Strips Manufacturing. A thesis presented to. the faculty of. In partial fulfillment. of the requirements for the degree. Master of Science.

Transcription

1 Comparison of Alternative Global Supply Chain Design Approaches for Blood Sugar Strips Manufacturing A thesis presented to the faculty of the Russ College of Engineering and Technology of Ohio University In partial fulfillment of the requirements for the degree Master of Science Jue Jiang December Jue Jiang. All Rights Reserved.

2 2 This thesis titled Comparison of Alternative Global Supply Chain Design Approaches for Blood Sugar Strips Manufacturing by JUE JIANG has been approved for the Department of Industrial and Systems Engineering and the Russ College of Engineering and Technology by Gürsel A. Süer Professor of Industrial and Systems Engineeering Dennis Irwin Dean, Russ College of Engineering and Technology

3 3 ABSTRACT JIANG, JUE, M.S., December 2016, Industrial and Systems Engineering Comparison of Alternative Global Supply Chain Design Approaches for Blood Sugar Strips Manufacturing Director of Thesis: Gürsel A. Süer This thesis discusses alternative global supply chain design approaches for blood sugar strips manufacturing. Three alternative approaches considered are: 1) three independent facilities in three regions to meet their own demand; 2) single location manufacturing facility to meet world demand; 3) two newly proposed hybrid supply chain strategies to merge different product families. Manufacturing facilities are designed under stochastic demand by considering layered cellular design. This approach allows three types of cells to be formed - dedicated cells, shared cells, and remainder cells. The main objective of this paper is to compare three alternatives with respect to the number of manufacturing cells and the number of machines needed. Statistical analysis and cost analysis are also included in the comparison. We will also discuss detailed operational control parameters in one of the three facilities and discuss simulation results to validate the results obtained by layered cellular design approach. Later, setup time is also considered in the simulation experiment to redesign the manufacturing system.

4 4 DEDICATION This thesis is dedicated to my parents and my love.

5 5 ACKNOWLEDGMENTS I would like to express my gratitude and appreciation to Dr. Gürsel Süer for his encouragement and help during my study and thesis work. Being an academic advisor, Dr. Gürsel Süer discussed many details about the thesis with me and guided me not only in the academic research but also in the life. Also, I would like to thank Dr. Dusan Sormaz for the help of the simulation model in this thesis. My thesis committee members: Dr. Tao Yuan and Dr. Ashley Metcalf. Thank you for your academic guidance and advice. I would also thank Dr. Ozan Ateş for his previous research which I follow for my thesis. Finally, I would like to express my thank you to my parents and my love for their endless support, trust and encouragement.

6 6 TABLE OF CONTENTS Page Abstract... 3 Dedication... 4 Acknowledgments... 5 Table of Contents... 6 List of Tables List of Figures Chapter 1 Introduction Background Manufacturing System Components of Manufacturing System Classification of Manufacturing System Cellular Manufacturing System Supply Chain Supply Chain Management Supply Chain Modeling Supply Chain Resource Management Heuristic Procedure... 25

7 7 1.5 Simulation Justification Research Objectives Organization of the Research Chapter 2 Literature Review Global Location Strategy Models Cellular Manufacturing Design Chapter 3 Problem Definition Blood Glucose Manufacturing Manufacturing Cell Design Alternative Supply Chain Designs Strategy 1: Independent Supply Chain Design Strategy 2: Single Manufacturing Facility Design Strategy 3: Global Supply Chain Design with Shared Cells Strategy 4: Global Supply Chain Design with Family Centralized Cells Chapter 4 Methodologies Used for Supply Chain Designs Independent Manufacturing Facilities Mean Capacity Requirements and Standard Deviation Demand Coverage Probabilities... 51

8 Expected Cell Utilizations Heuristic Algorithm for Layered Cellular Design Supply Chain Strategy 1: Host Market Production Strategy Supply Chain Strategy 2: Single Manufacturing System Design Combined New Hybrid Supply Chain Strategy Maximizing Number of Dedicated Cells Chapter 5 Simulation Experiment Simulation Experiment without Setup Time Simulation Experiment with Setup Time Chapter 6 Results Cell Comparison Comparison of Cell Utilization Values: Theoretical vs. Actual Cell Utilizations with Setup Time Statistical Analysis Statistical Analysis of WIP for Individual Facilities Statistical Analysis of WIP between Independent Design and Single Design Statistical Analysis of Average Flow Time Cost Analysis between Alternative Supply Chain Strategies Labor Cost... 93

9 Machine Cost Transportation Cost Total Cost Chapter 7 Conclusion and Future Work References Appendix A: Demand Coverage Probability Calculation of Fast System for China Region Appendix B: DCP and ECU of Fast System for China Region Appendix C: Decision Modules of Queue Size in China Facility

10 10 LIST OF TABLES Page Table 1.1 The Basic Correlation between the Arrangement of Manufacturing Elements and Type of Manufacturing System (Lenz, 2012) Table 3.1 Production Rates (per minute) in Fast System (Ateş, 2013) Table 3.2 Production Rates (per minute) in Slow System (Ateş, 2013) Table 3.3 An Example of Product-Machine Incidence Matrix (Ateş, 2013) Table 3.4 Product Families and Cells Table 3.5 Family vs. Cell Assignment Table 3.6 Family vs. Multiple Cells due to High Demand Table 3.7 Layered Cellular Design due to Stochastic Demand Table 4.1 Percentage to Arrange Standard Deviation with Mean Demand Table 4.2 Mean Demand and Standard Deviation by Family in Three Regions Table 4.3 Mean Capacity Requirements and Standard Deviation in Fast System Table 4.4 Mean Capacity Requirements and Standard Deviation in Slow System Table 4.5 Demand Coverage Probabilities of Fast System for China Region Table 4.6 Demand Coverage Probabilities of Slow System for China Region Table 4.7 Demand Coverage Probabilities of Fast System for Ireland Region Table 4.8 Demand Coverage Probabilities of Slow System for Ireland Region Table 4.9 Demand Coverage Probabilities of Fast System for Puerto Rico Region Table 4.10 Demand Coverage Probabilities of Slow System for Puerto Rico Region Table 4.11 Expected Cell Utilizations in Fast System for China Region... 57

11 11 Table 4.12 Expected Cell Utilizations in Slow System for China Region Table 4.13 Expected Cell Utilizations in Fast System for Ireland Region Table 4.14 Expected Cell Utilizations in Slow System for Ireland Region Table 4.15 Expected Cell Utilizations in Fast System for Puerto Rico Region Table 4.16 Expected Cell Utilizations in Slow System for Puerto Rico Region Table 4.17 Similarity Coefficients between Product Families (Ateş, 2013) Table 4.18 Cell Type in Fast System for China Region Table 4.19 Cell Type in Slow System for China Region Table 4.20 Cell Type in Fast System for Ireland Region Table 4.21 Cell Type in Slow System for Ireland Region Table 4.22 Cell Type in Fast System for Puerto Rico Region Table 4.23 Cell Type in Slow System for Puerto Rico Region Table 4.24 Mean Demand and Standard Deviation for Single System Table 4.25 Demand Coverage Probabilities of Fast System in Single Manufacturing Design Table 4.26 Expected Cell Utilizations of Fast System in Single Manufacturing Design. 64 Table 4.27 Cell Type for Fast System in Single Manufacturing Design Table 4.28 Cell Arrangement in Fast System for China Facility Table 4.29 Cell Arrangement in Fast System for Ireland Facility Table 4.30 Cell Arrangement in Fast System for Puerto Rico Facility Table 4.31 Cells in Fast System for China Region Table 4.32 Cells in Fast System for Ireland Region... 68

12 12 Table 4.33 Cells in Fast System for Puerto Rico Region Table 5.1 Inter-arrival Time for Each Product Family Table 5.2 Operation Times per Thousand Vials in Fast System Table 5.3 Simulation Results vs. Mean Demand of Fast System in China Region Table 5.4 Simulation Cell Utilizations of Fast System in China Region Table 5.5 Inter-arrival Time for Entities with Setup Time of Fast System in China Region Table 5.6 Simulation Results vs. Mean Capacity of Fast System in China Region Table 5.7 Number of Setups of Fast System in China Region Table 6.1 Comparison among Independent Supply Chain, Combined New Hybrid Supply Chain and Single Manufacturing System Table 6.2 Queue Size with Comparison of Cell Utilizations of Fast System in China Region Table 6.3 Another Queue Size with Comparison of Cell Utilizations of Fast System in China Region Table 6.4 Comparison of Demand Coverage Probabilities of Fast System for China Region Table 6.5 Comparison of Expected Cell Utilizations in Fast System for China Region. 85 Table 6.6 Comparison of Cell Type in Fast System for China Region Table 6.7 Working-in-Process in Each Manufacturing Facility Table 6.8 Means for WIP in Each Manufacturing Facility Table 6.9 Analysis of Variance... 88

13 13 Table 6.10 Fisher s Test Table 6.11 Working-in-Process in Two Manufacturing System Table 6.12 Means for WIP in Each Manufacturing Facility Table 6.13 Analysis of Variance Table 6.14 Average Flow Time (Mins) in Each Manufacturing Facility Table 6.15 Means for Average Flow Time in Each Manufacturing Facility Table 6.16 Analysis of Variance Table 6.17 Hourly Labor Cost (Ateş, 2013) Table 6.18 Number of Labor Required in the Fast System (Ateş, 2013) Table 6.19 Number of Labor in the Fast System for Independent Supply Chain Strategy Table 6.20 Number of Labor in the Fast System of China Facility for Combined Strategy Table 6.21 Number of Labor in the Fast System of Ireland Facility for Combined Strategy Table 6.22 Number of Labor in the Fast System of PR facility for Combined Strategy.. 95 Table 6.23 Number of Labor in the Fast System for Single Manufacturing Strategy Table 6.24 Number of Workers and Labor Costs for Three Strategies Table 6.25 Hourly Machine Cost (Ateş, 2013) Table 6.26 Number of Machine Required in the Fast Manufacturing System (Ateş, 2013) Table 6.27 Number of Machines in the Fast System for Host Production Strategy... 98

14 14 Table 6.28 Number of Machines in the Fast System of China Facility for Combined Strategy Table 6.29 Number of Machines in the Fast System of Ireland Facility for Combined Strategy Table 6.30 Number of Machines in the Fast System of PR Facility for Combined Strategy Table 6.31 Number of Machines in the Fast System for Single Manufacturing Strategy 99 Table 6.32 Number of Machines and Machine Costs for Three Strategies Table 6.33 Transportation Cost for Combined Supply Chain Strategy Table 6.34 Transportation Cost for Single Manufacturing Strategy Table 6.35 Total Cost for Three Strategies

