medium sized Make-To-Order company

(1)

medium sized Make-To-Order company

an empirical contribution

Input-output and the control functions for system adjustments [Wiendahl et al., 2001]

© 2005, W. Kooistra

Master Thesis, August 31, 2005

Supervisor: Dr. M.J. Land

Co-Assessor: Prof. Dr. Ir. G.J.C. Gaalman

issued orders

urgent orders

released orders incoming orders

store

adjusting screw

"time limit "

adjusting screw

"capacity" adjusting screw

"load limit"

inventory level

(2)

Acknowledgements

This article is the joint result of a research project executed by researchers of the University of Groningen. The article will summarize the research prior and during the implementation of Workload Control (WLC) ideas at E.A. Broekema, situated in Veendam, Holland. At the moment this thesis was completed the implementation at E.A.

Broekema was still proceeding.

First I would like to acknowledge my supervisor and co-assessor for providing me with useful feedback and the inspiration for completing my task. My supervisor, Dr. M.J.

Land, mainly provided me with very detailed feedback and accurate guidelines. My co- assessor, Prof. Dr. Ir. G.J.C. Gaalman, judged my work and supplied me with some useful insights to handle problems and issues differently.

Second I would like to thank the students Ron Apeldoorn and Wiebren Janssen for their preliminary investigation of the situation and outlining the issues of the (manufacturing) system for implementing WLC. Next I would like to thank Chris Bakker who simultaneous with my trajectory started the implementation of WLC. Together we conducted a detailed measurement for justifying our statements. The last student I would like to thank is Derk van Veen. Derk developed the information system to handle the WLC concept precisely.

Third I would like to acknowledge the board and employees of E.A. Broekema for making this research possible. E.A. Broekema invested people, time and money for extending the empirical research on WLC and to improve their manufacturing system’s performance.

Finally I would like to thank my girlfriend and family for supporting me throughout the project. I value their patience and respect in times I was not available for them.

(3)

Introduction

This article will describe the implementation of WLC ideas at E.A. Broekema, a producer of (crop) conveyor belts. Chapters I to IV will handle the current situation and performance at E.A. Broekema prior to the implementation of WLC ideas. In Chapter I the situation is outlined and an impression of the company is given. Furthermore the challenges the company is facing are described in Chapter I. In Chapter II a technical assessment of the manufacturing system is made. The current order and planning process will be discussed in Chapter III. The first three chapters are a descriptive diagnosis without a conclusion. Chapter IV describes the performance at E.A. Broekema prior to the implementation of WLC. The format and choices for implementation of the WLC concept get shape in Chapter V. The specific implementation difficulties and decisions for effective implementation of WLC will be handled in Chapter VI. Chapters IV to VI analyze the situation more detailed and design the new WLC system. For clarity purposes a conclusion is added to these three chapters.

Prior to the implementation of the WLC concept extensive measurements were made.

These measurements were collected in a five month period from the beginning of December 2004 till the end of April 2005. The data used in this article is extracted from those measurements. The used data will contribute to outline the scale of the issues at hand. The problem areas are identified by using the data. Furthermore the data from the measurements will act as a signal where a gap exists between the current situation and the planned end condition to be accomplished with the WLC concept.

(6)

Chapter I: Setting

§1.1 The company

E.A. Broekema was founded in the late 50’s. Today E.A. Broekema has grown to a main producer of conveyer belts for agricultural purposes, together with their matching drive components. The belts are custom made and are used on agricultural machines for harvesting crops as well as in washing and sorting installations for the food processing industry and fishery.

Figure 1.1: Example of a standard double-belted conveyor (with drive components).

In 1996 E.A. Broekema became part of the German firm Artemis. E.A. Broekema is situated in the north of Holland and has another production facility in Minnesota, USA.

Both production facilities are supported by sales offices in Holland, France and the USA and work with several agents all over the world. About 80 people are employed at the production facility in Holland of which a small part consists of temporary staff. E.A.

Broekema delivers to customers in more than 30 countries and their export share is 90%.

§1.2 The product

The conveyor belts are used on washing and sorting installations and on agricultural machines to harvest all sorts of crops. These crops vary widely, from potatoes to shells and from tomatoes to nuts. To serve this broad application the belts can be custom made varying with respect to several dimensions, see figure 1.2. In 2004 7369 orders were produced all with their own dimension specifications.

