Production Unit control

(1)

Master Thesis University of Groningen

Production Unit control

Order release and shop floor control at the

Copper Bar of Eaton Holec

Tijmen Essers

1402889

Primary Supervisor: Dr. J.A.C. Bokhorst

Secondary Supervisor: Dr. Ir. J. Slomp

(2)

Preface

Five months after my internship started at Eaton Holec, I have reached to write my master thesis. This master thesis concludes the master program “Technology Management”.

In this preface I would like use the opportunity to thank the people who supported me during my internship. First of all my supervisors from the University of Groningen, Dr. J.A.C. Bokhorst and Dr. Ir. J. Slomp, for their critical insights, useful comments, complicity and pleasant cooperation during my internship. Without their feedback I would not have been able to reach this result.

I would also like to thank my supervisor from Eaton Holec, J.M. Busschers, who has given me the opportunity to do my internship at Eaton Holec, who provided me with useful feedback and because of his support with the implementation of the developed concepts. I would like to thank the employees from the production unit Copper Bar, who helped me gathering the right information.

Finally, I would like to thank my father, who partly reviewed and corrected my English of my thesis.

Groningen, March 2008

(3)

Summary

The Eaton Corporation is a diversified worldwide industrial manufacturer with four distinct segments: Electrical, Fluid Power, Truck and Automotive. Eaton Holec forms part of the worldwide Eaton Corporation and is active in electrical energy at low and medium voltage level. PMC3 is an internal production line, part of the Eaton Electric General Supplies (EEGS), and delivers components to the three assembly lines of Eaton Holec. At PMC3 the operations punching, bending (three different bending machines), drilling, milling and bench working are active.

The buffer of orders in front of PMC3 is uncontrolled because all orders are released to the production floor. There is no sequence with which production order to start. A constraining Work In Process (WIP) system of PMC3, what is called the “TAKT system” was working quite well. An operator has to add a so called CONWIP card with a specific number, at each order which starts to be produced in PMC3. These numbers and related production throughput time are visualized on a screen. The TAKT system demands that once an order is taken into production, it has to be finished by the Copper Bar within the TAKT time multiplied by the maximum number of orders, which are 60 orders. However, eighteen percent of the orders had a too long production throughput time. In order to tackle the identified problems an Action Research approach was used.

Eaton structured its performance indicators in a circle of plan-do-check-act. This circle promotes the formulation of the vision of the company into actions. The vision of EEGS is “To be the fastest and most reliable supplier of parts for Eaton Holec”. However, the performance indicator “On Time to Request”, in relation with the performance indicator “Order Pattern” are not sharply enough defined in order to improve the performance of PMC3.

The workload concept, proposed for the release and production of orders, is dependent on the specific production characteristics of the production line PMC3. Complexity of low volume/high variety in products and the related different routings per product require that the production process is designed as a ‘job shop’ process. The high variation in production time per operation per order and the fact that more work stations are available than number of operators present, require that operators are multi trained to cope with the fluctuation in demand. Capacity can be constrained by either machines or by operators. The operators work in a two shift system. When at each shift at least 6 multi skilled operators are present, the punching machines are the bottleneck as they can handle on average 51 orders per day. When customer demand exceeds 51 orders per day, PMC3 is not able to fulfill the demand when capacity is not increased. Punching capacity can be increased by either an extra night shift, by working on Saturdays or by investing in an additional punching machine.

(4)

manage the production floor and the order release by the number of orders instead of by the required capacity per order, which makes the control easier.

Eighteen percent of the orders had a too long production throughput time. Flexibility operators, insufficient attention and technical problems represent most of the reasons for delay. The computer program for the TAKT system was developed by an external institution, and crashed on average once a month, resulting in a temporary uncontrolled production process and incorrect information about the production throughput time. The first part of the workload control concept is the modification of the TAKT system in order to tackle the discovered problems. The improvement lies in the connection with the “Enterprise Resource Planning” (ERP) system, with associated changes in the TAKT software, and the use of “Standardized Work”. Standardized Work is a method of communicating work procedures, which is used in the whole factory. The result of the modified TAKT system is a more stable and more transparent TAKT system due to the connection with the ERP system. Due to this connection, it is possible to view at all computers in the whole factory the progress of the Copper Bar at that moment. The prospects are that operators work more on the right order due to the Standardized work.

(5)

Chapter 1:

Introduction

This part aims to give insight in the organizational structure of the Eaton Corporation, how Holec is contributing to the Eaton Corporation, of which departments Eaton Holec consist, and the production line ‘Copper Bar’, in order to give the reader a better understanding where this research has been accomplished.

1.1 Eaton Corporation.

The Eaton Corporation is a diversified industrial manufacturer with in 2006 sales of $12.4 billion. Eaton is a global leader in electrical systems and components for power quality, distribution and control; fluid power systems and services for industrial, mobile and aircraft equipment; intelligent truck drive train systems for safety and fuel economy; and automotive engine air management systems, power train solutions and specialty controls for performance, fuel economy and safety. The Eaton Corporation gives work to 67,000 employees and sells products to customers in more than 140 countries.

Once known as a vehicle components supplier, the Eaton Corporation has diversified to include a broader industrial and commercial focus. Today, Eaton's businesses comprise four distinct segments: Electrical, Fluid Power, Truck and Automotive. A summary of these facts and the amount of contribution per segment is presented in figure 1.

Figure 1: Segments Eaton Corporation1

The vision and mission statements of Eaton are: To Be The Most Admired Company in Our Markets and To Be Our Customer's Best Supplier, respectively. The Eaton Business System provides both the means of achieving this mission and measuring the success doing so. The Eaton Corporation produced the Eaton Business System (EBS) to link Eaton's worldwide businesses and employees by providing a common set of values, philosophies, management tools and measures. The company claims that the EBS enables Eaton to systematically manage its businesses while

(8)

capturing the benefits of diversity, scale and rapid transfer of best practices. EBS positively impacts the results by improving working capital and operating margins and reducing costs through the following elements: Corporate Goals, Planning, Growth, Execution, Assessment and Tools. These elements are described in appendix 1 because they influence the way the people work.

1.2 Holec

On 3 February 2003 the activities of Holec Holland N.V. were transferred to Eaton Electric B.V. Holec is now called Eaton Holec and forms part of the worldwide Eaton Corporation. Already for a century Holec develops, produces and sells products for switching, distributing and protecting electrical energy on low and medium voltage level. The customers from Eaton Holec are presented in the picture below.

