Making CONWIP Work in High-Variety Manufacturing Erik H. Mol, Dr. M.J. Land (Supervision), and Dr. J. Riezebos (Co-assessor) Elaborating on a paper of Bart Dogger (2010)

(1)

-1-

Making CONWIP Work in High-Variety Manufacturing

Erik H. Mol, Dr. M.J. Land (Supervision), and Dr. J. Riezebos (Co-assessor)

Elaborating on a paper of Bart Dogger (2010)1.

Faculty of Economics and Business, University of Groningen, Master Thesis Operations and Supply Chains.

Abstract

In order to stay competitive many companies use a production planning and control (PPC) approach such as, the relatively simple, CONWIP to improve their throughput performance. The problem with this approach is its inability to deal with specific high-variety aspects of production environments, negatively influencing the throughput performance. More advanced approaches are available that deal with one or more specific high-variety aspects of a production environment. This paper tries to improve the performance of CONWIP by learning from other approaches.

First, possible high-variety aspects influencing a production environment are described. CONWIP, variants of CONWIP and PPC approaches similar to CONWIP are then described, analyzing their ability to deal with certain high-variety aspects. Based on this analysis a framework is designed with minimal adaptations to CONWIP. For each high-variety aspect a possible minimal adaptation is provided which improves the ability of CONWIP to deal with this high-variety aspect.

The use of this framework is demonstrated in three cases studies. Two of these cases explore high-variety aspects in different production environments. It shows what possible minimal adaptations should have been made when CONWIP was implemented, to improve the performance. The third case illustrates an actual implementation of CONWIP with minimal adaptations at a paint production plant. Several adaptations to CONWIP are made to deal with the specific high-variety aspects of the paint production process. The stepwise implementation decreased the throughput time, while remaining easy to understand for the workforce.

Keywords: CONWIP, Production Planning and Control, High-variety aspects, variability, minimal adaptations, implementation, evaluation, case-studies.

1. Introduction

In the past decades much research within the field of operations management focused on production planning and control (PPC) approaches. The approaches also had their impact in the managerial area, making it possible to set stricter criteria when it comes to speed, dependability, WIP and utilization. Implementing the right PPC approach can be the next step in a more efficient process and becoming more competitive. This is especially important when companies grow larger and deal with more and specifically demanding customers, which can result in high-variety production environments. High-variety production environments can be characterized by several variability aspects, such as inter-arrival time variability, variability of due date allowances, routing variability, routing sequence variability and routing length variability (Henrich et al. 2004). This high-variety production environment is the working field of Make-To-Order (MTO) companies, where generally customized products are made, instead of producing a standard product to keep up a stock level.

1_{Dogger, B., 2010. Making CONWIP Work in High-Variety Manufacturing. Preprints of the Sixteenth}

(2)

-2-

Over the years several PPC approaches, that limit the Work In Process (WIP), have emerged (see e.g Framinan et al. 2003 and Gonzalez et al. 2012) . CONstant Work In Process (CONWIP) (Spearman et al. 1990) is a pull-based approach, meaning that order release is based on a trigger. CONWIP was designed as a PPC approach that would be more applicable to high-variety production environments. With CONWIP only the Work In Process (WIP) is set to a maximum amount. This is represented by a maximum amount of orders in the process, which can be easily controlled and visualized by attaching a card to an order. When an order is finished, the card can be assigned to a new order, triggering a release into the system. The approach is known for its simplicity and is therefore easy to use. This form of limiting the workload in the process is the foundation of many approaches, which will be mentioned later in this paper. Limiting the workload results in less work on the shop floor, with better transparency and more predictability of its throughput performance. The problem with CONWIP is that it is unable to deal with specific kinds of high-variety aspects, as it has no workload balancing capability. This means that there is no balanced distribution among the workstations, which affects the throughput time performance of the system negatively (Germs and Riezebos, 2010). An imbalance of the workload can cause idle workstations in the system, even though orders for these stations are still available. This increases the throughput time, as the order will have to wait before released, which is discussed by Land and Gaalman (1998). High-variety aspects from production environments are a cause of these imbalances between workstations in a system.

There are more advanced approaches described in literature, which are designed to deal with more aspects of high-variety production environments (see e.g., Krishnamurthy and Suri, 2009). However, for the more advanced approaches few successful implementations are described (Stevenson et al. 2011) and implementation issues are common (Hendry et al. 2008).

The goal of this paper is to improve the ability of CONWIP to deal with high-variety aspects. By learning from the more advanced PPC approaches, minimal adaptations to CONWIP will be proposed to provide a better workload balancing capability when specific high-variety aspects occur. Available research on PPC approaches, with controlled order release, that limit the workload and their ability to deal with high-variety aspects is limited. Most research focuses on applicability of one approach (e.g. Henrich et al. 2004) or the comparison between approaches (e.g. Lodding et al. 2003, Germs and Riezebos, 2010). Therefore the paper will also provide a preliminary overview of high-variety aspects and the anticipation of PPC approaches on it, making a modest contribution to the knowledge about PPC approaches, variety aspects and how to deal with these aspects.

The structure of this paper will be as follows. Section 2 will continue with an overview of high-variety aspects in high-high-variety production environments. Section 3 describes the research methodology, indicating how the framework is set up and explaining the function of the case studies. Section 4 will give an overview of PPC approaches. In section 5 a framework with minimal adaptations to CONWIP is given. The use of this framework will then be explained by three case studies in section 6. Where two case studies will explore high variety aspects in two different production environments and a case will illustrate how CONWIP with minimal adaptations is used for implementation. Section 7 will discuss the use of the framework. Finally, in section 8 some conclusions will be drawn.

2. High-Variety Aspects in Production Environments

(3)

-3-

widely available, defining it e.g. as the state of non-uniformity of characteristics. Many PPC approaches, which make use of controlled order release and a maximum amount of WIP, are developed to deal with specific types of variability. As mentioned in the introduction, research on high-variety aspects and the anticipation of PPC approaches on these variability aspects is limited. In most of the research on PPC approaches, specific high-variety aspects and their appearance in particular production environments are mentioned (see, e.g. Kingsman et al., 1989 and Stevenson and Silva, 2008). However, little research has focused on the relationship between the effectiveness of PPC approaches and high-variety aspects.

This section will further elaborate on these types of variability, as they are the main input for high-variety aspects. A preliminary framework with key high-variety aspects will be provided, though it will not be complete, as it is a subject for research on itself.

