2011
University of Groningen Thijs Panneman
S1669680
Swedish Match Philippines Inc.
[REDUCING ORDER LEAD
TIME AND INVENTORIES AT SWEDISH MATCH]
Reducing the order lead time and reducing Work in Process with the implementation of transfer batching, a change of workstation design and changing the production planning and control system without radical investments necessary.
Author:
Thijs Panneman S1669680
Company:
Swedish Match Philippines Inc.
Laguna Technopark – SEZ 104 Technology Avenue Binan Laguna, 4024 Philippines
Coordinator SMPI Edgar V. Zulaybar
University Coordinators Dr. L. Zhang
Dr. X. Zhu
University of Groningen, April 4th 2011
Preface
In this report I show that improving production performance is possible without the need for major investments in equipment. Changing batching rules and workstation design do not necessarily involve large technical changes but have a relatively large influence on performance measures like inventory levels and order lead times which can be applied to many more organizations.
In this report I discuss a practical case at Swedish Match Philippines inc. where lead times and inventory levels were considered high and efforts to reduce these two measures were time consuming. The availability of internal reports about capacity consumption for all products per machine made it possible to gain insight in the order lead times and inventory built up and design a solution of three steps to reduce both the lead times and inventory levels.
For solution development, different literature is used ranging from scientific papers, to academic college books, internal data reports from Swedish Match and market research results published on the internet.
I would like to thank Swedish Match for giving me the opportunity to visit the Philippine plant for a period of 2 months, especially plant managers Pietjan Verhoog from Swedish Match Assen and Victor Bocalin from Manila. From the Philippine plant I would like to thank my supervisor Edgar Zulaybar for arranging meetings with many employees and arranging housing while I was at location. A very special thanks goes out to my colleague and house mate Anna Vizconde, who made my trip not only worthwhile business wise, but also on personal level.
Finally, I would like to thank my supervisors from the University of Groningen, Linda Zhang and Stuart Zhu, for making it possible for me to start the project earlier than normal and providing me with feedback to improve my research.
Abstract
Swedish match is a company in the tobacco industry which produces cigars, tobacco, matches and lighters. The market of lighters is growing, therefore Swedish Match wants to improve production throughput while minimizing costs. Lead times are considered long and inventories are considered high and vary in both size and location.
The objective of this project is to deliver a plan for production improvement to reduce the lead times and inventory levels. This is done by diagnosing the throughput of the production process at each process step, and consider adjustments in planning and routing rules to optimize the flow of products through the process.
The proposed optimization plan consists of three steps. First, workstation design of two production steps is altered. To generate a balanced flow, the number of machines per workstation should be changed from one to two at two locations. This will prevent inventory built up between the different operations in the future.
Second, the planning and control system should be changed to a kanban system in the second half of the production line. A Push‐pull interface should be implemented with some inventory to facilitate quick response from the downstream signals. By implementing kanban in large part of the
production steps, the flow of materials can be managed by pull production which automatically reduces the variance of inventory.
Thirdly, long lead times are reduced by implementing transfer batching at two locations in the process. By allowing 12 bins of lighters to move to the next operation where a production batch is 24 bins, waiting times are reduced.
The expected results from implementing the proposed 3 steps are as follows. Inventory levels will reduce by 24%, while lead times for three different product families will be reduced by 37%, 12% and 30%.
Content
Preface ... 3
Abstract ... 5
List of figures ... 8
List of abbreviation ... 9
1. Introduction ... 11
1.1 Company background ... 11
1.2 problem introduction ... 11
1.3 Project objective ... 11
2. Research design ... 12
2.1 Problem statement ... 12
2.2 Conceptual model ... 12
2.3 Research questions... 13
2.4 Methodology ... 13
3. Existing production process ... 14
3.1 Production layout ... 15
3.2 Production planning ... 15
4 Problem analysis ... 15
4.1 Variance of inventory ... 16
4.2 Long lead times ... 17
4.3 Expected capacity shortage in the future ... 20
5. Solution design ... 21
5.1 Workstation design at gas filling and printing. ... 23
5.2 Designing the production planning and control system ... 26
5.3 implementing transfer‐batching at two locations. ... 28
6. Quantification of benefits of the new design ... 33
6.1 Order lead time ... 34
6.2 WIP level and cost of WIP ... 35
7. Conclusions ... 36
8. Discussion ... 38
8.1 Dedicated lines ... 38
8.2 Changing layout of the plant ... 38
8.3 Processing time of the assembly operation for ED4 ... 39
References ... 40
Appendix 1: results of the SWOT analysis ... 42
Appendix 2: Capacity utilization estimations ... 43
Appendix 3: TOC vs. JIT approach ... 45
Appendix 4: Lead time simulations ... 46
Appendix 5: Simulation inventory behavior ... 49
Appendix 6: Calculations of WIP levels ... 50
List of figures
Figure 1: The conceptual model ... 12Figure 2: Current situation SMPI ... 14
Figure 3: Detailed model of production performance and indictors of buffering ... 16
Figure 4: Inventory levels and variance for lighter production batches ... 17
Figure 5: Processing times for the first batch for dedicated machines ... 18
Figure 6: Daily capacity in thousands operations 2011 ... 19
Figure 7: Capacity utilization for estimated 300mil in 2015 ... 21
Figure 8: Solution method in three steps ... 21
Figure 9: New process design production SMPI ... 22
Figure 10: Processing times in seconds per 1000 units, and utilization for dedicated machines. ... 23
Figure 11: From one to two machine workstation design ... 24
Figure 12: Processing times (per 1000) and number of machines for printing in different situations . 26 Figure 13: CT in hours for ED1 per production batch of 24k ... 28
Figure 14: WTTB per batch in current situation for ED1 ... 29
Figure 15: Transfer batching ... 30
Figure 16: Simulation cycle times with transfer batching for ED1 ... 32
Figure 17: Production flow for different transfer batches in hours, for ED1, ED0 andED4 ... 33
Figure 18: Production flow for ED0, ED1 and ED4 ... 33
Figure 19: Quantification order lead time improvement by new process design ... 34
Figure 20: WIP levels for current and proposed product in process ... 35
Figure 21: Breakdown cost of WIP per 1000 lighters for ED1, 2010 ... 36
Figure 22: Cost‐benefit analysis ... 37
Figure 23: Results SWOT analysis of Swedish Match Lighters ... 42
Figure 24: Daily capacity in thousands operations 2011 ... 43
Figure 25: Number of machines dedicated per product type ... 43
Figure 26: Capacity utilization for estimated 300mil in 2015 ... 44
Figure 27: Simulation lead times for ED0 ... 46
Figure 28: Simulation lead times for ED1 ... 47
Figure 29: Simulation lead times for ED4 ... 48
Figure 30: ED0 stock after humidification ... 49
Figure 31: Calculations WIP levels ED0 and ED1 ... 50 Figure 32: Calculations WIP levels ED4 ... 51
List of abbreviation
General abbreviations:
CT: Cycle time, the interval in which a batch can be produced at a workstation FGI: Finished Goods Inventory, the inventory waiting after the final production step LT: Lead Time, the time a product spends at an operation or in the entire factory.