(7)

Figure 1.2: Dimensions for a simple double-belted conveyor.

The conveyor belts consist of three main components: the rods, the traction belts, and extra or special parts. The output varies from conveyor belts with only two traction belts up to conveyor belts with four traction belts. For producing these conveyor belts the rods need to be fitted with the correct amount of flattened elements. The production distribution of these elements in the measured period is shown in table 1.1.

Rod type: # Orders % Orders 2 flattened elements 690 67%

3 flattened elements 272 27%

4 flattened elements 62 6%

TOTAL 1024 100%

Table 1.1: Distribution order set December 2004 till April 2005.

The first varying components are the rods. The elements of the rods which can vary are the rods length (B), the steel parameters (diameter (E) and composition), the number and specifications of the forged/flattened elements (A and H) and the physical appearance (e.g. cranking).

A second component that varies is the used traction belt. It varies from material specifications to thickness and wideness (C) and length of the belt (G). Furthermore the distance between the rods (F) is determined by the traction belts.

The last varying components are the extra or special parts that can be added to the conveyor belt. It is possible to cover the rods with tubes. These tubes serve as protection for the crops. Furthermore the tubes will reduce the rod gap clearance by increasing the rod diameter; this will determine the conveyor belt’s sorting properties. The same holds

A = Specification forged / flattened element B = Rod length C = Belt width

D = Conveyor belt width E = Rod (diameter) F = Inter rod distance G = Belt length H = Rivet hole

(8)

for vulcanized rods, rods completely covered with a synthetic material. The vulcanization operation is outsourced to a specialized company. Another variation is the fitting of flights or risers used for elevation of crops. The last three components, tubes, vulcanized rods and fitted special parts are not depicted in figure 1.2, because they will not appear on simple double-belted conveyor belts. For the WLC concept these components are important because they can make up a considerate amount of throughput time.

§1.3 The market

E.A. Broekema delivers to three categories of customers: machine manufactures (75%), mechanization companies (10%), contracting firms and end users (15%). In the first group most demand is initial demand. In the second group demand is split between initial and replacement demand. The third group consists of only replacement demand. Because of the heavy duty in which the conveyor belts are operating, the belts as well as the rods need to be replaced regularly. On average a rod needs to be replaced within two years.

E.A. Broekema delivers also to a subsidiary company in the United States. This company orders huge amounts of conveyor belt parts such as rods and belts for further assembly in their factory.

For the initial demand and the demand from the subsidiary company it applies that the delivery date is known in advance. These orders require flexibility of the manufacturing system on the longer term. The replacement demand is not known in advance. Capacity flexibility on the short term needs to cope with these urgent orders, because customers will not wait too long for a spare part.

Consequently three types of orders can be identified for the different customer groups:

• orders for complete conveyor belts

• orders for rods only

• orders for belts only

The last order type, orders for belts only, is not subject of discussion in this article because the belts department has enough capacity to produce these orders without obstructing the progress of the orders for the complete conveyor belts.

E.A. Broekema deals with a seasonal pattern in demand. In the peak season, January till July, they produce complete conveyor belts and are busy satisfying the replacement demand. In the off-season, August till December, E.A. Broekema produces for their subsidiary company in the United States and non-replacement items.

E.A. Broekema has no real threat in terms of new competitors or substitution products.

Within the market there is some competition and the competitors compete especially on delivery speed. E.A. Broekema’s negotiation position to suppliers as well as to customers is relatively weak. This position is a result of the size of most of their suppliers and customers. Consequently E.A. Broekema cannot influence the prices they are charged by suppliers and need to meet customer’s requirements, such as delivery speed and level of customization.

(9)

§1.4 The production process

In this section the production of the conveyor belt is described. The first sub-section handles the actual production followed by the consequences for the organization of the work in the second sub-section. The last sub-section discusses the used information system. How the production is organized and situated in the different departments will be discussed more detailed in Chapter II.

§1.4.1 Producing the conveyor belt

E.A. Broekema normally works with a five day workweek of eight hours. Adjustments in the normal workweek are possible depending on the situation the company is facing.

