CONSTRAINT-BASED SIMULATION OF

OUTFITTING PROCESSES IN SHIPBUILDING

AND CIVIL ENGINEERING
Markus König1, Ulrike Beißert1, Dirk Steinhauer2, Hans-Joachim Bargstädt1
1
Bauhaus-University of Weimar, Department of Civil Engineering,
99423 Weimar, Marienstrasse 7, Germany
2
Flensburger Schiffbau-Gesellschaft mbH & Co. KG, Production Technology,
24939 Flensburg, Batteriestrasse 52, Germany

[email protected]

Abstract

Currently, in shipbuilding as well as in civil engineering outfitting planning is not sufficiently

investigated. A multitude of requirements such as technological dependencies, resource and
work space assignment have to be considered. Outfitting processes in both domains are
distinguished by interferences, disturbances, great interdependencies and different
surrounding area requirements. In consequence, on production site an extensive coordination
effort is necessary to handle these problems. A realistic planning and detailed analysis will
help to reduce the on-site coordination effort and not to overrun the projected costs and time.
Appropriate tools have to be implemented to support planners and improve the outfitting
planning. Within the cooperation SIMoFIT (Simulation of Outfitting Processes in
Shipbuilding and Civil Engineering) a discrete-event simulation framework is developing to
support outfitting planning in shipbuilding and civil engineering. This paper focuses on using
a constraint-based simulation approach to detail outfitting tasks and their corresponding
restrictions and requirements. Typical outfitting restrictions and requirements are specified as
hard and soft constraints. This approach guarantees a high flexibility in modeling processes.
Further, outfitting processes can be specified more realistically. Thus, different practical
schedules can be simulated and evaluated in terms of work and material flow organization,
utilization of space and worker’s efficiency as well as process costs. The framework was lab
tested by both cooperation partners and proves its suitability to support the outfitting planning
process in shipbuilding and civil engineering.

Keywords: constraint satisfaction, discrete-event simulation, outfitting processes,