Figure 2: Customers Eaton Holec2

The picture shows that Eaton Holec is a partner for utilities, electrical contractors, light and heavy industry. Three product lines Medium Voltage Systems, Low Voltage Systems and Low Voltage Components assemble the products for these customers. The department Eaton Electric General Supplies (EEGS) is an internal supplier for these three product lines. The organization structure is presented in the figure below.

(9)

Figure 3: Organizational Structure Eaton Holec3

1.3 Eaton Electric General Supplies

EEGS is an internal supplier for the department Low voltage Components, Low Voltage Systems, Medium Voltage Systems and Service. It supplies these departments with components and semi products for their end products. EEGS has four product lines to serve its customers; Sheet Metal (PMC1), Turning and Milling (PMC2), Copper Bar (PMC3) and Punching (PMC4). The aim of EEGS is “To be the fastest and most reliable supplier of parts for Eaton Holec”. This research will focus on the operational planning and the shop floor control of PMC3 in order to strengthen the strategy of EEGS.

1.4 PMC3

At PMC3 the operations punching, bending (three different bending machines), drilling, milling and bench working are active. In order to give a better understanding of how and which operations take place, a value stream map is presented. A value stream map also gives a good first overview which processes take place and identifies where there is waste in manufacturing processes, and to help find ways to eliminate that waste.

(10)

Figure 4: Value Stream Map Copper Bar at the start of my internship (November 2007)

The square represents the operations which are performed at the Copper Bar. Four months before my internship started, a workload concept was introduced, under the name “TAKT” system (TAKT is derived from the German word TAKTzeit, which translates to clock cycle). This system reduced the work in process, and thereby the production throughput time, and demanded that once an order started production, the order had to be finished at the Copper Bar within twenty hours. The TAKT system increased the reliability and the transparency of the production process. However, eighteen percent of the orders exceeded the production throughput time of twenty hours. This indicates that the system was not working as it should.

(11)

(12)

Chapter 2: Research outline

2.1 Objective

To improve the performance of the production unit Copper Bar by modifying the TAKT system and introducing the order release system at the prevailing production and customer demand characteristics.

2.2 Conceptual model

The purpose of the conceptual model is to indicate which aspects need to be investigated in order to reach the objective. The objective is to improve the performance. The current performance has to be explored, and the desired performance has to be defined. To reach this performance, the TAKT system has to be modified and an order release system has to be designed, which together controls the workload. However, production characteristics and customer demand characteristics influence the way this workload control concept can be (re)designed. This leads to the conceptual model.

Customer demand:

* Order differences- influences the way the workload control concept can be used

* Demanding pattern- influences the way the order release concept can be used

Production characteristics: * complexity * uncertainty * flexibility Workload Control: Order release: TAKT system:

Flow and time of orders in the system

Performance:

(13)

2.3 Research questions and restrictions

The research questions are derivatives from the research model.

1. What is the current performance, what is the required performance and what are the comments on the current performance indicators?

2. What are the environment characteristics (production characteristics and customer demand)? 3. How are the TAKT system and the order release currently working?

4. What are the modifications of the TAKT system and the design of the order release taken into account the known environment characteristics?

5. How to use the workload concept when customer demand increases or decreases? 6. Which performance measurements should be used and who is responsible for what?

Research restrictions:

• The allocated time for this project is around 4 months.

• Only the department of Eaton Electric General Supplies will be taken into account, with focus on the product line Copper Bar.

2.4 Research approach

In this part it will be discussed how to obtain answers to the research questions. The answers will be gained by using an Action Research (AR) approach. “An action researcher is a participant in the implementation of a system, but simultaneously wants to evaluate a certain intervention technique. The action researcher is not an independent observer, but becomes a participant, and the process of change becomes the subject of research” (Benbasat & Goldstein 1987).

According to Coughlan & Coghlan (2002) Action Research focuses on research in action, rather than research about action. Second, AR is participative; members of the system which is being studied participate actively. Third AR is research concurrent with action. The goal is to make that action more effective while simultaneously building up a body of scientific knowledge. Finally, AR is both a sequence of events, and an approach to problem solving. These mentioned points what AR is all about suits totally the way this research will be accomplished.

(14)

Figure 6: Action Research Model, derived from Coughlan, P and Coghlan, D (2002)

In this model the one who is monitoring is the one who is doing the research. In this case I am the researcher. Some information was readily available; other information is gathered and processed into graphs, tables and figures. The gathered data will contain literature and information about the production process at the Copper Bar. The data feedback will be done by my supervisors and the concerning employees at Eaton Holec.

To determine what the required performance is, the strategy of Eaton Holec and EEGS will play an important role and data will be gathered about what is measured and their required targets. In order to determine what the environment characteristics are and to not exclude characteristics, the literature of Bertrand Wortmann and Wijngaard (1998) will help to make a holistic overview. Subsequently, I will determine which characteristics are relevant. The stored data of the company will be used to acquire the production and demand characteristics for the Copper Bar.

(15)

Chapter 3: Performance Indicators

This part aims to give answer to the first research question - What is the current performance, what is the required performance and what are the comments on the current performance indicators –. In order to give the reader a better understanding which subject is analyzed and how this relates to the remaining subjects, the conceptual model is displayed again.

Figure 7: Conceptual model, analyzed subject 3.1 Desired performance

Kaplan & Norton (1996), Globerson (1985) and others address that performance measurements should be derived from the strategy. The aim of EEGS is “To be the fastest and most reliable supplier of parts for Eaton Holec”. It is desired that a product will be delivered on the time the customer needs it (On Time to Request (OTR)), and the product has to meet the desired specifications. The OTR for 2005, 2006 and 2007 has been around 54 percent, which has to increase. In order to increase the OTR, internal processes have to be controlled with the focus on short throughput times and little work in process.

3.2 Relevance of performance indicators

“The traditional view is that performance measures are an integral element of the planning and control cycle. It is assumed that measurement provides a means of capturing performance data which can be used to inform decision making” (Neely et al 1997).

Performance measurements also have a behavioral impact. “Systems, especially systems involving humans, respond to performance measures. People modify their behaviors in an attempt to ensure a positive performance outcome” (Hopwood 1984). Good performance measures can motivate operators and increase thereby the desired performance.