In much of the research variability types are separated into two categories. For example Lu et al. (2011) mention process and demand variability, which is a basic distinction regarding the origin of the variability. It is either inside the production process, or a result from factors outside the process, such as customer demanding different products. Hopp and Spearman (2000: 261) also distinguish two categories, namely, processing time variability and flow variability. The first is focusing on the effective process time of a job at an individual workstation. This refers to the total time a job is tied to a workstation, no matter if it is being processed or not. When an order passes through multiple stations during production, high processing time variability will result in flow variability. Flow is referring to the transfer of an order from one station to another. The two variability types show a clear relation, as one affects the other.

Henrich et al. (2004) developed an overview of order characteristics for which the applicability of the PPC approach Workload Control (WLC) is checked. WLC is specifically designed for MTO companies, therefore the variability aspects from Henrich et al. will be used as input for the framework of high-variety aspects in this paper (figure 2.1). The first two types of variability mentioned are non-technical aspects of an order, namely inter-arrival time and variability of due date allowances. Inter-arrival variability can occur at the start of the process as incoming orders might not be distributed equally over time. For workstations individually it might also be a result of for example variable processing times of preceding workstations. Variability of due date allowances relates to the mix of urgent and non-urgent orders to process. The other variability types are technological aspects of an order, which are the result of e.g. specific operations and specific routing for a product between workstations. Processing time was already mentioned before. Secondly, there is the routing of a product. The routing can converge or diverge, and can be further characterized by routing length variability or routing sequence variability. The first is referring to a variable length of routings, as orders might need more or less workstations in their production routing. Routing sequence can vary, as a workstation could be a finishing operation for one order, but can also be the middle station for another order. These high-variety aspects are summarized in figure 2.1.

(4)

-4-

Fig. 2.1. – Framework to categorize high-variety aspects

3. Research Methodology

The main focus of this research is to improve the ability of CONWIP to deal with high-variety aspects, by making minimal adaptations. In order to do this, the next section will describe CONWIP and other PPC approaches. CONWIP will be described more extensively as it is the origin of the proposed approach. The alternative approaches in the next section are selected on a number of criteria. Firstly, these approaches are designed for high-variety manufacturing environments. Secondly, they function similar to CONWIP, controlling order release and limiting the workload. Thirdly, they can deal better with one or more specific high-variety aspects than CONWIP. Based on literature or argumentation this ability of a PPC approach to deal with specific high-variety aspects is described. These approaches are then used to learn from and to design minimal adaptations for CONWIP to anticipate on specific high-variety aspects. The function of the next section is not to give a comprehensive literature review of all PPC approaches available, as this is already done in other studies (see e.g. Gonzalez et al. 2012 and Stevenson et al. 2005). Describing the PPC approaches and their ability to deal with certain high-variety aspects serves as an input for the design of minimal adaptations to CONWIP. By learning from the alternative approaches in the next section, a minimal adaptation for each specific high-variety aspect from figure 2.1 is designed. A number of principles are used to design the minimal adaptations to CONWIP. Firstly, the minimal adaptation should improve the ability of CONWIP to deal with a specific high-variety aspect. Secondly, CONWIP controls order release and limits the workload, which has several benefits, as mentioned in the introduction. These benefits should be retained with the design of the minimal adaptations. Thirdly, the minimal adaptations should be designed as simple as possible. CONWIP is known for its simplicity. Applying minimal adaptations should not make the approach unnecessarily complicated.

The minimal adaptations are presented in a framework, which connects a minimal adaptation to a high variety aspect they deal with. By analyzing the high-variety aspects of a company, CONWIP with minimal adaptations can be applied based on the framework. In this way CONWIP can be used as a PPC approach which can be made more suitable for a production environment with one or more of the specified high-variety aspects.

Three exploratory case studies are described and their main function is to provide a better understanding of how to use the framework from CONWIP and minimal adaptations. There are several

High-variety aspects

Non-technological aspects Technological aspects

Inter-arrival variability Processing time variability

Variability of due date allowance

(5)

-5-

studies highlighting the value of case studies when exploring and elaborating on new theory (see e.g. Stuart et al. 2002 and Voss et al. 2002). The framework is first applied to two MTO companies. Both of the companies are currently using a different and advanced PPC approach, with controlled order release and a maximum amount of WIP, in order to accomplish short lead times and to be able to respond fast to specific customer demands. Therefore CONWIP with minimal adaptations will not be actually implemented in these companies. The two cases will be used to illustrate how to analyze the high-variety aspects from figure 2.1. Analyzing these aspects is required to make minimal adaptations to CONWIP for a particular situation. It will show which high-variety aspects are most dominant in the production environment, and therefore the need to adapt CONWIP to deal with this high-variety aspect. The same method to analyze the production environments is used for both companies. The companies were visited to discuss the current PPC approach used with the production managers and to observe the production process and the shop floor. It provided qualitative data about several factors, such as the initial motivation to use an advanced PPC, the influence of high-variety aspects on the production process and the possible improvements of the PPC approach used. Planning data were provided by the companies as quantitative data. These data are used to identify the high-variety aspects from figure 2.1. The raw planning data provides information such as production steps, routing information and processing times. The two companies both use more advanced PPC approaches, therefore planning data is available to analyze the high-variety aspects. After the analysis of the high-variety aspects the current PPC approaches of the companies can be compared to the possible outcome when CONWIP with minimal adaptations would have been applied. The companies also use two different advanced PPC approaches, therefore the differences can also be compared between the two cases.

A third case will illustrate an implementation of CONWIP with minimal adaptations in a paint production plant. The company used no PPC approach that controlled order release with a workload limit before the implementation of CONWIP with minimal adaptations, thus implementation is described from the start. The case shows how CONWIP with minimal adaptations can be used as a PPC approach on itself. CONWIP is implemented and adapted by making small steps at a time, in order to improve its ability to deal with high-variety aspects.

4. Production Planning and Control approaches

This section will give an overview of common PPC approaches that share the characteristic of order release based on a maximum WIP level. CONWIP will be described more extensively than the other PPC approaches in this section. For more extensive descriptions of these approaches we refer to the cited articles.

4.1 CONWIP

(6)

-6-

release. The part of the process for which a WIP cap is set is referred to as a control loop. CONWIP uses one loop over the whole process for controlling the workload. An easy way to control the WIP is by attaching cards to an order. Only a maximum amount of cards is allowed in the control loop. This use of WIP as a control variable has shown to be superior over approaches that use throughput as a control variable (Hopp and Spearman 1996, Suri, 1998). When focusing on these approaches that use WIP as a control variable, CONWIP is the most basic. The PPC approaches that will be described below make use of the same principle. They all set a maximum level of WIP to a specific loop, where orders that are not being processed will wait outside a loop until they are released in the loop. Figure 4.1.1 shows a simple production environment controlled by CONWIP. The control loop is represented by the black oval.

Fig. 4.1.1 – CONWIP controlled production environment.

