Accommodating the workload control concept to the dynamics of real life job shops

Accommodating the workload control concept to the

dynamics of real life job shops

Thesis

MSc. Double Degree in Operations Management

University of Groningen, Faculty of Economics and Business

Newcastle University, Business School

December 2013

Name: Victor Cransberg

Email: v.cransberg@student.rug.nl

Student number: 1766112 (University of Groningen) 120505276 (Newcastle University)

Supervisor: dr. M. (Martin) Land Co-assessor: prof. C. (Chris) Hicks

Preface

This paper has been written as the concluding chapter of my Dual Degree Master program in Operations Management. It has been a challenging but very rewarding journey, which could not have been completed without the help and assistance during this period of several people.

To start, I would like to thank Warse Klingenberg for supervising me during the first period of this project. Our interesting discussions led to the initial ideas for this paper. Unfortunately, Warse was not able to supervise me anymore after a few months due to personal circumstances and Martin Land became my new supervisor. He has guided me further through the process of writing my thesis and his input has been of great importance to the final result. Martin, thanks for the nice discussions we had and the interesting insights you gave me. Your enthusiasm about this field of research motivated me a lot. Furthermore, I would like to thank Chris Hicks as co-assessor from the University of Newcastle. During my visit in Newcastle, last October, your comments on my work were really helpful.

Finally, I would like to thank my supervisor Mark Pieter Alsem from the case company for giving me the opportunity of being an intern at the company and to conduct my research over there. I also would like to thank him and my other colleagues for the great intern period.

Groningen, December 2013

Abstract

Workload control (WLC) is a production and control method, primarily designed for job shop environments. The release decision is considered an important mechanism of the concept. This mechanism serves two important control functions, a timing function and a load balancing function. In real life job shops, the trade-off decision between these two functions is often complicated due to several additional complexities like sequence-dependent set-ups, process batches or desynchronized shift schedules. Each of these complexities needs its own trade-off considerations. Currently, the WLC concept is lacking clear guidelines on how to control the shop floor load in case of these complexities. To contribute to this gap, a decision framework is developed that can assist in deciding at which input level to control the load when considering the complexities. An explorative case study is performed to determine how the additional complexities can be considered in controlling the load. The results reveal that traditional considerations in controlling the load are no longer sufficient. It is shown how additional complexities can be taken in consideration to increase the practicality of the WLC concept for real life job shops.

Table of Contents

1. Introduction ... 5

2. Theoretical framework ... 6

2.1 Workload control ... 6

2.2 Traditional considerations in input control ... 7

2.3 Additional considerations in input control ... 8

2.4 Load control decision framework ... 9

3. Methodology ... 11

3.1 Data collection ... 12

3.2 Data analysis ... 13

3.3 Company description ... 13

3.3.1 Company overview ... 13

3.3.2 Demand data and current load ... 14

3.3.3 Current planning procedures ... 15

3.4 Remarks ... 16 4. Results ... 16 4.1 Sequence-dependent set-ups ... 16 4.2 Process batches ... 18 4.2.1 Sequential batches ... 18 4.2.2 Simultaneous batches ... 19

4.3 Desynchronized shift schedule ... 20

4.4 Additional complexities ... 21

5. Implications for WLC theory ... 22

6. Conclusions ... 24

7. Bibliography ... 26

8. Appendices ... 29

Appendix I ... 29

1. Introduction

Production planning and control (PPC) is a mature and well developed topic. Within this topic, workload control (WLC) has received increased attention in recent years. The purpose of this approach is to ensure that orders are completed by their due dates while fully utilizing work centre capacity (Fredendall et al., 2010). The WLC concept buffers the shop floor against the dynamics and variation of arriving orders by means of input/output control (Oosterman et al., 2000). According to Hendry et al. (2013), WLC research can be divided in three broad areas: (1) simulation studies (2) theoretical papers and (3) empirical papers. It is argued that insights into functioning of this concept in practice are lacking (Soepenberg et al., 2012a). This empirical study presents important insights into the causes of the difficulties that are often experienced by practitioners when applying this concept in real life job shops. In today’s marketplace, there is an increasing demand for customized products (Slomp et al., 2009). This results in an increasing number of make-to-order (MTO) companies, which are often characterized as high-variety/low-volume (Stevenson et al., 2005). Often, these companies are defined as job shops. In principle, WLC is an approach primarily designed for these job shops. This concept focuses on controlling the workload on the shop floor. The order release decision is the main instrument of this concept (Oosterman et al., 2000). This input decision serves two important control functions, a timing function and a load balancing function (Land, 2004). The timing function determines when a job in the pre-shop order pool is considered for release while the load balancing function avoids the shop floor of being overloaded. The two functions often counteract with each other and a trade-off has to be made between balancing the workload and the relative urgency of orders. Once released onto the shop floor, the throughput of orders is controlled at the dispatching level.

In a real-life job shop, the trade-off decision is often even more difficult due to additional complexities. For example, sequence-dependent set-ups require attention at both the release at the dispatching level (Thürer et al., 2012a). Releasing jobs onto the shop floor in a sequence based on relative urgency of individual jobs might lead to an inefficient sequence at the dispatching level. Moreover, this can lead to an uncontrolled buffer situation. Process batches and a desynchronized shift schedule are other contextual factors that cause similar difficulties to the trade-off decision. These complexities, which are often observed in practice, require a certain sequence of released or dispatched jobs. This sequence can often not be handled in the exact same order at other stations. Therefore, the complexities cause an increase in buffer sizes on the shop floor. It can be argued that the complexities tend to undermine the purpose of the WLC concept, namely to control the workload. Currently, clear guidelines on how to accommodate these dynamics are lacking. Therefore, the following research question is considered:

The structure of this paper is as follows. Chapter 2 presents a conceptual framework based on two dimensions that are considered important in deciding at which input level the load should be controlled in case of additional complexities. This can either be at the release or the dispatching level. Chapter 3 will provide the methods of data collection and analysis used during the field research, followed by a concise case company description. During the field research, it is examined how the complexities can be considered in controlling the load. The results of this field research will be discussed in chapter 4. Furthermore, other implications found in the case company that complicate the trade-off decision are discussed in this chapter. Implications for the current WLC concept are discussed in chapter 5, including the guidance for future research. This paper ends with a chapter that concludes on the findings and that states the research limitations.