PPC: Production Planning and Control system. The system of rules which defines how the flow of products between operations is defined and monitored
JIT: Just‐In‐Time, a PPC method
SMPI: Swedish Match Philippines Incorporated.
TOC: Theory of Constraints, a PPC method
WIP: Work In Process, a measure of inventory levels
WTTB: Waiting Time To Batch, a quantification of the time products are waiting, as part of the CT.
Swedish Match specific abbreviations:
Different product families are described as follows:
ED0: Round shaped, relatively large ED1: Relatively square shaped
ED4: square shaped with electric system C15: Gas dispenser for mosquito repellent
Production steps are sometimes referred to as follows:
M: Molding W: welding H: Humidification G: Gas filling A: Assembly Pr: Printing
L/W: Labeling & weighing Pa: Packaging
1. Introduction
1.1 Company background
Swedish Match produces products applicable in the tobacco industry in three product divisions:
Smoke‐free products, lighters and cigars. The lighting product division can be divided in two groups;
lighters and matches. This research is done in the Philippine factory, which is one of the three locations where the lighters are produced.
1.2 problem introduction
The products that are made by SMPI include three type of lighter products and a special product for 3rd party contract where the gas dispenser is used for mosquito repellent (C15). The ED0 lighters is a round turning wheel activated lighter, the largest of the three. The ED1 type is a more square formed lighter, also activated by a turning wheel. The ED4 lighter is an electric lighter with a push button to electrically light it instead of creating a spark with a turning wheel. The ED1 and ED4 both have different lengths within their family. The smaller lighters contain the same parts as the longer ones, but the gas‐reservoir is smaller. Within both the ED1 and ED4 family, there are also lighters with a child resistance strip available. To gain more insight in the products, a SWOT analysis on the SMPI lighters can be found in Appendix 1.
Swedish Match is unsatisfied with its current production performance because inventory levels vary in time and lead times increase whenever production output increase. The market for ED1 and ED4 products are expected to grow in the next few years, therefore Swedish Match expects both inventories and lead times to increase substantially if the situation does not change.
1.3 Project objective
The objective of this research is to redesign the production process to improve production performance to prevent production lead times and inventory levels to increase when production quantities increase.
The scope of this research can be defined in terms of products and process. Because ED1 product is the product with highest projected growth, this research is mainly based on improving ED1 production. In later stages, the proposed enhancements can also be applied to products ED0 and ED4, as will be described later in this report. This project focuses only on the assembly line of the lighter. The production of parts that are used in the assembly process is not included.
This report is structured as follows. Chapter two will describe the research design, chapter three the existing production process, chapter four the problem analysis and in the proposed solution will be given in chapter 5.
2. Research design
2.1 Problem statement
Within Swedish Match the following problem statement is defined:
Swedish Match should improve production performance to gain production flexibility and therefore respond to changing customer demand.
Production performance includes the two observations already described in the introduction, the high variance of inventories and the long lead times. Both these problems are related to production throughput, which will be shown in the conceptual model.
2.2 Conceptual model
To measure production performance at Swedish Match, performance indicators are defined. Figure 1 shows the conceptual model in which three parameters are defined which can be used to describe production performance at SMPI.
Figure 1: The conceptual model
According to literature, there is a trade‐off in using inventory, time and capacity as buffers to manage variability (Hopp and Spearman, 2000) and the production efficiency reduction because of these buffers (Nicholas and Soni, 2006). Variability degrades performance (Hopp and Spearman, 2000) and in the case of Swedish Match can be described as the difference in processing times at different operations. When customer demands changes, the consumption of capacity at each workstation changes. To prevent starvation, buffers in the form of inventories are built up. However, the
inventory buffers themselves also degrade production performance in terms of efficiency (Nicolas and Soni, 2006). The total lead time of a production batch on the shop floor increases due to the time the products spend in the inventory buffers.
These theories lead to the following conclusion. to increase throughput of the production line, Swedish Match should work on both minimizing the size of the buffers as well as minimizing the number of buffers needed to overcome variability in demand.
Inventory and lead time have a negative relation to production performance which is indicated with a minus‐sign in the figure (Nicholas and Soni, 2006). When the buffer of any of these variables increases, the total performance of the production environment decreases. Capacity has a positive relation to production performance which is indicated with a plus‐sign.