The conveyor belts are produced in four main departments, in which a set of operations are performed.

1. Rods department: the round steel rod is cut to length according to the conveyor width, the rod ends are heated, forged/flattened and rivet holes get punched at the ends.

2. Belts department: holes get punched in the belts and the punched belts are stocked. On customer order the belts are cut to length. Rivet retaining plates are placed between the traction belts underside profile and rivets are inserted into the plates and the belt. The rivet length relates to the belt thickness and to the rod diameter, e.g. forged thickness.

3. Special parts department: when special parts need to be fitted to the belt, in order to fulfill the customers demand, they will be produced and added to the rods and belts. The most common operation in this department is the fitting of flights or risers to the rods. Also the matching drive components are manufactured and enclosed.

4. Assembly department: rods from the rods department are fitted onto the rivet inserted belts from the belts department. These components are riveted into an integral package. The rivet head is absorbed into the countersunk retaining plate and a head is formed over the rod’s counter sunk upper surface. When the belt is assembled it will be coated with black paint for aesthetic reasons.

The machinery in these four departments is grouped according to their function. The layout of the manufacturing system will be described more detailed in Chapter II.

§1.4.2 Consequences for organization of work

The mainstream of products consists of the most common double-belted conveyor with limited special specifications. Nevertheless the production of the specialized conveyor belts is disrupting the production schedule. One consequence of the present product diversity is the variation in routings through the shop, because each order has several options for visiting mutually exchangeable machines. Another consequence of producing so much variation is that volume is generally low and production takes place in small batches based on customer orders. The small batch sizes and the product variety require general purpose machinery and mobility of the workforce.

Bertrand et al. [1998] distinguish two relevant factors that influence the logistical characteristics of the production situation. These factors are capacity complexity and

(10)

material complexity. At E.A. Broekema the capacity complexity can be considered as fairly high as a result of producing with several complicated and specific machines. The material complexity is considered low due to the limited number of input materials. The resulting production system is a job shop. E.A. Broekema is characterized by capacity requirements uncertainty and order priorities on the short term.

The product diversity causes the product movement from department to department in a variable but directed sequence. The directed character of the flow at E.A. Broekema for a general belt will be illustrated later in section 2.1. The directed flow is due to the assembly character of the production process, requiring a specific sequence of operations.

Due to the influence of the assembly character E.A. Broekema’s production system is not a ‘pure’ job shop.

Finally the consequence of working on customer specifications is excluding the opportunity for making to stock. Therefore the stock at hand at E.A. Broekema consists mainly of raw material inventory.

§1.4.3 Information technology

The production process is supported by two types of information systems. One information system in use is Exact for DOS, an ERP system. The applied version of Exact has limited opportunities for supporting planning decisions. The second type of information systems consists of the Microsoft office programs Excel and Access to generate overviews and support the planning decisions. The planning utilities for E.A.

Broekema of these information systems are limited. Currently planning needs to track approximately 300 orders simultaneously. Furthermore the orders follow many different routings and are processed on several machines. This results in uncertainty about capacity requirements and order priorities. Extra information support is needed to keep track of the orders and to release the right set of orders to keep the system efficiently loaded. The use of information systems will be dealt with in Chapter V and Chapter VI.

§1.5 The challenge

The objective for E.A. Broekema is to improve the delivery performance. Customers are mainly focused on speed instead of prices. For the customers it is more costly to wait for the parts than to pay little more for a faster delivery. Three elements of delivery performance will be distinguished: delivery time, delivery reliability and delivery flexibility.

The required delivery time for a standard double-belted conveyor belt is approximately two weeks. For the special conveyor belts (more than double-belted conveyor belts or conveyor belts fitted with special parts) a delivery time of three to six weeks is acceptable. The current average of little more than four weeks, see Table 1.2, is not satisfying. Especially if we bear in mind that 67% of the total number of orders consists of standard double-belted conveyor which require a maximum delivery time of two weeks. The orders for rods only, originating from the subsidiary company, have a longer delivery time mainly because they are ordered well in advance. In the analysis of the delivery time performance the result of required longer delivery dates by the customer is neglected.