15 15 LIST OF FIGURES Page Figure 1.1: Manufacturing system flow (Wright, 1990) Figure 1.2: Four types of manufacturing layout (Süer, Huang & Maddisetty, 2010) Figure 1.3: Host market production model (Dicken, 1992) Figure 1.4: Globally concentrated production model (Dicken, 1992) Figure 1.5: Regional/Global product specialization (Dicken, 1992) Figure 1.6: Semi-product specialization model Figure 1.7: Semi-host market model Figure 3.1: Independent manufacturing systems Figure 3.2: Global supply chain design with sharing cells Figure 3.3: Global supply chain design with family centralized cells Figure 3.4: Merged global supply chain design with family centralized cells Figure 4.1: Methodology flowchart Figure 4.2: Layered cellular design consisting dedicated, shared and remainder cells Figure 5.1: Arena sub-model of assign product families to cells Figure 5.2: Arena sub-model of cell processes Figure 5.3: Arena sub-model of product families dispose Figure 5.4: F2Create module Figure 5.5: BatchF2 module Figure 5.6: DecideF2 module Figure 5.7: AssignF2Cell4 module... 73

16 16 Figure 5.8: EnterCell module Figure 5.9: SeparateFs module Figure 5.10: StartProcess module Figure 5.11: Process module Figure 5.12: Value of working-in-process trend Figure 5.13: Arena sub-model of cell processes with setup time Figure 5.14: Setup? module Figure 5.15: SetupSeize module Figure 5.16: Setup module Figure 5.17: RecordSetup module Figure 5.18: SaveLastEntityType module Figure 6.1: Queue Size Arrangement in Family Figure 6.2: Boxplot of three facilities Figure 6.3: Fisher s plot of three facilities Figure 6.4: Boxplot of two facilities Figure 6.5: Boxplot of four facilities... 93

17 17 CHAPTER 1 INTRODUCTION This research focuses on designing alternative supply chain/ manufacturing systems for a global blood sugar strip manufacturer. There are three main supply chain system design approaches included in this research 1) independent supply chain system, 2) single manufacturing system, and 3) newly proposed hybrid supply chain designs. First, three independent manufacturing facilities are designed to meet the demand of three regions. Using cellular manufacturing concepts, the type and number of manufacturing cells in the manufacturing system are determined for each manufacturing facility considering their regional stochastic demand data. The next supply chain strategy is the one where all production is done in a single manufacturing facility. In order to minimize cost, different cells can be combined into one as long as capacity is not exceeded. Once many featuring processes are completed, products in these cells can be shipped to three regions. Another supply chain strategy is the combination of two newly proposed hybrid supply chain designs, which is named combined new hybrid supply chain strategy. In this approach, two comprehensive supply chain designs are combined based on the real demand data. By considering different methodologies, cell utilizations, work in process levels are compared to measure the usage of each cell. Meanwhile, cost analysis is conducted among the comparison between these three supply chain design approaches. Simulation is also an important method to verify the design and find the optimal parameters for operating conditions. Two additional issues are also studied in the thesis, a) setup time will be included in the designed system which will influence capacity and

18 18 consequently cell utilization, b) a simple procedure will be discussed by which the number of dedicated cells is increased. 1.1 Background Global manufacturing companies have various issues to consider in designing world-wide manufacturing systems. In the detailed design of a manufacturing system, similar products can be grouped into one product family. Each product family can be produced in their cell to minimize the number of duplicated machines and lower setups which thus increase cell utilization. The supply chain design includes what product families to produce in each manufacturing facility and which manufacturing cells to allocate each product family to reduce the cost and thus increase the total profit. 1.2 Manufacturing System From Kimemia and Gershwin (1983), a manufacturing system is the process in which raw materials are processed to produce unfinished or finished products via labors, machines and tools. Most of the products sold, from automobiles to airplanes, are finished goods of manufacturing. Nowadays, manufacturing tends to be globalized. Industries use raw materials from all over the world and the manufacturing is partially/fully outsourced into other countries to increase the profit and stay competitive Components of Manufacturing System Manufacturing systems usually refer to large scale industrial production. Wright (1990) proposed that manufacturing inputs, manufacturing processes, and manufacturing outputs are three components of a manufacturing system. Manufacturing inputs include raw material, knowledge, labors, machines and Financing (Wright, 1990).

19 19 Transformation technology and management technology are two types of manufacturing processes. Transformation technology deals with machine or other technical tools (Wright, 1990). Management technology is about research and other management processes. Manufacturing outputs are products, processing waste and so on (Wright, 1990). There is also information flow between manufacturing inputs and manufacturing outputs. Energy is also another input to the manufacturing system. Manufacturing Inputs Raw Materials Knowledge Labors Machines Financing Energy Manufacturing Process Transformation Process Management Process Figure 1.1: Manufacturing system flow (Wright, 1990) Manufacturing Outputs Products Services Waste Classification of Manufacturing System 1 The type of one manufacturing system mostly depends on the its layout. The layout is determined by the production process and production quantity. Evans (1987) proposed four major manufacturing types process layout, fixed layout, group layout and product layout. Based on Evans s research, Süer, Huang & Maddisetty (2010) adopted the name of cellular layout to replace group layout and listed three types of cells in cellular manufacturing, which are in Figure 1.2. Fixed layout is used when heavy 1 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

20 20 products staying in the same position. In these systems, machines, equipment and workers are brought to the product. Product layout deals with product which volume is high and variety is low. Product layout is usually quite efficient but inflexible. Process layout suits for high product variety systems with low product volume. These systems are not very efficient but very flexible. Cellular layout is more flexible than product layout. It is used for low to moderate demand with high product variety. Figure 1.2 shows the classification of four manufacturing layout types. In cellular manufacturing, there are three types of cells DC, SC and RC. A Dedicated Cell (DC) deals with one product family. A Shared Cell (SC) operates two product families, which have relatively similar operations. A Remainder Cell (RC) handles more than two product families. Figure 1.2: Four types of manufacturing layout (Süer, Huang & Maddisetty, 2010) Product layout, cell layout, process layout and fixed layout are presented by the increase of product variety in Table 1.1. When discussing the elements of each layout,

21 degree of automation, system redundancy and degree of dedication all decrease when product variety increases. 21 Table 1.1 The Basic Correlation between the Arrangement of Manufacturing Elements and Type of Manufacturing System (Lenz, 2012) Arrangement of element Product Cell Process Fixed Degree of product variety Low Medium High None Degree of automation Mediumhigh Medium Low Low System redundancy / Number of productioncapable entities High/ Low High-medium/ Medium-high Low/ High Low/ High Degree of dedication between workforce and machines Low-medium Mediumhigh Lowmedium Low Degree of dedication High Medium-high Low Med Cellular Manufacturing System 2 Cellular Manufacturing is based on the grouping of similar products with respect to common processes into one cell. In the real world, many uncertainties exist in the system such as demand uncertainty, supply uncertainty and processing uncertainty. These 2 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

22 22 uncertainties have been discussed in the related research. The uncertainties of product demand and processing times are considered (Süer, Huang & Maddisetty, 2010). By probabilistic market demand calculation, the part-family assignment is achieved (Süer, Huang & Maddisetty, 2010). Then, low utilized cells are grouped to increase the utilization of the system. 1.3 Supply Chain Supply chain is the network connecting suppliers, distribution centers, manufacturers and customers (New & Payne, 1995). When a manufacturing system is expanded, the connection between suppliers and manufacturing facilities becomes complex. Thus, the cooperation and resource exchange between companies become frequent. Therefore, the concept of supply chain is to connect all of the companies or manufacturers together by using quantitative methods. The purpose of supply chain is to satisfy the demand of each stage from raw materials to finished goods Supply Chain Management Supply chain management involves the planning of all activities in the supply chain. Importantly, supply chain management manages the integration between demand and supply. Therefore, supply chain also consists of manufacturing operations supply and marketing demand. Supply chain management developed from traditional business practices. Supply chain focuses on the long-term strategy, which includes all of the suppliers and consumers in the supply chain system as a whole. Designing, controlling and planning all the supply chain activities are used to create net value.

23 Supply Chain Modeling 3 Many supply chain models were discussed (Dicken, 1992). Among them, globally concentrated production model, host market production model and regional/global product specialization model are mentioned. Specifically, each of the geographic regions covers its own demand of that geographic region in the host market production model as shown in Figure 1.3. On the other hand, one manufacturing facility produces all the demand from all over the world in the globally concentrated production model which is shown in Figure 1.4. Another model is regional/global product specialization in which each product family is produced in each facility and may be shared in other regions which is shown in Figure 1.5. Figure 1.3: Host market production model (Dicken, 1992) 3 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

24 24 Figure 1.4: Globally concentrated production model (Dicken, 1992) Figure 1.5: Regional/Global product specialization (Dicken, 1992) Supply Chain Resource Management In supply chain resource management, three types of resource allocations are defined, which are strategic resource management, master resource management, and operational resource management (Süer, Huang & Maddisetty, 2010). In strategic resource management, products are grouped based on types to maintain the customer satisfaction (Süer, Huang & Maddisetty, 2010). For example, one company leaves the high technology products in the North America, and transfers low technology products overseas with low labor costs. By adapting this business strategy, this company is able to satisfy customers with low cost.

25 25 Master resource management focuses on finished products or semi-finished products, while operational resource management studies components and raw materials within individual factories. 1.4 Heuristic Procedure Heuristic procedure is a quick technique to solve complex problems. When problems cannot be solved in an acceptable time frame by using classical optimization techniques, a heuristic procedure produces a solution that is good enough because it does not require a remarkably long time to attain the results. This procedure may or may not achieve the optimal result. Heuristics may be applied independently, or may be adopted in association with optimization algorithms in order to improve their efficiency. The heuristic procedure used in this research is in Section from Süer, Huang & Maddisetty (2010). This heuristic algorithm is applied to assign manufacturing cells to product families in order to minimize the number of cells in each manufacturing facility. 1.5 Simulation Simulation is used for business decisions, engineering decisions, statistical decisions and other decisions. Simulation is a set of modules and conditional links between them by using digital computers to realize stochastic models (Kelton & Law, 2000). At the same time, simulation is a method used to validate theoretical results. An important step of simulation is to carefully state the decision problem in detail. Firstly, objectives are identified, which may vary in different problems and are complex in huge systems. Secondly, system and input data from proposed problems should be

