Kanban Systems- Exploring the boundaries in a High Variety Manufacturing Process

1

Kanban Systems- Exploring the boundaries in a High

Variety Manufacturing Process

MSc Thesis

MSc Double Degree in Operations Management

University of Groningen, Faculty of Economics and Business

Newcastle University, Business School

Student Name

_{Jenifer Onyebuchi Nwanze}

Student Number S3078361 University of Groningen B5059111 Newcastle University

Email j.o.nwanze@student.rug.nl

Groningen Supervisor Dr Jos Bokhorst (j.a.c.bokhorst@rug.nl)

2

Preface

This research was realised as the final requirement in fulfilment of the master programme in Technology and Operations Management program at the University of Groningen and Operations and Supply Chain Management at the Newcastle University.

My greatest gratitude is to God Almighty for his faithfulness and sustenance. I would like to thank Dr Jos Bokhorst from the University of Groningen and Professor Chris Hicks from Newcastle University for their constructive criticism, detailed feedback and guidance, throughout the realization of this research. Additionally, I would like to thank the case company for the wonderful opportunity to perform this research and experience first-hand the operation of a Kanban system. Furthermore, everyone who supported in one form or another, especially, my supervisor at the case company, the logistics engineers and MPU team for making me part of their team and also for the guidance and support. This aided in the detailed evaluation and understanding of the Kanban system, thus, resulting in the high standard of this research.

Finally, I would like to thank my parents, siblings, Spanish family and friends for the encouragement and prayers. I achieved this because you all believed in me.

Always with love

3

Table of content

2.2 Types of Kanban Systems ... 13

2.2.1 Single-Card System ... 13

2.2.2 Dual-Card system ... 15

2.3 Kanban Decision Board ... 16

2.4 Determination of the number of kanbans ... 18

2.5 Kanban Environment ... 20

2.6 Kanban Performance Measures ... 21

2.7 Variations of Traditional Kanban ... 24

2.8 Comparison of Control Systems applicable in High Variety Environments ... 26

3 Methodology ... 29

3.1 Research Model ... 29

3.2 Research Strategy ... 31

3.3 Data collection ... 31

3.4 Data Analysis ... 32

3.5 Reliability and validity ... 33

3.6 Ethical responsibility ... 33

4 Case Description ... 35

4

4.2 Overview of the Moulding Production Unit (MPU) ... 37

4.3 Production Planning Process ... 36

5 MPU Kanban System ... 39

5.1 System Operation ... 39

5.2 Daily Scheduling using the Kanban decision board ... 40

5.2.1Operating rules of the Kanban Decision Board ... 41

5.3 Comparison of types of Kanban Systems ... 43

5.4 Problem Definition ... 44

6 Findings and Discussion ... 45

6.1 Environmental Characteristics ... 45

6.1.1 Shop floor layout ... 45

6.1.2 Resources- (Machines and Equipment) ... 46

6.1.3 Resource Reliability ... 47

6.1.4 Setups and Changeovers ... 47

6.1.5 Worker (Flexibility and Skills) ... 48

6.2 Production Characteristics ... 48

6.2.1 Demand Pattern ... 49

6.2.2 Product mix levelling ... 50

6.3 Kanban Decision Board ... 50

7 Conclusion ... 53

Reference ... 55

Appendix A- Stakeholders Interview Questions ... 62

Appendix B- MPU Staff Interview Questions ... 62

Appendix C -Other Control Systems ... 64

5

8.2 Paired Cell Overlapping Loops of Cards with Authorization (POLCA) ... 66

8.3 Control of Balance by Card-Based Navigation (COBACABANA) ... 67

8.3.1 Concept of COBACABANA ... 67

8.3.2 Order Release Planning Board ... 68

List of figures

Figure 2: Pull Single-Card Kanban System ... 15

Figure 3: Dual-card Kanban system ... 16

Figure 4: Production Kanban decision board ... 17

Figure 5: Kanban Performance- Influencing Factors ... 30

Figure 6: Part Flow- Internal Supply Chain ODF ... 35

Figure 7: Variation in Parts ... 38

Figure 8: Variation in Trolleys ... 38

Figure 9: MPU Kanban System ... 39

Figure 10: MPU Kanban Board ... 41

Figure 11: Number of Moulding Machines ... 46

Figure 12: Concept of CONWIP ... 65

Figure 13: Concept of COBACABANA system ... 68

Figure 14: Planning board for order release COBACABANA system ... 69

List of tables

Table 2: Performance Measures ... 24

6

Table 4: Kanban Influencing factors and definitions ... 30

Table 5: Stakeholder Analysis ... 32

Table 6: Variation in High Variety Parts ... 37

Table 7: Kanban systems comparison ... 43

Table 8: Differences between TPS and MPU Kanban system ... 44

Acronyms

Abbreviation Explanation

APV Automated power circuit board

ASIC Application specific integrated circuit

BBA Basic body assembly

COBACABANA Control of balance by card based navigation system

CONWIP Constant work-in-progress

DUM Driving unit mechanism

EPEI Every part every interval

ETO Engineer to order

FCFS First come first serve

FKS Flexible Kanban system

GKS Generic Kanban system

HCA Hierarchical control architecture

JIT Just in time

MODD Modified operation due date

MOQ Minimum order quantity

MPS Master planning schedule

MPU Moulding production unit

MTS Make to stock

MTO Make to order

7

ODD Operation due date

ODF Order driven factory

PDF Process driven factory

POLCA Paired cell overlapping loops of cards and authorization

POK Production order kanban

PPS Periodic pull system

QRM Quick response manufacturing

SMED Single minute exchange of die

SPT Shortest processing time

SUA Shaving unit assembly

TPS Toyota production system

VK Virtual Kanban

WIP Work-in-progress

8

Abstract

The implementation of Kanban systems has proven successful in mostly repetitive manufacturing environments, characterised by standardized operations, low product variety, amongst other factors. However, with increasing customer preference, resulting in high product variety, manufacturing flexibility is paramount to fulfilling customer demand. Although a lot of research exists in the field of Kanban systems, focus has been on the influence of certain environmental factors and decision variables on the performance of the system. However, no work thus far exists on factors to consider specifically in a high variety manufacturing environment. A detailed case study analysis was performed on a Kanban system operating in a high variety setting, in which a research model was developed to illustrate factors to consider in such environment. Factors proposed are based on the environmental characteristics, production characteristics and Kanban decision variables. These served as the anchor for this research, as current practices of the system operation were analysed in relation to coping with high product variety. Analysis demonstrated major influencing factors on the system performance to include: worker (flexibility and skills); setups and changeovers; demand pattern; product-mix levelling; kanban card management; and production run. These findings provides valuable insights as to improvement areas for the case company. Similarly, the academic and managerial relevance of this research, stems from the research model, which highlights factors to be considered in the design and analysis of Kanban systems in a high variety manufacturing environment.