3.3 Relation between performance indicators

(16)

enough orders to meet the available capacity. The global relations between performance indicators are displayed in the figure below (Bertrand et al 1998).

Figure 8: Relationship Performance Indicators

3.4 Performance indicators Copper Bar

Eaton visualizes their performances in a structured way on a performance indicator board. Thereby it is easily possible for everyone to analyze and interpret the performance indicators. Eaton structured its performance indicators in a cycle of plan-do-check-act. This circle promotes to formulate the vision of the company into actions. It guarantees that research will be done to causes of deviations to the norm or target.

Plan Do Check Act cycle Copper Bar

The word “plan” is a derivative from the vision of the company, which is translated into measurable strategic directions. Also the vision of EEGS “To be the fastest and most reliable supplier of parts for Eaton Holec” is translated into measurable indicators and desired directions. These directions are customer intimacy (On Time to Request), profitability (productivity, amount of stock and the amount of waste) and operational excellence (Eaton Lean System (appendix 1) and Eaton Quality System; number of wrongly produced products + days without accidents + percentage absence through illness).

The word “do” is translated with an operator capability scheme and an operator tool scheme. The tools are Value Stream Map, 5S, Standardized Work, Error Proofing, Total Productivity Management, Set Up Time Reduction, Pull and Continuous Flow.

(17)

throughput time orders4_{, the quality in wrongly produced parts per million (discovered by the} customer), the amount of waste and the number of worked hours. Each of these performance indicators display historical values and indicate desired long term outcomes.

The word “act” is translated with reports of internal meetings and a 5S checklist.

Besides this Plan Do Check Act performance indicator board, there is live information about the production throughput time of orders in the Copper Bar, how many orders are produced that day, what the TAKT time is and what the lead or backlog is at that moment. This live information is visualized on a screen (see figure 15 page 39) at the Copper Bar and operators act on this information.

3.5 Comments on Performance Indicators

The way Eaton displays its performance indicators creates a shared understanding. This performance indicator board focuses on change efforts (initiatives drive desired long term outcomes) and permits learning at the management level. It is a holistic model which reflexes the strategy that

partly allows employees to see how they can contribute to organizational success. Partly because some performance indicators do not directly result from operators’ activities. The Performance Indicators for delivery reliability – On Time to Request and Order Pattern - , represent respectively the percentage of how many times an order was delivered at the time the customer requires it, and if the customer requested it on time. There is some fuzziness in these performance indicators.

The order pattern is ok if the entry date (the date that the customer requested the order) is before the original start date (the date on which the order should start). In 2007 the order pattern was 70 percent, which means that 30 percent of the orders were ordered late. The original start date is estimated without knowledge of the routing for 40 percent of the orders. This counts for customer specific orders. When there is no information about the routing, the throughput time and hence the original start date are hard to predict. A more detailed explanation about the order pattern is described in chapter customer demand.

The OTR is ok if the order is finished before the original end date (the date the customer desires his product). Though, 30 percent of the orders are requested with an original start date in the past. It is not possible to deliver these orders in time. An order which is delivered two weeks before the customer needs it is not on time, but is early. Though, it is measured as On Time to Request. In 2007 1,8 percent of the orders was delivered two weeks before the customer needed it. The OTR in 2007 was on average 54 percent.

OTR does not take the capacity into consideration. If all the customers demand on the same time their products, only a part of them can be finished on time (when not outsourcing). The production could have performed perfectly, but still results in a bad OTR performance.

Thirty percent of the orders have an operation on another production line or an operation that needs to be done by an external company. The operators of the Copper Bar have no influence on these processes. Therefore the OTR is based on quantities that can be partly influenced by the operators

(18)

from the Copper Bar. The indication how the operators from the Copper Bar perform in relation with the strategy should be improved and provide more accurate data about their performance.

Another comment on the performance indicators is that the TAKT screen visualizes an unfair lead/backlog of produced orders for that day. This number is a derivative from the TAKT time. This TAKT time is does not change within the day. The output generated has a high correlation with the number of operators present. This number of operators fluctuates throughout the day due to operators who work in shifts and by daytime. Therefore the TAKT time should change within the day. This issue, and other Performance Indicators, will be further discussed in the chapter Design of Performance Indicators.

3.6 Resume

This part indicates that the desired performance for the Copper Bar is to deliver parts fast with a high reliability. At the moment is this desired performance is not yet accomplished. The actual

performance indicators of PMC3 to indicate the delivery reliability are not sharply enough defined to reach the desired performance.

(19)

Chapter 4: Production characteristics

This part aims to give answer to the first part of the second research question - What are the environment characteristics (production characteristics and customer demand)–. In order to give the reader a better understanding which subject is analyzed and how this relates to the other subjects, the conceptual model is displayed again.

Order Release Performance KPI’s Customer Demand TAKT System Workload Control Production Characteristics

Figure 9: Conceptual model, analyzed subject

The modification of the TAKT system and the design of the order release are dependent on specific production characteristics of the production line Copper Bar. The way orders are released (in the introduced workload concept) and start at the TAKT system is e.g. dependent on the routing of the products, and the variation in production times of orders. In order to determine what the environment characteristics are and to not exclude characteristics, the literature of Bertrand et al (1998) will help to make a holistic overview. According to Bertrand et al (1998:44) production characteristics can differ in

complexity, uncertainty and flexibility. The flexibility of production has to deal with the complexity and uncertainty of production and customer demand. The customer demand will be discussed in the following chapter. First, the complexity, uncertainty and flexibility from PMC3 will be discussed in a detailed way, and the relevant aspects for the workload concept of these characteristics will be highlighted.

4.1 Complexity

Complexity is determined by variety in products, customers, operations, routings, number of productions per routing, resources, composition of the end product, and the amount of resources necessary at the same time per operation (Bertrand et al 1998:44). The customers of PMC3 are internal customers. The two major customers (LVS and MVS) represent 70 percent of the demand. Because PMC3 is not responsible for the end product, PMC3 has therefore not to deal with the complexity of composition of the end product. The different operations at PMC3 have already been described in the Value Stream Map, and shall therefore not be treated another time. The other factors which determine complexity will be discussed now.