2.3 Research questions
The research question is defined as follows:
How can production performance at Swedish Match be improved?
The following sub‐questions are based on the theory that both inventory levels and lead times influence the production performance at Swedish Match as described in the conceptual model.
A. What factors influence the levels of inventories within Swedish Match and how can these factors be changed to reduce the variance of the inventories?
B. What factors influence the production lead time within Swedish Match and how can they be adjusted to reduce this lead time?
2.4 Methodology
Various sources and techniques are used in this research to acquire information used to answer the research questions. Scientific Literature is used to define the conceptual model, to discuss different ways of production planning, distinguish different aspects of time and different types of batching. On general level, literature is used to structure the research analysis and the report.
To gain insight in the situation at Swedish Match and validate the perception gained from internal documents, both structured and unstructured interviews with experts within Swedish Match were conducted to gain insight in the production processes and validate research findings at different stages of the research project.
Most of the data used in this research was conducted from the interviews and validated by internal documents available at Swedish Match. Documents about processing times, portfolio distribution,
production defects, product cost overviews and more, were made available to analyze the current situation at Swedish Match.
3. Existing production process
In the current situation, the production steps to assemble the lighters at Swedish Match are defined as follows: Molding (M), Welding (W), Humidification (H), Gas Filling (G), Assembly (A), Printing (Pr), Labeling & Weighing (L/W) and Packaging (Pa).
Molding is where the body of the lighter is automatically attached to the bottom part. Welding is where the top plastics are automatically attached to the body. For ED1, these products are sold as Kits. To ensure quality the finished bodies are humidified before gas filling. The bodies are put in a container for 12 hours before they are further processed. At gas filling, gas is put into the lighters.
The assembly operation consist of automated lines where other parts are attached to the lighter. All activities are checked by sensors and the assembled product (naked product) is also checked on functionality by inspecting the size of the flame. The assembled lighters for ED1 are sold as naked‐
lighters. At printing, the lighters are printed with customer specific logo’s or images. After printing, barcode stickers are put on the products (Labeling) and the finalized product is being weighted to check the amount of gas in the finalized product. The final step in the process packaging, where the lighters are packed with 5 different colors, in quantities of 50 units (5 rows of 10). The 8 steps are visualized in Figure 2 below.
Figure 2: Current situation SMPI
The rectangles represent the different process steps (workstations), and the inventories between the operations are stocks are indicated with an letter ‘S’. There are two locations in the production line where stabilization periods are required to assure product quality. These are indicated with a letter Q and represent the cooling down period after molding and a 24 hour time window before weighing to detect gas‐leaks in the lighters.
3.1 Production layout
The layout of the shop floor can be described as functional (Nicholas, 1998). Each process step has its own production area and products move between these area’s on pallets.
The pallets used at Swedish Match are plastic carts on which plastic bins can be placed. Each production batch consists of one pallet with 24 bins. Because the products are of different size and weight, the number of products per bin differ. For instance, the ED0 product is relatively large, and has a batch size of 15.600 lighters, while the smaller ED1 product is produced in batches of 24.000.
3.2 Production planning
The production planning describes how batches of products move through the production process.
At Swedish Match, the pallets are produced using kanban cards between the last en the first operation, labeling and molding. Each time the labeling operation finishes a production batch, the kanban card is send to the molding operation, ordering a new batch of that same product.
The use of kanban cards in this way can be described as a Continuous work in process system (CONWIP), where a card is used to prevent excessive inventory build up by only allowing production at the first operation when the last operation has completed another batch (Spearman et al. 1990).
Since the cards move from molding down to packaging with the products batches as they are produced, this implementation of cards can be defined as a traditional push system (Nicholas, 1998).
Since the production quantities of the first operation is controlled by the kanban card, all other machines produce whatever is offered from the upstream operations.
Kanban is not introduced for all products at Swedish Match. Only product types that need all the production steps are regulated by the cards. Special orders, like the ED1 kits that only require molding and welding are pushed separately through these operations, without the use of Kanban cards.
4 Problem analysis
To analyze the different types of buffering, Lean literature is used to address different variables for each of these types. Within the three aspects of production performance measures, some possible indicators are shown in Figure 3.
Figure 3: Detailed model of production performance and indictors of buffering
The defined indicators of the different aspects inventory, lead time and capacity are related to some extent. For instance, buffers between the process steps (WIP) may lead to transport times between these process steps and this transport time might lead to a lower utility of individual machines. Also, the waiting time to batch at the buffers might lead to lower utility of the individual machines. The opposite result from the buffers might also be possible. This case, the buffers are optimal and would prevent starvation of the machines and therefore increase the utility.
Because of the interdependency of the different buffer types, measuring one of these buffers will give insight in the evaluation of the production process. To facilitate the optimization of the throughput of any production environment, different production planning and control (PPC) tools have been developed in the past century. All of these tools are designed to minimize Work In Process (WIP). Jodlbauer & Huber (2008) describe the use of Service level and WIP as a measure of performance of the PPC system and therefore for all buffers.
4.1 Variance of inventory
Ohne (1988) defines seven types of waste in his research on production efficiency. Inventory is one of them, where higher inventory levels lead to lower production efficiency and therefore lower performance. Ohne defines three types of inventory which should be analyzed independently. At the beginning of the production line, the Raw Materials Inventory is defined. Within the production process, both products that are worked on as well as buffers between the production steps are defined. This is defined as Work In Process (WIP). The Third type of inventory is defined at the end of the production process, the Finished Goods Inventory (FGI). All three types of inventory have an positive relation to the Inventory which is indicated with the plus‐symbol in Figure 3. This results for instance in an increase of inventory levels when the level of Finished goods increases.