shipbuilding and civil engineering

Presenting Author’s biography

Markus König is assistant professor for Theoretical Methods of Project
Management in the Department of Civil Engineering at the Bauhaus-
University Weimar, Germany. His research activities include simulation
in production and logistics, workflow and knowledge management,
process modeling, graph theory as well as mathematical aspects of
project and risk management.
 Engineering) a discrete-event simulation framework is
1 Introduction developing to support outfitting planning. SIMoFIT
was established between Bauhaus-University Weimar
Successful project realization is linked to quality, time and Flensburger Shipyard [2]. Outfitting processes in
and cost criteria. Often in project planning process it is shipbuilding and building industry bear a high
not possible to find optimum solutions to satisfy all resemblance to each other. The same circumstances
three criteria in equal measure. For example, an have to be considered such as dependencies between
exceeding quality leads to higher costs as normal. outfitting tasks, availability of resources and required
Thus, thorough and exact planning is necessary to work spaces as well as changing transport ways and
realize ship and building projects successfully. delivery dates. In addition the planners have to answer
Currently, in shipbuilding as well as in civil the same questions: find a practicable schedule with
engineering the planning of outfitting processes is not sufficiently utilized equipment and employees that
sufficiently considered. That is quite surprising, due to satisfies principal guidelines.
the fact that building installation and outfitting Therefore, a modular simulation toolkit is used and
processes represent about 38% of the construction adapted for outfitting processes. Flensburgers and
volume in building industry in Germany (2005) while their partners of the SimCoMar (Simulation
the main construction trade represents only about 30% Cooperation in the Maritime Industries) community
[1]. Outfitting processes in both domains are are developing the STS (Simulation Toolkit
distinguished by interferences, disturbances, great Shipbuilding) [3], [4]. The STS contains several
interdependencies and different surrounding area simulation components, for example, to model steel-
requirements. In consequence, on production site an prefabrication lines as well as material and assembling
extensive coordination effort is necessary to handle control components. These adjustable components can
these problems. A realistic planning and detailed be combined to a complete simulation model. Within
analysis will help to reduce the on-site coordination the SIMoFIT cooperation the existing simulation
effort. The question is: what can be done to improve experience of Flensburgers and the construction
the planning of outfitting process? Appropriate tools know-how of Bauhaus-University Weimar are
have to be found. combined to define and to develop further components
A multitude of requirements such as local, technical to model outfitting processes [5].
and project-specific have to be considered in outfitting
planning. In addition, the assignments of employees 2 Constraint Satisfaction
and equipment have to be regarded as well.
Constraint satisfaction is a powerful paradigm for
Consideration of all these different restrictions and
modeling combinatorial search problems [6].
requirements result in a wide choice of practicable
Conditions or restrictions of variables are dedicated as
outfitting schedules. Often these different solutions
their constraints. The problem consists in finding a
are not sufficiently analyzed. However, an in-depth
value combination for all variables, where all
investigation of the various solutions is very useful not
constraints are fulfilled [7], [8]. In this case,
to overrun the principal guidelines regarding costs and
simulation objects such as outfitting tasks are
time.
described by a set of variables. Relations between
In manufacturing industry such as steel prefabrication these variables like execution sequences or resource
and ship assembling simulations are used successfully requirements are described by constraints.
to improve production processes. These processes and Consequently, the constraint-based approach
hence their related simulation models are guarantees a high flexibility of modeling outfitting
characterized by static layouts, well-known and processes, if additions or new prerequisites in
limited process variations. Due to the fact that processing occur. The model can be easily adapted by
outfitting processes are more complex and dynamic, a simply adding or removing certain constraints. The
flexible simulation framework has to be developed. more constraints are specified, the more the solution
space is restricted (e.g., [9], [10]) and, consequently,
This paper highlights on a constraint-based simulation the more the multitude of possible schedules is
framework to model outfitting processes in
deducted.
shipbuilding and building engineering. The constraint-
based approach is proved to be appropriate to define Constraint satisfaction can be used effectively to
outfitting tasks and their corresponding restrictions model many forms of reasoning (e.g. temporal
and requirements. Using this constraint-based reasoning) and applied to many problem domains (e.g.
simulation approach different practical outfitting scheduling) [6], [11]. The classifying of constraints
schedules can be generated and evaluated in terms of into hard and soft constraints adds to models’ realistic
work and material flow organization, utilization of [12]. Hard constraints define stringent conditions in
space and worker’s efficiency as well as process costs. construction processes. They must be fulfilled before a
work can be started. Essential technological
Within the cooperation SIMoFIT (Simulation of dependencies and needed resource capacities to
Outfitting Processes in Shipbuilding and Civil execute a work are defined as hard constraint. Usually,
there are many requirements and preferences like Hard Constraints
functional or appropriate dependencies in outfitting
execution processes that have to be considered in the Capacity Amount and qualification of
planning process. As they don’t have to be fulfilled employees and equipment
completely they can be described by soft constraints Availability Supply of material linked to the
(e.g. [9], [13]). Consequently, the soft constraints can requirement of storage area
be violated to find possible variable configurations Safety Employees’ and equipments’
[14]. Extending the model on soft constraints allows Criteria protection
for discrimination among all the solutions which