(20)

The variety in products is high; in a record of nine months of orders in 2007, around 3,780 different articles have been produced of the 10,682 orders those nine months. This means that on average one article has been ordered three times in these nine months. The volume of an average order (which is regarded as the batch size) is 23 the same articles per order, which is low. PMC3 has to deal with a low volume/high variety demand.

The amount of resources necessary at the same time is moderate. The necessary resources are the copper bars, which are delivered by the warehouse, capable operators, functioning machines, the right drawing of the product and a right computer program. The drawing and the computer program are delivered by the department work preparation, which are sometimes mistaken or not available at the moment required. The predictability of delivery times of materials and components is determined by the Inventory. The inventory delivers material on average within one day to the production line Copper Bar. Unfortunately there is no data available how many times there was no material, or how long it took the inventory to bring the material.

Routing of orders

Most products do not need to pass all the operation stations and the sequence of operations can be 1 2 but also 2 1, therefore this process has to be designed as a job-shop process (Bertrand et al, 1998:158). All operations start at one of the punching machines, which have a great influence the way the order release system is designed. The design will be described in a later chapter.

On average an order has to follow 3.5 operations, where 2.8 of those operations are done at PMC3 and the remaining 0.7 operations are done by other PMC’s, powder coating or external manufacturers. 13.4 percent of the orders return to PMC3 for a second time after an operation outside PMC3.

A “from to” matrix in percentages is constructed of 10,682 orders which contain 24,050 operations, which visualize which operations sequence each other and the size of these streams. The operation bending Kennedy has been excluded from this Kennedy, as it represents les than 0.18 percent of the operations.

from/to punching bench working drilling bending Ehrt milling bending Promecam powder coating extern other PMC’s

punching 0,00 24,37 1,18 9,08 1,14 0,84 1,02 0,48 1,46 bench working 0,17 0,00 1,68 15,02 0,26 1,14 8,70 2,35 0,42 drilling 0,00 1,27 0,00 0,00 0,04 0,13 0,22 0,94 0,03 Bending Ehrt 0,00 11,81 1,64 0,00 0,12 0,16 0,37 1,53 0,12 milling 0,00 0,87 0,12 0,21 0,00 0,14 0,00 0,06 0,01 bending Promecam 0,00 1,20 0,28 0,08 0,02 0,00 0,00 0,69 0,20 Powder coating 0,00 1,28 0,02 0,00 0,07 0,00 0,00 1,87 0,00 extern 0,04 2,61 0,00 0,00 0,00 0,00 0,05 0,00 0,15 other PMC 0,14 0,60 0,16 0,62 0,01 0,37 0,05 0,38 0,00

(21)

This “from to” indicates that there are some logical steps which sequence each other, but there are also a high number of other routings. A lot of operations go from bench working to bending Ehrt and vise versa, which is also necessary at other operation stations. Therefore a job shop production is necessary and a line production is not an option.

4.2 Uncertainty

Uncertainty is created by unpredictability from the demand side and the process side. The demand side will be treated in a separate paragraph.

Uncertainty from the process side is related to the reliability of machines, fluctuation in production times, the degree of absence, the degree of technological control, predictability of delivery times of materials and components and the predictability of quality of materials and components (Bertrand et al 1998:47).

According to the department for quality assurance the uncertainty from fluctuation in quality of material is low and therefore not relevant. The degree of absence of operators has been improved from an average of 11 percent in 2005 to an average of 6 percent in 2007, which is low. There is little uncertainty from the factor absence of operators. The predictability of delivery times of components is high, because this is a well controlled internal process. The warehouse delivers the demanded products in a controlled way. There is little uncertainty in this factor. The workload control concept is not influenced by these factors. The other factors Bertrand et al (1998) describe will be discussed more in detail.

Reliability of machines

(22)

0 2 4 6 8 10 12 14 16 18 punc hing * * benc h wor king * drilli ng bend ing E hrt ** bend ing K enne dy millin g bend ing P romec am m in ut es

Figure 10: Average workload per machine per order 2007 * Six workplaces for bench working ** Two machines

(23)

0,00 2,00 4,00 6,00 8,00 10,00 12,00 14,00 16,00 18,00 20,00

TPM yearly inspection breakdowns

da ys o ut o f w or k Punching machine 1 Punching machine 2

Figure 11: Days out of work punching machines 2007

Punching machine 1 was not able to operate up to 25.7 days in 2007 and punching machine 2 was not able to operate up to 18.7 days in 2007 from the 250 operating days per year. On average the two machines were 8,89 percent ((25.7 + 18.7) / (2 machines * 250 operating days))*100 percent) of the time not able to operate, of which 62 percent was due to breakdowns.

Fluctuation production times on machines

The fluctuation in production times per order on a specific machine identifies the uncertainty that orders contain a certain amount of work on a specific machine. Insight in this fluctuation explains that the required capacity of production time on this machine from a random sample of orders in production will differ more with a high standard deviation. In table 2 is the total production time calculated by the (summation of orders ((batch size * up time single article) + positioning time)) / number of orders. percent of orders with an operation on Average batch size per operation Up time per single article

(min) positioning time(min) total production time (average/ operation) (min) Standard deviation/ Production time/ operation (min) Number of break-downs 2007 punching 98,64 22 0,81 14,56 31 46 15 & 34 bench working 78,60 20 0,73 14,05 28 38 3 drilling 11,39 24 0,94 19,80 43 62 2

bending Ehrt 58,18 16 0,81 15,44 29 41 30 & 17

bending Kennedy 0,18 16 1,10 12,00 29 11 1

milling 3,75 26 2,01 35,92 89 87 5

bending Promecam 6,20 37 0,64 30,02 54 40 13

(24)

The positioning time is relatively high; most of the time it represents half of the total production time. At punching the positioning time is due to changing tools. A new punching machine would increase the number of different tools available and would therefore decrease positioning time a bit. Standardizing products by design-for-manufacturing could reduce the number of changing tools per day. Though, realizing the standardization of products is outside the scope of this project.

Also the standard deviations of the operations are high. In order to meet the demand, (because there are more work stations than operators) operators have to be multi skilled. In the chapter customer demand, a more detailed discussion on how to deal with these high standard deviations in production times will be presented.