9

1 Introduction

The upsurge in global competition and rapid technological advancements among other factors have contributed to increased product variety experienced by manufacturers of consumer goods (Ramdas 2003). This diversity stems from variations in physical forms, product functions and features such as colour, size, shape and packaging (Ramdas 2003). As a consequence, manufacturers are faced with the challenge to adapt existing products and processes. Hence, several factors need to be considered with increased product variety including the need for efficient production control systems. This is pivotal since operational inefficiencies could be incurred whenever the production control system switches from the production of one item to another (Benjaafar et al. 2004). Therefore, the choice of a production control system becomes a strategic decision in any manufacturing firm.

Production control systems involve the coordination of material and product flow within a production process. The objectives of these systems include: (i) optimization of resources; (ii) organization of production schedules; and (iii) inventory management (Spearman & Zazanis 1992, Fernandes & do Carmo-Silva 2006, González-R & Framinan 2009). Production control systems are categorised as push, pull (Spearman et al. 1990) or hybrid. Hopp & Spearman (2004, p.142) defined a push production system as “one that has no explicit limit on the amount of work in progress that can be in the system” and a pull production system as “one that explicitly limits the amount of work in progress that can be in the system”. The work in progress (WIP) is due to partially finished parts awaiting completion. The fundamental difference between push and pull systems is centred on the WIP and throughput. Throughput is the amount of materials going through the manufacturing process. Push systems control the throughput and measure WIP, and the latter controls the WIP and measures throughput (Spearman et al. 1989). Additionally, in push systems, parts are pushed through the production system in anticipation of customer demand, whereas in pull systems, the occurrence of customer demand pulls parts through the production process. Hybrid systems are a combination of both pull and push mechanisms.

10

activity or process in the form of transport, inventory, motion, waiting, overproduction, over-processing and defects. Mura means process variation and muri is the overburden of equipment and people (Slack et al. 1998). JIT seeks to achieve zero or minimum inventory, as the required amount of material is produced and delivered to the right place at the right time. The most commonly implemented push and pull control systems are; Materials Requirements Planning (MRP), and Kanban respectively (Krishnamurthy et al. 2004). A Kanban system is based on the JIT approach, thus, the focus is on inventory control. The word Kanban in Japanese means “card or visual sign” (Gupta et al. 1999). It is an inventory control system that links upstream production to customer demand downstream. Several manufacturing firms around the world, especially in Japan have implemented the Kanban system following the success of Toyota (Rahman et al. 2013). Sugimori et al. (1977) posited three main reasons for implementation to include: reduced cost in processing information; precise and rapid acquisition of facts following the use of the cards; and reduction in over-production.

11

The scope of this research is contained within the moulding production unit (MPU) of a large consumer goods manufacturer. The study will only consider the MPU and will not go further into the internal supply chain, as it depicts a wider range of study. The MPU like many other manufacturing firms has experienced an increase in product variety over the last years. This has led to a strain on the operation of the system, resulting in a rippling effect in the overall production process. Such effects include complexity in production scheduling, and product mix levelling, amongst others. Hence, the realization of this research is of academic and practical relevance, as it provides insights to understanding the impact of increased product variety on a Kanban system. This is pivotal in determining the suitability of Kanban systems in high variety manufacturing environment. Subsequently, this research contributes to the limited literature on the operation and performance of Kanban systems in high variety manufacturing environment. The central research questions developed for the realization of this study include:

o How does increased product variety impact on the operational performance of a Kanban system?

o Can the boundaries established in the defined Kanban system suit a high variety manufacturing processes?

12

2 Theoretical Background

This section provides the background relating to the fundamentals of the traditional Kanban system. The concept and operation of the system is analysed and the parameters used to analyse the performance of several Kanban systems is provided. The literature is reviewed in relation to the variations of the traditional Kanban system, this provides an understanding as to why these variations exist, as the case Kanban system is considered a variation. The section will conclude by providing other production control systems that are applicable in environments characterised by high variety. The term Kanban represents both the card and the system, however, in this thesis, differentiation is established by using kanban, when referring to a card and Kanban with the capital “K”, when referring to the whole system. Furthermore, the words customer, process and workstation are used interchangeable to represent a stage in the production process.

2.1 Concept of Kanban

Kanban was developed by Toyota in the 1970s as an integral part of the JIT approach which is based on Lean Philosophy (Sugimori et al. 1977). The operation of the system is centred on a pull mechanism, which is grounded on the explicit control of WIP, contrary to a Materials Requirements Planning (MRP) system, that is based on a push mechanism. In MRP systems, no priori WIP limit is established, as production release is according to a master planning schedule (MPS) which drives how and when each part will be requested (Zijm 2000, Hopp & Spearman 2004). The decision making process in MRP systems is highly centralized compared to the JIT approach, as production runs are scheduled in advance. MRP systems require constant maintenance of the specified database (Bard & Golany 1991), as it is a computerised production planning and inventory control system. Instead, in Kanban systems, the control of production and WIP between each pair of workstation is achieved manually (Akturk & Erhun 1999).

13

information conveyed on the kanbans include: part type; part number; serial number of the card; order quantity to be produced; location in terms of supplier (pull from) and customer (pull to) (Chang & Yih 1994). The physical signal created by a Kanban system in notifying the replenishment of parts contributes to its effectiveness and simplicity in terms of material handling and replenishment. The signals produced could be in the form of a card, empty trolley or space on the shop floor, thus, making it difficult to ignore.

2.2 Types of Kanban Systems

Kanban systems can either be a single-card or dual-card, consisting of production order kanban (POK) and/or withdrawal kanban (WK) (Sugimori et al. 1977, Schonberger 1983). The POK defines the quantity of a specific part that should be manufactured to replenish used parts. On the other hand, the WK specifies the quantity that the succeeding workstation should withdraw from the preceding workstation (Akturk & Erhun 1999, Kumar & Panneerselvam 2007). Schonberger (1983) added that the POK serves the workstation replenishing the parts and the WK is for the workstation using the parts. However, this only applies in the dual-card system involving the use of both cards for its operation. The basic principle governing the single and dual-card systems is provided below.