CONWIP has several advantages which are also pointed out by Spearman et al. (1990) as it promotes an orderly shop floor with less clutter which will make environmental problems more noticeable. These advantages are mainly due to WIP control and thus advantages for all the approaches mentioned in this paper. The controlled order release by WIP control reduces the influences of both types of both non-technological aspects. As orders arrive they will first go into the order pool waiting outside the process for their controlled release as described above. This will smoothen the inter-arrival time variability. An important role from the WIP cap, as described by Hopp and Spearman (2000), is that it will stabilize the WIP. For example any disruption will not cause an extensive growth of the WIP, it will remain the same, making it easier to predict throughput times, resulting in a better predictability of the due dates. Orders can then be released on well-planned starting dates, and will only be released when they are more urgent then other orders. Therefore CONWIP is able to coop with variable due-date allowances. Again, other approaches also have these characteristics, but have been developed to deal with other types of variability as well.

(7)

-7- 4.2 Variants of CONWIP

Several variants of CONWIP have been developed in the two decades after the introduction of CONWIP (see, e.g., Framinan et al., 2003), the main focus of these variants was to overcome a specific issue of CONWIP. Rose (1999) presents CONLOAD, CONstant LOAD, which was developed to overcome the issue of different processing times at workstations. Different processing times can for example be caused by changes in the product mix that is produced. CONWIP only controls the number of orders and not the workload in processing times, this might cause fluctuating workload in the system and imbalance of the workload at individual workstation level. Therefore, with increasing processing time variability the effectiveness of CONWIP decreases. To deal with processing time variability Rose (1999) calculates the amount of load an order adds to the workload of the bottleneck workstation. This is calculated by the sum of bottleneck processing times from an order divided by the average throughput time of that order type. We argue that this calculation of bottleneck workload in relationship to the average throughput time of an order can also provide a partial correction for variable routing length. As the routing length increases the average throughput time is also likely to increase. The share of the bottleneck in the average throughput time is then likely to decrease, allowing workload for this bottleneck to enter the process.

M-CONWIP provides another way for better workload balancing. Germs and Riezebos (2009) criticize CONWIP for only having one control loop, which cannot balance work across individual workstations when routings differ per order. They discuss M-CONWIP as an alternative. This approach sets a control loop for every possible routing on the shop floor. In this way the M-CONWIP is able to smoothen the influence of high routing variability. Figure 4.2.1 shows a simple production environment with 3 different routings. In the study of Germs and Riezebos (2009) M-CONWIP is presented with a process that has diverging routings. Khojasteh-Ghamari (2009) provides an example of CONWIP with multiple loops in a situation with converging routings. Both researches show that M-CONWIP is able to provide a better workload balancing capability than CONWIP.

Fig. 4.2.1 – M-CONWIP controlled production environment.

(8)

-8-

provides a possibility to deal with bottlenecks, as these can be decoupled and uncoupled from a line as separate loops.

4.3 POLCA

Paired-cell Overlapping Loops of Cards with Authorisation (POLCA) is a PPC approach that aims at increasing the speed of job transfer between cells and a better balancing of the workload. It is a part of the companywide management philosophy Quick Response Manufacturing (Suri, 1998), which focuses on lead time reduction. POLCA attaches control loops to two subsequent workstations or cells. The order release and WIP control functions like a CONWIP loop, as a number of cards limits the number of orders within a loop. An order is only allowed in the production process when a card is available. Within routings that use more than two workstations, the loops overlap each other. In this way chains of loops can be made, and at the point of overlapping loops an order will need a card for both loops to start producing at that workstation. Therefore the approach is very suitable to deal with high routing variability. According to Vandaele et al. (2008) this makes POLCA more self-regulating and able to control every possible routing. Riezebos (2010) also explains that POLCA can deal with potential bottlenecks. We argue that POLCA can also deal with routing length and sequence variability. With every extra station in a routing, an extra loop is needed. A different sequence will result in loops between different workstations. So when both aspects get higher, the number of loops increases.

Figure 4.3.1- POLCA controlled production environment.

(9)

-9- 4.4 DBR

Perhaps the most well-known approach amongst the PPC approaches mentioned in this article is Drum-Buffer-Rope (DBR). It was developed by Goldratt and Cox (1984) in order to control the workload of a process with a stable bottleneck. DBR controls the number of orders or workload hours up to and including the bottleneck. Here the production rate of the bottleneck determines the speed of the production process. A time buffer is subtracted from the nominal start time of an order. The functioning is nearly the same as attaching a control loop from the start of the process to the bottleneck, where the capacity of the bottleneck functions as the WIP cap. Figure 4.4.1 shows a simple production environment with a control loop to the bottleneck.

Fig. 4.4.1 – Controlled order release for a bottleneck in a production environment.

If no attention is paid to the bottleneck, workload might congest in front of the bottleneck. Other stations can become idle as the workload limit is reached, and congested in front of the bottleneck, blocking the release of orders for the idle workstations. The bottleneck is not in the framework of high-variety aspects from figure 2.1, though as Stevenson et al. (2005) point out the presence of a bottleneck is typical for a production process. Stevenson et al. mention that when the types of routing variability increase it is likely that it will be harder to identify the bottleneck, as the bottleneck can move and occur anywhere. We argue that potential bottlenecks need to be controlled, though creating control loops to clear non-bottlenecks unnecessarily increases the complexity of a PPC approach.

4.5 WLC

Workload Control (WLC) is a PPC approach specifically designed for the needs of MTO companies (Stevenson et al. 2005). Like previous methods this is done with control loops with workload caps, in processing times, but with WLC each workstation will control its own workload. A control loop from order pool till the workstation is set, returning afterwards. With WLC orders can only go to the shop floor if they fit the norm that is set for the workstation. Orders that cannot go on the shop floor will remain in a pool. Figure 4.5.1 shows a simple production environment controlled by WLC.

(10)

-10-

Figure 4.5.1 – WLC controlled production environment

respectively from Bechte (1980), Bertrand and Wortmann (1981) and Tatsiopoulos (1983), but when it comes to order release they essentially follow the same approach. According to Oosterman et al. (2000) the difference between the three methods is the way the workload of the workstation is calculated. Oosterman et al. (2000) provide a fourth method that is modified to smooth the influence of routing length variability and routing sequence variability. This is done by making a correction for the position of the station in the routing. When the routing length increases and a station is at the end of this routing, dividing the workload by the stations positions will then allow more workload to enter, which will avoid stations from becoming idle. This modified method measures the workload per individual station, and divides this workload into a direct load and upstream load. Together this forms the aggregate load, to which a maximum is set. By controlling individual workstations WLC can deal with a changing bottleneck, which is typical for high-variety production environments with high routing variability (Stevenson et al., 2005). Land (2009) introduced a version of WLC using cards to control the workload, called COntrol of BAlance by CArd-BAsed NAvigation (Cobacabana). According to Hendrich et al. (2004) WLC is applicable to all high-variety aspects shown in figure 2.1. Though this sounds promising, WLC is still coping with many implementation issues (Hendry et al., 2008; Stevenson et al., 2011).

