|
Simulating Production Processes:
|
|
Shipbuilding Cost and Scheduling Tradeoffs
|
|
|
|
The US Navy (USN) shipbuilding program consumes a major fraction of the USN annual budget. Decision Dynamics, Inc. (DDI) developed an
advanced shipbuilding simulation model, called WorkFlow, to
help managers speed ship production and lower construction costs. PMS-400, the AEGIS Program Office,
contracted with DDI to validate WorkFlow’s behavior and to demonstrate model
utility. The test utilized data
supplied by Bath Iron Works (BIW) for panel production for DDG-51 Class
ships. Results validated model
operation by reproducing BIW manpower loading and production schedules. Results also demonstrated two unique aspects
of model utility. First, WorkFlow
simulations identified logical inconsistencies in the data that, when resolved,
improved the information used to support production planning. Second, WorkFlow simulations showed how the
use of overtime and/or overmanning on selected production tasks could shorten
schedules. Simulations also explored
how level labor loading could improve productivity on multiple panel
lines. Overall, WorkFlow proved itself
an extremely valuable tool to assist both the USN and shipyards in the complex
process of managing shipbuilding schedules and controlling ship production
costs.
|
|
|
|
This report is
divided into three sections. The first
section recreates the BIW baseline for producing three major panels in the ship
assembly process. Simulated schedules
match planned schedules and simulated labor (both shipfitters and welders)
match planned labor. Anomalies in
planned schedule and labor are identified to create a consistent baseline
model. The second section builds on the
baseline to demonstrate how the introduction of overtime and overmanning can
shorten schedules but increase costs. A
third section examines level labor loading alternatives to show how WorkFlow
could help planners improve productivity by maintaining a more constant work
force on each task.
|
|
|
|
Validating a Baseline
|
|
|
|
A Baseline Model was developed to illustrate that WorkFlow
can accurately simulate BIW’s panel fabrication process. For this demonstration, three major panel
units were modeled. The data used to
populate the baseline model included BIW information on labor resources
(manning), planned schedule and the basic work breakdown structure (WBS) for
the panel fabrication process.
|
|
|
|
The WBS represents the hierarchy of work tasks, from large
tasks (assemblies or products) down to smaller tasks that make up the larger
project. At the most detailed level in
the WBS, tasks were defined according to the type of work being performed, type
of labor and amount assigned to the task, and dependent relationships among
tasks. This level of detail allowed DDI
to validate the Baseline Model by comparing simulation results to BIW’s planned
manhour expenditures and schedules dates for individual panels as well as the
overall total results.
|
|
|
|
The initial simulation results indicated an inconsistency
between BIW’s planned task durations and BIW’s manning profile. The workday schedules (hours per shift) were
adjusted in the Baseline Model input data to resolve this anomaly.
|
|
|
|
Figure 1 plots the total manhours expended to complete the
three units within the 60-day schedule and the number of personnel
assigned. The simulation results
corresponded with the BIW’s projected calendar dates, manhour expenditures and
manning profile, thereby validating the accuracy of the Baseline Model to
reproduce an accurate representation of the panel fabrication process.
|
|
|
|
|
|
|
Figure 1: Baseline Total Manhours and Personnel
|
|
|
|
Shortened Schedules
|
|
|
|
To test alternative
“what-if?” scenarios, DDI first altered the
Baseline Model to plot profiles that more consistently reflect realistic
manhour expenditures; the workday schedule was defined as two 8-hour shifts,
five days a week. This data
modification caused a schedule extension in the Baseline Model from 60 days to
63 days.
|
|
|
|
To analyze the tradeoffs between alternative schedules, work layouts, and resource
allocation, an Overtime scenario was created that attempted to shorten the
Baseline Model panel fabrication process schedule from 63 days to 51 days. WorkFlow automatically generates the
schedule for the Baseline Model by calculating nominal duration required for
each task. During simulation, WorkFlow
multiplies the amount of desired labor times the number of work hours per day
and divides the product into the backlog of labor hours to get the number of
days required for each task. The
schedule acceleration was achieved by manually entering a completion date for
the project that decreased the nominal duration by 12 days.