2.2.1 Single-Card System

The single -card system is considered as the simplest implementation of the Kanban system, as it involves the utilization of only one type of kanban (Berkley 1992, Akturk & Erhun 1999, Schiraldi 2013). Most companies implement this system as the initial step in developing a dual-card system. The single-dual-card system is commonly applied in settings where the upstream and downstream workstations are physically close to each other (Schonberger 1983, Berkley 1992). As a result of the short distance, material handling is considered as instantaneous, resulting in shorter information lead times. Based on existing Kanban literature, the single-card system is categorised into: (i) hybrid strategy (Schonberger 1983); or (ii) pull strategy (Huang & Kusiak 1996), description of both systems is provided.

14

via a WK (Schonberger 1983, Akturk & Erhun 1999). In other words, a push strategy is used for production, i.e., parts are made to stock (MTS) and delivered based on a pull strategy, as the customer pulls the required amount of inventory at the time it is needed.

From figure 1, parts are delivered directly from workstation A to workstation B upon request, thus, no inbound buffer is required for workstation B. This is achievable as both workstations are closely linked together. The buffer at workstation A for produced parts, serves as an inbound and outbound buffer for the same workstation. The buffer in this system could be large, based on the MTS strategy at workstation A, as more inventory is held in the process while awaiting actual demand. The trigger for production is based on the central planning system, however, the trigger for withdrawal occurs when the WK from workstation B is returned to the Kanban board at workstation A.

15

In both single-card systems, when a trolley of parts is emptied at workstation B, the kanban card on the trolley is returned to the Kanban board at workstation A. This operation is performed periodically by an operator.

2.2.2 Dual-Card system

The dual-card system involves the use of two kanbans; Withdrawal kanban (WK) and Production order kanban (POK). The WK is used to withdraw parts from the preceding workstation, while the POK gives the authorization for replenishment of depleted parts (Yavuz. 1995, Akturk & Erhun 1999). The material handling operation is considered to be more complex and periodic than in the single-card system (Berkley 1992). This is based on the increase in distance between workstations, for instance; as seen in the Toyota Production System (TPS) by Sugimori et al. (1977).

16

A detailed description of the case Kanban system is provided in section 5, in addition, a comparison of the types of Kanban systems previously described is provided. Several benefits have been associated with the Kanban system based on the simple operating principle involving the use of cards. Typical benefits include: (i) inventory control, as the potential for over production and inventory obsolescence is minimized, thus, resulting in a reduction in inventory holding cost, floor space, quality issues, and obsolesce. Furthermore, inefficiencies in the production system such as: an unreliable production process is easily detectable and resolved; and (ii) flexibility in production in response to changes in demand is achieved, since the POK only triggers production based on the usage of parts from the downstream workstation.

2.3 Kanban Decision Board

The Kanban decision board provides a visual representation of the production sequence, as it displays the POK and WK (Murino et al. 2010). Similarly, a complete overview of parts is provided. The Kanban board can either be physical, electronic or virtual. The electronic Kanban board is a translation of the physical kanban on a computer (Schiraldi 2013). It can be deduced from the description of the dual-card system provided in the previous section, that production at the replenishing workstation is based on the POK placed on the Kanban board.

17

resource. The total number of POK for part A is 7 and part B is 4 respectively. The placement of these cards is done from top to bottom. The columns are grouped into three zones; green, yellow and red, indicating the level of production urgency; safe, moderate, and critical (Murino et al. 2010, Schiraldi 2013). The kanban in the red zone as shown in the diagram for part A implies that the buffer is almost empty and part A is likely to be requested by the succeeding workstation. Thus, production on part A should commence as soon as possible to avoid stock out. The red zone is considered as the safety stock, and is based on managerial decision (Kumar & Panneerselvam 2007). The safety stock considers unforeseen circumstances during production such as machine breakdown. The yellow zone is considered as the buffer stock and is required to accommodate variations in demand and supply. The green zone represents the cycle stock, defined as the average amount of inventory required to satisfy demand within a given time. Toyota developed a formula to calculate the number of kanbans which is described in the next section.

The priority rule used for issuing kanbans is commonly based on the first come first serve policy (FCFS) (Akturk & Erhun 1999). In manufacturing environments with dedicated resources (machines and equipment) production scheduling is easy to determine as illustrated in figure 4, as the production of part A and B can occur simultaneously. However, the decision-making process becomes more complex when a set of resources are required to produce a variety of products during the same time; i.e., multiple products on a shared resource. This is the current situation at the MPU and detailed description is provided in section 5. Therefore, it is essential that production levelling is achieved in the Kanban board operation to ensure constant flow of material on the shop floor.

Part A Part B

Machine 1 Machine 2

18

2.4 Determination of the number of kanbans

Diaz & Ali (2010) amongst several researchers posited the number of kanbans as a pivotal policy variable in JIT systems. The kanbans control the flow of parts entering each workstation to limit the amount of WIP in the system. Hence, establishing the number of kanbans is crucial to the manufacturing process, as it influences the WIP inventory and throughput time. Furthermore, the number of kanbans could significantly influence the load balancing between workstations, especially in environments with shared resources (machines and equipment). Sugimori et al. (1977), developed an empirical formula to determine the number of kanbans. It is expressed as a function of the demand for parts over the replenishment lead time as shown in formula 1 below.

𝑦 =

Where: y is the number of Kanban cards D is the demand per unit time

L is the lead time of one container a is the container capacity

∝ is the safety factor

The time between initiation and completion of a production process is known as the lead time. The manufacturing lead time takes into account the setup time, processing time, waiting time and kanban retrieval time (Philipoom et al. 1987, Kumar & Panneerselvam 2007). The kanban retrieval time is the time during which the kanban is returned to the beginning of the operation

19

size, product variety, as these differ between manufacturing processes; and (ii) the production and material handling lead times are considered as a function of the number of kanbans, i.e., the value of the lead time of a container, linearly increases as the value of y in formula (1) increases.

Furthermore, the number of kanbans is related to the inventory, i.e., if the value of “y” increases, the inventory of parts also increases and the converse also holds (Kumar & Panneerselvam 2007). An increase could result in idle inventory occupying space, while a decrease could potentially lead to backlogs, interuptions in the production process, idle workstations and workers. Nevertheless, the number of kanbans could be fine tuned to compensate for empty lanes from part shortage or high WIP. Hence, determining the optimum number of kanban for any production process is paramount. Several researchers have adopted other tools such as; simulation, mathematical models, and queuing models to determine the optimum number of kanbans (see:e.g. Panneerselvam 2007). The principles of the Kanban system based on curbing overproduction by limiting WIP, could be related to Little’s Law. This law is fundamental in queuing theory, as it defines the relationship between WIP, lead time and throughput.