Technological control

The degree of technological control can be determined by quality of products (the amount of wrongly produced products) and by the deviations in production throughput time. Wrongly produced products are discovered in the production line, or by the customer. The wrongly produced products which were discovered at the production line are around 0.9 percent of the total amount of products produced. The operation which represents most of the flaws (55 percent) is bending Promecam. The wrongly produced products which are discovered by the customer are around 1.2 percent of the total number of products produced. In total there are 2.1 percent of the products wrongly produced. The technological control in quality is well controlled.

The deviations in production throughput time have been quite high, eighteen percent of the orders exceeded the production throughput time of 20 hours. The reasons for this excessive throughput time will be discussed in the chapter TAKT system.

4.3 Flexibility

Flexibility of materials and capacity resources can be used to lower uncertainty and deal with complexity. Flexibility is related to the following aspects: Multi trained operators, low positioning time, commonality, short delivery times of components and materials, overcapacity, ease of changing capacity and stocks (Bertrand et al 1998:49). The delivery times of components and materials and the positioning time has been discussed in the previous part about uncertainty. Commonality (the use of the same parts in different assemblies) is interesting, but is out of the scope of this project. The other factors (Bertrand et al 1998) which determine flexibility will now be discussed.

Multi trained operators

There are more workstations than operators and sometimes operators have to switch workstations in order to finish an order. Incapable operators are a form of waste due to underutilization of equipment. The Copper Bar works with two shifts. Three operators work, due to physical reasons, only by day time, and participate partly in both shifts. The shifts change each week from working in the morning to working in the evening.

(25)

offer the required capabilities. E.g. when only one operator is skilled to punch, but 27 percent of the operation time per order is on one of the punching machines, there are probably too few operators skilled to punch. The average number of operators present per shift are calculated by the number of FTE’s per shift (derived from the capability scheme in appendix 2) times the factor 0.82. This factor 0.82 represent how often one FTE is present per factory production day (216 workdays per FTE – 6% illness) / 250 factory production days).

The number of operators which are skilled to do the operation punching, bending Ehrt etc. are derived from the capability scheme in appendix 2, and are also multiplied by the factor 0.82, represented by the Italic numbers. In table 3 are e.g. on average 7 operators present between 06:00 and 07:30 of whom 3.1 are skilled to punch.

In order to calculate the average necessary capacity per order, the necessary capacity per operation is calculated by dividing the production time of an operation ((number of articles * production time of an operation) + positioning time of an operation) by the total production time of an order. The average

necessary capacity per order is the summation of al these operation production times divided by the summation of all the production times of an order * 100 percent. This relation is presented below.

A ve ra ge n um be r of o pe ra to rs pr es en t pu nc hi ng E hr t (2 m ac hi ne s) be nd in g E hr t (2 m ac hi ne s) be nd in g P ro m ec am 50 0 be nd in g K en ne dy be nc h w or ki ng m ill in g D ril lin g O th er P M C 's + e xt er na l o pe ra tio ns S um m at io n av ai la bl e sk ill s pe r s hi ft average necessary capacity/order in % 27,4 14,4 3,2 0,2 25,3 2,9 4,2 22,4 available operators 06:00-07:30 7,0* 3,1 2,3 0,8 1,6 5,6 0,8 5,6 ? 19,8 available operators 07:30-14:30 9,1* 3,6 2,8 0,8 3,0 7,7 0,8 6,1 ? 24,8 available operators 14:30-16:00 7,6 4,3 2,7 0,0 2,7 7,6 0,0 2,7 ? 20,1 available operators 16:00-24:00** 5,5 3,9 2,2 0,0 1,4 5,5 0,0 2,2 ? 15,2 Table 3: Capability, capacity and flexibility scheme, Copper Bar, December 2007

*One operator works mostly for another PMC, but in this table the full time is attributed to PMC3. **On Friday they work till 18:30

(26)

sums the skills operators) is discovered that the operators from the second shift have less skills and are therefore less flexible. On average are 7.3 operators present per shift ((7.0+9.1+7.6+5.5)/4).

Capacity and ease of changing capacity

Capacity can be constrained by either machines or by operators. First I will discuss machines as the constraining factor for capacity, thereafter the available operators as the constraining factor.

As can be seen from figure 10, punching machines have the largest workload and are therefore the bottlenecks of PMC3. The bending Ehrt machines require the second most capacity. There are two bending Ehrt machines. If a bending machine is out of work due to technical problems, one of these machines will be the capacity constraining machine.

The available capacity for two punching machines, when operators work in a three shift, without breakdowns, preventive maintenance, missing material or pauses, 48 hours (2 machines * 24 hours). In 2007 the Overall Equipment Efficiency has been measured for the two punching machines. This measurement should give more insight how much time of the total production time the machines were able to contribute value to the product, and how much time the machines were not able to contribute value to the product due to different forms of waste. However, the value of these numbers is minimal due to too few measurements. The two punching machines have been measured only 25 hours and 31 hours respectively and the results from these measurements do not give a representative outcome. The first punching machine would have an effectiveness of 90 percent, lacking 5 percent of waste due to set up time, and 5 percent of waste due to missing materials. The second punching machine would have an effectiveness of 98 percent, lacking 1 percent of waste due to set up time, and one percent that there was no operator available. However, on average the set up time per order is 46 percent (derived from table 2 on page 22). Therefore, I regard the available OEE numbers as not representative and will not use them for further calculations.

From Monday until Thursday, the on average used capacity of these punching machines is 27.71 hours (2 machines * 16.3 hours (18 hours operators are present– 1.7 hour break)) * 0.85 productivity factor). The productivity factor is an estimate which represents the factor that a machine is able to operate during its total production time, which is lower than 100 percent. This is because of machine breakdown, TPM, planned preventive maintenance (these three together result in 8.9 percent downtime), wrong or missing construction drawing or computer program, which is estimated all together as 15 percent downtime. The productivity factor is therefore 0,85 ((100 percent – 15 percent downtime) / 100 percent). On Friday the average available capacity is 21.25 hours (2 machines * 12.5 hours (13.5 hours operators are present – 1 hour break) * 0.85 productivity factor). During a week the available capacity of the punching machines is 26.6 hours per day ((4 days * 27.71 hours + 1 day * 21.25 hours) / 5 days), with the assumption that there are enough skilled operators.

(27)

Regularly no other machine capacity would give a constraint with an output of even 95 orders. (26,6 hours bending per day / (0,48 bending Ehrt hours per order * 58,18 percent of the orders require a bending Ehrt operation)).