26 26 analyzed. Subsequently, a model is built and verified based on experimental data. Finally, scenarios will be run to generate simulation results. Arena simulation logics include entities, resources, queues, schedules and sets. Entities are used for the input storage, which include input data and data types. Sets define each source in a group, which will be easier to analyze utilization of each source. There are three categories of schedules - arrival schedule, resource schedule and implementing schedule. Arrival schedule is for fixed arriving process, which is different from the resource schedule in which capacity changes over time. 1.6 Justification Based on the research of Ateş (2013), three new extensions are proposed in this research combined new hybrid supply chain strategy, supply chain strategies including setup time, and increasing number of dedicated cells. In supply chain strategies, two new strategies are proposed. Regarding setup time, since setup time is not included in the simulation model of Ateş (2013), some studies that consider setup times are carried out. Finally a simple improvement procedure is proposed to increase number of dedicated cells. Dedicated cells are expected to be more efficient since they only process one product family to avoid setup times between product families. Combined new hybrid supply chain strategy includes two new supply chain strategies which are shown in Figure 1.6 and Figure 1.7. One model is semi-product specialization model in which family-based concentrated approach and product specialization approach are combined. For example, in Figure 1.6, product families 1 and 2 are produced in multiple manufacturing facilities and may be transferred to other

27 27 regions. However products of family 3 are produced in one facility and transfered to all others. Figure 1.7 shows the second model, semi-host market model. In this model, host market model and globally concentrated production are combined, which means some product families are produced in each manufacturing facility to meet the demand of their own region. However, some product families are produced in one facility and transferred to other regions. Figure 1.6: Semi-product specialization model Figure 1.7: Semi-host market model

28 Research Objectives There are three objectives in this research. The first objective is to evaluate the performance of newly proposed supply chain strategies. The demand data are obtained from Ateş (2013) based on a blood sugar strips manufacturer. Three manufacturing facilities China, Ireland and Puerto Rico are considered in the independent supply chain system to cover the demand from three regions Asia, Europe and North America. Demand coverage probability, expected cell utilizations and cell types are considered when analyzing them. Expected cell utilizations are calculated by using statistical methods. Then a heuristic algorithm is used to decide the cell types and conduct cell sharing in each facility. Later, simulation is used to verify the independent design. The second objective is to redesign the manufacturing system by including setup time. In this part, setup time is considered when processing different product families in the simulation experiment. After the simulation experiment, the number of setups in each cell will change the capacity when recalculating expected cell utilizations. By following similar statistical procedures, supply chain design with setup time will be different from the one without considering setup time. Another objective is to increase the number of dedicated cells. When the utilization of one product in the shared cell is already high and the other one is low, the product family with low cell utilization can be moved to other shared cells or remainder cells. By increasing the number of dedicated cells, we simplify planning process and also avoid family setup times. This return may reduce machine and labor costs.

29 Organization of the Research This thesis includes seven chapters. Chapter 1 first briefly introduces manufacturing systems, supply chain management, simulation and heuristic procedure. Then justification and research objectives are discussed. Chapter 2 includes literature reviews of global location strategy in the supply chain area and cellular manufacturing design/group technology. Chapter 3 discusses cell assignments and layered cellular design in manufacturing design. Then four comprehensive supply chain models are discussed based on the supply chain models in Chapter 1. Chapter 4 conducts the statistical calculations for different supply chain models which are discussed in Chapter 3. After the statistical calculations, a heuristic algorithm is implemented to decide the cell types. Independent supply chain system, single manufacturing system and one combined new hybrid supply chain design are discussed. Chapter 5 focuses on two simulation experiments. One is simulation without setup time and the other is simulation with setup time. This study has been carried out for independent supply chain strategy. Chapter 6 implements comparison between different results. First is the comparison of cell utilizations and cell types between the independent supply chain system and single manufacturing system. Then, simulation results are used to verify the results of independent supply chain design. Cell utilizations and cells types are calculated after considering setup time. At last, statistical analysis is conducted. Chapter 7 gives the conclusions and future work of this research.

30 30 CHAPTER 2 LITERATURE REVIEW 4 Literature regarding global location strategy models and manufacturing systems is summarized in this chapter. More specific areas of location strategy inside global supply chain and cellular manufacturing system are reviewed. 2.1 Global Location Strategy Models In global supply chain problems, decision of location strategy is an important part to decide the framework of global supply chain. Many works on location decisions have been proposed in the literature. Many supply chain models were discussed by Dicken (1992). Globally concentrated production model, host market production model and regional/global product specialization model are mentioned in Section Supply chain model includes three logistical drivers and three cross functional drivers (Chopra & Meindl, 2004). Three logistical drivers are facilities, inventory and transportation and three cross functional drivers are information, sourcing and pricing. Many factors affect a sophisticated network of multinational manufacturing facilities (Brush, Maritan & Karnani, 1999). This integrated network includes independent and integrated plant choices. This paper categorized the plant strategies into four groups integrated domestic, integrated foreign, independent domestic and independent foreign. Then, three-stage comparison was performed to determine the plant strategy. The first step was to compare the domestic and international plant determinants, along with comparison of the independent and integrated plant determinants. The second 4 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

31 31 step was to examine independent and integrated plants determinants for either domestic plant or international plant. Similarity, domestic and international plant determinants for either independent plant or integrated plant was examined. Finally, the importance of plant determinants ranking was examined to decide whether they would be changed. Besides considering facility selection, a facility location model was developed to study the location decision of high technology firms (Haug, 1992). The model identified the international manufacturing facility location based on domestic and potential international production markets, which allowed production to be transferred from domestic manufacturing facilities to foreign ones. One important assumption was generated about the unit material cost. Another assumption was transferring products from one manufacturing facility to another at the beginning of one year. The objective function identified the best location selection for production with minimum total manufacturing cost. However, there was no industrial application mentioned in this research. Arntzen, Brown, Harrison and Trafton (1995) introduced production quantities into facility selection. The decision variables were facility selection and production quantities. A unique feature of the model by Arntzen, Brown, Harrison and Trafton (1995) was its ability to combine both cost and time segment in the objective function. Time is measured to be the number of days needed to produce and transport between each functional facility in the supply chain model. Based on the consideration of facility selection and production quantities in the model of Arntzen, Brown, Harrison and Trafton (1995), the model in Vidal and

32 32 Goetschalckx (2001) added consideration of transfer price, transportation mode and allocation of transportation cost. It was an optimal model with a linear objective function, but constraints were both linear and bi-linear. The authors evaluated the heuristic with test problems, but no particular industry was identified as a basis in the creation of the computational examples. Unlike Vidal and Goetschalckx (2001), a two-phase approach including production capacity was developed. This approach incorporated uncertainty about exchange rates and its risk in an international production model (Lowe, Wendell and Hu, 2002). Information technologies were introduced in the global supply chain areas. National market was improved into the global supply chain market by considering connection among global markets (Canel & Das, 2002). The connections between several national markets were considered. One important point of the global decision is the connection between manufacturing and marketing. The manufacturing decision provided low costs and high efficiency. Marketing decision satisfied the customer demand. In order to solve the global manufacturing problems, an integrative mathematical model was developed to connect global manufacturing and marketing (Canel & Das, 2002). This model proved to maximize the total profit among global manufacturing. Cosner (2008) proposed a comprehensive framework to combine manufacturing facility locations, manufacturing system design and pricing strategy. Huang & Süer (2012) proposed one supplier selection framework depending on the demand and performance of suppliers. There were two parts of the framework single period optimization model and multi-period simulation model. Several objectives were

33 33 considered in the single period optimization model, such as minimizing costs, minimizing late delivery rate and minimizing the defect rate. In this model, individual performances of each supplier were generated for single period. Then, the performances are used in a simulation model, in which all the performances of multi-period were known. Ateş (2013) compared two alternative supply chain designs host market production model and globally concentrated production model in a blood sugar strip manufacturing company. Based on Ateş s (2013) research, Wilson (2015) proposed two different supply chain strategies one facility produces and ships product facilities to six regions, and expanding one facility to two. Multi-product and multi-level supply chain designs were implemented by Celikbilek & Süer (2016). Based on Ateş s (2013) research, this thesis expanded two more supply chain designs single manufacturing design and combined new hybrid supply chain design. Both of these two strategies are compared with the host market production model which is independent supply chain design in this thesis in Chapter Cellular Manufacturing Design Group Technology (GT) was introduced to improve productivity in the Cellular Manufacturing System (Hyer & Wemmerlov, 1982). When the total product categories vary and the size of each category is small, GT is used to group the products into product families based on similarities among parts. GT was adopted to decrease the cost and increase the productivity. Rajamani, Singh & Aneja (1990) also mentioned that GT played an important role in cellular manufacturing. They combined part families and machine groups to study the process plans rather than considering them separately.

34 34 Besides GT models, mathematical models were also proposed. A supplementary procedure was proposed to solve the limitation of Adaptive Resonance Theory (ART) (Chen and Cheng, 1995). They mentioned that the performance of ART depended on the initial matrix of bottleneck process. The proposed supplementary procedures could improve the reliability of results. Moreover, a new mathematical model based on cell utilization was conducted (Mahdavi, Javadi, Fallah-Alipour & Slomp, 2007). A comparison of part-machine grouping from this proposed method with the mathematical model from Chen and Cheng (1995) was tested. The model from Mahdavi, Javadi, Fallah-Alipour & Slomp (2007) tended to produce better results. A mixed integrated non-linear model was analyzed for CMS (Bulgak & Bektas, 2009). In this paper, the proposed model was an integrated approach to combine production planning and system reconfiguration. This CMS model was a new model, which includes sequence, duplicate machines, capacity of machines and lot splitting. The literature reviews discussed so far included the deterministic CMS problem. However, cellular manufacturing was difficult to design in the real world due to the uncertainty of the manufacturing process. Similarity coefficients are used to decide the type of machines for cell formation in the cellular manufacturing design. Süer & Ortega (1994) proposed a new similarity coefficient to conduct the type of machines along with the number of machines. Eğilmez, Süer & Özgüner (2012) introduced a hybrid similarity matrix to include both routes based and demand based similarities. Also, they proposed a stochastic mathematical model to design cellular manufacturing system by considering stochastic demand and processing time.

35 35 In order to deal with the uncertainty of product demand along with processing time, another research is proposed (Süer, Huang and Maddisetty, 2010). A heuristic methodology was conducted to distinguish cell types in the CMS - Dedicated Cell (DC), Shared Cell (SC) and Remainder Cell (RC). The product family configuration and cell allocation are accomplished by mathematical analysis. The designed manufacturing system turned to successfully solve the uncertainty of product demand and processing time through simulation method. The methodology conducted by Süer, Huang and Maddisetty (2010) is implemented in the current study for the purpose of designing the manufacturing system given the market demand, part-family formations, and the operations required to process the products. A normal distribution is applied to the demand and processing times to emphasize the effects of probabilistic parameters in the system. The details of the manufacturing system design analysis conducted in this research are presented in Chapter 4.