Little’s law:

𝐿 = 𝜆 ∗ 𝑊

Where: L= WIP 𝜆= Throughput W= Lead time

20

2.5 Kanban Environment

The performance of Kanban systems is influenced by the manufacturing environment in which it is applied (Krajewski et al. 1987). It is advocated to be an inventory control system that is suitable for a repetitive manufacturing environment (Andijani 1997, Akturk & Erhun 1999, White & Prybutok 2001). Repetitive manufacturing could be considered as a form of mass production, involving the production of standardized and identical product or families of products during a manufacturing cycle. Such manufacturing environments are characterised by continuous flow in large batches, short setup times based on the similarity of parts produced, thus, resulting in short lead times (Toni et al. 1993). It is essential that demand variability is minimal in the production process with limited number of parts to prevent proliferation of WIP inventory at each workstation. Several authors posit that the above conditions need to be satisfied to ascertain the performance of a Kanban system (Zijm 2000, Stevenson et al. 2005). A typical example of a repetitive manufacturing environment is the automobile industry, characterised by high volume and highly standardised products with relatively low variety, where the shop-floor configuration is typically product-oriented. According to Slack et al. (1998), in a product or line layout, the workstations are arranged in a sequence and are in close proximity, to facilitate a smooth flow of WIP.

White & Prybutok (2001) provided a clear distinction between the repetitive and non- repetitive manufacturing environments, as shown below.

Table 1: Difference between Repetitive and Non-Repetitive Settings (Adapted from White & Prybutok 2001)

Characteristics Repetitive Manufacturing Environment Non-repetitive Manufacturing Environment

Variety of Products Standardized Customized/Mixture

Demand Predictable Difficult to predict for product mix

Scheduling Fixed schedule Uncertain, frequent changes

Changeovers Few, Relatively simple, lost time and labour Multiple, Relatively complex, lost time, labour and raw material

Setups Very Few Multiple, different with jobs

WIP Low High

Finished goods High Low

Worker skills and flexibility Focus on skill specialization with no flexibility

21

Job shops are commonly associated with non-repetitive manufacturing environments (White & Prybutok 2001). These environments are typically characterized by high product variety requiring high WIP between workstations and low volume in terms of finished goods, based on the product variety provided. Additionally, Krajewski et al. (1987) in an experiment to determine the robustness of the Kanban system in diverse manufacturing environments in the US, posited that several factors impact on the systems’ performance and sensitivity. The performance of the Kanban system was defined in terms of average weekly labour hours, expressed as a measure of workforce productivity, average inventory investment, and average past due demand of all end items. The characteristics of environments analysed include: small lot sizes; no make to order (MTO) parts; low scrap rate; low equipment failure rate; flow shop layout; and minimum setup times. Analysis of the proposed factors was based on the established performance measures, in different settings; low setting representing favourable Kanban environment and high setting less favourable. From the analysis, the factors proven to have significant impact on the performance of the Kanban system are as follows: (i) inventory; (ii) process; and (iii) product structure. Parameters considered in the inventory include: inventory accuracy; end item lot sizing; component lot sizing; and setup times. These factors proved to have the largest impact on the defined performance measure. This was attributed to the systems’ sensitivity to lot sizes and setup times, based on its lean philosophy and JIT principles. Similarly, the process factors include: high scrap rates; low worker flexibility; high equipment failure rates; and high capacity imbalance. These factors influenced the service performance of Kanban systems, resulting in possible backlog and system bottleneck. According to Krajewski et al. (1987), the more attractive the settings for the process factors, the better the service level performance of the system. Furthermore, the product structure expressed in terms of the bill of material (BOM) and the degree of product customization proved to have the least impact.

2.6 Kanban Performance Measures

22

According to Moeeni et al. (1997), WIP and service level have been the most commonly utilized performance measures. These measures are grounded in the fundamentals of any Kanban system, as a relationship could be established with other performance measures. However, there is often a trade-off between both measures, as the operation of Kanban systems strive to limit the WIP while maintaining a high service level. An increase in the WIP results in excess inventory and increase in holding cost, however, it results in a more responsive system with a high service level. Other commonly used performance measures include; average flow time, mean cumulative throughput rate and weighted earliness of jobs (Kumar & Panneerselvam 2007).

Furthermore, several researchers have proposed distinct parameters to this effect. Deleersnyder et al. (1989) developed a discrete time Markov model to analyze a Kanban system and quantify the relationship between cost and service level, based on an operational control problem. It involved the interaction between production and inventory levels, factors such as stochastic demand and stochastic machine reliability were taken into consideration. The key performance measures defined include: (i) inventory level- expressed as a measure of the operating cost, and backlog level defined as a measure of the service level; (ii) job flow time - derived from the inventory level; and (iii) average duration and frequency of backlog- defined as the average backlog divided by the average production output. Berkley (1992) in a review of Kanban systems expressed performance measures in terms of average inventory and the ability to fulfil demand from inventory. Akturk & Erhun (1999) posited that for simulation models, commonly used performance measures include: machine utilization; inventory holding cost; shortage cost; and fill rate. They defined shortage costs in terms of backorder and fill rate as the probability of fulfilling demand through inventory. Gupta et al. (1999) in a study comparing traditional Kanban to flexible Kanban utilized four parameters to describe performance: WIP; average time spent by parts in the system; average order completion time; and the average number of back order units. Flexible Kanban is considered as a variation of the traditional Kanban system involving the manipulation of Kanban cards required to improve the performance of the system at a given time.

23

system. Li (2003) in a study to reduce setup times and processing time variability in a job shop, defined performance as three parameters: (i) average WIP inventory- as the total number of parts in the system averaged over a given period of time; (ii) average flow time- as the amount of time spent by a part from entry to completion of operation in the workstation averaged over all completed parts; and (iii) average setup time to processing time ratio- as the ratio of the total amount of time spent for processing parts averaged over all machines. Koukoumialos & Liberopoulos (2006) argued that production rate is the main performance measure, because it determines the maximum rate at which demand can be fulfilled. However, other performance measures proposed include: WIP; average number of finished parts at each workstation; average number of backlog demand; average waiting time; and percentage of backlog demand. The study was based on the evaluation of a multi-stage, serial, echelon Kanban control system. In the echelon Kanban system, when a part leaves the last workstation to satisfy a customer demand, a new part is requested and authorized to be released into intermediate workstation. Contrary to the traditional Kanban system, part production and release into a workstation is triggered when the part leaves the specific workstation. Diaz & Ali (2010) in a study of a dual- Kanban system in a non-repetitive manufacturing environment, measured performance by the average customer waiting time and total inventory. The total inventory is the combination of all WIP inventory and finished product inventory. Khojasteh-Ghamari (2012) described performance as average WIP and system throughput. WIP was defined as the sum of the average number of parts held in buffers and the parts being processed in the distinct workstations. The system throughput is expressed as a function of the average number of finished products per time unit.