Punching capacity can be increased due to 1) an extra nightshift, 2) working during the breaks 3) working on Saturdays and 4) an extra punching machine. I will shortly discuss each possibility. Operators have 50 minutes break.

1. An extra night shift at the punching machines will result in 37,8 punching hours ( 3 operators * (8.25 hours operators are present – (50 minutes break / 60minutes)) * 2 machines * 0.85 productivity factor). This is an increase of 11,2 hours per day (37,8 – 26,6).

2. Punching during breaks would increase capacity in a two shift 100 minutes per day and in a three shift 140 minutes per day. Operators are obliged to take a break; therefore it is not possible to demand from them to work during the breaks. Though, the operators (three in total) who work due to medical reasons on day time could start working on the punching machines during the breaks at daytime. Only one of these operators is trained to punch, the other two operators could be trained. I would not recommend this solution. The gain of each time 10 minutes or 30 minutes by working during the breaks has also a negative effect. The puncher works in a by him determined sequence of orders, depending on changing as few as possible tools at the punching machines. Another operator could easily mess up his

schedule, and therefore lose efficiency. Another factor for productivity is social cohesion. The social cohesion is for a large part formed during the breaks. Therefore I would not

recommend punching during the breaks.

3. Punching on Saturdays with the punching machines would increase capacity 9 hours ( 2 machines * (6 hours operators are present – 40 minutes break ) * 0.85 productivity factor ) 4. An extra punching machine would increase punching capacity at least by 13 hours in a two

(28)

Capacity can also be constrained by the number of operators available. There are more workplaces than operators available. This would imply that if there are more employees on the floor, more hours should be worked, resulting in a higher output and vise versa (when the punching machines are not the bottleneck). Because the order size fluctuates and there is free capacity on machines (up to a certain level) it is interesting to investigate the relationship between the number of operators and the amount of production.

The used MRP system is BAAN. BAAN stores the information which hours are produced on which day. Orders have a forecasted production, and set up time. With this information it is possible to determine how many hours are produced in a day. There is also information how many operators and for how long operators were present that day. The relation of the data from both man hours on the floor and booked hours on BAAN was showing an unstable pattern. This has several reasons. Sometimes there are a large number of orders which require the same articles. In such cases there is no set up time, though BAAN calculates that there is. Some operators help at other product lines if there is little work to do, and vice versa. It is not possible to book hours on BAAN between 23:00 and 02:00, which is done the day after. Because of these reasons the data has been smoothed with 5 days. After that, in order to visualize the relation between worked hours and booked hours on BAAN, the number of worked hours (form operators) has been put in an ascending sequence, keeping the fixed combination of coupled cells. Therefore is this graph not chronological. This information is captured in the graph below.

Baan produced hours versus hours of employees on the floor per day (January-november 2007; N=214 60 70 80 90 100 110 120 Days H ou rs

BAAN booked hours (smoothed 5 days) Hours employees (smoothed 5 days)

15 per. Mov. Avg. (BAAN booked hours (smoothed 5 days))

(29)

The amount of hours booked on BAAN represents the amount of work produced. This data is still fluctuating a lot. In order to visualize the average relation more clearly, a moving average of 15 days has been constructed.

This graph shows that on average there is a positive relation between how many hours of work have been done and how many employees are on the floor with a ratio of 0.85, till a certain amount. This is at the point where there are 105 hours of employees on the floor within one day. This is because the punching machine capacity is constraining the amount of output. Therefore it is not desired to have more employees on the floor which contribute to more than 105 hours a day, equally divided on the two shifts.

With more than 110 hours of employees (which is 5.8 employees per shift), even less output is produced. One of the reasons to explain this is that operators, who have no work, can keep other people from their work. More people result at this level in less production. In a two shift not more than 5.8 operators (110/19 hours present operators) should be present; productivity will not increase with more operators.

4.4 Resume

(30)

Chapter 5: Customer demand

This part aims to give answer to the second part of the second research question - What are the environment characteristics (production characteristics and customer demand) –. In order to give the reader a better understanding which subject is analyzed and how this relates to the other subjects, the conceptual model is displayed again.

There is a variety in customers and orders requested by these customers. Orders have a routing, and consist of an amount of the same articles. PMC3 is an internal supplier, mostly for the departments Low Voltage Systems and Medium Voltage Systems. These customers represent around 80 percent of the demand. Customers have, unfortunately, a fluctuating demanding pattern. Orders can be requested a long or a short period before the customer desire to have the order; the number of orders requested fluctuates and the size of orders fluctuates. The order differences have influence in the way the workload control concept can be used. The fluctuating demanding pattern has influence on how the orders can be released.

5.1 Order differences

There are four different order streams; Kanban orders, Customer Specific orders (PRP orders), regular orders (orders determined by the Manufacturing Resources Planning (MRP)) and Quick Service orders. There is a difference in the way these orders are released and a difference in how these orders are treated during production. The differences in the way orders are released will be explained in the section ‘The workload control concept, order release’. Briefly it will be explained which differences there are between these four different orders and how these orders are treated differently during production.

(31)

orders have a predicted production time, which is automatically determined for MRP and Kanban orders by (the number of articles * production time article) + positioning time. For PRP and Quick Service the department work preparation predict the order production time. On average per week the flux of the different streams are:

Average orders/week Average production time / order (minutes) St.dev production time (minutes) / order

MRP 107 139 213

PRP 108 91 181

KANBAN 44 145 118

Quick Service 10 ? ?

Table 4: Differences Order Streams

This table indicates that there are some small differences in the average production time and the standard deviation in the production time. The orders differ not in other production characteristics. Because the discussed differences are small, the different order streams will be treated the same in the workload concept.

Steering on the amount of orders

In order to manage the production it is interesting how many orders can be produced within a time frame. The customer is paying for the order, not for the amount of hours the production floor was busy to complete the order. The customer demands an order and not a part of the capacity from the production floor, but orders require a certain capacity from the production floor. It is clear that there is a relation between capacity of the production floor and the amount of orders which can be produced. Therefore it is necessary to discover the variation in production time of orders within a certain time frame. A time frame of a year will give a representative view of the reality.