2. Theoretical framework

This chapter examines the literature on the WLC concept and the current theory of this concept related to the complexities of real life job shops. Section 2.1 explains the concept of WLC, followed by section 2.2 that discusses the traditional considerations in input control in more detail. Section 2.3 provides the additional considerations that are considered relevant in controlling the load for real life job shop. Hereafter, a section will provide the theoretical background on the proposed decision framework. It is conjectured that two dimensions are of high relevance in determining in which input decision to address complexities, namely the position of the station in the routing of a job at which a complexity is found and the relative criticality of this complexity. This will be elaborated in section 2.4. It leads to a basic framework that will be presented at the end of this chapter.

2.1 Workload control

Production planning decides upon the use of production resources in order to meet future demand at a reasonable cost (Gelders & van Wassenhove, 1981). Choosing the appropriate PPC system can be seen as a strategic decision because this system serves functions as minimizing work-in-progress (WIP) and improving responsiveness to changes in demand (Stevenson et al., 2005). It can be especially challenging for MTO production systems to plan and control production (Hendry & Kingsman, 1989). This is due to varying order mixes and demand that can be characterized as volatile and difficult to predict. WLC is primarily designed for the MTO sector where job shop configurations are common (see e.g. Hendry et al., 2013; Land & Gaalman, 2009).

(e.g., Fredendall et al. 2010; Land, 2006; Thürer et al., 2010; Thürer et al., 2011). It can be argued that empirical studies are lacking due to the limited understanding about how the WLC concept can deal with additional complexities, which are often observed in real life job shops. Therefore, the current WLC concept should be accommodated to these factors in order to increase its robustness in practice.

According to Thürer et al. (2012b), WLC not only controls the WIP and reduces both throughput times and the number of tardy jobs but it also creates a smoother production flow. Eventually, this can lead to a ‘leaner’ shop floor and a just-in-time PPC system. Three input control decisions can be identified in controlling the load (Figure 1): (1) entry (2) release and (3) dispatching (Land, 2004). The entry decision involves accepting or rejecting the job and if accepted, determining the due date. Accepted orders arrive automatically in the pre-shop order pool. The job release decision determines in which sequence the jobs enter the shop floor and when this should take place. To make this decision, a trade-off has to be made between the balancing the workload between work centres and releasing urgent orders (Henrich et al., 2004). A job is identified as urgent if the planned start time has already been passed (Land & Gaalman, 1998). The final decision, dispatching, is made at each work centre and is based on some priority rule (e.g., First-in-First-out). The focus of this paper is on the input decision at the release and dispatching level, as indicated with the dashed red line in Figure 1. The reader is referred to Thürer et al. (2013) for an overview of different input decision rules at the acceptance level. Output control decides when an order is allowed to leave the queue, shop floor or system (Kingsman & Hendry, 2002). The next section will discuss the traditional considerations in making the input decision at the release and dispatching level in more detail. This section will be followed by a section on additional considerations in input decisions, caused by the dynamics of real life job shops.

2.2 Traditional considerations in input control

The main instrument to control the workload is the order release decision, which is an input control decision (Oosterman et al., 2000). The order release decision directly influences the amount of WIP on the shop floor. Furthermore, this function determines the balance on the shop floor (Bechte, 1988). The pre-shop order pool buffers the shop floor against dynamics like cancellation of orders or rush orders. Once the order has been released, it stays on the shop floor until all operations have been completed. Input control and especially the job

release decision can be seen as an important mechanism of the concept because it manages the transition of orders from the planning system to the execution stage (Bergamaschi et al., 1997). This means that arrival of jobs in the pre-shop order pool not necessarily involves direct release of the job onto the shop floor. This transition is facilitated by the order release decision. Especially in manufacturing systems with additional complexities, this decision can be complicated. It is questionable at which input level this decision should be made to control the load. The decision about the level of input control is from now one defined as the load decision.

The order release decision determines which jobs should be released from the pre-order pool (selection phase) and when this should take place (sequencing phase) (Land, 2004; Bergamaschi et al., 1997). The relation between the selection and the sequencing decision is illustrated in Table 1. Balancing the workload and releasing urgent orders are indicated as traditional considerations in input control. It is a constant trade-off between the two functions. In principal, orders in the pre-shop order pool are considered for release and if the next job considered fits in the predefined workload norms on the shop floor, it will be planned for release (Land & Gaalman, 1998). If one or more norms at the station the job passes in its routing is violated, the job remains in the pool and the next job will be considered for release (Thürer et al., 2012b). The orders in the pre-shop order pool are considered for release according to a simple rule (i.e., earliest planned release date or shortest slack). The objective of the release decision is to release the right amount of workload that fits within the workload norms while keeping the throughput times minimized. The release decision can either be taken periodically or continuously (Thürer et al., 2012b). Periodically decisions are taken each period, e.g. once a day or once a week. Continuously taken decisions are triggered when the workload at a station drops below a certain level. In that case, orders are released automatically onto the shop floor.

Phases Timing Function Load Balancing Function

Sequencing x

Selecting x

Table 1 - Function served by each phase of the release procedure (Land, 2004)

2.3 Additional considerations in input control

between load balancing and releasing urgent orders more complicated because non-urgent orders are pulled forward at the expense of urgent orders to form batches (Missbauer, 2002).

Another practical complexity to the trade-off decision is sequence-dependent set-ups. For this complexity, batches are created as well in a buffer to reduce set-ups. However, these batches require a certain sequence in which they are released or dispatched. This creates additional waiting time in a buffer to form this sequence. Eventually, this could lead to an uncontrolled load situation. This complexity can either be handled centrally or locally (Henrich et al., 2004). Organising this centrally means releasing the orders in the right sequence onto the shop floor in order to minimize set-ups. Dispatching rules on the shop floor organise this locally by reordering the sequence in the queue in front of a station. However, handling the set-ups centrally via order release leads to restricted queues and thus decreased effectiveness of the dispatching rule (Kim & Bobrowski, 1995). In practice, sequence-dependent set-ups are often considered locally (Fernandes & Carmo-Silva, 2011). However, this not always leads to the best performance. This is especially the case for tight workload norms which decrease the effectiveness of the dispatching rule. Coating is an example of an operation for which sequence-dependent set-ups are common. In planning jobs for this operation, jobs are sequenced in such a way that colour changes between batches are minimized.