36 36 CHAPTER 3 PROBLEM DEFINITION In this research, a blood glucose test strip manufacturing system is considered to study the alternative supply chain design approaches, namely independent facilities per region vs. single manufacturing facility and independent facilities per region vs sharing manufacturing resources among various manufacturing facilities located throughout multiple regions. The procedure used to decide shared manufacturing cells is explained in Section 3.2. Subsequently, comparison between independent, single manufacturing and global design with shared cells is conducted in Section Blood Glucose Manufacturing In recent decades, the number of people with diabetes has been growing. Diabetes develops when the body does not make enough insulin or is not able to use insulin effectively, or both. Insulin is a hormone made by beta cells in the pancreas. Diabetic people have high blood sugar. The blood sugar strips are used to monitor the blood sugar level by diabetic people. In 1962, Leland Clark and Champ Lyons developed the first glucose enzyme electrode, which depended on a thin layer of glucose oxidase on an electrode. Home blood sugar monitoring was demonstrated to improve blood sugar control of diabetes in the late 1970s, and the first monitors which are owned by Bayer and Roche were sold for home use around In this research, global supply chain problem regarding the Global Blood Sugar Monitoring market will be discussed. LifeScan Inc., Abbott Labs, Roche Diagnostics, and 5 This paragraph contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

37 37 Bayer Corp. are four main companies which produce blood sugar strips. These companies covered the majority of market share all over the world (Hughes, 2009). Customers from three regions are considered to be the most influential consumer force Europe, Asia and North America. Three manufacturing facilities are assumed to produce the products Ireland, China and Puerto Rico. The production data and manufacturing processes are gathered from Lobo (2006). All of the data are converted into common units by considering market share, revenue, and product price from Ateş (2013). Based on the different number of machines, there are two alternative manufacturing systems in the research fast system and slow system. There are eight or nine operations for each product family depending on different product families. Table 3.1 and Table 3.2 show operation rates (number of vials per minute) for each product family. Table 3.1 Production Rates (per minute) in Fast System (Ateş, 2013) Op1 Op2 Op3 Op4 Op5 Op6 Op Op7 Op8 Op9 Bottleneck F1 80 F F F F

38 38 Table 3.2 Production Rates (per minute) in Slow System (Ateş, 2013) Op1 Op2 Op3 Op4 Op5 Op6 Op Op7 Op8 Op9 Bottleneck F1 57 F F F F Manufacturing Cell Design 6 In order to decrease the production cost, many companies turn to the global manufacturing systems. In most manufacturing systems, different products require to be processed on different machines. Due to high product variety, products are grouped into several families based on similarity as proposed by the cellular manufacturing. Table 3.3 shows an example of product-machine incidence matrix. In this table, 1 in row i and column j indicates that product i needs to be produced on machine j. For example, Product 1 (P1) is processed on Machine 1 (M1), Machine 2 (M2) and Machine 3 (M3). 6 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

39 39 Table 3.3 An Example of Product-Machine Incidence Matrix (Ateş, 2013) M1 M2 M3 M4 M5 P P2 1 1 P3 1 1 P4 1 1 P5 1 1 P P7 1 1 Some products need similar machines. One can observe that products with similar manufacturing processes are grouped together. Table 3.4 shows families and their assignment in cellular manufacturing. For example, Product 1 (P1) is processed on Machine 1 (M1), Machine 2 (M2) and Machine 3 (M3). P2 and P3 are processed on M1, M3 and M2, M3, respectively. These three products are all included in the same family and the corresponding cell contains machines M1, M2 and M3. Table 3.4 Product Families and Cells Family Products Cell Machines in the Cell F1 P1, P2, P3 Cell1 M1, M2, M3 F2 P4,P5 Cell2 M4, M5 F3 P6,P7 Cell3 M1, M3, M5 In this study, it is assumed that cells are self-sufficient, in other words, all operations of a family are completed in their cell.

40 40 Table 3.5 shows how product families are allocated to different cells. For example, Product Family 1 (F1) is processed in Cell 1 (C1), and the other two product families are manufactured in Cell 2 and Cell 3, respectively. Table 3.5 Family vs. Cell Assignment C1 C2 C3 F1 1 F2 1 F3 1 However, in real life manufacturing systems, some product families may have quite high demand, which means one cell is not sufficient to meet the demand. Table 3.6 shows this multiple cell production system. For example, due to high demand, product families 1, 2 and 3 may need 2, 3 and 2 cells, respectively. Table 3.6 Family vs. Multiple Cells due to High Demand C1 C2 C3 C4 C5 C6 C7 F1 1 1 F F3 1 1 Yet another possibility is that demand values for product families follow a stochastic distribution. In some cases, expected utilization for some cells of families may be low. As a result, several product families may be expected to share one cell. In

41 41 general, dedicated Cell, Shared Cell and Remainder Cell may become necessary. A Dedicated Cell (DC) deals with one product family. A Shared Cell (SC) operates two product families, which have relatively similar operations. A Remainder Cell (RC) handles more than two product families. Both Shared Cells and Remainder Cells usually handle product families that have medium or low expected utilization values for some of its cells. Table 3.7 shows the cell sharing among three product families. For example, Cell 1 (C1) is Dedicated Cell for Product Family 1(F1). C2 is also Dedicated Cell for F2. C3 is a Remainder Cell to be shared by F1, F2 and F3. Finally, C4 is a Shared Cell between F2 and F3. Table 3.7 Layered Cellular Design due to Stochastic Demand C1 C2 C3 C4 F1 1 1 F F3 1 1 (DC) (DC) (RC) (SC) As mentioned in Figure 1.2, Cellular Manufacturing layout includes three types of cells Dedicated Cell (DC), Remainder Cell (RC) and Shared Cell (SC). In this research, we will deal with stochastic demand in multiple regions, namely North America, Asia and Europe.

42 Alternative Supply Chain Designs 7 In this section, three alternative supply chain design strategies will be discussed. Strategy 1 discusses the independent supply chain design which means the manufacturing facilities produce products independently in each region, namely North America, Asia and Europe by using three manufacturing facilities located in Puerto Rico, China and Ireland, respectively. Strategy 2 is the single location manufacturing system in which all of the products are produced in one location. Strategy 3 and Strategy 4 adopt sharing strategies, in which manufacturing resources are shared among three regions Strategy 1: Independent Supply Chain Design In this strategy, each region produces many types of products to meet its own demand. Products are produced independently in different facilities, which eliminates the need for transportation and information sharing between different regions. Figure 3.1 shows that the blood sugar strips are produced in three manufacturing facilities China, Ireland and Puerto Rico. The number 1 means corresponding Product family is allocated to the designated cell in that manufacturing facility. 7 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

43 43 Figure 3.1: Independent manufacturing systems Strategy 2: Single Manufacturing Facility Design In this strategy, one single manufacturing facility produces all the products. The location analysis of this facility is not within the scope of this study.

44 Strategy 3: Global Supply Chain Design with Shared Cells In the previous strategy, some cells may have high utilization. However, the low utilization cells may also exist in these facilities. It is important to improve the design for such cells. From this perspective, some product families can be combined together from different manufacturing facilities to process them together in one facility. Figure 3.2 shows that cell 4A and cell 5A in China facility have lower utilization. Same as cell 5B in Ireland, and cell 5C in Puerto Rico. Due to the low weight and volume of the product, transportation cost is less than cell operation cost. Based on this, cell 5C in Puerto Rico facility is merged into Cell4A in the China facility. Figure 3.2: Global supply chain design with sharing cells In this strategy, each family is produced in each facility. However, some of production may also take place on other facilities to lower manufacturing costs by

45 45 sharing cells. Obviously, this will lead to transportation costs. This has to be considered in the final evaluation. Cell 4A in China facility and Cell 5C in Puerto Rico can safely be combined since both cells process families 1 and 2. Similarly, Cell 5B in Ireland facility can be combined with Cell 5A in China facility as they process families 2 and 3. The next step is to allocate the combined cells to facilities considering both manufacturing and transportation costs Strategy 4: Global Supply Chain Design with Family Centralized Cells Figure 3.3 shows that each manufacturing facility has a dedicated cell which has low utilization. Cell 3A in China facility, Cell 5B in Ireland facility and Cell 3C in Puerto Rico facility produce product family 3. These three cells can be combined into one cell due to low utilization. Once again, we need to determine where that cell will be placed and this family will be produced. Figure 3.3: Global supply chain design with family centralized cells

46 46 In this case, one family (F3) will be produced only in one facility while others will be produced in all facilities. Assuming F3 will be produced in China facility, Figure 3.4 presents the new manufacturing cells and cell utilizations in each region. We can find cell utilization of Cell 3 in China facility is 0.71 ( ). Figure 3.4: Merged global supply chain design with family centralized cells

47 47 CHAPTER 4 METHODOLOGIES USED FOR SUPPLY CHAIN DESIGNS There are three supply chain strategies shown in Figure 4.1 Independent supply chain strategy, single manufacturing system design and combined new hybrid supply chain design. Independent supply chain strategy and single manufacturing design are similar about calculating demand coverage probability and expected cell utilization by following the flowchart in Figure 4.1. Combined new hybrid supply chain design is based on the expected cell utilization of independent supply chain strategy. Total products are divided into five different product families based on the similarity of product. In this study, product families are assumed to be already identified. Independent supply chain strategy is presented in Section 4.1 and Section 4.2. Expected cell utilizations are calculated by using cell capacity, product demand, and so on in Section Each cell capacity is assumed 2000 hours annually. Cell utilization captures the usage of each cell. By considering the cell utilization, different cells can be combined into one as long as capacity is available (Section 4.1.4). Single Manufacturing design is discussed in Section 4.3. Two newly proposed hybrid supply chain designs are combined into one since the data are limited which is shown in Section 4.4. This supply chain design is named combined new hybrid supply chain design. Section 4.5 discusses one attempt to increase the number of dedicated cells in order to increase the efficiency in the cell. 8 8 This paragraph contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

48 48 Independent Supply Chain Strategy (Section 4.1) Single Manufacturing System Design (Section 4.3) Identifying Mean Demand & Standard Deviation Identifying Mean Demand & Standard Deviation Calculating Capacity Requirements Calculating Capacity Requirements Computing Probability of Demand Coverage Computing Probability of Demand Coverage Combined New Hybrid Supply Chain Design (Section 4.4) Computing Expected Cell Utilization Computing Expected Cell Utilization Completing Design Completing Design (Section 4.2) Completing Design Evaluating Strategies (Chapter 6) Figure 4.1: Methodology flowchart

49 Independent Manufacturing Facilities 9 In this section, the methodology used for designing manufacturing facilities is discussed in detail Mean Capacity Requirements and Standard Deviation Historical demand values of four companies Roche, LifeScan, Bayer and Abbott from 2002 to 2010 are used to calculate the 2011 demand in the research by Ateş (2013). In this research, it is assumed that demand is normally distributed. Standard Deviation (σ) values for each product family are generated as a percentage of the mean demand (20% - 25%) as shown in Table 4.1. It is assumed that if Mean Demand is high, standard deviation is lower. From Ateş (2013), mean demand by family in all markets is shown in Table 4.2. Also, Standard Deviations (σ) in three regions are calculated in Table 4.2. Table 4.1 Percentage to Arrange Standard Deviation with Mean Demand Mean (10 3 ) Percentage(%) 0 5, ,000 10, ,000 15, ,000 20, ,000 25, , This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