To conclude, the PPC approaches mentioned above have been summarized in table 4.1, together with the high-variety aspects from the previous section. Checked cells indicate that the approach is able to smoothen the influence of a variability type.

5. Minimal adaptations to CONWIP.

(11)

-11- CO N WI P C O N LO A D M -C O N WI P PO LC A LB -PO LC A DBR WLC Non-technological aspects Inter-arrival variability √ √ √ √ √ √ √

Variability of due date

√ √ √ √ √ √ √ allowances Technological aspects Processing time √ √ √ √ variability Routing variability √ √ √ √ Routing length √ √ √ variability Routing sequence √ √ √ variability Potential bottleneck(s) √ √ √ √

Table 4.1 – PPC approaches and the high-variety aspects dealt with.

make CONWIP more suitable for a production environment with certain high-variety aspects. By analyzing the high-variety aspects in a production environment, adaptations to CONWIP can be made starting with the most dominant high-variety aspects. In this way the ability of CONWIP to deal with high-variety aspects is improved, also improving the throughput performance.

If processing times are variable a minimal adjustment to CONWIP is needed to deal with this high-variety aspect. As in CONLOAD, LB-POLCA and WLC the workload needs to be expressed in summed processing times, instead of number of orders. When there is a stable bottleneck in the process, the workload should be better expressed in processing times of the bottleneck workstation and a loop should be made to this bottleneck, as done in DBR to avoid idleness of the bottleneck. If there is no clear bottleneck, the workload can be expressed in total processing times of the orders. Expressing the workload in actual processing times instead of number of orders, will be a more realistic representation of the workload in the current shop floor situation.

(12)

-12-

the workstation, as in WLC. If routing sequences variability is high and the loops are attached to critical workstations, the workload also needs to be corrected for the workstation position. POLCA also corrects for this routing variability types by making overlapping loops. This is also an option to deal with length and sequence variability, though it will be more likely that this will result in more adaptations that need to be made.

As mentioned in the previous chapter it will become more difficult to identify a stable bottleneck in the process when other aspects of high variety production environments occur. It is more likely that the process will have different potential bottlenecks when the variability increases. It is important to control the direct loads in front of each potential bottleneck. When bottlenecks emerge, control loops need to be attached to these bottlenecks. Therefore the minimal adaptation to smooth the influence of these potential bottlenecks is to attach a control loop to every potential bottleneck and to return the loop immediately after the bottleneck. The workload is then calculated in units of bottleneck processing time.

An overview is given in the framework of table 5.1. This framework can be used to select a minimal adaptation to apply to CONWIP. Applying the minimal adaptation will make CONWIP better able to deal with high-variety aspects. By starting with CONWIP and adapting this approach for specific high variety aspects, a tailored approach can be implemented. After starting with CONWIP the production environment is checked for the high-variety aspects in the left column of table 5.1. When a high-variety aspect is present in a production environment then a minimal adaptation, shown in the upper row, is checked in the cell. This adaptation should be applied to CONWIP, to improve its performance, though the necessity of the adaptation depends on how strongly the variance aspects influence the production environment and its relation to the other aspects. The use of the framework will be illustrated with three cases in the next section.

Table 5.1 - Minimal adaptations required to CONWIP to deal with variability aspect.

6. Case Studies

Three case studies will be described in this section. First two MTO production environments will be analyzed. The first company uses POLCA, and the second company uses WLC as its PPC approach. Production planning data from both companies are used for the analysis together with the qualitative data from the production managers.

Limit workload (in processing

times)

Add loops per (critical) routing

Add loops per (critical) workstation or potential bottleneck Divide workload contribution by routing length or workstation position.

Processing time variability √

Routing variability √ √

Routing length variability √ √ √

(13)

-13- 6.1 Company B, hinges and light metal structures.

Company B is an MTO company that produces customer specific hinges and light metal structures, such as storage racks. The company serves industrial clients only, who use the end products from Company B for their own production. The company has over 500 active clients per year, resulting in more than 1.000 different products and 5.000 productions orders per year. Order sizes vary from 1 to 1.000 pieces, which is also the upper limit set by the management. In 2007 it started using POLCA as PPC approach and throughout the years this approach was refined and improved. POLCA was part of the new company strategy to focus on reducing lead times, reducing backorders, and therefore being able to respond faster on specific customer needs. With this new strategy the company was able to compete again by producing a high mix of products in low volume. POLCA is now implemented in an advanced stage. Although making products based on customer requirements, four main product families can be distinguished, which also make up the four main flows on the production floor. The production floor itself has a total of 105 machines, which has been redefined into ten production cells. This was already done in previous research, which is a complex process and part of the broader QRM strategy. It is based on several criteria, such as the proximity of resources, dedicated resources and product flow. The production cells where labeled with specific colors, for better visual representation on the shop floor. An overview of these cells is provided in table 6.1.1. A clear description of this production floor and the formation of production cells at this company is provided by Pejchinovska (2012). Data from this research is used to analyze the shop floor situation and its high-variety aspects. For the purpose of this research only the production process is taken into account, therefore work preparation for specific orders is excluded from this analysis. Planning data was based on the on the production cells, not on the individual machines.

Table 6.1.1 – Production cells at Company B.

6.1.1 Analyzing the possibilities for CONWIP with minimal adaptations.

Company B has to deal with highly variable demand and tight due dates. The majority of the orders is unique as only one client repeat his order every week and only a few repeat their order monthly. In this company the non-technological aspects, inter-arrival variability and variability of due date allowances, both have high variability. Together with the company strategy this a good reason for using a PPC approach with controlled order release and a workload limit. This offers a possibility to start with CONWIP. An analysis of the technological aspects will have to determine the need for minimal adaptations to CONWIP.

The first technological aspect from figure 2.1 is processing time. At Company B it is known that variability of the processing times is very high, as processing times can vary from a few minutes to

Cell name Color

Sheet metal preparation Orange

Tube Processing Ocher

Welding Lilac

Special Hinges Yellow

Piano Hinges Blue

Wet Processing Pink

Heavy Hinges Purple

Finishing Operations Green

Assembly and Quality Control Red

(14)

-14-

several hours. An example for this is the production cell yellow. In one month a total of 45 orders passed this production cell. These orders had an average process time of 432 minutes with a coefficient of variation (CV) of 1,08. According to Hopp (2003: 23) this is a high variability of processing times. If the workload is expressed in number of orders instead of processing time, this might for example result in a situation where the actual workload in processing time is very high and causing congestion. To deal with this the workload should not be limited in number of orders but in actual summed processing time.