satisfy the hard constraints [6]. Soft Constraints
Looking for solutions that fulfill all hard constraints Productivity Relation between workers’
and violate the soft constraints as little as possible is productivity and provided working
known as Constraints Satisfaction Problems (CSP). In space (e.g., [15], [16])
this case, Monte-Carlo simulation is used to solve the Strategies Predefined execution orders and
CSP: one simulation run calculates one possible value established process sequences
combination [7]. Allowing for search spaces’
dimension an adequate amount of simulation runs has Currently, only physical constraints and some enabler
to be performed in order to provide a significant set of constraints emphasized by Sriprasert and Dawood
solutions for the following investigation. [17], [18] are considered in the presented approach.
Looking for optimal solutions is known as Constraint 3.1 Technological dependencies
Optimization Problems (COP). Usually, the model Technological dependencies define stringent rules for
allows only “hard” Boolean-valued constraints but execution process, for example, definite sequences
adds a cost function over the variables that must be between construction tasks or work steps. Thus,
minimized. A constraint solver is used to find an technological dependencies are specified as hard
optimal assignment to all problem variables that constraints. For instance, the technological constraint
satisfies the constraints [6]. Using a constraint solver “work step A before work step B” means that work
such as a backtracking search algorithm to find step A has to be completed before work step B can be
optimal solutions is very time-consuming. In started. Also practical aspects can be described by
construction practice, an optimal solution often is not technological dependency constraints. For example, it
necessary. In fact, it is adequate to search or to is a good practice to achieve dimensional accuracy of
compute a few practicable schedules, reminding that a brick wall to first lay the cornerstones before
project constraints are changing rapidly and often building the intermediate wall sections.
defined construction schedules are valid only a couple
of days. Thus, it is more important to generate 3.2 Capacity
alternative practicable schedules fast to adjust The capacity constraint considers resource boundaries
occurring disturbances than to optimize an implausible like a limited amount of serviceable employees and
execution flow. equipment. Their quantity is finite. If the capacity is
The simulation runs can be analyzed afterwards with exhausted, no more work step can be started.
regard to different criteria. Preferred solutions can be Furthermore, the quality and quantity of resources can
visualized based on the simulation documentation. be determined. For example different skilled
Visualization increases the possibility to detect employees can be specified, if execution of different
potential improvements of the execution. Further, work steps requires various skilled employees.
current production states can be easily integrated into 3.3 Availability
the model. Based on the current execution state some
prognoses can be drawn. Availability represents material flows. For example
limited supply of material can be considered as well as
3 Outfitting Task Constraints the herewith linked requirement of storage area.
Following, availability is also defined as hard
The constraint-based approach is used to describe the constraint. The unavailability of material corresponds
outfitting tasks. Attributes of simulation objects and to the possibility of supply bottlenecks in production.
their relations are described by hard and soft Further, if suitable storage areas cannot be offered,
constraints. An overview of the defined outfitting more attention has to be paid to supply of material and
constraints is given in table 1. equipments’ disposability.
Table 1 Outfitting Process Constraints 3.4 Safety Criteria
Hard Constraints Safety criteria are protection criteria. Durations or
distances to protect employees and to assure the right
Technological Constructional and formal aspects
exposure to the equipment are specified. Safety
dependencies
criteria are specifying as hard constraints. Thus, if
prescribed safety criteria cannot be obeyed, work steps • Each work step has to be realized by its
cannot start. Typical safety criteria are distances required amount of employees, which cannot
between persons and machines, maximum working be deducted before finishing the work step
time for equipment and personnel or essential needed
working space (e.g., [15], [18]). • Each work step will be executed without a
change of the working position of employees
3.5 Productivity or equipments
The productivity constraint is a functional bench If a new event occurs, all not started work steps have
mark. It will be defined as a soft constraint. The to be checked. A work step can be executed, if all
complex coherence between free working space and associated hard constraints are fulfilled. Further, for
productivity of employees needs to be considered. For all executable work steps the fulfillment of soft
example, an employee only achieves 100 percent constraints has to be checked up. The executable work
productivity, if required work space is provided. steps are ordered by their percentage of soft
Productiveness will rapidly fall, if this operating range constraints’ fulfillment. Only the first in the list of
cannot be guaranteed [13], [16]. executable work steps can be started. If several work
3.6 Strategies steps fulfill their soft constraints in equal measure one
of them is chosen randomly.
Strategies are proven formal aspects and are defined
as soft constraints. Predefined execution orders and Each started work step presupposes certain material,
established process sequences can be simulated and resources and working space. The required objects
analyzed. They can extend the technological have to be locked during its execution. That means
dependencies but are not binding. For example, material, resources, equipment and working spaces
execution according to the length of walls or cannot be used by other work steps. After locking all
according to the longest distance between the working material, resources and spaces the work step state
groups can be modeled to assist the user on deciding changes from “not started” to “started”. Subsequently,
which assembling order is most useful (figure 1). all “not started” work steps have to be checked up
again on fulfillment of their hard and soft constraints
by going to step one until no more work steps can be
started at the current time (figure 2).