Finally, a desynchronized shift schedule (e.g. a station operating in three shifts, delivering to a station operating in a single shift) is often observed in practice. Shift scheduling is all about matching demand with resources, with resources being the employees to be scheduled and released orders as demand (Koole & Van Der Sluis, 2003). In a desynchronized shift situation, the amount of daily production hours differs per station. This is common in practice but neglected in the WLC literature, until now. This factor adds complexity in controlling the shop floor load, as it may strongly increase WIP at times when one station is active while another is not.

The WLC concept aims to create a continuous flow of jobs between stations on the shop floor. The above mentioned complexities tend to create a certain discontinuous flow due to specific set-up requirements or a desynchronized shift schedule. These contextual factors are dynamics of real life job shops and should therefore be well considered in controlling the load (Soepenberg et al., 2012a). The decision at which input control level these complexities should be taken in consideration is an important one. However, clear guidelines that can assist in this decision are lacking. Therefore, a conceptual framework is presented in the next section.

2.4 Load control decision framework

assist in the decision making on which input level (i.e. the release or the dispatching level) to control the load. The effectiveness of this framework in practice will be discussed later on in this paper.

The first dimension of the conceptual decision framework is the position of the complexity in the routing of a job. It is claimed by Soepenberg et al. (2012b) that only focusing on release is not sufficient for long and complex routings. Therefore, it is argued that the further downstream a station is positioned, the less powerful the input decision at the release level is. This is supported by Henrich et al. (2004), who stated that it is difficult to control the workload if routings become longer. On the other hand, Thürer et al. (2010) presented a prioritization rule in the order pool based on routing length to be able to better control the shop floor at release level. Breithaupt et al. (2002) proposed that downstream stations should be neglected in the release decision because these stations often cause difficulties in releasing larger orders. From the above it can be concluded that the further downstream the complexity occurs, the less powerful the release decision becomes. Hence, more attention should be given to controlling the load at the dispatching level. Moreover, the consideration at release may even needlessly disturb consideration of other factors which are important at the stage of release. Based on this reasoning, this dimension is included in the framework.

The second dimension of the decision framework is the criticality of the complexity. For this dimension, several sub-dimensions are identified. The first sub-dimension is whether the station at which the complexity occurs is a bottleneck or not. A bottleneck station is the most heavily utilized machine which acts to constrain throughput (Enns & Costa, 2002). A constant flow of work towards the bottleneck item is necessary. Therefore, the release decision plays an important role if the station is considered a bottleneck station. This is also confirmed by a study of Stevenson et al. (2011), which outlined the significance of bottlenecks and indicated the need for more research into the effects of bottlenecks on the performance of WLC. It is showed by Fernandes et al. (2013) that WLC can also be effective in unbalanced situations if norms are set appropriately. But not only control at the release level is considered important for critical stations. This is indicated by Stevenson et al. (2011), who found that bottlenecks often shift over time in job shops. This means that input control at the dispatching level is also be considered important because after the moment of release, a bottleneck may shift to a station in the routing of the released job. The jobs upstream of this bottleneck station should then be routed to other stations by means of the dispatching rule. This indicates the relative importance of control at the dispatching level. It can be said that the further downstream a complexity occurs, an increased need for a hybrid form of control at both the release and dispatching level exists.

arrives too late at this station, it might stay on the shop floor for a relative long time while it sits and wait for the next batch to be filled. In this case, criticality is considered high.

Based on the two dimensions, the decision framework is developed (Figure 2). This framework will facilitate decisions on which level to control the load. As can be seen in Figure 2, the two dimensions are outlined along the horizontal and vertical axis. Each dimension is split up into two values. A complexity can either be at an upstream or downstream station and the criticality of the complexity can either be high or low. A station is considered upstream if it is one of the stations that have their position in the upper half of the routing in a job. Otherwise it is considered a downstream station. In each quadrant, the proposed level of input control is provided. It should be noted that the proposed level of control should have the focus in controlling the load. Hence, it is not argued that control mechanisms at other input levels can be neglected.

Figure 2 – Conceptual decision framework

This chapter has discussed the WLC concept and its current difficulties in coping with contextual factors that add complexity to the trade-off decision between balancing the workload and on time delivery of individual jobs. Based on this discussion, a framework was presented. This framework can assist in the decision where (i.e. at the release or dispatching level) to take the additional complexities into consideration. An explorative case study is performed to research how these complexities should be taken in consideration in the load decision. The next chapter will present the methodology used in this paper, followed by a chapter on the results.

3. Methodology

(Voss et al., 2002). Field research is regarded appropriate when looking for practical solutions that can be translated in a theoretical contribution (DeHoratius & Robinovia, 2011). In field research, topics can be considered for which little prior theory exist (Edmondson & McManus, 2007). It is argued that only little is known about the applicability of the WLC concept in practice. In particular, little is known about the potential difficulties and complexities in considering implementation of WLC in practice.

Based on the existing literature, a conceptual framework is developed that can assist in the decision where to control the load. This can either be at the release or the dispatching level or at both levels. However, this framework does not provide specific guidance in the decision how to take the complexities in consideration in making the load decision. Therefore, an explorative case study is performed at a case company. Within the manufacturing system of this case company, all three previous mentioned complexities are found. Based on the examinations done in this company, lessons can be learned on how to take these additional complexities in consideration in controlling the load. The case company description will be provided after an explanation on the methods of data collection and analysis.

3.1 Data collection

The data for this paper is collected longitudinally for a period of six months on a full time basis within the case company. In the beginning of the six months, open-ended interviews were held with the head of planning and a second planner, who is responsible for the planning of several large production departments in which the complexities occurred. Daily observations on the production floor revealed the large amount of workload in several buffers. In combination with quantitative data from the ERP system, it became clear that the some complexities lead to an uncontrolled workload situation. Due to the presence of a scanning system on the shop floor, this data was available.

3.2 Data analysis

The start of the analysis was to understand the performance indicators of the production process and the challenges experienced within the planning department in making the trade-off decision. After this, the collected data was structured and organised based on the three complexities. For each complexity, the current practice in the case company was analysed to determine how this complexity was handled in the load decision. This analysis focused on daily planning procedures and the considerations in prioritizing orders in the pre-shop order pool. Other implications that added more complexity to the trade-off decision became visible during the observations, conversations and open-ended interviews. The proposed framework was discussed with several people within the organisation. Several meetings were organised with the plant manager, head of planning and several lean manufacturing coaches. Based on these meetings and the observations done, the practicality of the framework was evaluated and solutions were proposed to control the load. In follow-up meetings, it turned out that other implications made the trade-off decisions even more complicated. Evidence was gathered on these implications through informal meetings and observations.