24

Factors Authors

Inventory level Deleersnyder et al. (1989), Berkley (1992), Gupta et al. (1999), Diaz & Ali (2010)

Backlog level Deleersnyder et al. (1989), Gupta et al. (1999), Koukoumialos & Liberopoulos (2006)

Quality Berkley (1992), Takeda et al. (2000)

Machine utilization Akturk & Erhun (1999)

Holding cost Akturk & Erhun (1999)

Shortage cost Akturk & Erhun (1999)

Production cost Takeda et al. (2000)

WIP Gupta et al. (1999), Li (2003), Koukoumialos & Liberopoulos (2006), Khojasteh-Ghamari (2012)

Throughput Takeda et al. (2000), Khojasteh-Ghamari (2012)

Service Level Deleersnyder et al. (1989)

Fill rate Berkley (1992), Akturk & Erhun (1999)

Lead time Takeda et al. (2000), Li (2003),

Setup time Li (2003)

Waiting time Koukoumialos & Liberopoulos (2006)

Production rate Koukoumialos & Liberopoulos (2006), Diaz & Ali (2010)

Table 2: Performance Measures

2.7 Variations of Traditional Kanban

In recent years, the Kanban system has become an area of interest for researchers, as several modifications have been proposed to analyse its applicability in varying manufacturing environments. However, these variations are developed to suit specific characteristics in certain production processes. Hence, generalization is difficult since no production process is considered the same, and no framework exists by which these systems could be compared. Nevertheless, it is important to consider the reasons as to why these variations exist. Complexity involved in applying the traditional Kanban system in diverse situations has led to the development of several variations, to fulfil specific requirements. A comprehensive review is provided by Lage Junior & Godinho Filho (2010) in which 32 variations of the traditional Kanban system between 1985 to 2002 were identified. The categorization of these variations is based on the following factors:

25

based on the inventory level or scheduling of the last workstation; decentralized control of shop floor and WIP control;

(2) Operational differences between the variations and the traditional Kanban system. This factor reflects the way the traditional Kanban was modified in relation to material release, operating concept, variable maximum inventory level, information gathering and usage relating to inventory level and demand, signal usage;

(3) Advantages over the traditional system, several advantages were taken into consideration based on the limitation of the traditional Kanban system. Flexibility of the replenishment cycle; introduction of new products, applicability in manufacturing environments with high variety, complex material flow, and automated operations were some of the stipulated advantages;

(4) Disadvantages over the traditional system were defined as: delay in the transmission of information; average inventory level; and movement of workers and trolleys;

(5) Practical variations developed or applied in real situations and theoretical variations developed using mathematical or simulation models. The number of theoretical variations were more than the practical variations.

26

developed a theoretical model called Virtual Kanban (VK), which was applied in an application specific integrated circuit (ASIC) manufacturing facility. The facility was characterised by a complex factory with multi-stage production system where products were custom-made on a large scale. Gao et al. (2008) developed a multi-triangular Kanban by using integer programming in determining the batch size, inventory control limit and the number of kanban card. The system was applied in an electronic connector plant with a manufacturing environment characterised by high product variety, long machine setup time for multiple machines to multiple assembly lines. The multi-triangular Kanban is considered the same as the POK in authorizing the replenishment of depleted parts.

2.8 Comparison of Control Systems applicable in High Variety Environments

This section provides a comparison of control systems applicable in a high variety manufacturing environment. Constant-work-in-progress (CONWIP) proposed by Spearman et al. (1990), Paired cell overlapping loops of cards and authorization (POLCA) developed by Rajan (1998), and Control of balance by card-based navigation system (COBACABANA) by Land (2008). The emergence of these systems is based on the limitation of Kanban systems in production processes with increased variation and complexity (Spearman et al. 1990, Krishnamurthy & Suri 2009). In general, these systems involve the use of cards for operation, where uncertainty in the production process is reduced by using feedback loops based on the current situations on the shop floor. The underlying principle governing the operation of these systems is inventory control achieved by controlling the WIP level and input/output of task. The input/output control system aligns the input and output of work by adjusting the output or controlling the input of the workstation or shop floor (Thurer et al. 2016). (See Appendix C for detailed description of the aforementioned systems).

27

Table 3: Comparison of Card- based control systems

A major advantage of Kanban systems is the simple, effective and visual means of inventory control (Sugimori et al. 1977). However, since Kanban systems are designed for environments where the use of small intermediate stocks applies, an increase in product variety, could result in large number of different bins or card loops and increased machine setup time. The combination of both effects could result in large WIP (Germs & Riezebos 2010). CONWIP a hybrid system is based on Hierarchical Control Architecture (HCA), which is a combination of JIT and hierarchical control structures (Spearman et al. 1989). In this system, a WIP cap is placed on the amount of inventory at a production line at any given period. Once this limit is reached, the entrance of inventory to the system is blocked until a demand removes the corresponding amount of inventory from the production line (Prakash & Feng 2011). It is assumed that the total process time at the bottle neck for each container is the same and also the WIP inventory is constant at workstations, thus, CONWIP flowlines are predictable (Spearman et al. 1990). A superior advantage of CONWIP over the Kanban is the applicability in manufacturing environments characterised by high product variety (Kabadurmus 2009). The CONWIP strategy offers a high degree of system predictability by maintaining WIP, production quota and safety capacity. Additionally, Slomp et al. (2009) argued that based on the working principle of CONWIP, the system is considered as the simplest pull control system in comparison with Kanban. The differences between CONWIP and Kanban include: (i) when the WIP is used to determine the sequence of parts, CONWIP outperforms Kanban at comparable WIP levels; (ii) a single set of global cards is used for parts in the production line in CONWIP rather than individual part number as in Kanban; and (iii) jobs flow through the workstations in series upon authorization at the beginning of the line in CONWIP (Spearman et al.

System Philosophy/Emphasis WIP Cap Point of establishment Push/Pull Environment

Kanban JIT/ Throughput Control Per station Between two workstations Pull Repetitive

CONWIP HCA/ WIP regulation Shop-floor

inventory

Between entry and exist

workstations

Hybrid Non-repetitive

POLCA QRM/ Utilization Per cell Between two cells Hybrid Non-repetitive

28

1990). Furthermore, in CONWIP, a job is pulled only at the beginning of the production line, while Kanbans pull work between every workstation.

A distinct characteristic of POLCA compared to Kanban and CONWIP is that two cells are required to make a loop in which a tighter WIP control is based on overlapping cells (Liu & Huang 2009). The ability of POLCA systems to offer different routings between cells and workstations makes it suitable in environments with high product variety compared to Kanban and CONWIP. However, in manufacturing environments with high routing mix, the use of POLCA cards is complicated, as it involves the use of card loops for each possible combination of work centre (Land 2008).