Inventory levels at SMPI
From the interviews of multiple managers the complaint arises that inventory levels vary daily. The manufacturing manager and employees spend time and effort in reducing the amount of inventory every week. Sometimes inventory levels before one workstation increase to twice the daily capacity of that machine.
To validate the observations, an analysis is performed to calculate the fluctuations of the inventory using data from internal reports. Results of the performed data analysis of the ED0 and ED1 products are show in Figure 4. ‘Variance high’ describes the average variance on inventory levels above the mean inventory levels, where ‘variance low’ describes the average variance of inventory below the mean.
ED0 ED1 ED4
Mean: 475579 1733649 973526
Variance High 18% 31% 59%
Variance Low 11% 27% 65%
per batch 15600 24000 20000
# batch 30.5 72.2 49
Figure 4: Inventory levels and variance for lighter production batches
Both the mean number of batches in the process and the variance of inventory levels differ per lighter. The ED4 figures show a relatively large variance which can be explained by the interference of the employees to reduce the high stock levels in the time span of which the data was recovered.
The average number of ED4 batches in production is 49, while the highest number, was 78.
4.2 Long lead times
Buffering time is embedded in the lead time which is the time it takes for a lighter to be produced from raw inventory to end product. Ohne (1988) describes transport‐ and waiting times as two types of waste which both increase the cycle time (the + symbols in the figure) of the process and therefore decrease production performance.
Transport times include the time it takes for products, product batches or materials to be moved from machine to machine or between production lines. The waiting time is defined by Ohne (1988) as the time the machine has to wait for products or parts, set‐up times, or for repair after a breakdown. Reducing these times would decrease lead times and therefore increase production performance.
A third way of reducing cycle times would be to reduce processing times which is defined as the time it takes to perform a certain task on a product (Nicholas, 1998).
Lead times at SMPI
The observation of lead time can be validated with the data available from internal reports.
Comparing the actual lead time with the best possible theoretical lead time validates the observation of ‘long’ lead times.
The best possible cycle time (CTBEST) is the sum of process times of all operations (Hopp and Spearman, 2000). Using the theory on production performance, both transport and waiting times should be reduced to zero. The best possible order Lead time (LTBEST) is equal to CTBEST, meaning that orders (which contain one type of product) should be produced without interference of other orders.
Using the capacity data from all the operating machines from the internal reports at SMPI, the CTBEST for one production batch can be calculated as shown in Figure 5. Assuming an order of one million lighters is produced at ones, multiple production batches will be produced one after the other. The number of batches needed to fill one order of a million lighters are also mentioned in the figure.
ED0 ED1 ED4
# batch per order: 64 42 50
Processing times (minutes):
Molding 449.3 432 576
Q1 720 720 720
Welding 320.8 314 274.3
Gas filling 249.6 636.6 303
Assembly 224.6 314 320
Printing 187.2 426.8 355.7
Q2 1440 1440 1440
Labeling / Weighing 204.1 314 261.7
Total in minutes 3795.6 4324.4 4250.1
Total in Hours 63.3 72.1 70.8
Figure 5: Processing times for the first batch for dedicated machines
The difference in lead times for each batch as well as the difference in processing times at operations are due to the different sizes of the lighters and the number of lighters in one batch. At printing for instance, ED1 and ED4 are printed at the same speed. However, because one production batch of ED1 contains 24k products while an ED1 batch contains 20k per batch, the processing time per batch differ.
The next step in calculating the best possible order lead time is to determine the pace in which all other batches are produced. In an environment where inventory is minimized, this means that the operation with the longest processing time determines the pace of production output (Nicholas
1998), which for both product types is the molding operation. The total cycle time for one order of a million lighters can be calculated as follows:
ED0: CTBEST‐ORDER = 63.3h + (#batch‐1 * CTMOLDING) = 535 hours = 22.3 Days ED1: CTBEST‐ORDER = 72.1h + (#batch‐1 * CTMOLDING) = 367 hours = 15.3 Days ED4: CTBEST‐ORDER = 70.8h + (#batch‐1 * CTMOLDING) = 541 hours = 22.6 Days
The best possible Cycle Time for one order (CTBEST‐ORDER) equals the best possible cycle time for the first batch plus the total number of batches minus the first batch (#batch‐1) times the pace in which the batches are produced, which is the molding operation (CTMOLDING).
The results of the three calculations for ED0, ED1 and ED4 are consistent with the results gathered from the interviews about the estimated order lead times.
To identify whether or not the lead time is relatively long, the current order lead time is compared to the capacity charts of the different machines. The capacity of each machine is lighter dependent and can differ per operation. For instance, the daily capacity for the molding of bodies is 50k, 80k and 50k respectively to the types ED0, ED1 and ED4. The capacity figures are summarized in Figure 6 and are based on actual capacity with an efficiency rate of 95%.
Capacity ED0 ED1 ED4
Operation Cap. Cap. Cap.
Molding 100 720 100
Welding 140 770 105
Gas filling 180 665 95
Assembly 100 550 90
Printing 120 567 567
Labeling / weighing 770 770 770
Figure 6: Daily capacity in thousands operations 2011
To calculate the best theoretical cycle time (and optimal lead time), multiple workstations can be used per operation and both waiting‐ and transport times are reduced to zero. This is done by taking the operation with the lowest capacity and the order size of one million by this number. The total production time needs to be increased with 36 hours of stabilization between molding and welding (12h) and after assembly (24h), to guarantee quality.
ED0: 1m / 100kd = 10 days + 1.5 days of stabilization = 11.5 days ED1: 1m / 550kd = 1.8 days + 1.5 days of stabilization = 3.3 days ED4: 1m / 90kd = 11.1 days + 1.5 days of stabilization = 12.6 days
Comparing these results with the actual lead times calculated before, it can be concluded that order lead times can be reduced drastically. An explanation for the relatively large difference between ED1 and the other products is that ED1 has more dedicated machines to use.