3.3 Company description

Due to the explorative nature of this field study, a concise description of the case company is required. This description will be provided in the next section. The structure of this description is adapted from the structure that Stevenson & Silva (2008) used in describing their cases. Therefore, this chapter is divided in the following three sections:

1. Company overview

2. Demand data and current load 3. Current planning procedures 3.3.1 Company overview

The company produces aluminium profiles and is located in the north of the Netherlands. Although, part of a larger group, the company can be considered as a medium-sized enterprise (MSE) with an annual turnover of 50 Million Euros and around 150 employees. In terms of Amaro et al. (1999), the company can be described as both a versatile manufacturing company (VMC) and a repeat business customizer (RBC). In a VMC, each order is unique and a RBC handles orders that are repeatedly ordered by a customer. Within the case company, each order requires a unique mould so for one-off orders these moulds are engineered-to-order. On the other hand, the company is also involved in repeat production of products over the length of a contract with specific customers. This makes managing the shop floor a relative difficult task because as a RBC, speed can be seen as the most important order winning factor while flexibility is more important for a VMC.

located are England, Switzerland and French. The goal for 2013 is to produce 16000 ton aluminium profiles.

Industry Tonnage Percentage

Transport 1.388 10%

Engineering 976 6%

Building & Construction 9.364 64%

Furniture & Interior 611 4%

Retailers 866 6%

Others 1.404 10%

Total 14.609 100%

Table 2 - Overview of different industries the case company serves (2012)

3.3.2 Demand data and current load

Within the company, several large departments make up the production organisation (see Appendix I for the shop floor lay-out). These are: extrusion, anodizing, coating, packaging and wrapping. Other smaller departments (i.e. correction, thermal break, foiling, sawing and mould fabrication & correction) complete the production organisation. In between the departments, the aluminium profiles are transferred in cradles. The extrusion press is always the initial station in the routing of a job. After being extruded, simultaneous batches are filled which are processed in the ageing oven. This operation provides the aluminium profiles with the right material hardness.

The shop can be characterized as a job shop with some characteristics of a flow shop. Firstly, a dominant flow direction can be distinguished. Products that follow this routing pass the following three stations: extrusion, ageing oven and packaging. In general, the routings are relatively short. An overview of the different routings and their occurrence can be found in Appendix II. Secondly, the flow has the same direction on the shop floor, which is another characteristic of a flow shop (Oosterman et al., 2000). Nevertheless, the shop is considered a job shop because of the routing sequence variability and the variation in routing lengths.

In 2012, the arrival intensity of orders is considered high with an average of 95 orders per day. As indicated by Henrich et al. (2004), high order arrival intensity and relatively short processing times are desirable when considering WLC. In this case, processing times range from several minutes to almost a whole day. The process can be characterized as divergent, meaning that the job can follow different routings after being pressed (Dogger et al., 2010). The three complexities (i.e. sequence-dependent set-ups, process batches and a desynchronized shift schedule) are all found in one routing. This routing involves the extrusion, ageing, coating and wrapping step (Figure 3). Therefore, the remainder of this paper focuses on these stations.

Currently, the focus of the company is to reduce its overall WIP level in the factory and to increase its on-time-delivery from 90% to 95%. Several improvements projects are currently running in order to achieve this goal. The next section will describe the current planning system, followed by a section about the variety in job sizes.

3.3.3 Current planning procedures

The planning department consists of five people, each with their own responsibilities. To start, there is one head of planning who oversees the whole planning process. Next to this person, one person is responsible for the planning of jobs on the extrusion press and one for the planning of the packaging, anodizing and coating department. The fourth and the fifth planner are responsible for the planning of the smaller production departments and transportation of finished goods, respectively. The order lead time depends on the product family and is listed in Table 3. A product family is defined as those orders that follow the same routing. The different product families can be found in Appendix II (Table 6). The order lead time is build up from the pre-shop throughput time and the shop floor throughput time.

Product family Pre-Shop LT (days) Shop floor LT (days) Total LT (days) St. Dev. LT (days)

BRUUT 12 4 16 10

RAL 15 9 24 13

VOM1 14 8 22 11

Table 3 - Order lead times per product family

As can be noted from Table 3, the standard deviation of the total lead time is relatively high. This is one of the reasons why this company is having difficulties with delivering all orders on time. The lead time is determined by the head of planning, based on calculations of the ERP system. The ERP system calculates this based on the finite-capacity scheduling approach, in which the system in actively searching for an optimal schedule (Hopp & Spearman, 2008).

3.4 Remarks

From the above, it can be concluded that orders are pushed into the production system based on the planning made by the extrusion planner. At this moment, almost none communication exists between the extrusion planner and the planner of the other departments about what kind of orders are considered for release. This causes often an imbalance of workload between stations. A lot of parameters have to be considered at the release level, making this a complicated task. For example, the type of the alloy of the aluminium, the mould at the extrusion press, the routing of the job and the relative urgency of that job are considerations in the release decision. From a practical point of view, it is questionable which factors are considered most important at the release or dispatching level and how to consider these factors to control the load. In the next section, the examinations on how to consider the complexities in the load decision are discussed. Also, other implications found during the field research that add complexity to the trade-off decision are discussed.

4. Results

To accommodate the WLC concept for real life job shops, the framework is used to determine at which input level to control the load for each of the three complexities. Hereafter, initial ideas are proposed about how to consider the complexities in the load decision. This chapter is divided into four sections. The first three sections cover the three complexities. They are structured as follows. Firstly, the complexity as examined in the company will be analysed. Secondly, the consequences of the complexity in the current situation are discussed. Thirdly, the decision framework will be tested to determine at which level the load should be controlled for that complexity. Finally, the consequences of applying this framework are discussed. In the fourth section, the remaining issues as observed in practice that complicate the trade-off decision will be presented. In the next chapter, the implications for the current WLC theory are discussed. This paper ends with a conclusion.