29

3 Methodology

The previous section has provided an in-depth analysis of the fundamentals of the traditional Kanban system. This section provides the methodology used for the realization of this research. This is grounded in the basics of the research onion proposed by Saunders et al. (2015).

3.1 Research Model

A research model was developed based on review of relevant literature and the case production process. The model served as the anchor for investigating the defined research questions;

o How does increased product variety impact on the operational performance of a Kanban system?

o Can the boundaries established in the defined Kanban system suit a high variety manufacturing processes?

30

overall system stability. Hence, it is pivotal that the influence of the production characteristics is critically analysed especially in a high product mix manufacturing environment.

Kanban Decisions Environmental Characteristics Production Characteristics KANBAN PERFORMANCE

Figure 5: Kanban Performance- Influencing factors

Table 4: Definition of Kanban performance influencing factors

Therefore, in addition to the existing environmental and Kanban decision variables proposed in literature, the model implies that the environmental and production characteristics influence the decisions made in a Kanban system, which in turn impacts on its performance. The research model is explored by analysing and evaluating the Kanban system at the case company.

Factors Variables

Environment • Shop floor layout

• Resources- Shared and Dedicated (Machines, Equipment) • Resource Reliability

• Setups and Changeovers • Worker (flexibility and skills)

Production • Demand Pattern (Volume, Variability) • Product mix levelling

• Production Interval calculation for machines - Every part every interval (EPEI) • kanban card calculation

Kanban Decisions • Priority rules • Withdrawal cycle • Transfer policy

• kanban card management • Production run

Kanban Performance

31

3.2 Research Strategy

The strategy employed in this research is a single case study performed in a large consumer goods manufacturer, where production control and material handling is achieved using a Kanban system. In comparison with other research strategies such as experiment, or survey, a case study proved adequate in this study, as: (i) the Kanban system was closely examined in a real operating context, which eliminated any form of manipulation, thus, increasing the validity of findings; and (ii) it aided in gaining an in depth understanding of the system, especially since the boundaries between the Kanban system and the high variety manufacturing environment were not evident (Yin 1981). However, a major drawback of the single case study is the inability to draw generalizable conclusions (Fletcher et al. 1997, Karlsson 2009). The research is considered as exploratory, as it is aimed at investigating the current state, to gain insights from analysing the operation and performance of the phenomenon. Additionally, the research is explanatory, as it examined the context at a surface and deep level in order to explain the phenomenon (Fletcher et al. 1997). Similarly, it explained the implications of the factors proposed in the research model in relation to the increased product variety.

3.3 Data collection

In this study, multiple sources of data were utilised and triangulated (Karlsson 2009).

Primary data was obtained from a single researcher stationed full time in the company over a

period of 4 months between August and December 2016. Techniques employed include interviews and direct observation.

32

Table 5: Stakeholder Analysis

Detailed information regarding the operation of the MPU Kanban was obtained from individuals at distinct levels of the order driven factory (ODF). This stage involved semi-structured interviews with the MPU staffs including: production manager; assistant production managers; team leads; changeover specialists; and machine operators. At the start of each interview, several predetermined open- questions were asked, which led to follow-up questions to obtain in-depth information and insights about the production process and Kanban system operation (see appendix B).

(ii) Direct observation: The researcher worked in several shifts at the MPU to experience first-hand the daily activities and decisions relating to the Kanban operation. Similarly, the researcher made several tours around the shop floor, directly observing the operations between the supplier (MPU), its customers and the buffer. Detailed notes were made on relevant subjects relating to the daily operation of the Kanban board including kanban card management, production process, and material flow from the supplier to the customer.

Secondary data was obtained from documents (articles, and text books), archival records (company

reports on earlier studies introducing the Kanban system into their production process). Quantitative data from the company’s SAP-system and excel files were also used to strengthen the validity of information from direct observation and interviews. The focus was on analyzing the planning process and production data of moulded parts.

3.4 Data Analysis

The data obtained during this research was analyzed in different ways. The key points of each interview were documented, interviewees reviewed and confirmed response to ensure the

Primary Stakeholders Secondary Stakeholders

Management Lean

MPU Staff Demand planner

Internal customers Maintenance

33

reliability and validity of adopted statements. Explanations documented during the direct observation process were reaffirmed by the logistics engineer and production team of the MPU. Quantitative data from 2013 and 2016 was analyzed using Microsoft Excel, to gain insights in the case situation regarding the increase in product variety and the production process. The analysis was used to validate to the research model, as it demonstrated factors in the current state that influenced the performance of the Kanban system in a high variety manufacturing environment. Detailed description is provided in the chapters ahead. Expert opinion from the logistics engineer responsible for the MPU was required in identifying and eliminating data of any phased-out part, to ensure validity of data.

3.5 Reliability and validity

The reliability and validity was ensured based on the different dimensions mentioned by Karlsson (2009): (i) by using multiple sources of data, i.e., interviews, direct observation, qualitative, and quantitative data, construct validity was realized; (ii) internal validity was attained as interpretation of information relating to the Kanban system in this study was confirmed by the logistics and the production team at the MPU; and (iii) external validity relating to the generalization of this research was difficult to achieve, based on the limitation of performing a single-case study. Nevertheless, this research contributes to the limited research on Kanban system operation in a high variety manufacturing environment. Similarly, it highlights factors to be considered in the design and analysis of the system performance in such setting.

3.6 Ethical responsibility

34

35

4 Case Description

This section provides a schematic representation of the material flow within the order driven factory (ODF), its production planning process is summarized and an overview of the MPU manufacturing environment is provided.

4.1 Material Flow

Figure 6 illustrates the flow of parts within the internal supply chain of the ODF. The blocks represent distinct processes; however, emphasis is on the processes in the blue blocks representing the MPU and its internal customers. The process of material handling and production control between the MPU and its internal customers is achieved via a Kanban system, as depicted in the figure below. Master Production Schedule Reflow Lacquering KANBAN MPU

APV Driving Unit

Mechanism Basic Body

Final Assembly Sol gel KANBAN Laser Printing Cutting Unit Shaving Unit Assembly PDF KANBAN KANBAN KANBAN KANBAN KANBAN KANBAN

Figure 6: Part Flow- Internal Supply Chain ODF

36

the Driving Unit Mechanism (DUM), the engine and gears are attached. The next step involves merging the complete power supply to a basic body housing, this is performed in the Basic Body Assembly (BBA). The semi-finished products are sent to the final assembly, where the required parts for completion of the end product are received. The MPU is responsible for the delivery of several parts to multiple internal customers including: BBA; Lacquering (involves the use of a special type of lacquer on the moulded parts); Printing; Sol gel; Cutting Unit; Shaving Unit Assembly (SUA); Process Driven Factory (PDF) and Final Assembly. It can be deduced from figure 6, that the interaction between the MPU and its internal customers is not considered to be sequential.