4.3 Expected capacity shortage in the future
The capacity of a plant is dependent on the capacity of the individual machines used in the different process steps. As defined earlier, the bottleneck is defined as the process step or machine with the lowest throughput (Hopp and Spearman, 2000). The bottleneck should therefore set the pace in which products can flow through the plant without adding more inventory (Nicholas, 1998). More theory about pace‐setting will be discussed in a later section. At Swedish Match, focus is put on increasing the efficiency of capacity use, to postpone investments in new machinery as long as possible.
Nicholas (1998) defines the need to increase utility of the bottleneck machines. Since the bottleneck has the largest impact on total throughput, production planning should be focused on the bottleneck.
Lean literature shows that the availability of machines can be improved to improve capacity use (Nicholas & Soni, 2006). Preventing breakdowns and reducing set‐ups are examples of ways to improve availability.
A third factor of capacity, identified by the researcher, is efficiency. This describes the ratio of products per machines usable for the next operation divided by the total number of products produced on a machine. By reducing the number of errors, or unusable products, the efficiency of the machine increases and therefore the capacity increases.
Capacity at SMPI
The ED0 and ED1 product are expected to grow in sales over the next 5 years. Using the expected sales numbers the ED1 product will be the only lighter in 2015 that has to little capacity with the current machines available. The only operation with enough capacity for the ED1 product is gas filling. This can be derived from the calculations described in Appendix 2 of which the results are shown in Figure 7. Derived from the conceptual model, utilization is used as a variable for capacity.
When no changes are made at machine level all operations but gas filling show utilizations higher than 90% for ED1 product. Utilizations higher than 100% indicate lack of capacity. Since 10% of machine capacity is reserved for (preventative) maintenance, all operations with utilization higher than 90% should be improved.
Operation ED0 ED1 ED4 C15
Molding 79 % 108 % 63 % 127 %
Welding 57 % 101 % 60 % 118 %
Gas filling 44 % 88 % 67 % 88 %
Assembly 79 % 107 % 71 % 88 %
Printing 97 % 97 % 97 % ‐
Labeling / weighing 95 % 95 % 95 % ‐
Figure 7: Capacity utilization for estimated 300mil in 2015
The printing and labeling machines are shared between the different products. Therefore improving the utilization rate for printing ED1 will also improve printing for ED0 and ED4.
5. Solution design
Using the conceptual model, variables were changed to reduce both the variance of inventories and the lead time. The proposed solution consist of three steps shown in Figure 8.
Figure 8: Solution method in three steps
First, The workstation design of the operations printing and gas filling are adjusted to reduce the lead times of these operations and generate a balanced production line. The variance of inventories are caused by the difference in processing times of the workstations and by changing the workstation design of gas filling and printing, these processing times will be altered. Creating a more balanced flow will not only reduce the inventories, but also make the implementation of a kanban system easier.
In step two, the planning and control system are changed to a Kanban system from Labeling up to Humidifying to automatically manage inventory levels between these stages. The first three operations, molding, welding and humidification, will still be pushed. A push‐pull interface is necessary between humidification and gas filling to link the push and the pull system efficiently. To
minimize the amount of cards only two types of cards are used. One Card for the steps printing and labeling and one card for the production steps gas filling and assembly. There are two reasons why the old production planning and control tool is not functioning optimally at SMPI. First, WIP levels are both high and variable while sequence of processing times should facilitate zero inventory between the operations for ED0. Second, kanban cards should facilitate Just‐in‐Time production resulting in minimal inventories, yet the employees have to spend time reducing the inventory levels.
Third, transfer batches of 12 bins are implemented between gas filling and assembly and between printing and labeling. Waiting times between these operations will reduce drastically when bins are allowed to move to the next operation without waiting for another 12 bins of products. By reducing the waiting time, the total lead time will be reduced.
The new process design to solve the problems defined earlier is shown in Figure 9. The square blocks represent the eight production steps molding, welding, humidification, gas‐filling, assembly, printing, labeling &weighing and packaging. The push‐pull interface is shown with a dotted between humidification and gas filling. The choice of location for the push‐full interface is determined because of the complexity of the molding and humidification operations, and are explained in section 5.2. The blue arrows describe the production control flow, where a push flow is shown between molding and humidification (moving downstream) and a pull movement between gas filling and assembly (Kanban cards moving upstream). There are two locations where Kanban cards are used; one between gas filling & assembly and one between printing & labeling. The two locations where transfer batching is allowed are indicated with the letters TB.
Figure 9: New process design production SMPI
5.1 Workstation design at gas filling and printing.
Figure 10 shows the processing times and the utilization for the different operations of ED1 product.
Focus is on the ED1 product since this is the only product which is expected to have a shortage of capacity in 2015, as described in the portfolio analysis. The expected utilization is described earlier (in figure 7). Operations with expected utilization of over 90% indicate a lack of capacity and therefore require improvement.
Operation:
Processing times ED1
Utilization ED1
Molding 1080 108 %
Welding 785 101 %
Gas filling 909 88 %
Assembly 785 107 %
Printing 1067 97 %
Labeling / Weighing 785 95 %
Figure 10: Processing times in seconds per 1000 units, and utilization for dedicated machines.
In Figure 10, processing times differ from one operation to another. Molding, gas filling and printing each have a processing time of around 1.5 times the other three operations. Because the longer processing times and the shorter processing times change from one another one‐by‐one, the production flow is unbalanced (Chakravorty and Atwater, 1996). Humidification is not mentioned in this table because this operation is done in cells in which up to 4 batches are humidified at ones. The processing time is 24 hours whether there is one bin of 1000 products in the cell or 4 pallets of each 24 bins (= 96.000 units). The average processing time cannot be calculated because batches of products can be combined in one humidification cell.