4.1 Sequence-dependent set-ups

The consequence of the above is a large amount of workload in the buffer of this station to make the most optimal sequence possible. The WIP level in the buffer of this station fluctuates strongly, with an average buffer size of four days of workload and a standard deviation of 1,5 day (Figure 4). This is considered an uncontrolled WIP situation by the management of the plant. The goal is to reduce the WIP by half, at least. In 2013, 22% of the orders had this station in their routing (Appendix II). The position of this station can be considered downstream because it is the third station of the four stations in total. Next to this, this complexity is also considered critical because of the need for an optimal job mix at this station to reduce set-ups. Furthermore, this station turns sometimes into a bottleneck due to the push philosophy in planning the jobs on the extrusion station. Due to the unpredictability of this shiftiness, increasing the capacity of this station to remove this bottleneck is often a reactive action.

Figure 4 - Fluctuations in the WIP level of the coating station in 2013

Control of the load of this station is rather difficult because it is a downstream critical station. The sequence-dependent set-ups make this control even more complex. Based on the proposed framework, it is argued that the input control decision should be at both the release as dispatching level. To do this, sequencing orders in the pre-shop order pool should be done based on an extra criterion. This criterion is the colour of the orders that needs to be coated. In this way, not only the dispatching rule controls the load in the buffer of this station but also the release decision contributes to this. Currently, this criterion is neglected at release. By considering a colour category at release level, the load in the buffer of this station can be reduced. For example, lighter colours can be released at one day while releasing darker colours the following day. In this way, the load of this station’s buffer is more controlled that it currently is. The dispatching rule, based on differences in colour, then ultimately sequences the order in the optimal order.

Fully controlling the load at the release level is found unsuitable for this station. This is due to uncertainty upstream of the coating station, mainly caused by the possibility of defects produced at the extrusion station. If the jobs would have been released in an optimal sequence to reduce set-up times at the coating station and a job would be extruded with quality problems, the coating line would suffer from this in terms of reduced output. This, in turn, affects the total output of the production process. A small buffer in front of the coating station

should cover the risks of defects upstream. Therefore, the sequence of jobs for a downstream station with sequence-dependent set-ups should be considered at the release level in terms of general product characteristics, while an optimal sequence is created at the dispatching level. 4.2 Process batches

Next to the sequence-dependent set-ups, process batching is found within the production process as another complexity to the load control decision. Two types of batching are found, simultaneous batching and sequential batching. Each of the two complexities will be discussed in the next sections.

4.2.1 Sequential batches

Within the case company, sequential batching is examined at the extrusion press (Figure 5). As can been seen in the fourth column, the different orders require the same mould (the same mould-id applies to all of them, i.e. 263515). Therefore, the four orders are combined in one sequential batch. In an ideal situation, the LPS (first column) is leading. However, as indicated above, aspects like maximum machine utilization and minimum set-ups need to be considered as well. Due to these factors, orders are pulled forward (sometimes for several weeks) to be combined in one batch with orders that require the same mould. This creates a certain imbalance on the shop floor because less attention is given to balancing the load between stations. Furthermore, non-urgent orders are pulled forward at the expense of urgent orders. It is not considered an option by the plant manager to release individual jobs if they could have been batched because of the potential loss of money due to an increased number of set-ups.

Figure 5 – Sequential batching at initial station

Next to the imbalance on the shop floor that is created by this complexity, other issues arise. The non-urgent orders that are produced too early are stored in the buffers on the shop floor, sometimes for several weeks. This increases the probability of quality problems, e.g. due to dust. Furthermore, it happens too often that jobs are ‘lost’ because of a lack of feedback from the shop floor about the location at which the job is stored. This results in unnecessary re-production of that order which is at the end more costly than a change-over at the extrusion station.

the sequence in which the jobs are processed. This indicates the relative importance of the release decision.

As described above, the potential costs of storing non-urgent orders on the shop floor are considered higher than the cost of a set-up at this station due to the relative high cost of re-producing a job. However, removing this complexity is not an option. Therefore, a solution should be found to consider this complexity at the release level while minimizing additional re-production costs. A possible solution for this is to include a time limit in the pre-shop order pool that prevents non-urgent orders that fall outside this time limit of being pulled forward to be batched. In this way, sequential batches can still be formed while keeping control of potential additional re-production costs. It depends on the product family how long this time limit should be. Specifically, it depends on the routing length of the product family and the possible presence of other complexities in this routings. For example, the time limit can increase in sequence-dependent set-ups are also needed to be considered at the release level. For a family of products that require only a small number of production steps, this time limit may be shorter. For each product family, a fixed time limit can be set. This time limit serves as a simple planning rule which should be followed in planning the jobs for release. 4.2.2 Simultaneous batches

Another form of batching as examined in the case company is simultaneous batching, which is found at the ageing oven for heat treatment. This process step provides the extruded aluminium profiles with the right material characteristics (i.e. the hardness of the material). The ageing time depends on the aluminium alloy (Table 4). In the case company, almost 90% of the orders are F22 or F25. Therefore, only a relative few orders have a twelve hour cycle time. Jobs that are coming from extrusion station are batched on lorries and when the lorries are full, the whole batch is processed in the ovens.

Type of alloy Hardness Temperature (C) Cycle Time (h)

AlMgSi 0,5 F22 185 6

AlMgSi 0,5 F25 185 6

AlMgSi 0,7 F27 175 10

AlMgSi 1,0 F28 175 12

AlMgSi 1,0 F31 175 12

Table 4 - Overview cycle time for simultaneous batching in case company

Due to the potential problems, caused by this complexity, and the position of this station in the routing of a job, this station is considered a critical upstream station. According to the framework, the load should be controlled at the release level. This is confirmed by the current practice in the case company of releasing the jobs onto the shop floor. The company has designed its own practical solution to release the jobs in the right sequence to avoid an overload in the buffer of this station. Planning is restricted to release the higher alloys (i.e. F27, F28, and F31) at the end of the week. During the remainder of the week, F22 and F25 can be released. This results in another consideration in the release decision, next to the relative urgency of individual jobs (translated in the LPS), the type of mould to combine individual orders in sequential batches and the colour categories to reduce set-ups at the coating station. This makes sequencing at the release level a challenging task.

To execute this task, it is proposed that the complexity that is considered most critical, is translated into an objective. This objective should be achieved by releasing jobs in the right sequence while less critical complexities are bounded to several planning rules. For the case company, the sequential batching complexity is considered most critical due to value that is added at this station to the end product. Therefore, the objective at release would be to minimize set-ups at the extrusion station by creating sequential batches with jobs that require the same mould. This objective is subjected to a strict planning rule, namely to sequence the higher alloys at the end of the week to create simultaneous batches. It is not allowed to deviate from this rule without permission of higher level management. Another, less strict, rule is to take the predefined time limit per product family into account in creating sequential batches. It should be aimed to follow this rule, however, it should not stop jobs from being released. Finally, another planning rule that should be followed but which should not prevent jobs from being released is to sequence jobs, that have the coating station in their routing, based on colour category. In this way, all complexities can be considered at release.