Figure 1 Examples of working strategies

4 Simulation Concept
The constraint-based approach is used within a
discrete-event simulation concept. That means, only
points in time are inspected at which events occur.
Typical events are, for example, a work step is
completed or a material element is entering a storage
area. After an event occurs, it has to be investigated, if
new time points have to be generated or existing time
points have to be moved. Thus, the simulation time
leaps from event to event.
This presented discrete simulation approach focuses
on simulating single work steps. Each task, e.g.
erecting a partition wall, is decomposed into work
steps such as plastering or installing a stub. A work
step has a current state – “not started”, “started” and
“finished” – and requires a certain execution time. The
execution of a work step is bound to some general
restrictions:
• Each work step has to be executed without Figure 2 UML activity diagram for starting
any interruption a new work step
The simulation time is continuously checked during
the simulation run. Every started work step exhibits a
determined execution time. If the remaining time is 5.2 Resource Administration
expired, the work step is marked as finished. Its
The resource administration component was
locked resources, equipments and working spaces will
implemented to manage, assign and release the
be unlocked and can be used by other work steps.
required resources of work steps. Further, the
The simulation will be repeated until all work steps administration records each access to an employee or
are finished. Events such as starting and finishing of work equipment. Resources are, as per description,
work steps as well as locking and unlocking of employees or work equipment. Currently, only
material, resources, equipment and working space are employees and their technical skills are considered.
recorded. Thus, the simulation run can be investigated Movable equipment such as welding apparatus or
afterwards. One simulation run calculates one erecting scaffoldings will be implemented next.
practical execution schedule, one material flow as well
5.3 Spatial Management
as utilization of employees and equipment. The
quantifying of simulated solutions is not intended by An important objective of this simulation approach is
the simulation itself. The overall goal is to simulate the consideration of required work and storage spaces
different practicable solutions, which can be analyzed as well as transport ways. Therefore, a special cell-
regarding principal guidelines such as time, cost and based spatial component is developing to manage
quality. available, required and locked spaces on production
sites. Each production site, such as buildings or ships,
5 Implementation is divided into levels. A production level describes, for
example, building storeys or fire zones of ships.
The presented constraint-based simulation approach is Currently, all production levels have to be defined
implemented by extending the Simulation Toolkit manually. Each level is modeled by a regular
Shipbuilding (STS) of the SimCoMar community. The rectangular grid (e.g. [21]). The cells of the grid have
STS uses the discrete-event simulation program Plant a certain state – “unlocked” or “locked” – and special
Simulation provided by UGS Tecnomatix [20]. This attributes such as “generally locked for storage” or
simulation framework enables modeling, simulating “generally locked for transportation” (figure 3).
and visualizing of production systems and processes.
Some important features are [20]:
• object-oriented, hierarchical models of
logistic and production processes
• graphs and charts for analysis of throughput,
resources and bottlenecks
• comprehensive analysis tools, including
automatic bottleneck analyzer, sankey
diagrams and Gantt charts
• 3D online visualization and animation
• open system architecture supporting multiple
interfaces and integration capacities (CAD,
Oracle, SQL, XML, etc.)
To generate project-specific constraint-based
simulation models for outfitting processes in
shipbuilding and civil engineering the following
simulation components are extended or implemented.
5.1 Material Administration
The material administration component of the Figure 3 Production site, production levels, cell-based
simulation model manages the material elements of all spatial component and locked cells
outfitting work steps. All material elements are 5.4 Transport Control
generated by a special supplier component of the STS
based on a material sheet. Every outfitting element The transport control component of the STS manages
such as a panel, plate or plasterboard is registered at transport equipments and transport requests. Typical
the material administration component by committing equipments are cranes or lift trucks. If a transport job
its current storage position. Thus, if a work step is requested, the transport control component provides
requests a certain material element, the material a method to find and lock required transport
administration reports whether the requested element equipment and possible transport ways considering
is available or not and submits its current storage current available spaces. Possible transport ways are
position. detected by using graph search algorithms. Currently,
an A* search algorithm (e.g. [22]) or a Dijkstra associated work area, appropriate transport jobs have
algorithm (e.g. [23]) can be used to calculate shortest to be assigned. The material is marked as used. The
paths. Before a work step can be started, resources as well as the work area are locked. Work
corresponding transport processes of requested steps’ start is suspended until all material and
material or equipment have to be initiated. equipment were transported. Finishing transport jobs
and work steps, respectively, generate new simulation
5.5 Constraint Management
events. If a new event occurs, the assembling
The constraint management component stores all component controls which resources and spaces have
outfitting work steps and their associated hard and soft to be unlocked and further requests a new executable
constraints. Work steps as well as the technological work step.
dependency, capacity, availability and strategy
5.7 Data Management
constraints are generated by using predefined
templates. For example, a typical template for For building up a practicable simulation model the
technological dependency in drywall assembling is: definition of input data is very time-consuming. The
U-channels on floor and ceiling have to be fixed, work steps and their execution times have to be
before C-stubs can be installed. These templates are specified as well as work spaces relative to production
specified manually. areas, site layout with spatial restrictions, material
sheets and detailed assembling positions. Some input
If a new simulation event occurs, all currently not
data can be transferred directly or simply adapted
started work steps have to be checked up on
from CAD-systems; other data have to be defined
fulfillment of their constraints. Subsequently, a list of
manually.
next executable work steps is generated, ordered by
the percentage of their soft constraints fulfillment. Currently, the major problem consists in: data of
Now, the first work step in the list is submitted to the available CAD-systems often do not contain the
assembly control component. required details for simulation models. Typically, in
building industry production objects such as drywalls
According to the specified constraint types different
are specified by a boundary representation model. But
framework components are used. For example, the
for a detailed assembly simulation construction details
fulfillment of availability constraints only can be
such as channel, stud or plasterboard objects are
checked by using the material administration. The
needed. A practical solution is to implement special
relations between different constraint types and the
purpose data generators. Such data generators define
implemented components are shown in figure 4.
all required input data of a certain outfitting process.
Within this research activity a first prototype of a
drywall generator is implemented. Depending on the
length of drywalls and the desired distance between C-
studs, all work steps, material sheets and assembly
positions can be generated. Currently, Flensburger
Shipyard is developing an appropriate data structure to
manage simulation input data in shipbuilding.
5.8 Visualization and Animation
Know-how exchange assistance within a company and
between partners respectively is a convincing
argument for using visualization in outfitting
planning. A detailed visualization and animation of
execution sequences is a valuable communication
basis for planners and executers. Especially, to see
construction processes from a different angle may help
to improve projected execution strategies.
Figure 4 Relations between constraint types and Furthermore, using visualization to depict problems
simulation components might be helpful to argue for exceptional facilities and
5.6 Assembly Control extraordinary expenses.