Gas filling and printing workstations are changed in the new process design, where two machines will act as one workstation. In the current design, every workstation contains one machine. Molding workstations are not changed because the molding machines are relatively big, expensive and they are difficult to move because they are connected to the nylon inventory.
Gas filling
To balance the production flow, the workstations of the gas filling operation in the new process design are changed from one to two machines per workstation. This will cut processing times in half since production batches will be split over two machines. This solution is viable for two reasons. First, the operation has overcapacity. As the results in Figure 7 show, the predicted utilization for 2015 is 88%, while the processing times is one of the longest. This is a result of the SMPI focus on capacity instead of processing times or flow which resulted into investing in new machines without looking into the possibilities to improve the production flow. Second, no investments in faster machinery is needed because the current machines can be used to design the new workstations.
From one‐ to two machine workstations
For tracking and tracing reasons, every batch is fully handled by one individual machine, as shown in situation 1 in Figure 11 (this holds for all operations, not just gas filling).
Figure 11: From one to two machine workstation design
By changing the number of machines per workstation to two, as shown in situation 2 in Figure 11, no changes are needed for the individual machines though the processing time of the operation will be reduced by roughly 50%. This leads to three desirable changes. First, the production flow will be more balanced, leading to less inventory variance. Second, a reduction of Work In Process (WIP) is realized and third, since the processing time of the gas filling operation is reduced, the total order lead time will also be reduced. In the current situation each gas filling machine has one production batch of WIP while in the new workstation design every two machines will have one batch of WIP.
Since there are 10 gas filling machines, this can be quantified as a reduction of inventory of 5 production batches.
Validation of viability using capacity and set‐up times
Since there are ten gas filling machines in 2010, the new workstation design would have 5 work stations where gas filling is done with two machines each. One disadvantage of the new workstation design is the increase in set‐ups. Because the current system requires no investment in gas filling capacity, the future state should ideally also be viable without adding capacity.
The five gas filling workstations will feed eight assembly workstations, therefore changeovers of the gas filling workstations are needed to feed the assembly machines with different types of lighters.
Assuming the processing time of gas filling is reduced from little over 900 to less than 500 and the assembly operation has a processing time of little less than 800 (see Figure 10), the difference of roughly 300 seconds can be used for machine changeovers. The available time to set‐up the two gas filling machines for one workstation is 12 x 300 seconds, because there are 24k in a production batch which are split between the two machines. This equals 3600 seconds, which is one hour. According to the operators, this available set‐up time is acceptable.
Important to note is that the capacity number on which these calculations are based already include
set‐up times for different products within a product type (like ED1). These however, do not include the set‐ups to other product families since all machines are dedicated in the current situation.
Printing
The changes of the workstations proposed for gas filling also apply to printing. In the new design one workstation consists of two printing machines, reducing the processing times drastically to facilitate a more balanced production flow. New machinery should be built or bought because printing has an expected shortage of capacity in the near future. In the old situation, there are two types of machines which operate at different speed. One is built in the Philippine factory while the other has to be shipped from the Netherlands. The following trade‐off should be considered: (1) buy one fast (large) machine for ED4 and use the 8 Slower machines to built 4 workstations for ED1. Or (2) buy two slower machines to built 5 workstations for ED1 and ED4 together with the possibility to change between the product types.
Currently, there are nine printing machines, 1 with a capacity of 120k per day and 8 with the capacity of 81k per day. The faster machine is used for the ED0 line. this machine does not need changeovers since it has enough capacity to print all the ED0 lighters that need printing. This fast printing machine is produced in the Swedish Match factory in the Netherlands and therefore relatively expensive to built. A second downside of the fast machine is the size. It is more than twice the size of the other machines while its capacity is only roughly 1.5 times higher.
The slower machine is built in the Philippines and since knowledge is present within the plant, it is relatively cheap to built extra machines. However, since the processing time of this machine is over 1000 seconds (see again Figure 10), it is a less desirable machine when it comes to balancing the production flow.
The two investment possibilities:
Within Swedish match, the costs of the machines are estimated as follows:
Printing machine CAP = 120k, (incl. transport from Holland) investment = $100.000 Printing machine CAP = 81k, investment = $ 48.000
Set‐up times for printing are relatively low, which makes changing the definition of the workstation in this operation a option. By combining two machines in one workstation, the processing time is reduced by 50% which makes it less than 600 seconds per thousand lighters.
The faster, larger but more expensive machine has a processing time of 720 seconds per one thousand lighters while the combination of two slower machines has a processing time of 600 seconds. Figure 12 show the results of the analysis in which option 1 represents the scenario where two smaller machines are built and option 2 the situation where one faster machine is bought.
Printing ED0 ED1 ED4 Total
Capacity 120k 81k 81k
Processing time 2010
720 1067 1067
# Machines 2010 1 6.5 1.5 9
2015 option 1:
Capacity 120k 81k 81k
Processing time 720 600 600
# Machines 1 8 2 11
2015 option 2:
Capacity 120k 81k 120k
Processing time 720 600 720
# Machines 1 8 1 10
Figure 12: Processing times (per 1000) and number of machines for printing in different situations
In the current situation the one fast machine is dedicated to one product family. Due to size and average capacity, the number of machines between ED1 and ED4 will be 6.5, 1.5 respectively.
Internal data at SMPI showed that sharing of capacity is done between ED1 and ED4.
Option 1 describes the scenario where two slower machines are built. This way all ED1 and ED4 products are printed at the same speed (Balanced) and changeovers between the two product families are minimized. This solution leads to a number of machines of 1, 8, 2 for ED0, ED1 and ED4 respectively.