(32)

Graph 2: Fluctuation total production time orders Copper Bar 2007

The graph indicates that the average production time is 116 minutes per order, and the variation in production time of orders is moderate. Some orders have an exceptional high production time, which increase the standard deviation enormously. These orders mostly consist of a high number of articles. In order to better control the system by number of orders, the number of the articles per order from the orders with a higher production time than 1000 min (which were 93 orders in 2007) should be decreased. This would lead to a lower average production time and a lower standard deviation respectively 103 and 111. This lower standard deviation is desirable when managing the order release by the number of orders. Therefore these large orders should be divided in two separate orders.

(33)

Graph 3: fluctuation total production time of 50 orders Copper Bar 2007

This graph indicates that with the same capacity on the production floor some fluctuation can be expected in the number of orders produced within a day. The fluctuation in necessary production time for 50 orders is now relatively low and more balanced. If the standard deviation in production time of 50 orders was still high, the fluctuation of amount of orders produced within one day would also have been high.

(34)

Graph 4: fluctuation punching production time of 1 and of 50 orders Copper Bar 2007

This graph indicates that on average 32.3 punching minutes are necessary for one order, and 25.5 punching hours are necessary for 50 orders. The excessive production time of orders (represented in the ellipse) will be partly compensated by the low production time from other orders within a day. In 4 percent of the days the necessary production time of the punching machines will be twice as high as the average production time per day and in 3 percent of the days the necessary production time of the punching machines will be twice as low as the average production time per day, with an average output from the punching machines of 50 orders per day. When steering on the amount of orders, this insight helps to determine how big the buffer at the punching machines has to be.

On average, when producing in two shifts, there are 26,6 hours punching capacity per day. In 35 percent of the days there will be too few punching capacity to process all 50 orders. The conclusion from these graphs is that the production floor can be managed by the number of orders when the production from the previous day will be taken into account.

5.2 Order pattern

The date on which a customer makes a hard reservation (SFC) for a product is called the entry date. The customer requires the product to be finished on a certain date, which is called the original end date. BAAN determines for the MRP and KANBAN the throughput time and in relation with the by the customer determined original end date, the original start date. The throughput time is estimated by the number of operations and the identity of an operation.

(35)

preparation added the necessary routing to produce the product, with no information about the necessary operations to complete the order. Therefore 40 percent of the orders has an original start date which could be wrong to produce the product on time.

For the operations punching and bench working on PMC3, BAAN calculates for each operation one day. For the operations drilling, bending, milling and bending Promecam on PMC3, BAAN calculates for each operation half a day. For the operations powder coating is on average 2.8 days calculated, depending on what kind of treatment and with or without bench working as part of the operation powder coating. For another operation on another production line, BAAN calculates on average one day. For the operations which have to be done by an external supplier BAAN calculates, depending on the external supplier, three to fifteen days. These production throughput times for single operations are relatively large, as products have on average 3.5 operations and are finished on average within 13 hours and 50 minutes.

In general, if the original start date is before the entry date, the order is demanded late. There is not enough time to produce the product before the customer needs it5_{. The OTR and the order pattern} declines. From the MRP, PRP and Kanban orders between 15-06-2006 till 29-11-2007, were 30 percent ordered late, with an average of 16.8 days. These were MRP and PRP orders, of which 80 percent of those late orders were PRP orders. Kanban orders are ordered on the same day as the start production date. An order pattern of 100 percent indicates that all orders were ordered before the start production date. Because 30 percent of the orders were ordered late, and 18 percent of the orders were ordered on the same day as production should start, the order pattern was 52 percent (100 percent – 30 percent – 18 percent). The percentages and standard deviations of orders which enter and have to start production within a certain amount of days are presented in the table below. Entry Date - Start Original prod. Date % cumulative standard deviation

negative 29,90 29,90 19,11 0 17,52 47,42 10,42 1 3,45 50,88 4,31 2 3,49 54,36 3,70 3 3,15 57,51 3,58 4 2,04 59,55 1,44 5 2,53 62,08 2,31 6 2,66 64,74 2,27 7 5,29 70,03 2,91 8 2,13 72,16 2,39 9 1,33 73,49 1,20 10 1,92 75,41 1,06 11 2,02 77,43 1,20 12 2,65 80,08 2,38 13 1,34 81,42 2,21 >13 17,70 99,12 2,79

Table 5: Order demand (days before production should start) Copper Bar 2007

(36)

With an average of 50 orders entering a day, on average 15 orders should have started production before the entry date, which is not possible. This is because customers order their products too late. These high standard deviations indicate that a lot of data points are far from the mean, which represents a high uncertainty. An uncertainty to predict how many orders will enter that day which should start production within a certain number of days.

This information determines that there is a lot of uncertainty which amount of capacity on the work floor should be reserved for hard orders which haven’t been requested yet. Insight in these fluctuations indicates that it is hard to predict how many orders with a certain start production date will enter per day.

The historical fluctuation in number of customer demand is graphed below:

0 100 200 300 400 500 600 700 800 1 _{6 11 16 21 26 31 36 41 46 51 5 10 15 20 25 30 35 40 45 50 4 9 14 19 24 29 34 39 44} orders completed new orders av. orders cpl. av. new orders min orders completed max orders completed min new orders max new orders

2005 2006 2007

Graph 5: Chronological order demand and production 2005, 2006 and 20076

The amount of new orders increased from 2005 to 2006, and from 2006 to 2007. The increase in orders is a result of a higher demand from customers and partly a result of the decrease of order size. Production has been working a lot more hours to cope with this higher demand (on average 268 orders have been produced per week in 2007). This year (2008) will probably show another increase

(37)

in orders. Order sizes will decrease more and the customer LVS, representing 70 percent of the total demand from PMC3, announced to have a 25 percent sales increase. This implies that either the output has to increase, or a part has to be outsourced.

5.3 Resume

(38)

Chapter 6: The workload control concept

This section describes how the workload was controlled at the moment my internship started (30-10-2007). First I shall give a small introduction about the TAKT system, the order release and the scope of the concept. After that it will be described how the TAKT system was working, what the discovered problems are and, in order to tackle those problems, what the modification of the TAKT system is. The same methodology will be used for the order release, first I shall describe how the orders were released, what the discovered problems are and, in order to tackle those problems, what the design of the order release is.

A constraining Work In Process (WIP) system what is called the “TAKT system” was working quite well. Orders are coupled to a constant work in process (CONWIP) card and the production throughput time for these orders is visible. On average an order takes 13 hours and 50 minutes from the first production process till the last process in PMC3.