50 50 Table 4.2 Mean Demand and Standard Deviation by Family in Three Regions China Ireland PR F Mean % STDEV Mean STDEV Mean STDEV 1 1,422, ,571 1,422, ,731 1,337, , ,098, ,703,565 7,101,377 1,704,330 6,672,986 1,601, ,711, ,610,741 6,714,438 1,611,465 6,309,389 1,514, ,313, ,105,784 22,000,00 4,620,000 22,856,82 4,799, ,137, ,363 3,138, ,715 2,949, ,377 Having mean values and standard deviations, the mean capacity requirements by product family are calculated by using Equation 4.1 in Maddisetty (2005). Bottleneck Processing Time is defined by the bottleneck machine as the longest processing time in the cell. ( 4.1 ) In this study, both slow and fast manufacturing systems were considered. For example, Mean Capacity Requirements for Product Family 1 in Fast manufacturing system of China region is decided by Mean Demand of Product Family 1 in China region which is 1,422,286. BPT (Bottleneck Processing Time) is 1/80 = min for fast Manufacturing system in China region. (This is also the µ) (This is also σ) The results of Mean Capacity Requirements and Standard Deviation for different regions are shown in Table 4.3 and Table 4.4. The difference between fast and slow

51 system which is shown in Section 3.1 depends on the number of duplicate machines which is discussed later in details in Section Table 4.3 Mean Capacity Requirements and Standard Deviation in Fast System China Ireland PR Family MCR STDEV MCR STDEV MCR STDEV pit Table 4.4 Mean Capacity Requirements and Standard Deviation in Slow System China Ireland PR Family MCR STDEV MCR STDEV MCR STDEV pit Demand Coverage Probabilities The demand coverage probability shows the probability that a given number of cells will meet the demand. In this paper, the number of cells to process the particular family of products is unknown. At the same time, demand is assumed to follow the normal distribution. The demand coverage probability is assumed to be 99.99% in this research. It could be other values as well. The annual work time in one cell is 2000 hrs. Mean Capacity Requirement (MCR) is calculated in Section Demand Coverage

52 Probability (DCP) for a family and cell combination is calculated by Equation 4.2 in Maddisetty (2005). 52 ( 4.2 ) For Cell 1 of Product Family 1 in Fast system for the China region, Demand Coverage Probability for a given number of cells is decided by Mean Capacity Requirement and Standard Deviation. Mean Capacity Requirement for Product Family 1 in Fast system for China region is 296 which is shown in Table 4.3. Standard Deviation for Product Family 1 in Fast system for China region is 74, which is also shown in Table 4.3. Based on these values, the Demand Coverage Probability for the first cell is 99.99%. In other words, only one cell is sufficient to cover demand almost fully for Family 1. ( ) All the results of Demand Coverage Probabilities for different regions are shown in Table 4.5 and Table For family 2, one cell will cover demand 93% of the time. By adding a second cell, the Demand Coverage Probability jumps to 99.99%. Table 4.5 Demand Coverage Probabilities of Fast System for China Region China & Fast Cell Family

53 53 Table 4.6 Demand Coverage Probabilities of Slow System for China Region China & Slow Cell Family Table 4.7 Demand Coverage Probabilities of Fast System for Ireland Region Ireland & Fast Cell Family Table 4.8 Demand Coverage Probabilities of Slow System for Ireland Region Ireland & Slow Cell Family

54 54 Table 4.9 Demand Coverage Probabilities of Fast System for Puerto Rico Region PR & Fast Cell Family Table 4.10 Demand Coverage Probabilities of Slow System for Puerto Rico Region PR & Slow Cell Family Expected Cell Utilizations Expected Cell Utilization is determined by using Demand Coverage Probability, Mean and Standard Deviation from Equation 4.3 to Equation 4.6 by Süer & Ortega (1998). (4.3 ) where E(C=X) P (CR>X) Expected cell utilization for Xth cell in a product family Probability that the number of cells required is greater than X

55 55 PU 1 Percentage utilization of Xth cell when CR > X, PU 1 =1.0 P (X-1 CR X) PU 2 Probability that CR between X-1 and X Percentage utilization of Xth cell when CR between X-1 and X P(CR<X-1) Probability that CR < X-1 PU 3 Percentage utilization of Xth cell when CR < X-1, PU 3 = 0.0 PU 2 is solved by Equation 4.4. ( 4.4 ) where y f(y) A Variable representing CR Probability density function for CR Probability that CR between X-1 and X f(y) and A are calculated by Equations 4.5 and 4.6. ( 4.5 ) ( 4.6 ) Since the distribution for demand is assumed to be normal distribution, the probability density f(y) will follow a normal distribution. For example, in the fast system for China region, Expected Cell Utilization of Cell1 in Product Family 4 is determined by three segments. The first segment is the probability that the required number of cells greater than 1. In this case, cell utilization is 100 percentage. The second segment is the probability that the required number of cells between 0 and 1 multiplied by the corresponding percentage utilization of the first cell.

56 56 The third segment is the probability that the required number of cells is less than 0 multiplied by 0. The following two equations show the calculations of the percentage utilization of the first cell when CR between 0 and 1 and expected cell utilization. Similar calculation is done in the following equation for the second cell. All the results of Expected Cell Utilizations for different regions are shown from Table 4.11 to Table 4.16.

57 57 Table 4.11 Expected Cell Utilizations in Fast System for China Region China & Fast Cell Family Table 4.12 Expected Cell Utilizations in Slow System for China Region China & Slow Cell Family Table 4.13 Expected Cell Utilizations in Fast System for Ireland Region Ireland & Fast Cell Family

58 58 Table 4.14 Expected Cell Utilizations in Slow System for Ireland Region Ireland & Slow Cell Family Table 4.15 Expected Cell Utilizations in Fast System for Puerto Rico Region PR & Fast Cell Family Table 4.16 Expected Cell Utilizations in Slow System for Puerto Rico Region PR & Slow Cell Family As one may observe, the slow system design needed the same or more cells than the fast system design as expected.

59 Heuristic Algorithm for Layered Cellular Design After computing Expected Cell Utilization, Dedicated Cells (DC), Shared Cells (SC), and Remainder Cells (RC) are identified. The heuristic algorithm is used for identifying cell types (Süer, Huang and Maddisetty, 2010). When all the Expected Cell Utilizations in three regions are calculated in Section 4.1.3, manufacturing cell types are determined by following the heuristic algorithm. Tabulate the expected cell utilization (ECU) and similarity coefficient values Sort ECU in the decreasing order Consider a coverage segment with highest ECU Is ECU = 100%? Yes Allocate a cell to the family No No Is ECU >=50% No Decrease ST & consider existing cells for possible assignment (if capacity is avaliable) Yes No Allocate a cell to the family Are all coverage segments assigned? No Remaining segments from a remainder cell No Can the cell be utilized by other coverage segments? Yes Set similarity threshold (ST) & compare similarity between original family allocated to cell and other families Yes Allocation of families to cells is complete Is the similarity coefficient between the family/families >= ST? Yes Yes Assign the family to the cell similar to the original family Are there any other segments for possible assignment? Figure 4.2: Layered cellular design consisting dedicated, shared and remainder cells (Süer, Huang, & Maddisetty, 2010)

60 60 Expected Cell Utilizations are sorted in decreasing order with the highest Expected Cell Utilization considered. If the Expected Cell Utilization is 100%, this cell is considered to be a Dedicated Cell (DC). If the Expected Cell Utilization is larger than 50%, a cell will be allocated to a product family. Then other similar product families are allocated to the cell to make the cell coverage close to 100% by considering similarities among families. These cells are named Shared Cells (SC) if they process only two product families. If the Expected Cell Utilization is smaller than 50% and cannot be merged with existing cells, these cells will be grouped together to form a Remainder Cell (RC). Typically, Remainder Cells will process three or more product families. Table 4.17 Similarity Coefficients between Product Families (Ateş, 2013) Family Table 4.17 shows the similarity of machines and operations between different product families. The threshold value is the lowest acceptable similarity coefficient that allows two families to be grouped in a cell. The Similarity Threshold is set to 77% in this research. For example, when sorting the ECU in the fast manufacturing system for China region, the highest ECU is 100% of Product Family 4 in Cell 1. Then Product family 4 will be allocated to Cell 1. When second cell with 0.96% utilization considered, it is allocated to a new cell Cell 2. Table 4.17 will be used to search the other similar

61 61 Product Families with Product Family 4. From Table 4.17, Families 1, 2 and 5 can be considered to share a cell with Product Family 4 since merging this cell with family 1 and Family 5 will exceed 100% utilization. The only option is to merge Cell 2 (1% utilization) of Family 2 with Family 4. Table 4.18 Cell Type in Fast System for China Region China & Fast Cell Family (DC) 1 (SC) 2 (DC) 3 (RC) 4 (SC) Table 4.19 Cell Type in Slow System for China Region China & Slow Cell Family (DC) 1 (DC) 2 (DC) 3 (SC) 4 (DC) 5 (SC) 6 (RC) In China Fast case, there are two Dedicated Cells, two Shared Cells and one Remainder Cell. Similar distributions occur in all cases.

