Page | 2
Table of Contents
Executive Summery 4
Goals 5
Expected Output 5
Process Flow Modeling 6
Part Numbers 8
Kitting 8
Animation 8
Logical Diagram 8
Empty System Timer 10
Part Request & Release and Blackbox of Factory 10
Plane Creation (Release Parts for Production & Kitting) 10
Batching and Loading 10
Transportation Modules 10
Assumptions 11
Arrival Rates 11
Number of Planes 12
Number of Parts 12
Number of Kits 13
Number of Modules 13
Matrix 13
Second Issues 14
Finished Goods Inventory 14
Part Processing/Flow times 15
Transportation Module Flow 16
Distance Factors 17
Speed Factors 17
Loading/Unloading Delay Factors 17
Costing 17
Regular and Overtime Costs 18
Second Issue Costs 18
Finished Goods Costs 18
Raw Material Safety Stock 19
Control Model 19
Control Model Verification 20
System Inputs and Initial Results 21
Variable Models 22
System Outputs 22
Data Results 22
Parts Completion Percentage 23
Finished Goods Inventory Costs 24
Conclusions 24
Appendices 25
Executive Summery
At the beginning of the spring 2009 school semester our group from Scott Metlen’s business simulations class was assigned a project with Boeing Skin and Spar in Frederickson, Washington. The nature of the project involved utilizing the simulations programming skills we would acquire throughout the semester on a real life, real time, and real money situation.
The group was comprised of five seniors from the University of Idaho Production/Operations Management program: Dustin Smith, Jeremy Wemple, John Noble, Nic Pentzer, and Shane Wemhoff. We traveled to Boeing at the beginning of the class at met with Boeing’s Industrial Engineering group to discuss project scope and expectations. Our assigned group sponsors were Rick Jones and Don McCart.
We utilized Rockwell’s Arena simulations software to complete the project modeling Boeing’s skin and spar production facility in order to determine an appropriate level of finished goods inventory. Each student had access to personal laptops for use, while the main model was ran through a personal high capacity desktop computer owned by one of the students.
We reported our discoveries on May 1, 2009 to Boeing. This report is a follow up to the presentation and its purpose is to relay all the research and conclusions we came to as a team, as a deliverable document for Professor Metlen Ph.D and Boeing’s Industrial Engineering group.
We have accumulated a lot of knowledge throughout the semester on process management and the value of simulation as a powerful tool to assess and improve any process in any organization. We would like to thank Boeing for the opportunity to work with them, especially Rick and Don. We are confident the following information will aid the process improvement situation at Boeing skin and spar.
Goals
The purpose for modeling the production, kitting, and transportation system for the Frederickson plant is to determine adequate levels of safety stock either in raw material or finished goods inventory. The objective is to minimize factory disruption caused by independent demand defined here as additional demand occurring from customer requests or parts rejected by quality control procedures. The objective arises from five observations taken directly from the factory:
1. Direct module loading philosophy minimizes finished goods inventory. This observation occurs as part of a broader company initiative to implement lean manufacturing principles and subsequently eliminate most finished goods inventory. The production system is intended to produce parts with a mindset analogous to a just-in-time philosophy. In theory, when parts need to be transported they should be finishing the manufacturing process with the timing set to almost eliminate inventory. This process is effective at reducing inventory but it leaves the system highly vulnerable to disruptions caused by rejected parts or additional unforeseen demand.
2. Lack of floor space impacts the ability to store finished goods. After lean practices were implemented in the Frederickson factory, the additional floor space that was created from removing finished goods inventory was used to house additional production equipment. This creates a dynamic that limits the ability of the factory to utilize a blanket solution calling for the increase in the general finished goods inventory for all parts. This problem creates a need to generate optimal safety stock for a select number of parts that are most affected by second issue demand.
3. There is a lack of visibility for independent demand. The unpredictable nature of the additional demand creates added pressure on a highly nervous system. While average part scrap rates can be computed and integrated into the production schedule, outlier demand can still have a profound impact on the system. This observation further highlights the need for an optimization of finished goods safety stock.
4. The factory reacts to quality rejections rather than anticipating them. Currently, quality rejections have a material impact on the production system. Instead of building a buffer against average scrap rates in anticipation of rejected parts, a scrapped part will require priority rework causing disruption of other scheduled parts and decrease in normal flow resulting in large amounts of overtime.
5. The factory buffers safety stock items by purchasing additional raw material. The raw material kept on hand to satisfy second issue demand does not guard against disruption from parts that need to reenter the production process. To optimize the system, the model will explore efficient levels of safety stock of finished goods.
Expected Output
1. Transport tool utilization
2. Cost comparison between the base model vs. experimental models.
3. Excess capacity created by eliminating scrap.
4. Throughput of parts, planes, kits and batches
The project will be organized into two reporting periods. All preliminary work will be completed after the assignment has been delivered by Boeing. The first reporting period was scheduled for mid-April during which all simulation work was critiqued by the project sponsors, Rick Jones and Don McCart. The final reporting period was scheduled for May 1 at Boeing’s Skin and Spar Frederickson facility.
Process Flow Modeling
We modeled the flow of spar and stringer parts through the entire production system from raw materials to finished goods. From there we modeled the shipping and unloading phase of the kitted parts in order to integrate the timing data back into the model. Greater detail will be placed on the finished goods required to reduce overtime and ensure on time shipment of full kits. We first modeled the existing production system which will be referred to as the base model. Afterwards we created a series of experimental models representing changes in safety stock, overtime usage, and theoretical changes to scrap rate. This two model approach allowed us to see the impact of potential changes
The model utilized a variety of inputs.