Starting and stopping of work steps is the essential The 2D visualization and animation concept of the
function of the assembly control component. used simulation application is based on icons for static
Primarily, after receiving the next executable work and moveable object as well as animation points.
step from the constraint management component the Every simulation object has a special icon
assembling control checks the current storage position representation. During a simulation run objects can
of the presupposed material and equipment. If the change their positions, for example, employees are
material or equipment is not located nearby the working at certain places or material elements are
assembled at certain positions. To assign a new
position to an object its associated animation point has Drywall Constraints
to be moved. This only can be done, during events’
Constraints for assembling a drywall have to be
processing. To implement continuous transport
specified in the next definition step. Within a drywall
animations between two successive events appropriate
template technological dependencies are defined
animation polygons can be specified. In addition, a
generally, as shown in figure 6. Sticking an
simplified 3D visualization also is possible. Each 2D
intumescent strip and a U-channel together, for
icon can be linked to some 3D visualization objects
example, needs drywall calibration as finishing work
implemented as cuboids. For the 3D animation the
step.
same animation points are used.

6 Examples
The research approach and the implemented
simulation components are validated by the
cooperating partners in both industrial sectors.
6.1 Example 1: Assembling drywalls
The Bauhaus-University Weimar investigates drywall
construction processes in office buildings using the
developed framework. In this example thirteen
drywalls have to be installed (figure 5). The material
sheet of a drywall includes the number of intumescent
strip rolls, U-channels, C-studs, plasterboards, loft
insulating rolls and plaster bags.