4.3 Desynchronized shift schedule

The final complexity that is discussed in this paper is a desynchronized shift schedule. This factor is found at the coating station, which normally produces in one shift per workday while the up- and downstream stations produce in three shifts. Due to the current high load in front of this station to handle the sequence-dependent set-ups, this complexity does not receive particular attention in the case company. Nevertheless, if the company achieves a substantial load reduction by reducing the set-up times at the coating station, this complexity will require attention.

It seems interesting to focus on the station upstream of this station. In terms of criticality, that station becomes critical because orders should be dispatched in a sequence to supply the next station always with enough workload. This can either be at the beginning of the shift or it can be that load arrives in the buffer during the one shift. The other sixteen hours, the jobs upstream of this complexity should be dispatched to other stations. In this way, an overload of work in front of the station for which a single shift is scheduled can be avoided. For the case company, this seems difficult because the station upstream is the ageing oven that produces in simultaneous batches. The station starts processing when this batch is filled (see section 4.2). On average, this is four times a day but this could also be three or even two times. This variability is caused by the size of the aluminium profiles and the performance of the extrusion press. Therefore, to avoid starvation, a certain amount of workload is required in front of this complexity. In general, it can be stated that in case of a desynchronized shift schedule, control at the dispatching level at the station upstream of the station with less shifts scheduled becomes critical.

4.4 Additional complexities

Next to these above discussed complexities, other factors are found that add complexity to the trade-off decision between balancing the workload and release of urgent orders. These factors are rush orders and a relative low order position. These factors will be discussed in this section.

To start, rush orders play a disturbing role in controlling the load. Controlling this complexity at release is found difficult because the relative long time of preparing a mould. This makes control at the release level rather inflexible. Therefore, the focus of considering this complexity in controlling the load should be on the dispatching level. Once the job is released onto the shop floor, the job can be speeded up by control at the dispatching level. This causes difficulties if this rush order has the coating operation in its routing. Due to the presence of sequence-dependent set-ups, rush orders can create problems in dispatching an optimal mix of orders. For example, the rush order might to be coated in a dark colour while an optimal order mix in lighter colours was planned to be dispatched. Disturbing this optimal mix can result in other orders to be delivered late (in case of equal coating capacity), depending on the remaining slack of these orders. To consider this complexity at the dispatching level, the effects of disturbing the optimal mix in terms of orders potentially being delivered late should be made visible. For example, it can be possible that one large order forces three other orders, ordered by three different customers, to wait for several days before an optimal mix can be made again. For each of these three smaller orders, the remaining slack should be made visible. Based on this analysis, it can then be decided whether it is economically feasible to process the rush order or not.

relative long period, waiting in the buffer of downstream stations. This counteracts with the timing function of the WLC concept because this function aims to only release urgent orders to avoid pushing orders into the system that are not desired yet by the customer. However, a low order position creates a situation in which this goal is difficult to achieve. As described at the end of section 4.2, the case company’s objective is to reduce set-ups subjected to several restrictions that aim to balance the workload. In case of a low order position, this objective might shift more to the timing function of the WLC concept. This means that prioritizing jobs in the pre-shop order pool based on their LPS becomes more important.

In previous chapter, the proposed decision framework was used to determine at which level the load should be controlled in case of complexities to the trade-off decision (Figure 6). Next to this, several possibilities have been provided that can assist in deciding how different complexities can be considered in control of the load. These possibilities are based on the examinations done during the explorative case study. Furthermore, additional complexities as examined in the case company were discussed. A discussion of the findings of this paper and the accompanied implications for the current WLC theory is provided in the next chapter.

Figure 6 - Applied decision framework in case company

5. Implications for WLC theory

The results of the field research show several aspects of real life shops that complicate the functioning of the WLC concept in practice. Therefore, the trade-off decision between balancing the load and the relative urgency of individual jobs is often more difficult than indicated in the current WLC theory. The implications for the WLC theory based on the empirical findings are discussed in this chapter. This paper will end with the conclusions.

to a strong deterioration of the shop floor performance. The dispatching rule is considered most important in controlling this complexity in case of a downstream station. However, the set-ups should also be considered at release to avoid an overload of work in front of this complexity. It is argued that sequence-dependent set-ups should be considered at release based on the product category. On the dispatching level, an optimal sequence should be found. This should result in a controlled shop floor load situation. More research is required to provide insights into how to consider product categories at the release level while optimizing the sequence at the dispatching level.

Creating sequential batches is one of the characteristics of high volume job shops. These batches negatively influence the goal of the WLC concept, namely controlling the workload (Missbauer, 2002). However, this factor cannot be neglected in real life job shops that have the same process characteristics as the case company. As examined, sequential batches are considered more important than releasing individual orders. Therefore, more empirical research is required on a decision framework that can assist in deciding when it is appropriate to create sequential batches. The two dimensions of the proposed framework in this paper can serve as a starting point and these might be extended with more parameters, e.g. workload norms and a maximum time limit in the pre-shop order pool.

Simultaneous batches create a certain discontinuous flow on the shop floor because the station only starts to run when the batch is filled (Hopp & Spearman, 2008). This makes controlling downstream buffers difficult. A buffer after this complexity is always required to prevent downstream stations from starvation. Also, a buffer in front of this station is required to allow filling up the simultaneous batch. In case of an upstream station at which this complexity occurs, the load should be controlled at the release level. This control can be translated in a strict rule in sequencing the jobs for release. This rule allows only the products that are designated to the same simultaneous batch to be released in a predefined time frame during the week. In this way, this complexity can be taken in consideration in the trade-off decision. Further research is required to investigate the effects on the control of the load if this complexity is located more downstream.