For routing variability aspects a total of 1643 orders from a three-month period are analyzed. From these orders 17 were deleted as production information was unknown or incomplete. The orders result in a total of 108 unique routings. The resulting routings give a clear overview of the positions of all the production cells in the routings (see appendix 1), which are summarized in figure 6.1.1.1. This is used

Figure 6.1.1.1 – Frequency of production cell positions in routings.

as input for analyzing the high-variety aspects Company B has to deal with. The analysis of the orders and their production planning shows that routing variability is high (appendix 1). A routing has an average frequency of fifteen times in the three months period (CV = 2,63). In contrast with the other high variety aspects it shows a relatively large dispersion of this frequencies per routing. According to framework from table 5.1 loops should be added per (critical) routing or station. The number of different routings is high at Company B, much higher than the number of production cells. Even when looking at the most occurring routings (15% of the routings cover 80% of the orders), there are more routings than production cells. Therefore applying control loops to the production cells instead of loops per routing is a simpler adaptation. The routing variability can be further characterized by routing length and routing sequence variability. Compared to each other, routing length seems to be most variable from these two aspects. On average a routing consist of four production cells, but the routings vary from one to ten cells. Most production cells have a relative stable position within the routing, except for red, black and green (appendix 1). The adaptation provided according to the framework here would be to correct the workload

0 500 1000 1500 2000 2500 3000 1 2 3 4 5 6 7 8 9 10 Totaal F re qu ency o f po sit io n Position

Positions of production cells in routings

(15)

-15-

for the position of the production cell. The consequence of this will be that when a production cell has more cells upstream, the workload calculated for this production cell will be less. Therefore more workload can enter the process.

Dealing with potential bottlenecks would require loops per bottleneck station, but in this case there was not enough data available to determine potential bottlenecks. This could have lowered the number of control loops to production cells, as it is likely that not all production cells will be a bottleneck for the production process. When making minimal adaptations checking for potentials bottlenecks seems a logical first step. When applying adaptations to potential bottlenecks the further adaptations will be smaller than when applying adaptations to every workstation available.

6.2 Company M, sheet processing.

Company M started as a small company mainly producing agricultural machine equipment. Throughout its 91 years of existence it grew to a company employing 120 employees. The company separated into two businesses, in this case the focus is on the sheet processing compan. Company M switched from producing in large series to production in smaller series, focusing on shorter lead times and high delivery reliability. This resulted in an MTO company that specializes in sheet metal processing, tube laser cutting, frame construction and assembly. In 2012 the company had 278 different clients and processed approximately 6.500 orders. In 2011 the company started implementing WLC as a PPC approach. The motivation for implementing WLC was to improve the business performance of the company. At Company M a total of 163 different activities can be performed, which are arranged into functional departments. Machines that perform similar activities are clustered into a functional department. An overview of the shop floor is provided in figure 5.2.1. The departments consist of eighteen capacity groups, which is the level to which the production planning and control is applied. An extended

Fig. 6.2.1 – Shop floor at Company M.

description about the company and the shop floor, together with a description of the start of the WLC implementation at Company M is given in the research of Poppinga (2011).

6.2.1 Analyzing the possibilities for CONWIP with minimal adaptations

(16)

-16-

The production planner from Company M identified five capacity groups as potential bottlenecks in the production process. Therefore these five capacity groups are also treated as bottlenecks in the planning. Control loops are made from order release till these bottlenecks and returning afterwards. The remaining minimal adaptations to make are applied to these five bottlenecks. These five bottlenecks are (see appendix 2):

- Bystronic (laser cutting); - Press Brake;

- Welding Robot; - Tube Laser; - Assembly.

The first technological aspect is the processing time. The processing time from Company M is based on the setup time plus cycle time. The capacity group Bystronic will be taken as an example to indicate the processing time variability. This capacity group is a production step in 919 orders. Most of the time is spend on setup. The average processing time is 5,3 hours, with a standard deviation of 9,7 hours. The coefficient of variation (CV) for processing times is 1,84. According to Hopp (2003) this is very high, which is true in the case of the Bystronic. This is partly due to the different production quantities, which are also highly variable. Highly variable production quantities are not mentioned in the framework from table 5.1. This variability type results in different processing time and thus different workloads. Therefore measuring the workload in actual processing times is the minimal adaptation to apply in this case.

Figure 6.2.1.1 – Frequency of capacity group position in routings.

0 100 200 300 400 500 600 700 800 900 0 1 2 3 4 5 6 7 8 F re qu ency o f P o sit io n Position

Position of capacity groups in routings

(17)

-17-

For the routing variability a total of 1.669 orders from the planning data are analyzed. The positions of the bottlenecks in the orders’ routings are summarized in figure 6.2.1.1. The orders have a total of 3.656 operation steps. About 63.8% of these operations are performed at the bottlenecks, where only one operation step is performed. The average routing consists of two operations steps, although routings range up till eight operations. The positions of the five capacity groups seem relatively fixed, though the welding robot (cv = 0,95) and assembly (cv = 0,65) show larger variations of their position in the routings, compared to the other three capacity groups. To make a minimal adaptation to deal with routing length, the workload of the bottleneck should be corrected for the position of this bottleneck in a routing. This is because loops are applied to bottlenecks (capacity groups) and not to complete routings. This adaptation also corrects for routing sequence variability, although routing sequence variability does not seem very high in the case of Company M. In this case adaptations for all high-variety aspects from table 5.1 are made.

6.3 Company A, paint production.

Company A is a paint plant that mainly produces ‘specialties’, products at the beginning or the end of a lifecycle and test products. These specialties often have an unstable and therefore unpredictable demand. The formal delivery time, as suggested by the plant itself, varies from three to six weeks. However, the practically agreed delivery time is often much shorter and due dates are tight. The production process consists of two stages (see figure 1). Between those stages the paint is temporarily stored in tanks or mobile vessels. The first stage is the manufacturer of the paint (adding the raw materials, mixing, beat milling and color adding). In this stage order sizes range from 1 liter to 10.000 liters, with an average of 1.000 liters. After manufacturing a quality check of the paint is performed. If the paint is not within

Picking raw materials Raw Materials Mixing Colour adding Beat milling Quality checking Adjusting Filling Final product Packa ging Distribution rework rework Line 1 Line 2 Rest Line 3 Inter- mediate storage

(18)

-18-

specification, it is rejected and returns to the manufacturing stage for rework. On average 15% of the paint needs rework. If the paint is approved it moves on to the second stage, the filling of the paint in different sorts of packaging. In this assembly type of stage cans, lids, labels, boxes and paint come together per order. Three main filling lines perform the filling operation in this plant. Two filling lines perform long runs of products with large order sizes. The other filling line performs short runs of small order sizes. In this stage order sizes ranges from 1 piece to 20.000 pieces, with an average of 400 pieces. The total number of orders per year is 6.000 with an average repetition rate of 5 orders per year for similar products.