The old order release: the customer demanded an order on a certain time. The production order was automatically printed three days before the start date and the inventory department started to collect the material which it transported to the buffer at the punching machines. At this buffer the product remained till the puncher started on this order.

The scope of the workload concept is displayed in the figure below.

Figure 13: Overview Scope Workload Control Concept7

(39)

6.1 TAKT system

This part aims to give answer to the third and fourth research question regarding the TAKT system. – How are the TAKT system and the order release currently working -, and - What is the modification of the TAKT system, the design of the order release taken into account the known environment characteristics -. In order to give the reader a better understanding which subject is analyzed and how this relates to the other subjects, the conceptual model is displayed again.

6.1.1 Current working of the TAKT system

After the orders are released, they have to start being produced. At the shop floor a workload and (partly) sequencing control system was introduced which was operating for four months and it was called the ‘TAKT’ system. The TAKT system controls the amount of work in process and makes the production process more transparent. It assures that when an order enters the TAKT system, the order will leave the TAKT system within a certain amount of time. The sequence of processes an order follows in the TAKT system will be described in order to give a better understanding how the TAKT process works. To visualize the process, I refer to the displayed Value Stream Map on figure 4, page 9.

Orders wait in the buffer before the punching machine. The puncher determines with which order he starts. He will first finish all the orders placed on one barrow, before starting on a new barrow. Which barrow he chooses next depends on the amount of priority orders on the barrow. If there is no priority order, the puncher will choose the vehicle with the orders he likes most.

(40)

Figure 15: TAKT Screen

There are sixty cards available in the TAKT system, which means that the workload of the TAKT system has a maximum of sixty orders. The system is active from Monday till Thursday from six in the morning till midnight, on Friday from 6 till 18.30 and on Saturdays from six till midday.

After punching, the order is either finished or has to go to the following operation, described on the routing. If the following operation is not located at PMC3, the operator has to dismiss the CONWIP card from the TAKT system after finishing his operation. The throughput time will stop and is stored in the TAKT system. The dismissed CONWIP card is available for a new order. After each operation the operator has to fill in BAAN the number of wrongly produced products. A dismissed order can either go to the customer, or has to follow some more operations outside the Copper Bar. An order could enter the TAKT system again and a new CONWIP card will be added to this order when the next operation starts.

(41)

In order to know which operator has to work on which order, one of the operators makes a list with all the “red”, or almost “red” orders. This list contains information about where the orders are and sometimes which operations of the routings have to be followed.

Overall the TAKT system has reduced the production throughput time, it has made the process more reliable, the workload is more controlled and it has made the process more transparent. Problems are easier detected and adequate actions can be taken. The TAKT system is definitely an improvement.

6.1.2 Discovered problems of the TAKT system

The control of the throughput time at the TAKT system has been recorded. If the throughput time of an order in the TAKT system exceeds twenty hours, an operator has to give a reason for this. Eighteen percent of the orders exceeded the production throughput time of twenty hours. As can be seen from graph 6, flexibility operators, insufficient attention and technical problems represent most of the reasons for delay.

Graph 6: Reasons Exceptional Throughput Time Copper Bar (3-7-07 till 5-12-07)

The category flexibility operators represent operators’ lack of skill to do the required work.

Insufficient attention refers to operators having worked on the wrong order. The operator who makes the list with the “red” orders informs the other operators of his shift where the orders are which require the most attention. The operators who work on daytime (and not participate in a shift) take no part in this discussion. This is not desirable. And even when operators know on which orders they should work, it is not guaranteed that operators work on these orders due to personal preference. In the design chapter a short discussion will be presented how to deal with this problem.

(42)

This graph indicates that too large production time per order is not a significant reason for excessive production time. This indication supports the conclusion of the chapter Production Characteristics that the fluctuations in orders are small enough to manage the TAKT system and the order release by the number of orders.

Some rules about the TAKT system are not clear.

• The moment that an order should be added to the TAKT system. One operator adds the order to the TAKT system after the process punching and having the material checked on flaws, the other operator adds the order before the punching operation starts. The two supervisors have not the same thoughts about what is right. Sometimes orders which only have to be punched never enter the TAKT system because of this.

• It happened two days that orders were bundled, e.g. fifteen orders enter the TAKT system with only one CONWIP card. These orders are bundled because they contain the same articles, and are bundled by the department work preparation to save positioning time. • Orders enter the TAKT system without a CONWIP card. These orders might get after a while

a CONWIP card, and some might not at all. This results in wrong production throughput time feedback and more WIP. There is no data available how often this happened.

Besides these not clearly defined rules some other problems are discovered.

• The machine process bending Promecam is not located in the area of the Copper Bar. The operator has no vision on the TAKT screen. This means that if there is a queue of orders at this machine, the operator doesn’t know which order has priority above the other order. • The machine bending Promecam is used for PMC1 and PMC3. Only one shift an operator

who works for both PMC3 and PMC1 will do some orders from PMC3.

• The computer program for the TAKT system was developed by an external institution, and crashed on average once a month, resulting in a temporary uncontrolled production process and incorrect information about the production throughput time. Therefore does the data captured of the TAKT system, correspond not with the data captured from BAAN. (e.g. the amount of orders which passed the TAKT system).

6.1.3 Modification of the TAKT system

The TAKT system is re-designed in order to tackle the described problems.

(43)

The old stand alone TAKT system has been replaced for a new TAKT system which is connected with BAAN. The planner and I specified what the requirements of the system should be. The captured data of the TAKT system correspond now with the captured data from BAAN. An extra feature of this connection with BAAN is that the order routing can be visualized by clicking on the production order. The IT department is working on a screen-visualization of production drawings. This will solve the problem that sometimes production drawings are not present. It is possible to view this screen at all the computers in the company which are connected to the network of Eaton Holec. Therefore it is easier for the supervisors to view how PMC3 is performing and operators can use more computers to add or dismiss an order to the TAKT system. Also the operator at bending Promecam has now a view at the TAKT screen, because he has a computer which is connected with the network. Thereby this operator knows now which operation should be performed first. Though, he still has to walk to the central point to get a CONWIP card.

The products are visualized on a slightly different way compared to the old TAKT screen as can be seen in figure 16.

Figure 16: New TAKT screen

Production Unit control

Master Thesis University of Groningen