(11)

# Orders Average delivery time

Overall 1024 20,4 manufacturing days

Orders for complete conveyor belts 925 16,9 manufacturing days Orders for rods only 99 53,5 manufacturing days

Table 1.2: Average delivery time December 2004 till April 2005 in manufacturing days.

E.A. Broekema tries to realize a delivery reliability of 95%. Over 2004 the delivery reliability was 97,5%. The percentage tardy is measured by dividing the orders that exceed the delivery date by the total of processed orders in 2004. The current level is more than satisfying and could signal that the production process is well ordered and able to process the accepted amount of orders. Another reasons for this high reliability percentage could be the result of setting delivery dates with too much redundancy or the fact that in 2004 the order portfolio was slightly lacking behind the expected amount of orders. In general E.A. Broekema makes delivery dates on time only the exact reason is not known in this stage.

The rush orders should be produced within one week to meet the current customer wishes. E.A. Broekema tries to realize that in the near future 5-10% of all orders can be completed within one week. The implemented WLC concept should create enough flexibility to realize a delivery date of one week for the rush orders.

To improve the delivery performance, especially the delivery time, first the causes of current delays and waiting times need to be known. One cause could lie in the pre- production throughput times of an order. E.A. Broekema already tries to limit and speed up the pre-production throughput time. Another cause could lie in the production throughput time. The production throughput time consists of waiting time, transportation time and processing time. The production throughput time will be subject to improvement given the capacity restrictions currently present at E.A. Broekema. The last cause could be the transportation time to the customer due to the global activities of E.A. Broekema.

This transportation time can make up a considerate amount of time. Nevertheless the transportation time is neglected in this article. To the customers a production end-date is promised. The customer can than calculate when the products are approximately available for further processing.

To accomplish the required delivery performances, especially the targets set for delivery time, the ideas of the WLC concept are implemented. The WLC concept tries to control queue length and balance the released load for each capacity groups by using input/output decisions. The reduced and controlled queue length should result in shorter and less variable throughput times.

The WLC concept for E.A. Broekema is fitted into an information system, programmed in Visual Basic, and provides real time feedback to the department planning. The information supported WLC concept helps to enable planning at E.A. Broekema in the following manner:

1. The WLC concept tries to reduce planning complexity by regulating the working methods, both for sales and for planning. Sales will be responsible for the acceptance of simple orders in the future, while planning can focus on the current (rush) orders

(12)

and the more complex orders. When orders are accepted the WLC concept will limit the load to be released by planning for each critical capacity type on the shop floor.

2. The WLC concept reduces queues of jobs on the shop floor in front of each critical capacity type. These reduced queues of jobs prevent excessive amounts of work to built up on the shop floor. The limited amount of work enables E.A. Broekema to switch quickly between products. The quick switches between products contribute to the delivery flexibility. Even the delivery reliability can be improved. When E.A.

Broekema is able to change quickly between products they are also able to process the urgent orders sooner. Only the capacity is already assigned to the production of products with standard urgency. E.A. Broekema needs to reserve capacity for completing these urgent orders besides the schedule for the products with standard urgency.

How WLC specifically realizes the objective will be discussed in Chapter V and VI.

(13)

Chapter II: The manufacturing system

§2.1 The physical structure

The order fulfillment process for E.A. Broekema is visualized in figure 2.1. Production takes place on customer order. Most raw materials are in stock, because the nature of demand is sufficiently predictable. Only the timing and volume of most orders needs to be determined, except for the highly specialized custom orders. The customer order decoupling point (CODP) is positioned at the raw materials inventory. For the production of the belts the customer order decoupling point is positioned after the first operation.

This is because E.A. Broekema stocks all the belts partly prefabricated. Some highly customized conveyor belts may require procurement on customer order. Figure 2.1 emphasizes the make-to-order nature of the production system.

Figure 2.1: Manufacturing system E.A. Broekema.

The rods department and the assembly department are identified as constraining departments with respect to capacity. Those two departments have the most influence on the throughput time because both departments are located sequential and determine the longest production route. The capacity of the belts department is primarily not constraining. Only the operation rod covering, performed in the belts department, sometimes needs to process a considerable high amount of work. Due to this high amount of work the operation rod covering could influence the throughput time heavily. The belts department will be included to depict the operation rod covering.