62 Supply Chain Strategy 1: Host Market Production Strategy In this section, the results of layered cellular design is discussed for host market production strategy, namely each facility builds products for their markets, namely Ireland & Puerto Rico is addition to results given for China in the previous section. Similar designs happen in Ireland and Puerto Rico regions. Regarding to fast manufacturing system, all three facilities have totally 5 cells 2 dedicated, 2 shared and 1 remainder cell. Table 4.20 Cell Type in Fast System for Ireland Region Ireland & Fast Cell Family (DC) 1 (SC) 2 (DC) 3 (RC) 4 (SC) Table 4.21 Cell Type in Slow System for Ireland Region Ireland & Slow Cell Family (DC) 1 (DC) 2 (DC) 3 (DC) 4 (DC) 5 (SC) 6 (RC)

63 63 Table 4.22 Cell Type in Fast System for Puerto Rico Region PR & Fast Cell Family (DC) 1 (SC) 2 (DC) 3 (RC) 4 (SC) Table 4.23 Cell Type in Slow System for Puerto Rico Region PR & Slow Cell Family (DC) 1 (DC) 2 (DC) 3 (DC) 4 (DC) 5 (SC) 6 (RC) Supply Chain Strategy 2: Single Manufacturing System Design 10 In this section, all the processes are similar with Section 4.1. The total mean demand values are presented in Table Standard deviation values are calculated based on standard deviation values from different regions which is shown in Equation 4.7. ( 4.7 ) 10 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

64 64 Table 4.24 Mean Demand and Standard Deviation for Single System Family Mean Demand STDEV Demand 1 4,182, , ,872,551 2,893, ,735,250 2,735, ,170,087 8,393, ,225,828 1,332,189 Table 4.25 and Table 4.26 show demand coverage probabilities and expected cell utilization values for single manufacturing system. Table 4.25 Demand Coverage Probabilities of Fast System in Single Manufacturing Design Total & Fast Cell Family Table 4.26 Expected Cell Utilizations of Fast System in Single Manufacturing Design Total & Fast Cell Family

65 65 Table 4.27 shows cell type for single manufacturing design after heuristic algorithm. In this design, the total number of cells turns out to be 14 which is less than the sum of three independent supply chain designs. Table 4.27 Cell Type for Fast System in Single Manufacturing Design Total & Fast DC 1 DC 2 DC 3 DC 4 DC 5 DC 6 DC 7 DC 8 DC SC 10 SC 11 SC 12 RC 13 SC Combined New Hybrid Supply Chain Strategy There are two supply chain strategies mentioned in Section 3 dealing with sharing cells, which are Strategy 3 global supply chain design with shared cells and Strategy 4 Global Supply chain design with family centralized cells. This section combines these two strategies to deal with cell combination in the fast system for three regions. As shown in Table 4.28, cell utilization of Family 4 in Cell 5 for China region is In Table 4.30, cell utilization of Family 4 in Cell 5 for Puerto Rico region is These two cells can be combined into one cell in China facility based on strategy 2 in the following Table Meanwhile, due to low cell utilization, all of the Family 5 in China, Ireland and Puerto Rico facility, which is shown in Table 4.28, Table 4.29 and Table 4.30, can be combined into one cell (Cell 6) in China facility based on strategy 3 in Table 4.30.

66 66 Table 4.28 Cell Arrangement in Fast System for China Facility China & Fast Cell Family (DC) 1 (SC) 2 (DC) 3 (RC) 4 (DC) 5 (DC) Table 4.29 Cell Arrangement in Fast System for Ireland Facility Ireland & Fast Cell Family Table 4.30 (DC) 1 (SC) 2 (DC) 3 (RC) 4 (DC) Cell Arrangement in Fast System for Puerto Rico Facility PR & Fast Cell Family (DC) (SC) (DC) (RC)

67 Maximizing Number of Dedicated Cells In order to maximize the number of Dedicated Cells, an improvement procedure is applied to the results obtained in Section 4.1. For example, in Cell 2 of China region, Product Family 2 is shared with Product Family 4. Since Cell 2 for Product Family 4 is already in high utilization, there is no need to share it with Product Family 2, which increases the setup times needed in the cell. This will adversely affect capacity utilization. There is a danger that capacity is exceeded in that cell. From this perspective, Product Family 2 moves to Cell 5 since Cell 5 is a Shared Cell. Table 4.31 shows the change in Fast System for China region. The same procedure is applied in Table 4.32 and Table 4.33 for other regions. This design increased the number of dedicated cells and consequently led to higher number of remainder cells. Table 4.31 Cells in Fast System for China Region China & Fast Cell Family (DC) 1 (DC) 2 (DC) 3 (RC) 4 (RC)

68 68 Table 4.32 Cells in Fast System for Ireland Region Ireland & Fast Cell Family (DC) 1 (DC) 2 (DC) 3 (RC) 4 (RC) Table 4.33 Cells in Fast System for Puerto Rico Region PR & Fast Cell Family (DC) 1 (DC) 2 (DC) 3 (RC) 4 (RC)

69 69 CHAPTER 5 SIMULATION EXPERIMENT In this chapter, two simulation models are developed for fast manufacturing systems in China region. Section 5.1 presents details of the simulation model which does not consider setup time. Section 5.2 discusses the simulation model which considers setup time. Both simulation models are run 10 replications. 5.1 Simulation Experiment without Setup Time 11 In this section, a simulation model for fast system in China region is developed to compare the results with the Expected Cell Utilization in Section The running time is assumed to be 2000 hours in a year. So the total time is 2400 hours by considering 400 hours as warm up periods. Figure 5.1 shows the sub-model of Assign Product Families to Cells, in which each product family arrives in certain inter-arrival time (Table 5.1) calculated from mean capacity by Equation 5.1. Before assigning product families into cells, product families are held until they are grouped into three units which is shown in Figure 5.5. Lot sizing is important when considering setup time in Section 5.2. After decision modules, product families are assigned to different cells. Figure 5.2 shows the sub-model of cell processes. In each cell, product families have several operations processed in different machines. The number of machines and processing times on each machine are included based on different product family types. Table 5.2 presents the operation time of each operation in the Fast system. Figure 5.3 shows the sub-model of Product Families Dispose, in which product families leave the system when they finish all the operations. 11 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

70 70 F1Enter BatchF1 AssignF1Cell4 0 0 F2Enter BatchF2 DecideF2 0 Tru e AssignF2Cell Fa l s e AssignF2Cell2 F3Enter 0 BatchF3 0 DecideF3 0 Fa l s e 0 Tru e AssignF3Cell3 RouteToCells AssignF3Cell4 F4Enter BatchF4 DecideF4 AssignF4Cell1 0 0 El s e NQ(Qu e u e 6 1 )< =1 0 0 NQ(Qu e u e 6 2 )< =1 0 0 NQ(Qu e u e 6 5 )< =1 0 0 AssignF4Cell2 AssignF4Cell5 AssignF4Cell4 F5Enter BatchF5 AssignF5Cell5 0 0 Figure 5.1: Arena sub-model of assign product families to cells 0 EnterCell SeparateFs StartProcess Process Batch End Figure 5.2: Arena sub-model of cell processes exitcell Separate NumOut Figure 5.3: Arena sub-model of product families dispose Leave 0

71 71 ( 5.1 ) Table 5.1 Inter-arrival Time for Each Product Family Mean Demand Family (thousands) Inter-arrival Time (mins) Table 5.2 Operation Times per Thousand Vials in Fast System Op Op Op Op Op Op Op Op Op9 Bottleneck (Mins) Family Famiy 12.5 Famiy Famiy 12.5 Famiy The module of F2Create decides the inter-arrival time of product family 2. Figure 5.4: F2Create module

72 72 BatchF2 combines several number (Batch Size) of F2 together then assigns them to the cell. Figure 5.5 shows how to define the variable of BatchSize. Unit transfer was converted 1000 vials to be able deal with memory issues in simulation software. Figure 5.5: BatchF2 module Figure 5.6 shows DecideF2 module and decides in which cell F2 will be processed, which is based on the queue size in cell4. All the definitions of queue size are attached in Appendix C.

73 73 Figure 5.6: DecideF2 module Figure 5.7 shows how to define module AssignF2Cell4. When product family 2 is assigned to Cell 4, it will wait in the queue of cell 4 which is named F2Sequence4. In the Cell4, there are 8 stations. One machine processes the product families with different operation times. Figure 5.7: AssignF2Cell4 module

74 The module of EnterCell in Figure 5.8 stores all the stations and machines in the manufacturing system which totals 43 stations and 43 machines. 74 Figure 5.8: EnterCell module After entering the cells, grouped product families will be separated to be processed as shown in Figure 5.9. Figure 5.9: SeparateFs module

75 Different machines will be used to produce specific product families as shown in Figure 5.10 and Figure Figure 5.10: StartProcess module Figure 5.11: Process module

76 76 The results of the Arena model are presented in Table 5.3. Ten replications are run in each experiment. Figure 5.12 shows that value of working in process in the simulation experiment trends to be stable, which means the number of warm up periods is reasonable. The comparison of average Number Out and Mean capacity is shown in the same table. The maximum deviation is 2.7%, which indicates that demand in China region is covered reasonably well. Figure 5.12: Value of working-in-process trend Table 5.3 Simulation Results vs. Mean Demand of Fast System in China Region Family Number Out (K) Mean Demand (K) Deviation (%) F F F F F Cell utilization is also an important index in a manufacturing system. Table 5.4 shows the comparison of average cell utilizations in the simulation model.

77 77 Table 5.4 Simulation Cell Utilizations of Fast System in China Region Cell Simulation Cell Utilization Expected Cell Utilization Deviation (%) Cell Cell Cell Cell Cell

78 Simulation Experiment with Setup Time So far, setup time has been ignored. Setup time exists in the real world manufacturing systems between producing different products. The total productive capacity will decrease when setup times exist. The simulation results with setup time will be used to provide number of setups for the statistical calculations (Expected Cell Utilizations). In the first step, simulation model will be run to meet the mean demands by considering setup time (20 mins) in the system. The number of setups for different product families in each cell will be used for calculating Expected Cell Utilizations in the second step. Entry and dispose sub-models are same as discussed in Section 5.1. Only one submodel of cell processes is different as shown in Figure When the product family type processed is different from the previous type, setup time is considered. 0 Tr ue 0 EnterCell Setup? StartProces s Proc e s s Sa v e L a s t Se pa ra tefs Ba tc h End En tity Ty pe 0 False 0 0 SetupSeiz e Se tu p Re c o rd Setu p 0 Figure 5.13: Arena sub-model of cell processes with setup time Before processing product families, if entity type is different from last entity, setup will happen in each machine as shown from Figure 5.14 to Figure 5.16.

79 79 Figure 5.14: Setup? module Figure 5.15: SetupSeize module Figure 5.16: Setup module

80 RecordSetup module is used to record the number of setups in order to apply these numbers to calculate expected cell utilization. 80 Figure 5.17: RecordSetup module Figure 5.18 shows how to store the last entity type. Figure 5.18: SaveLastEntityType module In this simulation model, in order to increase the accuracy of the result, 400 hours warm up period is adopted. Since more time is used for setup,more capacity (2080 hours) is set for running this simulation model. Total running time turns to be 2480 hours. Also, in order to match the NumberOut with mean demand, inter-arrival time of five entities needs to be modified as shown in Table 5.5. The comparison of NumberOut and mean demand is shown in Table 5.6. The number of setup is presented in Table 5.7.