6.1.2 Implementation of CONWIP and adaptations

The analysis of Company A, before the implementation of CONWIP, showed that no feedback between the output of the process and order release existed. An increase in demand was directly translated to an increase of the release rate. The varying release rate led to a varying amount of orders in the production process. This in turn caused a high variability of throughput times and in periods with high demand led to a decreasing due date adherence. To overcome these issues it was decided to limit the amount of orders in the process by controlling the order release. Limiting and stabilizing the WIP-level should enable a better throughput time prediction, which in turn should lead to an increase of the due date predictability.

The main sources of variability that the plant needed to cope with were an unstable and unpredictable demand and varying due dates. As indicated in table 4.1 all PPC approaches are designed to smoothen variable arrival patterns and are able to cope with variable due date allowances. CONWIP, because of the simplicity of the approach, was therefore chosen as a starting point for the implementation.

Several brainstorm sessions with the employees indicated that limiting the amount of orders for the entire process was not necessary. For only one filling line the required capacity exceeded the available capacity in certain periods. In this case a simple adaptation to CONWIP, as indicated by our scheme, is to control the amount of orders in the loop of this critical workstation. A more in-depth analysis of this critical workstation showed that processing times at this workstation were variable; they vary from 15 minutes to several hours. Our scheme shows that this can be anticipated for by expressing the workload in bottleneck processing times.

Suggestions from the employees were used to come to a simple, useful and supported implementation. The following method was proposed and implemented: Upon release the planner calculates the amount of processing time needed at the critical filling line and registers the order and corresponding bottleneck processing time in a simple spreadsheet. If an order is completed, the planner deletes the order from the spreadsheet. A new order is only allowed to enter the process when the current load plus the load introduced by the new order is less than a given norm value. Each time an order enters the process, the load is increased by the order’s load contribution, and each time an order leaves the process it is decreased by the same amount. The norm value is the maximum amount of hours in the process from release until completion at critical filling line. The initial norm value was determined by analyzing a queuing network model of Company A.

After this analysis the adapted version of CONWIP was implemented in the paint factory. This approach led to a decrease of the average throughput time and to a decrease of the variation of throughput times. The simplicity of this approach and the involvement of the employees, led to understanding and support by the employees.

(19)

-19-

employees and the simplicity of the PPC approach made that this adaptation was implemented in just a few weeks.

One of the current problems is the increased risk of idle time at the critical workstations, while work for these workstations is still present in the process. The main reason behind the varying direct load is the fact that the routing length per order varies. Some of the orders need rework while other orders do not. Since CONWIP will release the same amount of orders, independent if the order needs rework or not, it does not control the direct load of the critical workstations. Therefore the tracking of the direct load of the filling lines is improved. The introduced method makes use of the existing production order forms. These order forms travel with the order from the moment of release till the completion of the order at the filling line. Every day an employee checks all workstations in the process and registers the amount of bottleneck hours at every workstation. With this information the planner is able to speed up orders for the filling line, if there is a high risk of idle time at this workstation. This approach however demands an extra intervention in the process. These extra interventions can be prevented by a further adaptation of the CONWIP approach. As indicated by our scheme, the varying routing length can be anticipated for by correcting for the position of the critical workstation in the order’s routing. For most of the orders it is known in advance whether there is a high chance that the order will need rework or not. The proposed adaptation is therefore to divide the workload contribution of the order by its total throughput time based on historical data.

7. Discussion.

The high-variety aspects that characterize the production environments from the three companies are summarized in table 7.1. The checked cells mark the high-variety aspects for which a adaptation to CONWIP is made.

Table 7.1 – High-variety aspects at the case study companies.

Com pan y B C om pan y M C om pan y A Non-technological aspects Inter-arrival variability √ √ √ Variability of due date

(20)

-20-

In all three cases the companies have to deal with inter-arrival variability, variability of due date allowance and processing time variability. Therefore COWNIP can be applied, and workload should be measured in actual processing times instead of number of orders. The most important difference between Company B and the other two companies is focus on the bottlenecks. In Company M and Company A this focus on the bottlenecks reduces the amount of control loops to apply and as a result of the focus on these bottlenecks there are no loops applied to routings. All the companies have varying routing lengths, therefore a correction should be made for the position of the workstation. As the loops in all three cases are added to workstations, the correction for workstation position to deal with routing length variability also deals with routing sequence variability. The three cases show that making minimal adaptations to CONWIP provides the possibility to suit the PPC approach to a specific production environment.

Company B has implemented POLCA. The motivation for this is to shorten the lead times of orders and to be able to respond fast to specific customer needs. When analyzing the high-variety aspects that characterize the production environment of Company B, this is a good choice. The high-variety aspects most important for Company B are routing variability and processing time variability. In order to deal with the processing time variability POLCA cards should represent actual processing time instead of a complete order, which is done by LB-POLCA. CONWIP can also be adapted to deal with this. The difference in the final result of POLCA and CONWIP with minimal adaptations is the number of loops. As routing length or sequence variability will be higher, the number of loops will increase with POLCA. CONWIP with minimal adaptations does this by making a correcting the workload contribution for the position of the workstation.

The formation of production cells in this case however is a big influence on the final result. For example it will automatically make it more interesting to apply loops to critical workstations instead of routings as there are far less critical workstations. It is also hard to detect bottlenecks in this process as capacity is unknown, but the final result might be less loops when some workstations are not critical to the process. Another aspects that is influencing the case of Company B is the nesting decision. In order to make maximum usage of sheet metals, orders need to be nested together before they go to the laser machines (orange). For some orders this will result in a earlier release to the shop floor and a higher level of work in process. This aspect is not reflected in the framework of high-variety aspects and needs further research.

At Company M, WLC is used as a production planning and control approach, for the same reason as Company B uses POLCA. Based on the high-variety aspects and the ability of WLC to coop with these aspects, this is a good choice. Like the production cells from Company B, Company M uses capacity groups for planning. Five of these groups are identified as critical, therefore the production planning focuses on these five groups. Like with Company A this has an important impact on the remaining process. The next adaptations can be applied to these bottlenecks, so the choice for focussing on these critical groups reduces the total amount of loops to be applied. The remaining adaptations made are also present in WLC, and thus accounted for in the current situation at Company M.