Figure 6 General technological dependencies between

Figure 5 Building storey with drywalls drywall work step types
Drywall work steps Generally, certain material and resources are required
to execute a drywall work step. For example, to
The assembling process of a drywall consists of eight execute the work step “sticking strip and U-channel
work step types: calibrating the wall, sticking together”, an employee with the skill “drywall
intumescent strips and U-channels together, fixing U- constructing”, an intumescent strip roll and a U-
channels at ceiling and floor, installing C-studs, fixing channel element are needed. For each drywall work
plasterboards first side, filling loft insulation material, step type the corresponding availability and capacity
fixing plasterboards second side and plastering constraints have to be specified.
drywall. Currently, work steps like cutting material
and mixing plaster are not considered. For each Resource definition
material element, single work steps and their Resources have to be defined manually for each
execution positions have to be calculated based on simulation experiment. In the presented example only
these eight work step types. For example assembling a technical skills and workers are specified. To execute
drywall of length 4 m and distance of 0.625 m drywall work steps the following skills are specified:
between the C-studs consists of seven work steps calibrating, drywall constructing, filling insulation
installing C-studs and eight work steps fixing material and plastering. Based on these skills different
plasterboards, amongst others. The execution time for types of employees are defined: foreman (all skills),
each work step is calculated based on well-known worker (all skills excluded calibrating) and laborer
working time standard values (e.g., [24]). For (filling insulation material and plastering skill).
example, generally a worker needs about 0.1 h/m² to
fill insulation material.
Strategy constraints After running all experiments the minimal working
time and labor costs are evaluated as shown in table 3.
Within this example two different global execution
In this scenario, the following wages per hour are
strategies are specified as soft constraints manually.
assumed: foreman 28 EURO/h, worker 21 EURO/h
The first global strategy combines certain drywalls
and laborer 11 EURO/h. The shortest working time is
execution groups ordered by priority. Drywalls within
given for the assignment of one foreman, two workers
a group are assembled randomly. The second strategy
and two laborers by using the drywall group strategy
sorts all drywalls relative to their wall length in
(i.e. experiment 3).
descending order.
Simulation and Evaluation
Table 3 Working time and labor costs results of
Three employee variations were simulated for each
drywall execution experiments
global execution strategy. The different employee
variations of foremen, workers and laborers are shown Results → Min. time [h] Costs [€]
in table 2.
Drywall group strategy
Table 2 Employee variations
Experiment 1
152 9120
Employees → Number Number Number (Variation 1)
of of of Experiment 2
Variation ↓ foremen workers laborers 120 9720
(Variation 2)
Variation 1 1 1 1 Experiment 3
112 10304
Variation 2 1 2 1 (Variation 3)
Variation 3 1 2 2 Drywall length strategy
Experiment 4
Different simulation runs were performed for each 144 8640
(Variation 1)
employee variation and global execution strategy.
Within these simulation runs the identical drywall Experiment 5
136 11016
work steps, hard constraints and material sheets are (Variation 2)
used. For each simulation run the work step schedule Experiment 6
and the workload of employees are recorded to 128 11776
(Variation 3)
evaluate them afterwards. Furthermore, every
simulation run was animated using the discrete-event To find a suitable group combination, not only
simulation framework and can be used for visual working times and costs are important but also
control of the execution progress. Some parts of the workloads have to be regarded. The workloads of the
simulation model and a snapshot of a simulation step defined employees are shown in table 4.
are shown in figure 7.
Table 4: Workload results of all experiments

Workloads
Foreman Workers Laborers
[%] →

Drywall group strategy

Experiment 1 73 95 24

Experiment 2 46 94 78 24

Experiment 3 52 98 82 26 12

Drywall length strategy

Experiment 4 79 90 34

Experiment 5 42 73 16 16

Experiment 6 44 78 16 17 4
Figure 7 Snapshot of a simulation step
The results of this simple drywall example illustrate
that working times, labor costs and utilization vary
according to the number of appointed employees and practicable execution order within the fire zone rooms
used strategy. Now, considering project-specific time is guaranteed (figure 8).
and costs restrictions the planner can choose one of
these employee and strategy combinations or define
new constraints and resources to find other solutions
with more steady work flows or more efficient
employee utilizations. Currently, this constraint-based
simulation approach cannot guarantee to find optimum
solutions. However, a multitude of practical schedules
can be generated to find an optimized solution
manually.
6.2 Example 2: Ferry outfitting
Currently Flensburger Shipyard is building so-called Figure 8 Visualization of the fire zone
RoPax ferries for cars, trucks and passengers. More
than ten years ago the production of deckhouses was Global constraints between fire zones consider the
outsourced. Therefore, experience to control outfitting dependencies between neighboring compartments. For
processes in passenger or crew areas are marginal. example, ceiling outfitting in a fire zone cannot start
Thus, there is an increasing requirement for a tool to before floor straightening work in the zone above is
evaluate these complex production flows. not completed. Another example, floor pavement only
can be floated after ships’ launching due to the
Simulation has been established at Flensburger as a inclination angle of the slipway. These global
practicable tool for production development and constraints are defined generically as dependencies
production planning in steel production, already. between work steps of assembly tasks in different fire
Several simulation-based applications support the zones.
daily work of planners and foremen. However,
outfitting processes, particularly in passenger areas, By using simulation models different assembling
have not been analyzed sufficiently yet. A prototypic strategies can be analyzed regarding the required
simulation model of a passenger cabin area, developed execution time, workers efficiency and space
some years ago, turned out to be not sustainable to utilization. In future, constraint-based outfitting
model outfitting tasks. Though, it provided interesting simulation is planned to be used for production
cognitions for the further research. planning of further passenger ferry projects.