4.2 Production Planning Process

37

4.3 Overview of the Moulding Production Unit (MPU)

Over the last years, several changes have occurred in the business strategy of the organization, resulting in an increase in product variety. Hence, data from distinct periods was compared to depict certain variations in relation to the increase in product variety. However, focus is on the current state, as the Kanban operation has remained the same over the years. Similarly, the shop floor layout has also stayed the same, the layout is considered as functional, where moulding machines are grouped together. The MPU is responsible for the manufacture of a variety of parts that differ mostly in colour and specification, i.e., geometry. The raw material used in the production of moulded parts is known as Resins. The production of these parts involves a single step before the parts are placed in a buffer, where they await retrieval by customers. Hence, a single route is involved in the MPU production process. The MPU is considered as a non-repetitive manufacturing environment, based on the definitions proposed in table 1.

Important variations to consider include:

1) High Variety Parts- Variety experienced at the MPU, stems from certain parts which include: front shell; deco panel; deco frame; scarf; and front grips, required by internal customers. Although there are multiple customers of the MPU, some require the high variety parts, while others require generic parts or a mixture. For instance; BBA and PDF require mostly generic parts, while the sol gel, lacquering and final assembly require a mix. Table 6 illustrates variation in these parts in 2013 and 2016 respectively, from the 1/1/13 to 1/12/13 and 1/1/16 to 1/12/16. Also the available resource (mould), variations in color and the total volume of moulded parts are illustrated in the table.The total volume depicts the increased complexity in relation to the limitation in moulds required for the production of these parts.

Table 6: Variation in High Variety Parts

Year 2013 Year 2016

High Variety Parts Moulds Colors Total Volume Moulds Colors Total Volume

Front Shell 4 10 2.745.998 6 15 4.495.304

Deco Panel 3 6 2.646.031 2 13 3.300.942

Deco Frame - - - 4 10 3.268.686

Scarf - - - 1 6 1.746.974

38

It can be deduced that the current state in comparison to the former, has an increase in complexity, thus, requiring more fire-fighting skills to ensure a smooth production process.

2) Moulded Parts- It is pivotal to consider the number of moulded parts produced in the distinct time frames, as it has impacted on the operation of the system, specifically in terms of machine and moulds. In 2013, the total number of moulded parts was 95 and 2016 accounted for 145, resulting in a 65% increase.

Figure 7: Variation in Parts

3) Buffer Space- This is defined in relation to the required buffer between the MPU and its customers. Following the increase in parts, thus, resulting in an increase in the number of trolleys required, the floor space is considered as limited. This is attributed to the fact that the size of the buffer and ODF has remained the same over the last years. For instance, in 2013, the buffer could accommodate the trolleys, however, in the current state, there is approximately a percentage increase of 82%. The trolleys are used for the storage of moulded parts, while awaiting customer retrieval. As a consequence, some trolleys are placed in a warehouse, relatively close to the MPU. This could be considered as a form of waste in relation to excessive movement lost in transportation.

Figure 8: Variation in Trolleys

39

5 MPU Kanban System

This section provides a detailed description of the MPU Kanban system, its operation and problem definition. Operation of the system involves kanbans representing a variety of parts, produced on shared and dedicated resources (machines and moulds). The service level and WIP are considered as the key performance measures of this system from an operational perspective. These parameters are of strategic importance to the MPU, as the aim is to maintain a high service level while limiting the amount of WIP in the system.

5.1 System Operation

Operation of the system is based on a pull strategy, where production control and material handling involves the use of production order kanban (POK) and withdrawal kanban (WK). Figure 9 provides a schematic representation of the MPU Kanban operation. The system comprises of: (i) Kanban board of approximately 30m long, which is used for production planning and display of kanbans; (ii) a pair of POK and corresponding WK for each part; and (iii) a buffer that serves as the outbound buffer for the MPU and the inbound buffer for the customer.

MPU CUSTOMER

Kanban Board

Flow of WK and POK Kanban Retrieval Board

Inbound buffer for produced parts Outbound buffer to Customer

WK POK WK POK WK POK WK POK WK POK WK Material flow 1 POK WK 2 POK 3 5 4 POK

WK Full trolley with WK and POK attached

POK WK

40

When a customer pulls a full trolley with parts from the buffer as seen in step 1, the POK on the trolley is placed on the Kanban retrieval board at a short distance to the customer workstation, as in step 2. The Kanban retrieval board mitigates excessive movement between the customer and the MPU. Retrieval of these cards is performed by an operator at intervals of 2 hours. In step 3, the POK is scanned by an operator at the MPU to register the parts received by its customer, thereafter, the POK is placed at the bottom of the Kanban board, while it awaits the corresponding WK. In step 4, the WK is placed on the Kanban retrieval board only when the trolley of parts is emptied at the customer workstation. Retrieval of the WK also occurs at intervals of 2 hours. However, unlike the POK that is placed at the bottom of the Kanban board, in step 5, the WK is placed on the coloured zones (green, yellow or red) of the Kanban board. The WK acts as a trigger for planning the next production period, i.e., planning occurs every 2 hours, based on the arrival of the WK at the MPU Kanban board. The minimum production run time for every part is 8hours, and production will occur, when the planning process has occurred and the prioritization format established. Thus, the production period is considered to be a minimum of 8 hours.

The placement of the POK over its corresponding WK implies that production is bound to occur. This step should also fix the production run size for that period, i.e., only cards (pair of WK and POK) which are on the Kanban board at that moment should be produced. At the start of production, the POK is removed from the Kanban board and placed on an empty trolley beside the producing moulding machine. The corresponding WK remains on the Kanban board, until the trolley is filled with parts, i.e., one full trolley is equivalent to 1440 parts, and this is applicable to every part. In which case, the WK is removed and placed with the corresponding POK on the trolley. A periodic transfer is applied, as replenished trolleys are placed in the outbound buffer of the MPU. This reduces the number of downstream trips between the MPU and its internal customers. The next section explains the operation of the Kanban board.

5.2 Daily Scheduling using the Kanban decision board

41

the interpretation of the Kanban board is uniform across all shifts and the rules of the system are adhered to.

5.2.1 Operating rules of the Kanban Decision Board

Figure 10 illustrates a typical Kanban board at the MPU which displays the WK and POK. There are three zones on the Kanban board that signify production: green; yellow; and red. The placement of the kanbans for all part starts from the green zone to the red, however, production begins from the red to the green zone.