As can be seen in Figure 6, several complexities should be taken in consideration in making the release decision. However, optimizing all complexities is not an option. Therefore, these complexities should be translated into objectives and planning rules. This can be done by prioritizing the examined complexities based on the criticality of each complexity. The complexity that is considered most critical should be translated into an objective while the other complexities are bounded to several planning rules. This is seen as an important contribution to the current WLC theory because this increases the robustness of the concept for practical application. Further empirical studies should be conducted to determine the practicality of this theoretical contribution.

critical. This dispatching rule should dispatch an amount of work to supply the station at which the complexity occurs with enough workload for the scheduled number of production hours to avoid starvation of this station. During the remaining hours, the load should be dispatched to other stations to avoid blocking in the buffer of the complexity. This implication can be a starting point for simulation studies to design a new dispatching rule that can consider this complexity at the dispatching level.

Furthermore, it became clear during the field research that rush orders play a disturbing role in the effectiveness of the dispatching rule. At the release level, the pre-shop order pool buffers the shop floor against rush orders (Oosterman et al., 2000). However, it can be concluded that rush orders have a high impact in case of sequence-dependent set-ups. Control at the dispatching level is then found most appropriate. Therefore, a relation between the degree of impact and the level of control can be found. It is argued that control at the dispatching level becomes more important in case of a relative high impact of the rush order on the output of the process. Empirical studies are required to research this relationship.

The last implication to the current concept is the role of the relative order position of the job shop. It turned out that in case of a low order position, non-urgent orders are pulled forward because of a lack of urgent orders. Therefore, it is argued that the timing function of the load decision becomes increasingly important in case of a relative low order position. In case of busier period, the balancing function becomes more important. This is another important implication for the current WLC theory. It is claimed that the dominance in controlling the load shifts between the balancing function and the timing function of the WLC concept, dependent on the current order position of the job shop. Further research is required into the effects of this function shiftiness on controlling the load in real life jobs shops.

This section has presented the implications of this research for the current WLC theory. It opened-up a discussion on the current shortcomings of the concept. Several future research directions have been provided. The last chapter will provide the conclusions of this paper.

6. Conclusions

The presence of sequence-dependent set-ups turned out to be very disturbing in controlling the load. This study showed that control of this complexity should be considered at both the release and the dispatching level to be able to control the load. In doing so, jobs are combined in broad product categories at release while at the dispatching level, an optimal sequence is dispatched. As a consequence, a relative small buffer on the shop floor is required to be able to dispatch such a sequence. To handle the sequential batches, it is proposed in this paper to set a certain time limit in the pre-shop order pool per product family. This time limit depends on both the routing length and the possible presence of other complexities in the routing of the product family. Furthermore, this paper highlighted the creation of a discontinuous flow, caused by the simultaneous batching complexity. Due to the disturbing character of this complexity, this study proposes that this complexity should be considered at release as a strict rule in releasing orders. Next to this, the results of this study revealed the criticality of the dispatching rule upstream of the desynchronized shift schedule complexity. This dispatching rule should avoid the station at which this complexity occurs of starvation or an overload of work. The optimization of all complexities that need to be considered at the release level turned out to be unrealistic. Therefore, this paper proposes to transpose the complexity, which is considered most critical, into an objective of the release function while the less critical complexities are bounded to different planning rules. In this way, several complexities can be considered at release in the most effective way.

7. Bibliography

Amaro, G., Hendry, L. & Kingsman, B. (1999) Competitive advantage, customization and a new taxonomy for non make-to-stock companies, International Journal of Operations &

Production Management, Vol. 19 (4), p. 349-371.

Bechte, W. (1988) Theory and practice of load-oriented manufacturing control, International

Journal of Production Research, Vol. 26 (3), p. 375-395.

Bergamaschi, D., Cigolini, R., Perona, M. & Portioli, A. (1997) Order review and release strategy in a job shop environment: A review and a classification, International Journal of

Production Research, Vol. 35 (2), p. 399-420.

DeHoratius, N. & Rabinovich, E. (2011) Field research in operations and supply chain management, Journal of Operations Management, Vol. 29 (5), p. 371 – 375.

Dogger, B., Land, M.J. & Van Foreest, N.D. (2010) Making CONWIP work in high-variety manufacturing, Proceedings of the International Working Seminar on Production Economics, p. 133-143.

Edmondson, A.C. & McManus, S.E. (2007) Methodological fit in management research,

Academy of Management Review, Vol. 32 (4), p. 1155-1179.

Enns, S. T. & Costa, M. P. (2002) The effectiveness of input control based on aggregate versus bottleneck work loads, Production Planning & Control, Vol. 13 (7), p. 614-624.

Fernandes, N.O. & Carmo-Silva, S. (2011) Order release in a workload controlled flow-shop with sequence-dependent set-up times, International Journal of Production Research, Vol. 49 (8), p. 2443-2454.

Fernandes, N.O., Land, M.J. & Carmo-Silva, S. (2013) Workload control in unbalanced job shops, International Journal of Production Research

Fredendall, L.D., Ojha, D. & Patterson, J.W. (2010) Concerning the theory of workload control, European Journal of Operational Research, Vol. 201 (1), p. 99-111.

Gelders, L.F. & van Wassenhove, L.N. (1981) Production planning: A review, European

Journal of Operational Research, Vol. 7 (2), p. 101–110.

Hendry, L.C, Huang, Y. & Stevenson, M. (2013) Workload control: Successful implementation taking a contingency-based view of production planning and control,

International Journal of Operations & Production Management, Vol. 33 (1), p. 69-103.

Henrich, P., Land, M.J. & Gaalman, G. (2004) Exploring applicability of the workload control concept, International Journal of Production Economics, Vol. 90, p. 187 – 198.

Hopp, W.J. & Spearman, M.L. (2008) Factory Physics, New York: McGraw-Hill.

Karlsson, C. (2009) Researching Operations Management, Oxon: Taylor & Francis.

Kim, S.J. & Bobrowski, P.M. (1995) Evaluating order release mechanisms in a job shop with sequence-dependent set-up times, Production and Operations Management, Vol. 4 (2), p. 163-180.

Kingsman, B.G. (2000) Modeling input-output control for dynamic capacity planning in production planning systems, International Journal of Production Economics, Vol. 68 (1), p. 73-93.

Kingsman, B. & Hendry, L. (2002) The relative contributions of input and output controls on the performance of a workload control system in Make-to-Order companies, Production

Planning and Control: The Management of Operations, Vol. 13 (7), p. 579–590.

Koole, G. & Van Der Sluis, E. (2003) Optimal shift scheduling with a global service level constraint, IIE Transactions, Vol. 35, p. 1049-155.