In the sub-sections 2.1.1 to 2.1.3 the three departments, their machinery, the material flows, the waiting times (WT) and the inventories are charted. The special parts department will not be discussed in detail. Although the special parts department is essential to complete the customized conveyor belt it is not constraining with respect to capacity. In the following sub-sections the departments will be visualized graphically.

The symbols in use are explained in figure 2.2.

(14)

Figure 2.2: Legend

The material flow is divided into three groups and depicted as in figure 2.2: a dominant flow, i.e. a flow which processes 50% or more of the total of orders processed; an average flow, i.e. a flow which processes between 10% and 50% of the total of orders processed;

an insignificant flow, i.e. a flow which maximally processes 10% of the total of orders processed.

Three types of waiting times are distinguished in figure 2.2. Congestion waiting times typically appear when subsequent processing times differ, when arrivals patterns are irregular, when processing times are long, when utilization levels are high and when sequence dependent set up times are apparent. Products need to wait for available capacity and queue up in front of the work station. Assembly waiting time is due to products waiting for other (semi-finished) parts. Batch waiting times occur when machines only start to operate if there is enough work to make production economical and efficient. Only the most frequently occurring waiting times will be depicted in each picture of the departments.

The depicted material flow and waiting times will be discussed in section 2.2.

§2.1.1 The rods department

In the rods department four operation types: cutting, bulging, pressing and tempering can be performed, see figure 2.3.

Figure 2.3: Rods department.

(15)

For cutting the steel three machines are available, one for coil cutting and two for bundle cutting. The steel is delivered on coil or in bundles and needs to be cut to the desired length. The operation coil cutting is not constraining in terms of capacity. The bundle cutting can become a constraining operation.

Shortening the coil into (connection) rods can be executed on the coil cutting machine.

The coil cutting machine can also produce steel on the same lengths as delivered by bundle. These lengths can than be processed on the bundle cutting machines. The bundle cutting machines use bundles of steel which they transform into (connection) rods. E.A.

Broekema tries to produce as much as possible on the coil cutting machine but this is limited to approximately 50% because not all steel types can be produced on this machine. The coil cutting machine normally has relative high sequence dependent set up times. E.A. Broekema already spend some effort to change over the machine more quickly and is able to change over the machine without completely processing the coil.

The operation type bulging is for strengthening rods with more than two forged/flattened elements on the spot where an extra forged/flattened element needs to be placed. This process is depicted in figure 2.4.

Figure 2.4: The formation process of a bulge to strengthen the rod.

The machine that creates a bulge heats up the rods on the spot where an extra forged/flattened element needs to be placed and presses the rod together until a bulge is formed, see figure 2.4 (Bulg-1). Rods for conveyor belts with four belts pass this machine twice and get two bulges (Bulg-2).

209 orders needed to be bulged out of a total of 1024 orders. A small amount of the 209 orders, 32 orders, where bulged twice. In total just under 116.000 rods where processed on the bulging machine, see table 2.1.

Machine # Orders # Rods

Bulg-1 177 85% 99391 86%

Bulg-2 32 15% 16408 14%

TOTAL 209 100% 115799 100%

Table 2.1: Output of the bulging machine December 2004 till April 2005.

The operation type pressing is performed on four flattening machines. There are two presses used for the rods with two flattened/forged elements, i.e. to be used for a double- belted conveyor. These machines are the flattening-2 and the 100 Ton. For the rods with three flattened elements the 160 Ton is used. For the rods with four flattened elements a separate manual machine is used. The flattening capacity used for the rods with four flattened elements is not constraining. Because the overcapacity the machine is not

(16)

depicted in figure 2.3 but will be referred to as flattening-4. Recently the company purchased a new press; a 400 Ton. This machine will be able to process all types of rods.

The process for most of these machines is the same; the rods are heated on the position(s) where the flattened/forged elements need to come, the press flattens the rod on these position(s) and rivet holes are punched in the flattened/forged element. Some of these machines have relatively high set up times between batches, especially the machines indicated 100 Ton and 160 Ton. These set up times are the result of the amount of bolts in the flattening block, these bolts need to be unscrewed and screwed. Another constraint changing over a press quickly is the temperature of the parts that need to be replaced. A high temperature does not allow changing over directly after an operation.