1. Raw Materials. The average stock on hand used to cover second issue demand.
2. Fabrication inputs.
- Part numbers
- Process times and rates
- Resource utilization
- Queue size
- Failure rates
- Available up and down times
- Part arrival rates
3. Kitting inputs
- Frequency statistics
- Production rates
- Resource utilization
4. Costs
- Raw material costs
- Raw material inventory cots
- Finished goods inventory costs
- Production costs of people
- Scrap costs
Our model will begin with the raw material and milling step in the production system. Rather than simulating the entire production facility in order to provide part arrivals for the purpose of kitting, we will simulate the production system using a limited number of processes by using average flow times and scrap rates associated with each part number. In order to simulate the milling and fabrication times for “black-boxing,” we will use flow times for each part based on its duration time distribution Individual raw material stock may be simulated if Boeing wants to include that analysis in the project scope for purposes of reducing or optimizing raw material safety stock.
Figure 1 represents a simplified map of the manufacturing process at the Frederickson plant. A variety of wing skins and spars move through the system based on higher level demand. For the purposes of this project we will focus on the final kitting process outlined in red. We will reduce the manufacturing process flow to a “black box” model that will provide an accurate output of parts without modeling the individual processes.
The second phase of the current process consists of the transportation of the completed parts and kits to locations in Renton and Everett, WA. The availability transportation tools can affect the timing of delivery and will be included in the modeling process.
Part Numbers
Part numbers will be inputted into the system from specific parts derived from the multiple parts lists furnished by Boeing. Demand for specific parts will be generated by historical demand for entire planes. For instances of second issue demand, or additional demand caused by scrapped parts, the part will be rereleased into the system and will take immediate precedence over existing production that covers scheduled demand.
Kitting
After the individual parts are produced they are collected into various groups called kits. The kits are arranged in a way that allows them to be shipped to the final assembly locations at either Renton or Everett and be easily assembled by the on-site wing facilities. Based on the type of airplane more or less kits may be required to assemble a completed wing. Smaller parts are loaded into final storage racks by hand and larger parts are handled by a series of overhead cranes and hoists.
Animation
We will use animation in the model in order to assist in increasing the ability to visualize the modeled processes. Our intention is to use a basic layout of the factory as a background for the model to show how the processes beginning with milling and continuing to kitting and transportation interact with the entire system. We believe the ability to picture the system as a whole will increase understanding and further aid in optimizing the level of safety stock required for efficient operation. Depending on time constraints additional animation may be added to further add to the understanding of parties that may not be fully versed in Arena’s functionality. While animation is purely cosmetic and does not affect the inner workings of the model, the added benefit of ease of understanding it could create for third parties means that it may be a valuable addition to the final product.
Logical Diagram
In this diagram the model is brought to its context level showing the purpose of each major section of the model and its key operations.
Empty System Timer
In all models there needs to be a cutoff point that the model runs too. We wanted to show the completed transportation modules, so the model runs until the transportation has completed for the high level demand.
Part Request & Release and Blackbox of Factory
In order for a part to obtain a part number, Arena assigns an Index that is correlated to the physical part number used to refer to the part. Once Arena assigns the part an Index that index is used to refer to the part in the remained of the model using a data set. In order to evoke a process on a part it needs to be created in Arena, which is the job of the Assign function. To Process the Part in the factory a calculated delay of manufacturing time is applied to the parts duration in the “Blackbox” Process.
Plane Order Creation
To simulate the demand of a part included in a plane the model requires that “Plane Order” be addressed. To Simulate production of plane types the high level demand of the planes was determined by the historical averages of planes per model produced.
Plane Creation (Release Parts for Production & Kitting)
For a kit of parts to be created it needs a certain criteria or parts that make up the kit per plane per version. The process was devised to be made up of multiple modules that include all of the plane versions, because all the planes need different kits. One problem that this method addresses is the inherent modeling barrier of duplicate parts going to different kits.
Batching and Loading
In this segment of the model, the kits are held to get ready to go in to the transportation modules. Each batch of kits require a specific module to be loaded in. the loading of kits into modules occurs in this step through the Batching Process.
Transportation Modules
After the parts have been collected into kits they are loaded onto a transportation module which is attached to a semi-trailer for shipping. The demand for transportation modules is based on the demand schedule based on assembly of planes in Everett or Renton and their associated kits. In addition to the kit attached to the module, additional parts may be attached as necessary to cover second issue demand. In a fully functioning system, transportation modules will be routed from their destinations and return to the factory in time to load another set of wing kits.
Assumptions
One of the first things needed to create a working model is a time period for which the model will run. We chose to model an entire year’s worth of plane demand and production for the Frederickson facility. Thus, we needed to determine the total amount of hours the Frederickson facility is running in a year. The data was given to us for normal non-overtime production days in a month and the hours for each day. Some overtime data was given to us, but after reviewing the total overtime costs in a year for Frederickson, we used some of the data given and the overtime costs to work backwards and establish the number of overtime hours used across the year. For model running purposes, it is a lot easier to take total productions hours in a year and divide that across 365 days equally, rather than setting up more complex schedules to control operating time. This was the main driver for our purpose of setting up the data this way. Data and numbers used for determining total operating hours in a year for the current/control model are listed below. Those for the variable models have overtime hours at 50%, 25%, and 0% of that of the current/control. As the overtime reduces, the total hours in a year reduce as well, so when divided by 365 the hours per day becomes less as well. The total hours in a year and the hours per day for the variable models are displayed in Exhibit 1 and Exhibit 2.