(21)

-21-

approach. Gradually the performance of CONWIP was improved, resulting in a more complex approach that is able to cope with several high-variety aspects, which CONWIP could not have done.

The implementation of CONWIP with minimal adaptations also shows a difference with the PPC approaches currently used by Company B and Company M. POLCA and WLC immediately start with a maximum amount of loops when implemented, while CONWIP with minimal adaptations does the opposite of this, as it start in its most simple form and is then adjusted and becoming more complex to improve the performance. So the implementation process of CONWIP with minimal adaptations is different, which might provide new insights into the implementation issues PPC approaches deal with.

POLCA is one of the PPC approaches in this paper that can deal with all high-variety aspects mentioned in figure 2.1. Though its role in the design of the minimal adaptations is minimal. It is the only PPC approach that makes use of overlapping loops to control. We argued that this is likely to cause more control loops and thus not the simplest adaptation to make. POLCA is still able to deal with the high-variety aspects in a different way than then other approaches, especially for routing length and sequence variability. Therefore it is worth to look how POLCA can improve the minimal adaptations for CONWIP.

8. Conclusion

The goal of this paper was to improve the ability of CONWIP to deal with high-variety aspects. This is done by learning from other PPC approaches. First, high-variety aspects based on order characteristic were identified. Based on the ability of other PPC approaches to deal with these high-variety aspects, specific adaptations for CONWIP are proposed in a framework. The PPC approaches shared the characteristic of controlled order release and a maximum amount of workload control, like CONWIP. The minimal adaptations to CONWIP should allow a better workload balance among the workstations in a production environment, decreasing idleness of workstations or congestions on the production floor, improving the throughput performance and throughput predictability. The paper contributed to knowledge about PPC approaches and their relation to high-variety aspects. A limitation of the framework was the focus of the high-variety aspects on order characteristics. There are more high-variety aspects to describe. Examples are material requirements, variability of set-up times or machine breakdowns. However, this also provides future research possibilities. Expanding the framework, identifying more high-variety aspects and minimal adaptations to deal with these aspects.

The analysis of the high-variety aspects itself also provides options for improvement. Now mostly the coefficient of variation was used to indicate the variability of the aspects, or when there were no quantitative data available it was based on discussion with the production planner. So far there is no general guideline in measuring these high-variety aspects and thus this offers possibilities for future research. This should also provide better insight into the question when it is worth to make a minimal adaptation to CONWIP, so that it will provide a significant improvement on the performance of the approach.

(22)

-22- References

Bechte, W., 1980. Steuerung der Durchlaufzeit durch belastungsorientierte. Auftragsfreigabe bei Werkstattfertigung. Dissertation. VDI-Z Reihe 3, 70

Bertrand, J.M.W. and Wortmann, J.C., 1981. Production control and information systems for component-manufacturing shops. Dissertation. Elsevier, Amsterdam, The Netherlands.

Fernandes, N.O. and Do Carmo-Silva, S., 2006. Generic POLCA – A production and materials flow control mechanism for quick response manufacturing. International Journal of Production Economics, 104(1): 74-84.

Framinan. J.M., González, P.L. and Ruiz-Usano, R., 2003. The CONWIP production control system: Review and research issues. Production Planning & Control: The Management of Operations, 14(3): 255-265.

Gaury, E.G.A., Pierreval, H. and Kleijnen J.P.C., 2000. An evolutionary approach to select a pull system among Kanban, Conwip and Hybrid. Journal of Intelligent Manufacturing, 11(2): 157-167. Germs, R and Riezebos, J., 2010. Workload balancing capability of pull systems in MTO production.

International Journal of Production Research, 48 (8): 2345-2360. Goldratt, E.M. & Cox, J. 1984, The Goal, New York: North River Press.

González, P.L., Framinan, J.M. and Henry, P., 2012. Token-based pull production control systems: an introductory overview. Journal of Intelligent Manufacturing, 23(1): 5-22.

Henrich, P., Land, M.J. and Gaalman, G. 2004. Exploring applicability of the workload control concept. International Journal of Production Economics, 90(2): 187-198.

Hendry, L., Land, M., Stevenson, M. and Gaalman, G., 2008. Investigating implementation issues for workload control (WLC): A comparative case study analysis. International Journal of Production Economics, 112(1): 452-469.

Hopp, W., 2003. Supply Chain Science. Wallace J. Hopp.

Hopp, W.J. and Spearman, M.L., 2000. Factory physics, foundations of manufacturing management. (2nd ed). Burr Ridge, IL: Irwin/McGraw-Hill.

Land, M.J. and Gaalman, G.J.C., 1996. Workload control concepts in job shops A critical assessment. International Journal of Production Economics, 46-47: 535-548.

(23)

-23-

Land, M.J., 2009, Cobacabana (control of balance by card-based navigation): A card-based system for job shop control. International Journal of Production Economics, 117(1): 97-103.

Lödding, H., Yu, K. and Wiendahl, H., 2003. Decentralized WIP-oriented manufacturing control (DEWIP). Production Planning & Control: The Management of Operations, 14(1): 42-54.

Lu, J., Yang, T. and Wang, C., 2011. A lean pull system design analysed by value stream mapping and multiple criteria decision-making method under demand uncertainty. International Journal of Computer integrated Manufacturing. 24(3): 211-228.

Khojasteh-Ghamari, Y., 2009. A performance comparison between Kanban and CONWIP controlled assembly systems. Journal of Intelligent Manufacturing, 20(6): 751-760.

Kingsman, B.G., 2000. Modelling input-output workload control for dynamic capacity planning in production planning systems. International Journal of Production Economics, 68(1): 73-93. Kingsman, B.G., Tatsiopoulos, I.P. and Hendry, L.C., 1989. A structural methodology for managing

manufacturing lead times in make-to-order companies. European Journal of Operational Research, 40(2): 196-209.

Oosterman, B., Land, M.J. and Gaalman, G.J.C., 2000. The influence of shop characteristics on workload control. International Journal of Production Economics, 68(1): 107-119.

Pejchinovska, J., 2012. Excelling the game of POLCA cards – faster and better. Unpublished Master Thesis.

Poppinga, 2011. A lean implementation of Workload Control at Meijer Sheet Processing. Unpublished Master Thesis.

Riezebos, J., 2010. Design of POLCA material control systems. International Journal of Production Research. 48(5): 1455-1477.

Rose, O., 1999. CONLOAD - A new lot release rule for semiconductor wafer fabs. Proceedings of the 1999 Winter Simulation Conference.

Spearman, M.L., Woodruff, D.L. and Hopp, W.J., 1990. CONWIP: a pull alternative to kanban. International Journal of Production Research, 28(5): 879-894.

