zero delay (In the theory of scheduling, the formal
2
definition of the problem is n j1 j prec j Cmax [Conway et al. 1967])
From equation (12), it is easy to see that in general, H has a and solved using the algorithm in Lawler (1973). The structure of
larger effect than L and DL has a larger impact than DH. Figure 5 the algorithm is the following.
shows how DH and DL affect the makespan with different H and
1. Let the time t be the sum of the lifting times of all the blocks.
L values.
2. Consider the blocks which do not have any predecessors
among the blocks that are not scheduled. Schedule the one
3.2. Short joining times with the lowest cost if done at the time t to the time t.
Reduce t by the lifting time of the block.
If the joining times are short compared with the lifting time,
3. Repeat Step 2 until all the blocks are scheduled.
the problem of scheduling becomes quite simple. This model
becomes practical, e.g., if the joining times of the blocks are parts For a practical model, the considered lifting time has to be the
of the lifting times and there are no actual joining constraints. sum of the lifting time and the maximum of the joining times.
Fig. 5 Effect of DH and DL on the makespan, m. The results are generated using equation (12)
Fig. 6 The effect of L on the makespan with different constraints when the Coffman-Graham algorithm is used to calculate the makespan. H is
set to 3. DL ¼ DH ¼ p ¼ 1. L varies from 1 to 8
joining time DH. It is better to reduce one unit from the horizontal In the case of similar lifting and joining times, it is easily seen
joining time, DL, than one unit from the vertical joining time, that the starting of the building and ending of the building are the
DH, when (L – 1)/2 2H þ 1 (flat ship), even if DL > DH. cases in which the joining times have an effect. After the first
If DH > DL, reducing DH does not give any benefit. However, couple of blocks, there are always blocks that can be erected and
although horizontal joining times have a greater effect than verti- joined to the blocks of which the joining times have passed. The
cal ones, the number of blocks in the length of a ship does have a same can occur when the last blocks are inserted. From the results
smaller effect than the number of blocks in the height of a ship. in Fig. 6 and Fig. 7 we can see that starting from the middle of
The reason for this is the no-skipped-blocks constraint, which was the ship is the better solution. The reason for this is that it allows
used when the equation was being formulated. the ship to be built in two different directions, and blocks can be
If the joining times are short, the scheduling is shown to be erected in one direction while the joining in the other direction
easy because all feasible solutions give a minimal or near-minimal is completed, which reduces the effect of the joining time at the
makespan. In that case, we should clearly focus on other things beginning and end of the erection schedule. Stability does have
than the makespan. However, if the joining times can be made a significant impact with ships with a small number of horizontal
short, the result shows that the scheduling has significant flexi- or vertical blocks.
bility, because changes in the schedule do not incur any penalty if It should be noted that the solutions do not cover all possible
the schedule is feasible after the changes. joining time and lifting time combinations. Figure 8 shows the
areas where the analysis of the article is valid and where the
algorithms described should be used. The equations with short
lifting times, which make the construction form a pyramid, are
valid if the lifting times are short compared with the joining times.
All schedules give a good makespan if the lifting time is the
bottleneck. The Coffman-Graham algorithm can be used when
the lifting and joining times are equal or the lifting times are
multiples of the joining times. If both times are short, then it is
unlikely that the erection scheduling is the problem in the ship-
building and other things should be optimized.
As shown in Fig. 8 and stated in Section 3.4., there exists a
significant area where the algorithms do not work, which is when
the lifting time is significant but shorter than the joining times.
For these cases, the model in Section 2 can be used as a MILP
model and solved optimally or a suitable heuristic could be used.
The algorithms in this article that are optimal for the other cases
could also be used as heuristics for that case.
5. Conclusion