Flensburger Shipyard uses the adjustable components 7 Conclusion

developing in the SIMoFIT cooperation to build-up
simulation models of ferry outfitting. One fire zone of Outfitting execution processes are very complex. A
a ferry that is erected currently on the slipway is multitude of requirements such as technological
specified as a prototype simulation model. Fire zones dependencies and safety criteria have to be considered
are compartments of a ship in a passenger deck The planners must exhibit high competence in order to
separated by fire protection doors and walls. A fire take all different effects on execution into account.
zone contains several rooms and is also planning unit Definitely, it is very time-consuming to find
for outfitting activities. In the presented example the practicable schedules manually where equipment and
fire zone contains a snack bar, lounge areas and employees are sufficiently utilized. This paper
restrooms. Thus, many different aspects of outfitting introduces simulation as an appropriate instrument to
and furnishing have to be considered. Logistic support the planning process. Constraint-based
processes were not taken into account because they simulation models are highlighted due to the fact that
were analyzed in an earlier simulation project, they allow modeling dynamic structures. Therefore,
intensively. requirements can be easily defined or adapted by
adding or removing constraints.
The product data for the simulation model was
generated from drawings and part lists. In future Currently, in order to model outfitting processes, the
research activities will focus on suitable tools in order following constraints have to be considered:
to provide the required data and the as-is state in time technological dependencies, capacity, availability,
to reduce the data mining work to a minimum. safety criteria, productivity and strategies.
In the first step three different types of outfitting Within the cooperation SIMoFIT a constraint-based
processes with different requirements are analyzed: simulation framework is developing to assist the
drywall, ground floor and ceiling construction. outfitting planning. Therefore, a modular discrete-
Processes and outfitting sequences were derived from event simulation toolkit is used. In result, different
manual scheduled activities. Sequences were practicable schedules for execution can be generated.
transformed into technological dependency constraints The simulated schedules can be evaluated regarding
associated to the outfitting work steps. Thus, a worker’s efficiency, utilization of space as well as
process costs, afterwards. Lab tested outfitting [10] P. van Hentenryck, V. Saraswat, et al. Strategic
applications of both partners are presented. Directions in Constraint Programming. ACM
Computing in constraint programming, 28(4), pp.
Future works concentrate on defining further outfitting
701-726, 1996.
constraints. Appropriate methods especially to
describe soft constraints have to be researched. [11] J. C. Beck and M. S. Fox. A Generic Framework
Definitions of the highlighted constraints safety for Constraint-Directed Search and Scheduling,
criteria, productivity and strategies are still in AI Magazine, winter 1998, pp. 103-132, 1998.
progress. Some of the considered soft constraints like
[12] J. Sauer. A Multi-Site Scheduling System. Proc.
strategies can be violated infinite. Others like
Artificial Intelligence and Manufacturing –
productivity constraints describe previous sections to
Research Planning Workshop – State of the Art
what it is satisfied [14]. Currently, the ability of
and State of the Practice, Albuquerque, AAAI-
several methods to describe various constraints types
Press, pp. 161-168, 1998.
like fuzzy, weighted or semiring-based constraints is
investigating. [13] J. C. Beck and M. S. Fox. Supply Chain
Coordination via Mediated Constraint
Further simulation components are projected, for
Relaxation. Proceedings of the First Canadian
example, an extended storage area control component
Workshop on Distributed Artificial Intelligence,
to broaden the models’ realistic behavior. The
Banff, Alberta, 1994.
projected component estimates production sites
regarding outfitting execution in order to find suitable [14] H. Fargier, J. Lang, R. Martin-Clouaire and T.
storage areas. Adequate storage areas are important to Schiex. A constraint satisfaction framework for
guarantee an undisturbed execution flow. decision under uncertainty. Proceedings of the
11th International Conference on Uncertainty in
8 References Artificial Intelligence, Montreal, Canada, 1995.
[1] IBW – Deutsches Institut für Wirtschafts- [15] B. Akinci, M. Fischer, R. Levitt and R. Carlson.
forschung. Strukturdaten zur Produktion und Formalization and Automation of Time-Space
Beschäftigung im Baugewerbe – Berechnungen Conflicts Analysis. Journal of Computing in
für das Jahr 2005, http://www.bmvbs.de, 2006. Civil Engineering, 16(2), pp. 124-134, 2002.
[2] http://www.simofit.com. [16] Z. Mallasi. Identification and Visualization of
Construction Activities’ Workspace Conflicts
[3] http://www.simcomar.com. Utilizing 4D CAD/VR Tools. 1st ASCAAD
[4] D. Steinhauer. Simulation in Shipbuilding – International Conference e-Design in
Supporting Shipyard Planning, Production Architecture, Dhahran, Saudi Arabia, pp. 235-
Planning and Product Development. 12. ASIM 253, 2004.
Fachtagung Simulation in Produktion und [17] E. Sriprasert and N. Dawood. Genetic
Logistik - Sigrid Wenzel (Hrsg.), Society for Algorithms for Multi-Constraint Scheduling: An
Modeling and Simulation International and SCS Application for the Construction Industry.
Publishing House e.V., San Diego/Erlangen, pp. International Council for Research and
1-14, 2006. Innovation in Building and Construction – CIB
[5] U. Beißert, M. König and H.-J. Bargstädt. w78 conference, Construction Informatics
Constraint-Based Simulation of Outfitting Digital Library (available at
Processes in Building Engineering. CIB 24th http://itc.scix.net/paper w78-2003-341.content),
W78 Conference, Maribor, Slovenia, 2007. 2003.