Figure 10: MPU Kanban Board

42

producing machine. The reasoning behind this practice is unknown to the production team; or (ii) planning is yet to occur, in which case the corresponding POK is at the bottom of the board in the white zone. Furthermore, the white zone at the top section of the board, is used to display other relevant information, which aid in the production process, for instance; the production sequence represented by the yellow triangular cards.

The following rules apply in the operation of the Kanban board at the MPU:

1. The kanban sizing for a part within a given period is represented by the colored zones as: (i) green- is the cycle stock expressed as average forecasted demand; (ii) yellow zone- considered as the buffer stock and covers any deviation in the green zone. This is the difference between the maximum and average demand; and (iii) red zone- is the maximum WIP/ safety stock. A 15% safety factor based on managerial decision is included in this system;

2. The trigger for planning occurs when the WK of a part is returned to the Kanban board. However, production will occur when the POK is placed over the corresponding WK, resulting in a fixed production run;

3. Production of a part should occur when the minimum run time of 8 hours is achieved. A fixed production run should be completed regardless of the number of returning kanbans for that specific part. However, there is an exception, for instance, if part A and B are scheduled to be produced on the same machine or mould and the current production run is fixed for part A. Production of the returning kanban for part A should only be included in the current production run, one at a time if the minimum run time of 8 hours of part B is yet to be reached. Otherwise, a changeover is necessary to begin production on part B. Changeovers are in terms of colour and/or mould;

43

5. The sequencing strategy defined for moulds with multiple colours, is to begin production from the lightest to the darkest colour, i.e., from the left to the right side of the Kanban board. The goal is to minimize the amount of waste and quality issues resulting from colour change-overs.

5.3 Comparison of types of Kanban Systems

Table 7 summarizes the key points of the previously described Kanban systems, these include: single-card Kanban (hybrid and pull), dual-card Kanban and the MPU Kanban. This is essential in establishing the type of Kanban system implemented in the case company. Although the MPU Kanban involves the use of POK and WK respectively, the basic operating principle is grounded in the single-card Kanban system that is based on a pull strategy.

Table 7: Kanban systems comparison

Therefore, the MPU system is considered as a single- card system, however operating with the POK and WK. The MPU Kanban is also compared with the Toyota Production System (TPS), as shown in table 8. The TPS is considered as a dual-card system, involving the use of both cards in its operation.

Single-card Single-card Dual-card MPU Kanban

Operating principle Hybrid Pull Pull Pull

Number of buffer One (Inbound) One (Inbound and outbound)

Two (Inbound and outbound)

One (Inbound and outbound)

Distance between two processes Small Small Large Small

WIP between two workstations Large Small Moderate Large

Production Control MRP POK POK POK and WK

Material Handling WK POK WK WK

44

Factors TPS MPU System MPU Challenge

Demand Variability Low Variable Variability in Takt time and unlevelled production schedule

Product Variety Low High High inventory holding costs and complexity

Production Run Long with stable batch size

Varying with fixed batch size

Difficulty in planning and unlevelled schedule

Cycle time Fixed Differs per part Complexity in workload levelling and increased production lead time

Changeovers Few Multiple and complex Complexity in changeover levelling

Table 8: Differences between TPS and MPU Kanban system (Adapted from Lander & Liker 2007)

5.4 Problem Definition

45

6 Findings and Discussion

This section provides the findings from analysing and evaluating current practices in relation to variables proposed in the research model. This is essential in answering the research questions and establishing their implication in coping with high product variety. It is important to note that high levels of WIP inventories were used to buffer the effects of uncertainties in the MPU production process. As a high amount of WIP between the MPU and its internal customers allowed for maintaining a high efficiency, thus, resulting in a high service level.

6.1 Environmental Characteristics

The environmental characteristics are defined based several factors proposed by Krajewski et al. (1987) and White & Prybutok (2001). The factors considered are shop floor factors that influence the operation of the Kanban system, these include: shop floor layout; changeovers and setups; worker flexibility and skills; resources (machines and moulds); and resource reliability.

6.1.1 Shop floor layout

46

mix and improve the throughput time. Nevertheless, a disadvantage of group technology in the MPU, could imply frequent rearrangement of machines. This is attributed to the flexibility in the planning process in relation to variation in daily product mix.

6.1.2 Resources- (Machines and Equipment)

Resources are defined in terms of the machines and moulds. In 2013, the total number of machines was 62, of which dedicated machines accounted for 52, resulting in 16% of shared machines. However, in 2016, 28 machines were dedicated and 31 shared, resulting in 53% of shared machines. This is illustrated in the figure 11 below. A dedicated machine is defined as a one with a single mould for parts of the same geometry, however, with multiple colours, i.e., only colour changeovers occur on dedicated machines. Shared machines involve multiple moulds for parts with distinct geometry and multiple colours, i.e., changeovers involve colour and/or mould. It can be deduced that the MPU has a high percentage of shared resources in terms of machines and moulds. Thus, the increase in product variety and number of moulded parts, has led to a constraining effect and complexity in production scheduling. This problem occurred whenever a set of common resources were required for the production of different parts during the same period. Production scheduling involved operational decisions regarding the production process, such as machine production sequence displayed on the Kanban board.

Figure 11: Number of Moulding Machines

0 10 20 30 40 50 60 2013 2016 Moulding Machines

47 6.1.3 Resource Reliability

In an ideal Kanban environment, it is assumed that machines are 100% reliable, i.e., with no machine failure (Krajewski et al. 1987). However, this is far from the reality, as machine failure is bound to occur in all manufacturing environments (Prakash & Feng 2011). Similarly, in the MPU, machines and moulds are subject to failure, even more so with the continuous adaptions required to accommodate the production of more parts. According to Savsar (1997), the performance of Kanban systems is affected by the failure rate of machines, as machine utilization is high when preventive maintenance (PM) in addition to corrective maintenance (CM) is applied. Thus, the maintenance strategy implemented at the MPU is a combination of both. Preventive maintenance is considered as planned, with the aim of minimizing the frequency at which breakdowns occur. Corrective maintenance are unplanned activities that mitigate the severity of breakdown on resources and the overall system performance. Additionally, in a high product mix environment, the service level and amount of WIP is influenced by the maintenance strategy applied, i.e., the level of WIP is highest when no failure occurs, thus, resulting in a high service level and is lowest with increased occurrence of CM. Hence, the MPU strives for frequent PM, in order to reduce the occurrence of CM and the risk of unfulfilled customer demand.

6.1.4 Setups and Changeovers