Land, M.J. (2004) Workload control in job shops: Grasping the tap, Ph.D. Thesis, University of Groningen, Ridderkerk: Labyrint Publications.

Land, M.J. (2006) Parameters and sensitivity in workload control, International Journal of

Production Economics, Vol. 104 (2), p. 625–638.

Land, M.J. & Gaalman, G.J.C. (1998) The performance of workload control concepts in job shops: Improving the release method, International Journal of Production Economics, Vol. 56-57, p. 347-364.

Land, M.J. & Gaalman, G.J.C. (2009) Production planning and control in SMEs: time for change, Production Planning & Control: The Management of Operations, Vol. 20 (7), p. 548-558.

Missbauer, H. (2002) Lot sizing in workload control systems, Production Planning &

Control, Vol. 13 (7), p. 649-664.

Slomp, J., Bokhorst, J.A.C. & Germs, R. (2009) A lean production control system for high-variety/low-volume environments: a case study implementation, Production Planning &

Control: The Management of Operations, Vol. 20 (7), p. 586-595.

Soepenberg, G.D., Land, M. & Gaalman, G. (2012a) Workload control dynamics in practice,

International Journal of Production Research, Vol. 2 (15), p. 443-460.

Soepenberg, G.D., Land, M.J. & Gaalman, G.J.C. (2012b) Adapting workload control for job shops with high routing complexity, International Journal of Production Economics, Vol. 140, p. 681-690.

Spearman, M.L. & Zazanis, M.A. (1992) Push and pull production systems: Issues and comparisons, Operations Research, Vol. 40 (3), p. 521 – 532.

Stevenson, M. & Hendry, L.C. & Kingsman, B.G. (2005) A review of production planning and control: the applicability of key concepts to the make-to-order industry, International

Journal of Production Research, Vol. 43 (5), p. 869-898.

Stevenson, M. & Silva, C. (2008) Theoretical development of a workload control methodology: evidence from case studies, International Journal of Production Research, Vol. 46 (11), p. 3107-3131.

Stevenson, M., Huang, Y., Hendry, L.C. & Soepenberg, E. (2011) The theory and practice of workload control: A research agenda and implementation strategy, International Journal of

Production Economics, Vol. 131, p. 689-700.

Thürer, M., Silva, C. & Stevenson, M. (2010) Workload control release mechanisms: from practice back to theory building, International Journal of Production Research, Vol. 48 (12), p. 3593-3617.

Thürer, M., Silva, C. & Stevenson, M. (2011) Optimising workload norms: the influence of shops floor characteristics on setting workload norms for the workload control concept,

International Journal of Production Research, Vol. 49 (4), p. 1151-1171.

Thürer, M., Silva, C., Stevenson, M. & Land, M. (2012a) Controlled order release: a performance assessment in job shops with sequence-dependent set-up times, Production

Planning & Control, Vol. 1, p. 1-13.

Thürer, M., Stevenson, M., Silva, C., Land, M.J. & Fredendall, L.D. (2012b) Workload control and order release: A lean solution for Make-to-Order Companies, Production and

Thürer, M., Stevenson, M., Silva, C. & Land, M (2013) Towards an integrated workload control (WLC) concept: The performance of due date setting rules in job shops with contingent orders, International Journal of Production Research, Vol. 51 (15), p. 4502-4516.

Voss, C., Tsikriktsis, N. & Frohlich, M. (2002) Case research in operations management,

International Journal of Operations & Production Management, Vol. 22 (2), p. 195-219.

8. Appendices

Appendix I

Figure 7 - Lay-out shop floor case company

The lay-out of the shop floor of the case company is provided in Figure 7. I1 and I2 are the locations in the factory where raw aluminium material enters the shop floor. The produced orders leave the shop floor at O1 or O2, depending on the product family. A description of each capacity centre is provided in Table 5, including the relevant capacities. Note, the capacity is only provided for the stations considered most relevant for this research. The capacity is calculated based on a 5-day working week, 46 weeks per year.

Centre Description Capacity (kg/year) Shifts/day

1 Extrusion press 2 and ageing oven (VO) 1 & 2 10.488.000 3 2 Extrusion press 3 and ageing oven (VO) 3 & 4 8.004.000 3

3 Packaging 11.095.200 3 4 Sawing 5 Correction 6 Foiling 7 Anodizing 3.829.500 3 8 Coating 3.220.000 1 9 Thermal-break 10 Wrapping 4.554.000 3

Appendix II

Based on the capacity centres provided in Table 5, the possible routings through the factory are determined. In the first half of 2013, a total of 11423 orders are produced with a total weight of 8621 ton aluminium. To evaluate the different routings, a Pareto analysis is performed (Figure 8). A total of 71 different routings are recognized. As can be noted from the figure, more than 80% of the orders follow only a few routings. This is illustrated in more detail in Figure 9. It becomes clear that more than 80% of the total orders follow only three routings. These are BRUUT, RAL, and VOM1. The details per routing are given in Table 6. In Table 7, a From-To matrix is provided for these routings. The most notably observation from Table 7 is that almost all orders have the ageing oven in their routing. The effects of this are described in section 4.2.2.

Figure 8 - Pareto analysis (100% - overview)

Figure 9 - Pareto analysis (80% - detailed)

Routing Stations # Orders Percentage (%)

BRUUT Press – Ageing Oven – Packaging 4765 41.71

RAL Press – Ageing Oven – Coating – Packaging 2539 22.23

VOM1 Press – Ageing Oven – Anodizing - Packaging 1615 14.14

Table 6 - Most performed routings (2013)

0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 0 1000 2000 3000 4000 5000 6000 BR U U T L_ RAL ISOR V L_ IIS LVO M1 BR U BV BR ALI U P AB BIS O BR U _ I BU LB BLB Z PIZLI _PIRA L B R B AB _BIBI LBEW U LVOM Product Family Cum. Percentage 80% 0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 0 1000 2000 3000 4000 5000 6000

BRUUT RAL VOM1 BRU_B

From To Press Ageing Oven Anodizing Coating Quality Inspection Packaging Press 0 0,96 0 0 0,02 0,01 Ageing Oven 0 0,01 0,18 0,27 0,12 0,42 Anodizing 0 0 0 0,01 0,01 0,17 Coating 0 0 0,01 0,02 0,02 0,28 Quality Inspection 0 0,01 0 0,03 0 0,12 Packaging 0 0 0 0 0 0