|
|
|
|
WorkFlow compares the actual
start and completion dates for each task against the nominal schedule. If the days required to complete the task
exceed the days remaining in the schedule, schedule pressure begins to
build. Decreasing the duration in the
first two tasks of the Baseline Model triggered a response from WorkFlow that
the tasks were behind schedule causing schedule pressure to build. WorkFlow has several management functions
that allow the user to specify how to respond to various conditions and
situations including schedule pressure.
|
|
|
|
In the Overtime 'what-if?'
scenario, two of WorkFlow’s management functions were applied to allow for overtime
work to occur and to include a minor influence of fatigue resulting from
overtime work. As displayed in Figure
2, the increase in schedule pressure caused overtime work for key labor trades
over the course of the project. This
allowed the work to be finished 12 days earlier, as expected when the forced
deadline was set for 51 days instead of the Baseline Model of 63 days. However, due to fatigue from overtime work,
slight productivity losses occurred increasing total manhours in the overtime
scenario by 50 hours.
|
|
|
|
|
|
|
Figure 2: Schedule Comparison - Overtime
|
|
|
|
When comparing the
Baseline Model and Overtime scenario results, one of the factors to consider is
the difference between the cost of standard manhours (MHs) and overtime
MHs. In Table 1 assume that a standard
MH equals $20 and an overtime MH is equivalent to $30.
|
|
|
|
|
Total MHs
|
Overtime MHs
|
Overtime Cost
|
Total MH Cost
|
|
Baseline
|
2,773
|
|
|
$55,460
|
|
Overtime
|
2,823
|
55
|
$1,650
|
$57,110
|
|
|
|
|
Table 1: Schedule Comparison – Manhour Cost
|
|
|
|
It is obvious from
the table that the Overtime scenario has a higher total MH cost. The tradeoff now concerns comparing the
savings in scheduled workdays against increased MH costs.
|
|
|
|
A second “what-if?”
scenario was simulated to compare the Overtime scenario to an Overmanning
scenario. This scenario applies the
overmanning management function that increases the level of labor working on a
task (with possible productivity losses due to overmanning) in response to the
experienced schedule pressure. The plot
in Figure 3 indicates the Overmanning scenario also achieved a 12-day
acceleration, however, the total number of manhours significantly increased due
to much higher productivity losses.
|
|
|
|
|
|
|
Figure 3: Schedule Comparison - Overmanning
|
|
|
|
When building the
Baseline Model, a minimum, desired and maximum labor level was designated by
BIW for the specific labor on each task. If the current level of labor is greater than or less than the desired
labor level, the productivity of the labor working on the task may increase or
decrease relative to 'normal' productivity depending on the assumptions made in
the management functions. In the
Overmanning scenario, this function assumed that when the maximum labor level
is achieved, there are too many people working on the task to be fully productive.
|
|
|
|
Figure 4 plots the
comparison of total personnel assigned (labor) for both the Baseline Model and
Overmanning scenario. As Figure 4
depicts, during simulation WorkFlow assigned additional persons to the tasks
experiencing schedule pressure in order to accelerate the Overmanning scenario
schedule. The maximum number of persons
assigned to any task was defined as 12 due to physical limitations.
|
|
|
|
The model behavior
in this 'what-if?' comparison reflects what one would expect in a real-world situation. As schedule pressure increases during the
simulation, the number of persons assigned increases to accelerate the work
accomplished. In the first eight days
of the simulation, one to two additional persons are assigned in response to
early signs of schedule pressure. As
the schedule pressure becomes more significant in days eight through 16, the
model responds by doubling the number of additional persons assigned (up to
four people).
|
|
|
|
|
|
|
Figure 4: Manpower Profile Comparison - Overmanning
|
|
|
|
Again, assuming a
standard MH equals $20, a comparison can now be made between the cost of the
Baseline, Overtime and Overmanning scenarios. As Table 2 clearly states, overmanning is the most expensive option due
to the total number of MHs.
|
|
|
|
|
Total MHs
|
Cost of MHs
|
|
Baseline
|
2,773
|
$55,460
|
|
Overtime
|
2,823
|
$57,110
|
|
Overmanning
|
3,306
|
$66,120
|
|
|
|
|
Table 2: Schedule Comparison – Manhour Cost
|
|
|
|
In the Baseline
Model it was assumed that productivity losses due to overmanning tasks were
more significant than losses incurred from overtime work. When comparing the simulation output
results, it is apparent that the Overmanning scenario required additional MHs
to complete the same work backlog because of greater productivity losses. WorkFlow helps program managers identify and
test sensitive elements in their system. A new scenario that contains a different assumption about productivity
losses can be created and simulated in seconds. By offering a control structure to test the consequences
of alternative 'what-if?' assumptions and scenarios, WorkFlow can help program
managers avoid ineffective actions, reduce program risk, lower support costs,
and develop an affordable strategy to meet program requirements.