(24)

-24-

Stevenson, M. and Silva, C., 2008. Theoretical development of a workload control methodology: Evidence from two case studies. International Journal of Production Research¸ 46(11): 3107-3131

Stevenson, M., Huang, Y., Hendry, L.C. and Soepenberg, E., 2011. The theory and practice of workload control: A research agenda and implementation strategy. International Journal of Production Economics, 131(2): 689-700.

Stuart, I., McCutcheon, D., Handfield, R., McLachlin, R. and Samson, D., 2002. Effective case

research in operations management: A process perspective. Journal of Operations

Management 20(5): 419–433.

Suri, R., 1998. Quick response manufacturing: a companywide approach to reducing leadtimes. Portland, OR: Productivity Press.

Tatsiopoulos, I.P., 1983. A microcomputer-based interactive system for managing production and marketing in small component-manufacturing firms using a hierarchical backlog control and lead time management methodology. Ph.D. Thesis, University of Lancaster.

Vandaele, N., Van Nieuwenhuyse, I., Claerhout, D. and Cremmery, R., 2008. Load-Based POLCA: An Integrated Material Control System for Multiproduct, Multimachine Job Shops. Manufacturing Service Operations Management, 10(2): 181-197.

(25)

-25- Appendix 1.

In the table above the position of each production cell from Company B was counted, which resulted in an average position, a standard deviation and coefficient of variation. For example workstation ‘Red’ was counted 855 times on position 3 in the routing of an order. There was a total of 1.626 routings, in which workstation red occurred 2.530 times. The average position of Red was 3,65. Compared to the other workstations, the spread of the position of red was higher, as the coefficient of the variance was 0,49. In the second table the routing length and frequency of routings was analyzed. For example, there were 439 routings with a total of 3 workstations. Routings occurred on average 15,06 times.

Position Red Black Orange Ocher Lilac Pink Purple Yellow Green Blue Total

(26)

-26- Appendix 2.

This table shows the workstations from the capacity groups which are potential bottlenecks for Company M. It is read in the same way as the table in appendix 1.

Description Position Total Av er a g e St. Dev . Va ria nce CV Bystronic 1 2 3 4 5 6 7 8 Bystronic Laser 648 49 0 697 1,07 0,26 0,07 0,24 Bystronic Laser NF 200 8 8 216 1,11 0,42 0,17 0,37 Bystronic Nesting 6 0 0 6 1,00 0,00 0,00 0,00 Total 854 57 8 919 1,08 0,30 0,09 0,28

Press Brake (PB) 1 2 3 4 5 6 7 8 Total

Eccentric 0 9 0 0 0 0 9 2,00 0,00 0,00 0,00

Hydraulic Speed Press 0 5 0 0 0 0 5 2,00 0,00 0,00 0,00

Pressfit PB 0 3 6 2 1 0 12 3,08 0,86 0,74 0,28 Pressfit PB NF 0 2 2 0 0 0 4 2,50 0,50 0,25 0,20 NF PB with 2 men 6 6 1 0 0 3 16 2,44 1,80 3,25 0,74 Plates Roller 1 54 10 1 1 0 67 2,21 0,56 0,31 0,25 Folding 0 6 0 0 1 0 7 2,43 1,05 1,10 0,43 PB 9 366 88 7 0 0 470 2,20 0,48 0,23 0,22 PB NF 3 61 64 6 0 0 134 2,54 0,62 0,38 0,24 PB with 2 men 0 24 15 5 0 0 44 2,57 0,69 0,47 0,27 Total 19 536 186 21 3 3 768 2,30 0,62 0,38 0,27

Welding Robot (WR) 1 2 3 4 5 6 7 8 Total

WR Installation 11 3 0 0 1 0 0 0 15 1,47 1,02 1,05 0,70

WR Programming 1 0 0 0 0 0 0 1 2 4,50 3,50 12,25 0,78

WR Progr. Hall 02 0 0 1 0 0 0 0 0 1 3,00 0,00 0,00 0,00

Total 12 3 1 0 1 0 0 1 18 1,89 1,79 3,21 0,95

Tube Laser (TL) 1 2 3 4 5 6 7 8 Total

TL LT652 (2D) 60 0 60 1,00 0,00 0,00 0,00

TL LT8 8 0 8 1,00 0,00 0,00 0,00

TL LT813 (3D) 5 0 5 1,00 0,00 0,00 0,00

Nesting General 25 0 25 1,00 0,00 0,00 0,00

Nesting TL LT8 2 0 2 1,00 0,00 0,00 0,00

Nesting parts Moba frames 6 0 6 1,00 0,00 0,00 0,00

Nesting Supermarket 13 31 44 1,70 0,46 0,21 0,27

Total 119 31 150 1,21 0,40 0,16 0,34

(27)

-27-

Assembly 1 2 3 4 5 6 7 8 Total

Aluminium Tig Welding 4 0 3 0 0 0 0 7 1,86 0,99 0,98 0,53

Autogenous Welding 1 0 0 0 0 0 0 1 1,00 0,00 0,00 0,00 CO-2 Welding A 8 0 0 0 0 0 0 8 1,00 0,00 0,00 0,00 CO-2 Welding B 135 11 24 4 0 0 0 174 1,41 0,81 0,66 0,57 Hand Tools 0 2 1 1 0 2 0 6 3,83 1,67 2,81 0,44 Welding G-F 8 0 0 0 0 0 0 8 1,00 0,00 0,00 0,00 Spotwelding 26 1 0 0 0 0 0 27 1,04 0,19 0,04 0,18

Stainless Steel Tig Welding 52 3 10 5 0 0 0 70 1,54 0,98 0,96 0,64 Stainless St. Tig Welding M-F 10 0 0 0 0 0 0 10 1,00 0,00 0,00 0,00

Clean Assembly 74 3 4 3 3 0 0 87 1,37 0,97 0,95 0,71

Steel Installation Workplace 1 0 0 0 0 0 0 1 1,00 0,00 0,00 0,00

Steel Grinding 0 24 12 2 0 2 1 41 2,71 1,17 1,38 0,43

Steel Tig Welding 18 4 8 2 0 0 0 32 1,81 1,01 1,03 0,56

Stud Welding 0 1 0 0 0 0 0 1 2,00 0,00 0,00 0,00

Stud Welding Non-Ferrous (NF) 0 0 0 1 0 0 0 1 4,00 0,00 0,00 0,00

Tig Robot 0 0 2 0 0 0 0 2 3,00 0,00 0,00 0,00

Assembly KIT-Packing 1 0 0 0 0 0 0 1 1,00 0,00 0,00 0,00