81 81 Table 5.5 Inter-arrival Time for Entities with Setup Time of Fast System in China Region Family Inter-arrival Time F F F F4 2.1 F Table 5.6 Simulation Results vs. Mean Capacity of Fast System in China Region Family Number Out (K) Mean Demand (K) ǀ Deviation ǀ (%) F F F F F Table 5.7 Number of Setups of Fast System in China Region Family Cell Setup F1 C4 192 F2 C4 9 C2 14 F3 C3 33 C4 16 C1 8 F4 C2 206 C5 11 C4 5 F5 C5 14

82 82 CHAPTER 6 RESULTS In this chapter, five sections are included. The cell comparison among independent supply chain design, single manufacturing system design and combined new hybrid supply chain strategy in section 6.1. Expected cell utilizations before considering setup times are compared with simulation results in section 6.2. Expected cell utilization calculations after considering setup times are presented in section 6.3. Statistical analysis between independent supply chain design and single manufacturing design is discussed in section 6.4. Cost analysis is conducted among three supply chain strategies in section Cell Comparison 12 The cell comparison among three designs is presented in Table 6.1. It shows single manufacturing facility can produce product families more efficiently which means cells have higher utilization in single manufacturing system compared to multiple independent plants and combined new hybrid supply chain approach. Seven parameters are compared among these three approaches. The number of dedicated cells is higher in combined new hybrid supply chain and single manufacturing design. The number of total cells in single manufacturing system is less than other two designs. Transportation fees are included in the cost analysis, which has important influence in the final decision. 12 This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

83 83 Table 6.1 Comparison among Independent Supply Chain, Combined New Hybrid Supply Chain and Single Manufacturing System Independent Supply Chain China Ireland P.R Total Combined New Hybrid Supply Chain Single # of DCs # of SCs # of RCs # of total cells # of Cell Util > # of workers # of machines Comparison of Cell Utilization Values: Theoretical vs. Actual 13 The comparison between theoretical cell utilization and actual simulation values is discussed in this section. Only fast manufacturing of China facility is conducted. Table 6.2 shows the queue size to arrange each product family in different cells. Also Table 6.2 shows the comparison between simulation cell utilizations in Table 5.4 and expected cell utilizations in Section The maximum deviation is around 5.0%, which indicates that simulation model realizes expected results reasonably well. The example of queue size in product family 4 is shown in Figure This section contains adapted writing directly from a conference paper written by the thesis author, Jue Jiang (Jiang & Süer, Alternative Global Supply Chain Design Strategies for A Blood Sugar Strip Manufacturer Considering Layered Cellular Design, 2016)

84 84 Table 6.2 Queue Size with Comparison of Cell Utilizations of Fast System in China Region Cell Simulation Cell Utilization Expected Cell Utilization Deviation (%) F F F F F Figure 6.1: Queue Size Arrangement in Family 4 When finding the optimal queue in the simulation experiment, many attempts have been tried. Table 6.3 lists one of these attempts which shows more gap between simulation cell utilization and expected cell utilization in product family 4. Table 6.3 Another Queue Size with Comparison of Cell Utilizations of Fast System in China Region Cell Simulation Expected Deviation Cell Cell (%) Utilization Utilization F F F F F

85 Cell Utilizations with Setup Time With the similar methodology used in Section 4.1, part of the calculations are presented in Appendix B. Mean Capacity Requirements and Standard Deviation of fast system for China region stays the same in Table 4.3. All the results are presented from Table 6.4 to Table 6.6. The results show that the cell utilizations are higher in the strategy before considering setup times. Table 6.4 Comparison of Demand Coverage Probabilities of Fast System for China Region With Setup Times Without Setup Times Cell Family Table 6.5 Comparison of Expected Cell Utilizations in Fast System for China Region With Setup Times Without Setup Times Cell Family

86 86 Table 6.6 Comparison of Cell Type in Fast System for China Region With Setup Times Without Setup Times Cell Family (DC) 1 (SC) 2 (DC) 3 (RC) 4 (SC) 5 (DC) 1 (SC) 2 (DC) 3 (RC) 4 (SC) Statistical Analysis Statistical Analysis of WIP for Individual Facilities After running the simulation models, the number of work in process (WIP) is recorded of each manufacturing facilities in Table 6.7. There are 10 replications in each simulation experiment. There is one hypothesis that the mean values of working-inprocess in three manufacturing facilities are same which is shown in Equation 6.1. ( 6.1 )

87 87 Table 6.7 Working-in-Process in Each Manufacturing Facility Replication China Ireland PR By using One-way ANOVA in Minitab, means and analysis of variance are conducted in Table 6.8 and Table 6.9. Boxplot of three facilities is shown in Figure 6.2. As shown in Table 6.8, p-value is almost which is less than significance level 0.05, it means the null hypothesis is rejected. The conclusion is that working in process levels in these facilities are not same. Table 6.8 Means for WIP in Each Manufacturing Facility Factor N Mean StDev 95% CI China (336.1, 426.7) Ireland (217.26, ) PR (194.88, )

88 Data 88 Table 6.9 Analysis of Variance Source DF Adj SS Adj MS F-Value P-Value Factor Error Total Boxplot of China, Ireland, China Ireland PR Figure 6.2: Boxplot of three facilities In order to classify three facilities, Fish s text is conducted and the results are shown in Table 6.10 and Figure 6.3. From the statistic results in Figure 6.3, the difference of mean values between Ireland and Puerto Rico facility is small (near 0), which means these two facilities can be classified into one group.

89 89 Table 6.10 Fisher s Test Factor N Mean Grouping China A Ireland B PR B Figure 6.3: Fisher s plot of three facilities Statistical Analysis of WIP between Independent Design and Single Design Since the comparisons of cells between independent manufacturing design and single manufacturing system are discussed in Section 6.1, this section will analyze the WIP of them. Table 6.11 shows working-in-process (WIP) of the total value of three independent facilities along with WIP of single manufacturing system. The hypothesis is that mean values of working-in-process in independent manufacturing facilities design and single manufacturing system are same which is shown in Equation 6.2. ( 6.2 )

90 90 Table 6.11 Working-in-Process in Two Manufacturing System Replication Sum Single Means and analysis of variance are also conducted in Table 6.12 and Table Boxplot of two facilities is shown in Figure 6.4. Since p-value is less than significance level 0.05, that leads to rejecting the null hypothesis. In other words, mean working-inprocess inventory levels of both systems are not same. Table 6.12 Means for WIP in Each Manufacturing Facility Factor N Mean StDev 95% CI Sum (821.6, 929.3) Single (315.67, )

91 Data 91 Table 6.13 Analysis of Variance Source DF Adj SS Adj MS F-Value P-Value Factor Error Total Boxplot of Sum, Single Sum Single Figure 6.4: Boxplot of two facilities Statistical Analysis of Average Flow Time In this section, statistical analysis of average flow time is discussed. Table 6.14 shows average flow time in four facilities China, Ireland, PR and Single manufacturing system. There are 10 replications in each simulation experiment. There is one hypothesis that the average flow time in four manufacturing facilities are same which is shown in Equation 6.3. ( 6.3 )

92 92 Table 6.14 Average Flow Time (Mins) in Each Manufacturing Facility Family China Ireland PR Single By using One-way ANOVA in Minitab, means and analysis of variance are conducted in Table 6.15 and Table Boxplot of four facilities is shown in Figure 6.5. As shown in Table 6.16, p-value is which is larger than significance level 0.05, it means the hypothesis is correct. The conclusion is that these four facilities are similar. Table 6.15 Means for Average Flow Time in Each Manufacturing Facility Factor N Mean StDev 95% CI China (1272, 3181) Ireland (128, 2038) PR (18, 1927) Single (1173, 3082) Table 6.16 Analysis of Variance Source DF Adj SS Adj MS F-Value P-Value Factor Error Total

93 Data 93 Boxplot of China, Ireland, PR & Single China Ireland PR Single Figure 6.5: Boxplot of four facilities 6.5 Cost Analysis between Alternative Supply Chain Strategies In this section, three types of costs are discussed labor cost, machine cost and transportation cost. Since independent supply chain strategy deals with independent manufacturing facilities, only labor cost and machine cost will be included in this strategy. All three costs will be discussed in combined supply chain strategy. Since all of the product families are processed in China facility in the single manufacturing system strategy, labor cost and machine cost of China facility along with the transportation cost from China to other two regions need to be included Labor Cost Hourly labor cost and the number of workers in each product family is shown in Table 6.17 and Table 6.18 which is from Ateş (2013). In the research of Ateş (2013), the labor rates were obtained from official newpapers or announcements.

94 94 Table 6.17 Hourly Labor Cost (Ateş, 2013) Labor China Ireland PR Hourly $1.95 $20.94 $21.49 Table 6.18 Number of Labor Required in the Fast System (Ateş, 2013) Family Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total in Table For independednt supply chain strategy, the number of labor in each cell is shown Table 6.19 Number of Labor in the Fast System for Independent Supply Chain Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total For the combined supply chain strategy, the number of workers in each cell is shown from Table 6.20 to Table 6.22.

95 95 Table 6.20 Number of Labor in the Fast System of China Facility for Combined Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total Table 6.21 Number of Labor in the Fast System of Ireland Facility for Combined Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total Table 6.22 Number of Labor in the Fast System of PR facility for Combined Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total Table For single manufacturing strategy, the number of labor in each cell is shown in

96 96 Table 6.23 Number of Labor in the Fast System for Single Manufacturing Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total in Table The total number of workers and labor costs for all of three strategies are shown Table 6.24 Number of Workers and Labor Costs for Three Strategies Independent Supply Combined Supply Chain Single Manufacturing Labor Chain Strategy Strategy Strategy (If in China) Number Cost Number Cost Number Cost China 82 $319, $374, $846,300 Ireland 82 $3,434, $3,182,880 - PR 82 $3,524, $2,664,760 - Total 246 $7,278, $6,222, $846,300

97 Machine Cost Hourly Machine cost and the number of machines in each product family is shown in Table 6.25 and Table 6.26 which is from Ateş (2013). In the research of Ateş (2013), the machines were assumed to run five years. Then hourly machine cost was calculated by the overall machine cost and 2000 hours operation time per year. Table 6.25 Hourly Machine Cost (Ateş, 2013) Machine Mach1 Mach2 Mach3 Mach4 Mach5 Mach6 Mach7 Mach8 Fast $50 $30 $25 $15 $20 $40 $25 $20 Table 6.26 Number of Machine Required in the Fast Manufacturing System (Ateş, 2013) Family Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op For independent supply chain strategy, the number of machines in each cell is shown in Table 6.27.

98 98 Table 6.27 Number of Machines in the Fast System for Host Production Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total For combined supply chain strategy, the number of machines in each cell is shown from Table 6.28 to Table Table 6.28 Number of Machines in the Fast System of China Facility for Combined Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total Table 6.29 Number of Machines in the Fast System of Ireland Facility for Combined Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total

99 99 Table 6.30 Number of Machines in the Fast System of PR Facility for Combined Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total in Table For single manufacturing strategy, the number of machines in each cell is shown Table 6.31 Number of Machines in the Fast System for Single Manufacturing Strategy Cell Op1 Op2 Op3 Op4 Op5 Op6 Op7 Op8 Op9 Total The total number of machines and machine costs for all of three strategies are shown in Table 6.32.