In option 2, the one printing machine currently used for ED4 is replaced by one faster machine which would have sufficient capacity to print all ED4 and a desirable processing time. By adding one fast machine to the printing department, the slower machines currently used for ED4 can then be used for ED1.
The difference between option 1 and two are as follows: will one fast machine be bought to handle all ED4 products, or will two small machines be built to handle them. More information is needed in this topic to be able to decide which of the scenario’s is most desirable.
5.2 Designing the production planning and control system
Using the methods described in the conceptual analysis, a hybrid method of TOC and JIT is used to design the proposed kanban system. Since multiple stock locations are allowed (JIT) and the system
constraints are tackled one by one (TOC). More details about this choice of PPC system can be found in Appendix 3.
The new production planning and control system can be described using a Push Pull interface between humidification & gas filling and two kanban cards pulling batches to packaging from printing and to printing from gas filling. These are shown in Figure 9.
The push‐pull interface
The proposed PPC system can be described as the hybrid push/pull system (Nicholas, 1998). In the first half, orders are pushed through the process while in the second half the orders are pulled. This way the capacity use of the bottleneck operation molding (longest CT) and the constrained humidification cells will be maximized because they can be planned in advance while the other operations will be order driven.
There are three reasons why the push pull interface is implemented after humidification in the new process design. First, the molding operation is done in 3 batches of the same color at ones. One production run at molding will always produce 3 batches of one color because of the high set‐up time and waste it produces during set‐ups. The packaging department would prefer one batch of each color to repack them into batches with 5 colors per pack. The second reason is the cooling down period after molding. Inventory must be held there to maintain product quality. Basically, molding and cooling should be considered as one operation which makes the difference with the processing time of welding relatively large. This makes a production flow between these two operations with minimal inventory rather complex. The third reason is capacity restriction of humidification cells.
Four batches fit in one humidification cell and the cell should stay closed for the entire processing time. Maximizing capacity also makes it difficult to create a production flow. Humidification is done based on product type. It is possible to combine multiple colors in one humidification cell.
By implementing a Push‐pull interface, the number of set‐ups of both molding and welding operation can be minimized. This results in the possibility to plan this capacity as efficient as possible. In the long run, investments in new machinery will be necessary to increase current capacity.
The Kanban cards
The locations of the two Kanban cards are determined using the theory of production flow (Nicholas, 1998) and is used to minimize the inventories should to keep the system as simple as possible.
Although the operations gas filling & assembly and printing & labeling are not physically connected,
the transfer batching system described in the next paragraph links them from the PPC perspective.
To manage the stock levels needed on three locations in the production cycle, the minimum number of cards is two. Each combination of operations will pull batches from stock. There is a stock location at the push‐pull interface to supply gas filling & assembly, and one before printing & labeling. The third stock location (before packaging) is needed because 5 colors need to be combined into the order‐batches which the ladies can pull directly from stock. This will trigger the Kanban system. The two kanban cards that trigger the previously defined parts of the production process are also shown in Figure 9.
The new production process can be cut in four parts. Figure 13 shows both the cycle times per operation as well as the new design in which operations are linked in terms of production planning.
All figures are obtained from the documentation of actual performance in the past, measured by Swedish Match. The bottom row of Figure 13 shows the cycle times for producing one batch of 24k in each of the four parts. The first combination of operations represents the push side of the process.
The latter three combined processing times represent the pull side of the process. The use of transfer batching between gas filling & assembly and between printing & labeling are also included in these calculations. Within the pull side of the process, cycle times go down when orders move upstream, meaning that upstream operations should always be able to deliver the products ordered by the downstream workstation.
M C W H G A Pr L Pa
Current
CT (h) 7.2 12 5.2 10 3.3 5.2 4 5.2 7.7
New
CT (h) 34.4 6.9 7.2 7.7
Figure 13: CT in hours for ED1 per production batch of 24k
5.3 implementing transfer‐batching at two locations.
In the new process design there are two locations where transfer batching is allowed. The choice of locations where transfer batching is allowed and the size of the transfer batch are determined with the two research goals in mind; to reduce waiting times to improve order lead times and to improve production flow to reduce inventories.
Waiting times
As stated in the Figure 3, three variables influence the cycle times and therefore the order lead time:
processing time, waiting time and transport time. The variance in processing times is already described, shown in Figure 10 and is reduced by changing the workstation design. The transport times are negligible, because the actual movement between operations can be less than one minute.
The one variable left in the cycle time area is the waiting time quantified above, which can be reduced by transfer batching.
Waiting time within a process is captures in the so called Waiting Time To Batch (WTTB). The WTTB can be defined as the time one product waits to be processed while the batch is already in process plus the time a product waits after being processed until the rest of the batch is processed (Hopp &
Spearman, 2000). In a batch of 24k, the first lighter has to wait 23.999 times the processing time after it is processed while the last lighter has to wait 23.999 time the processing time before it can be processed. A trade‐off is defined between number of set‐ups and the waiting times. The larger the production batches, the smaller the number of set‐ups but the higher the waiting times and vice versa. The implementation of transfer batching reduces the waiting times.
Figure 14 shows the results of calculations made to show the waiting times for different products at SMPI.
Process time
per bin
WTTB per batch
WTTB per batch
Minutes minutes in hours
Molding 18.00 414.00 6.90
Welding 13.08 300.92 5.02
Humidifying 0.00 0.00 0.00
Gas Filling 15.15 348.45 5.81
Assembly 13.08 300.92 5.02
Printing 17.78 409.02 6.82
Labeling 13.08 300.92 5.02
Total 90.18 2074.22 34.57
Figure 14: WTTB per batch in current situation for ED1
These results show a WTTB of more than 34 hours for a batch of ED1 product. By reducing this WTTB both sub‐problems defined in this research are party solved. The WIP will go down because less products are ‘waiting’ within the process and the Lead time is reduced because it takes less long for a batch to be pulled through the process. Directly linked to the WTTB is the batch size and the way a batch moves through the process. The choice of definition for these batching‐rules can therefore explain the waiting times and changing the batching rules will reduce the waiting time.