|
|
|
|
The tradeoff
between MH cost and schedule can be clearly seen in Figure 5, which plots a
comparison of the Overtime, Overmanning and a third scenario, Overtime and
Overmanning. In the Overtime and
Overmanning scenario, both overtime work and overmanning were applied to a
schedule that was accelerated by 20 days.
|
|
|
|
|
|
|
Figure 5: Schedule Comparison – Overmanning and Overtime
|
|
|
|
Again, the tradeoff concerns the additional MH cost
associated with shortening the schedule. Depending on the situation a shorter schedule may be worth the
additional cost.
|
|
|
|
Level Labor Loading
|
|
|
|
Figure 6 plots a comparison between the Baseline Model and a
scenario that more evenly distributes labor usage. The Baseline Model assigns labor according to the type of task
and work backlog causing an irregular distribution in the manpower profile with
several peaks and dips over the course of the project. The Level Labor scenario assigned the same
number of persons to each task. Although there is clearly a significant improvement in the Level Labor
scenario manpower distribution, the schedule is extended eight days because
less labor is assigned to the tasks with larger backlogs.
|
|
|
|
|
|
|
Figure 6: Manpower Distribution Comparison – Level Labor
|
|
|
|
Figures 7 and 8 show variations of other production
scenarios involving multiple items produced on two parallel lines. In Figure 7, a second Baseline Model
production line was started on project day 16 to use the manpower resources
available during the dips in the Baseline Model manpower profile. No additional resources were added to the
Baseline Model labor pool of shipfitters and welders. Although the manpower profiles retain the irregular distribution
of the Baseline Model, there is enough labor available during the dips in Line
One to complete both Line One and Line Two in about 78 days.
|
|
|
|
|
|
|
Figure 7: Manpower Distribution Comparison – Multiple Production Lines
|
|
|
|
The ability to quickly incorporate changes in the scenario
and compare results is easily demonstrated in the next scenario. Figure 8 plots the results of operating the
two production lines using the level labor manpower profile.
|
|
|
|
|
|
|
Figure 8: Manpower Distribution Comparison – Multiple Level Labor Production Lines
|
|
|
|
|
|
|
Figure 9: Total Manhour Comparison – Multiple Production Lines
|
|
|
|
It is obvious from the output in Figures 8 and 9 that the
Baseline Model option finishes earlier than the Level Labor scenario with no
significant difference in total manhours. The program manager can now compare the advantages of a level manpower
profile against the savings in the project schedule.
|
|
|
|
Conclusion
|
|
|
|
WorkFlow can provide unexpected insights into program
management alternatives and support management decision-making. Options become clearer when all assumptions
are considered and quantified and when all parties to the decision agree on the
same assumptions. When differences arise,
alternative values can be tested for their impact on model behavior and
outcome. If the results remain
unchanged then differences over assumptions become unimportant; if results
vary, then the differences can be resolved by further analysis and data. This ability of the model to help focus
discussion on key issues while demonstrating the marginality of other issues
serves to quickly build consensus for effective decision-making.
|
|
|
|
WorkFlow provides program
managers and design engineers with the ability to successfully develop a
strategic plan by integrating and managing the multitude of functions that are
key to the shipbuilding production process. The results achieved and the output available from simulating with the
model include:
|
|
|
- schedules for all tasks and assemblies;
- overall schedule;
- labor manning (by shift and by trade);
- labor hours for all tasks and assemblies; and
- total labor hours.
|
|
|
|
WorkFlow
offers an innovative simulation tool for quantifying manhour cost and schedule
tradeoffs, tracking changes in productivity due to internal and external
conditions, and tracing the impact of design changes and delays on ship cost
and schedule. Program managers and
design engineers can use the model to define and test alternative “what-if?”
scenarios to search for ways to shorten production times and increase
manufacturing productivity.
|
|
|
|
|
|
|