100 100 Table 6.32 Number of Machines and Machine Costs for Three Strategies Independent Supply Combined Supply Single Manufacturing Machine Chain Strategy Chain Strategy Strategy Number Cost Number Cost Number Cost China 66 $4,030, $4,900, $11,220,000 Ireland 66 $4,030, $4,030, PR 66 $4,030, $3,160, Total 198 $12,090, $12,090, $11,220, Transportation Cost Transportation cost exists in the combined supply chain strategy and single manufacturing strategy when products need to ship from one region to another. In this section, we consider maritime transportation. Ateş (2013) decided each container could stock 172,000 products and the unit cost is for each container which varies from different transportation routes. He also proposed the unit cost of container to transport containers from China to Ireland is $ and the unit cost to transport containers from China to Puerto Rico is $ which are shown in Table The total number of product families shipped from China to the other regions are from Simulation results for the combined supply chain strategy.

101 101 Table 6.33 Transportation Cost for Combined Supply Chain Strategy China - Ireland China - PR Total Number of Units 3,125,000 7,094,000 Capacity/Container 172, ,000 Number of Container Unit Container Cost $4, $2, Transportation Cost $93, $110, Total Transportation Cost $203, (based on 2013 dollars) Transportation cost for single manufacturing strategy is calculated in Table Total number of shipments are based on the toal mean demand of each region in Table 4.2. Table 6.34 Transportation Cost for Single Manufacturing Strategy China - Ireland China - PR Total Number of Units 40,377,603 40,125,799 Capacity/Container Number of Container Unit Container Cost $4, $2, Transportation Cost $1,154, $615, Total Transportation Cost $1,770, Total Cost By considering labor cost, machine cost and transportation cost, total cost comparison is conducted in this section. From Table 6.35, total cost is lower in single manufacturing strategy among three strategies.

102 102 Table 6.35 Total Cost for Three Strategies Host Production Cost Strategy Combined Supply Chain Strategy Single Manufacturing Strategy Labor $7,278, $6,222, $846, Machine $12,090, $12,090, $11,220, Transportation - $203, $1,770, Total $19,368, $18,515, $13,836,514.05

103 103 CHAPTER 7 CONCLUSION AND FUTURE WORK This chapter discusses the overall conclusion and future work of this research. This research deals with manufacturing system design and three alternative supply chain designs independent supply chain system, single manufacturing system, and combined new hybrid supply chain design. Since demand data are obtained to calculate the demand coverage probabilities and expected cell utilizations, independent supply chain design and single manufacturing system design are conducted to meet the demand. Based on the expected cell utilizations in independent supply chain design, two newly proposed hybrid supply chain designs are combined into one new hybrid supply chain system design. Comparison of expected cell utilizations and statistical analysis are conducted between independent supply chain design and single manufacturing design. Single manufacturing system can produce product families more efficiently if located in China. Also, cost analysis is conducted among three alternative supply chain designs. By considering labor cost, machine cost and transportation cost, the total cost in single manufacturing strategy is the lowest due to low labor cost in China region even though this strategy includes higher transportation costs. Simulation is used to verify the independent supply chain design. By comparing the expected cell utilizations and number out between the results of simulation and theoretical results, the strategy turns out to be correct and reasonable. Then, setup time is considered in independent supply chain strategy. When running simulation experiment, setup times are included between different product families. Then, the results of simulation experiment show the number of setups in each

104 104 cell. These numbers are used to calculate the new expected cell utilizations in the manufacturing design. The consideration of setup time is important in this thesis. The first step is to use simulation to estimate the number of setups. The second step is to dertermine available capacity for production. The last step is do the theoretical calculations for more accurate design. One of the attempts in this thesis is to increase the number of dedicated cells. By merging the product family with low cell utilization in the shared cell to another remainder cell, the original shared cell turns into a dedicated cell with high cell utilization. The results show that planning process is simplified and family setup times are avoided or reduced. This return may reduce machine and labor costs. Based on the methodologies and experiments completed in this research, there is still some future work that can be done. The analysis of different types of transportation can be incorporated. Based on this, the total cost comparisons can be made among these three approaches. In this thesis, another attempt is made to increase the number of dedicated cells. However, more further work is needed to evaluate this methodology. By comparing this methodology with original independent supply chain design, results can prove some improvements or shortages. After merging, some cells still can have low utilization. Then pricing strategy can be adopted. Pricing strategy can be used to increase the cell utilization. When some cell utilization is smaller than 20% (for example), pricing strategy can be used in this product

105 105 family. By reducing the product price, the demand will increase. This will lead to the increase of the cell utilization. This thesis deals with one single period of manufacturing design and supply chain design. There is some further work can be done in multiple periods to evaluate the performance of the system.

106 106 REFERENCES Ateş, O. (2013). Global supply chain and competitive business strategies: A case study of blood sugar monitoring industry (Doctoral dissertation). Retrieved from Brush, T. H., Marutan, C. A., & Karnani, A. (1999). The plant location decision in multinational manufacturing firms: an empirical analysis of international business and manufacturing strategy perspectives. Production and Operations Management, 8(2), Bulgak, A. A., & Bektas, T. (2009). Integrated cellular manufacturing systems design with production planning and dynamic system reconfiguration. European Journal of Operational Research, 192(2), Canel, C., & Das, S. R. (2002). Modeling global facility location decisions: integrating marketing and manufacturing decisions. Industrial Management & Data Systems, 102(2), Celikbilek, C., & Süer, G. A. (2016). Supply Chain Design Approaches for Dual Demand Management Strategies. In R. Addo-Tenkorang (Ed.), Supply Chain Strategies and the Engineer-to-Order Approach, (pp ), IGI Global. Chen, S. J., & Cheng, C. S. (1995). A neural network-based cell formation algorithm in cellular manufacturing. The International Journal of Production Research, 33(2), Chopra, S., & Meindl, P. (2004). Supply chain management: Strategy, planning & operations (second edition). New Jersey: Prentice Hall, Upper Saddle River.

107 107 Cosner, J. (2008). Comparison of alternative manufacturing systems for global supply chain business strategies in blood sugar monitoring industry (Master s thesis). Retrieved from Dicken, P. (1992). Global shift: the internationalization of economic activity (second edition.). London: Paul Chapman Publishing. Eğilmez, G., Süer, G. A., & Özgüner, O. (2012). Stochastic Capacitated Cellular Manufacturing System Design with Hybrid Similarity Coefficient. In F. A. Aziz (Ed.), Manufacturing System, (pp ), Rijeka: Intech. Evans, J. R. (1987). Applied production and operations management. West Publishing Company. Haug, P. (1992). An international location and production transfer model for high technology multinational enterprises. The International Journal of Production Research, 30(3), Huang, J., & Süer, G. (2012). Long-term sourcing model for multi-objective supplier selection under uncertainty. In Proceedings of 42nd International Conference on Computers & Industrial Engineering. Hughes, M. D. (2009). The business of self-monitoring of blood glucose: a market profile. Journal of Diabetes Science and Technology, 3 (5), Hyer, N. L. (1982). Education: MRP/GT: a framework for production planning and control of cellular manufacturing. Decision Sciences, 13(4), Jiang, J., & Süer, G. (2016). Alternative global supply chain design strategies for a blood sugar strip manufacturer considering layered cellular design. Paper presented at

108 108 Global Joint Conference on Industrial Engineering and Its Application Areas, Istanbul, Turkey. Kelton, W. D., & Law, A. M. (2000). Simulation modeling and analysis. Boston: McGraw Hill. Kimemia, J. G., & Gershwin, S. B. (1983). An algorithm for the computer control of production in flexible manufacturing systems. Massachusetts Institute of Technology Laboratory for Information and Decision Systems Report. LIDS-P Lenz, R. (2012). The Impact of Product Design and Product Development on the Production System Design. Examicus Publishing. Lobo, R. (2006). Comparison of connected vs disconnected cellular systems using simulation (Doctoral dissertation, Ohio University). Retrieved from Lowe, T. J., Wendell, R. E., & Hu, G. (2002). Screening location strategies to reduce exchange rate risk. European Journal of Operational Research, 136(3), Maddisetty, S. (2005). Design of shared cells in a probabilistic demand environment (Thesis, Ohio University). Retrieved from Mahdavi, I., Javadi, B., Fallah-Alipour, K., & Slomp, J. (2007). Designing a new mathematical model for cellular manufacturing system based on cell utilization. Applied Mathematics and Computation, 190(1), Min, H., & Zhou, G. (2002). Supply chain modeling: past, present and future. Computers &Industrial Engineering, 43(1),

109 109 New, S. J., & Payne, P. (1995). Research frameworks in logistics: three models, seven dinners and a survey. International journal of Physical Distribution and Logistics Management, 25 (10), Rajamani, D., Singh, N., & Aneja, Y. P. (1990). Integrated design of cellular manufacturing systems in the presence of alternative process plans. International Journal of Production Research, 28(8), Roberts, P. W. (1999). Product innovation, product-market competition and persistent profitability in the US pharmaceutical industry. Strategic Management Journal, 20(7), Süer, G. A., Huang, J., & Maddisetty, S. (2010). Design of dedicated, shared and remainder cells in a probabilistic demand environment. International Journal of Production Research, 48 (19), Süer, G., & Huang, J. (2012). Long-Term Sourcing Model For Multi-Objective Supplier Selection Under Uncertainty. Computers and Industrial Engineering 42. Süer, G. A., & Ortega, M. (1994). A machine level based-similarity coefficient for forming manufacturing cells. Computers & industrial engineering, 27(1-4), Süer, G., & Ortega, M. (1998). Flexibility considerations in design of manufacturing cells. In A. K. Kamrani & P. Logendran (Eds.), Group technology and cellular manufacturing: Methodologies and Applications (pp ). New York: Gordon and Breach Science.

110 110 Wilson, D. (2015). Comparison of alternative manufacturing systems for global supply chain business strategies in blood sugar monitoring industry (Master s thesis). Retrieved from Wright, R. T. (1990). Manufacturing Systems. Chicago, II: Goodheart-Willcox Co.

111 111 APPENDIX A: DEMAND COVERAGE PROBABILITY CALCULATION OF FAST SYSTEM FOR CHINA REGION For Family 1 in Fast system for China region: For Family 2 in Fast system for China region: ( ) ( ) ( ) For Family 3 in Fast system for China region: ( ) ( ) For Family 4 in Fast system for China region: ( ) ( ) ( ) ( ) ( ) For Family 5 in Fast system for China region: ( )

112 112 APPENDIX B: DCP AND ECU OF FAST SYSTEM FOR CHINA REGION Demand Coverage Probability Calculation of Fast System for China Region: For Family 1 in Fast system for China region: For Family 2 in Fast system for China region: ( ) ( ) ( ) For Family 3 in Fast system for China region: ( ) ( ) For Family 4 in Fast system for China region: ( ) ( ) ( )

113 113 ( ) ( ) For Family 5 in Fast system for China region: ( ) Expected Cell Utilizations Calculation of Fast System for China Region: For Family 1 in Fast system for China region:

114 APPENDIX C: DECISION MODULES OF QUEUE SIZE IN CHINA FACILITY 114

115 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Thesis and Dissertation Services!