Creating Flow with transfer batches
The results from Figure 10 show the difference in cycle times for each operation. Since they differ greatly from each other, the old situation does not resemble a desired production flow. To create a desirable flow in a pull production process, the upstream machine should always be able to meet demand asked by the following downstream machine. Therefore the lead time of the downstream machine should always be larger than the lead time of machine upstream machine.
In the top row of the previously discussed Figure 13 can be seen that the current situation where batches are transferred of 24 bins only, this constraint is not met: the assembly operation is slower than printing. With the use of transfer batching, the cycle times are changed because of reduction of waiting time, while the processing times stay the same.
In literature, distinction is made between two types of batching: production batches and transfer batches (Nicholas, 1998). The size of production batches size equals the number of products that are produced in one machine‐run. Transfer batches are described as the de number of products which is allowed to be moved between operations. Figure 15 shows an example of a situation where transfer batching is put to practice. Where the production batch size stays the same in both situations (left and right, 4 products per batch) the right situation allows products to move to the next operation when they are with 2. In this example the Lead time is reduced from 40 to 30 seconds, which is 25%
without changing the machines or the production planning.
Figure 15: Transfer batching
Theoretically, linking two operations into one machine would result in the lowest WTTB because transfer batches would be reduced to 1 while production batches stay the same. This creation of a one‐piece flow (Nicholas and Soni, 2006) would require the most changes in the manufacturing facilities. First, every line would need a dedicated workstation for each operation (which requires
investments). Second, by reducing the lead time this much, the packing ladies need to work much faster and much more flexible to be able to cope with this faster production lead time. Because of work hours and contracts of employees the packaging should probably also be done automatically, which requires another investment.
Reducing the waiting time to almost zero is not a practical solution, since high investments to physically link the machines are needed.
Transfer batching locations
There are two locations where transfer batching is implemented in the new process design, between gas filling & assembly and between printing & labeling. A part of the simulation for ED1 product is shown in Figure 16. In each combination the first processing time is shorter than the second which prevents the second operation from idling. Transfer batching will not gain any advantage in the Push part of the process because of the cooling down time after molding and capacity constraint on humidification. The 24 hours of waiting time to assure product quality between assembly and printing makes it counterproductive to speed up the flow between them.
# Bins Transfer batch size 6 bins 12 bins 24 bins
Transfer batch size: 6 12 24
Operation ED1 ED1 ED1
Molding 25920 25920 25920
Cooling 43200 43200 43200
Welding 18840 18840 18840
Humidify 36000 36000 36000
First batch Humidified (LT): 123960 123960 123960
Hours 34,4 34,4 34,4
# Batch inventory 5,7 5,0 4,0
Gas filling 3000 6000 12000
Assembly 4710 9420 18840
First Batch Assembled (LT): 21840 24840 30840
Hours 6,1 6,9 8,6
# Batch in inventory (24h) 3,85 3,32 2,60
Printing 3600 7200 14400
Labeling / Weighing 4710 9420 18840
First Batch Labeled (LT): 22440 26040 33240
Hours 6,2 7,2 9,2
Stock 5 colors 5 5 5
Packing (5 to 1 batches) in hours 7,7 7,7 7,7
Order one million
# Batch 41,67 41,67 41,67
Rounded up 42 42 42
Figure 16: Simulation cycle times with transfer batching for ED1
Figure 16 also shows a trade‐off between lead time and inventory levels. Between the different production parts, inventory is held to prevent starvations. By keeping the LTN lower than LTN+1 this stock is minimized. The speed in which these inventories can be replenished determines the size of the inventories. The higher the difference in Lead time between 2 production parts, the higher the inventory needs to be. The three different transfer batches lead to inventories of 15, 14, and 13 for Transfer Batches of 6, 12 and 24 bins respectively for ED1, divided in three locations.
The first inventory is before packaging, and needs to be 5 batches for ED1, one batch of each color.
This is necessary because the packaging ladies can only start packing an order when all five colors are present. The second stock is the stock after assembly. Looking only at processing times, no stock is needed because the combination of printing & labeling is faster than gas filling & assembly. However, there is a quality constraint as described previously. Each batch has to stabilize for 24 hours before it is weighed at the labeling station. The size of the stock can therefore be calculated with the following formula, which leads to a stock of 4.85 = 5 batches for ED1:
Inventory Size (#2) = 24 / CTG+Pr
The third stock location is the largest. This is due to the push‐pull interface placed, where the push side takes over 34 hours to produce one batch, although it produces 2 at ones. To validate the safety stock of 8 pallets after Humidifying for ED0, another simulation is built and shown in Appendix 5.
Transfer batch size
The simulation of the production process shown is figure 17 is also built for ED1 and ED4, which shown in Appendix 4. Because all production batches contain 24 bins, three situations are included in the simulation for each type of lighter, transfer batches of 24 bins (current situation), 12 bins, and 6 bins. Figure 17 shows the cycle times for the 4 parts of production.
ED1 M C W H G A Pr L Pa
TB = 24 34 8.6 9.2 7.7
TB = 12 34 6.9 7.2 7.7
TB = 6 34 6.1 6.2 7.7
ED0 M C W H G A Pr L Pa
TB = 24 35 5.9 6.5 5.2
TB = 12 35 4.8 5.0 5.2