[6] F. Rossi, P. van Beek and T. Walsh. Handbook [18] E. Sriprasert and N. Dawood. Requirements
of Constraint Programming, Foundations of identification for 4D constraint-based
Artificial Intelligence, Elsevier, Amsterdam, construction planning and control systems.
2006. International Council for Research and
Innovation in Building and Construction – CIB
[7] V. Kumar. Algorithms for Constraint Satifcation w78 conference, Construction Informatics
Problems: A Survey. AI Magazine, spring 1992, Digital Library (available at http://itc.scix.net),
pp. 32-43, 1992. 2002.
[8] E. C. Freuder and R. Wallace. Partial constraint [19] B. Akinci, M. Fischer and J. Kunz. Automated
satisfaction. Artificial Intelligence, 58:21–70, Generation of Work Spaces Required by
1992. Construction Activities, Journal of Construction
[9] M. S. Fox and S. F. Smith. ISIS – a knowledge- Engineering and Management, 128(4), pp. 306-
based system for factory scheduling. Expert 315, 2002.
Systems Journal, 1(1), pp. 25-49, 1984. [20] http://www.ugs.com/products/Tecnomatix.
[21] S. Wolfram. A new kind of science. Wolfram
Media, Section 5: Two Dimensions and Beyond,
pp. 169-221, 2002.
[22] R. Dechter and J. Pearl. Generalized best-first
search strategies and the optimality of A*".
Journal of the ACM, 32 (3), pp. 505 – 536, 1985.
[23] T. H. Cormen, C. E. Leiserson, R. L. Rivest, and
C. Stein. Introduction to Algorithms, MIT Press
and McGraw-Hill, Section 24.3: Dijkstra's
algorithm, pp.595–601, 2001.
[24] IZB - Institut für Zeitwirtschaft und Betriebs-
beratung Bau. Trockenbau-Planzeiten Wände:
methodisch ermittelte Zeitwerte für die Planung,
Ausführung und Entlohnung, Neu-Isenburg -
Zeittechnik-Verlag, Karlsruhe, 2002.