Machines of this type need some amount of work available to combine orders economically. The process for the 400 Ton will be a little different because it heats up the rods with induction instead of gas. This means that the machine can be changed over quicker because the machine parts will not become as hot as with the other machines.

Most orders, 641 orders, are for rods for a double-belted conveyor. The next most produced rod is the rod for a triple-belted conveyor with 265 orders. Only a small part of the orders, 46 orders, is for rods for a conveyor belt with four belts. The remainder is processed manually. In total just under 582.000 rods where pressed. Figures about the amount of orders processed on the 400 Ton are not yet available.

100T 544 53% 246292 42%

160T 265 26% 153846 26%

Flattening-2 97 9% 152609 26%

Flattening-4 46 4% 18520 3%

Remainder 72 7% 10689 2%

TOTAL 1024 100% 581956 100%

Table 2.2: Output of the different presses December 2004 till April 2005.

The fourth operation tempering relates to a heat treatment of the rods in an oven which makes the rods stronger and more durable. Only a small part of all rods is processed on this machine. After being tempered the rods have to wait five and a half hours for further processing. The rods cool down slowly to get the best quality of the steel properties. The tempering machine is only started when there is a significant amount of work. It is inefficient to put the oven into production for small batches. Finally the finished rods in de rods department are delivered to the assembly department.

Not all types of rods go directly to the assembly department. Some rods are sent to the special parts department to fit flights or risers, other rods get covered with tubes in the belts department. Not al equipment is depicted in figure 2.3. There are also some tools for manual processing. These tools will only be used for very small batches or highly specialized customer demands, e.g. cranked rods. Furthermore the production of some specialized rods is outsourced to other companies. An example of outsourced production holds for vulcanized rods.

(17)

§2.1.2 The belts department

The belts department is illustrated in figure 2.5. Belts arrive in this department on big rolls. When capacity is available holes get punched into the belts. After that the punched belt is stocked. This stock point is a customer order decoupling point (CODP). On customer order the belts are cut to length. Rivet retaining plates are placed between the traction belt’s underside profile and rivets are inserted into the plates and belt. The belt is now semi-finished.

Another operation performed in this department is the covering of the rods. First the tubes are cut to length on customer order. Then they are than fitted mechanically on to the rods. Only a small amount of all rods need to be covered with tubes, but when the rods need to be covered, this can make up a considerate amount of work and can slow down throughput time. Totals of orders and rods processed are not measured in this department because the belt production is not constraining in terms of capacity.

Figure 2.5: Belts department.

§2.1.3 The assembly department

The assembly department, depicted in figure 2.6, processes the semi-finished belts and rods into a finished conveyor belt.

The first operation is connecting the belts from the belts department. Hinges are mounted and a connection rod is added (after being processed in the special parts department).

The second operation in this department is riveting the finished belt and the matching rods into a complete conveyor belt. There are five riveting-2 machines to produce a double-belted conveyor, two riveting-3 machines to produce a triple-belted conveyor and one riveting-4 machine to put together a conveyor belt with four belts. The riveting operation is the capacity constraining operation in the assembly department. Each riveting machine can become a bottleneck and delay the throughput time.

(18)

After riveting the components into a complete conveyor belt it is coated with black paint for aesthetic purposes. Before shipment the matching drive components from the special parts department are added and the package is completed.

Figure 2.6: Assembly department.

Because the riveting machines are capacity constraining an overview of their output is depicted in table 2.3. The riveting-2 machine processes the majority, 630 orders, of the total of the orders. The riveting-3 machine processes a considerate amount of rods. And the riveting-4 machine processes only 56 orders. In total little more than 398.000 rods where processed on the riveting machines.

Riveting-2 630 68% 310366 78%

Riveting-3 239 26% 77916 20%

Riveting-4 56 6% 10391 3%

TOTAL 925 100% 398673 100%

Table 2.3: Output of the different riveting machines December 2004 till April 2005.

(19)

§2.2 The material flow

At E.A. Broekema the machines are grouped in a functional lay-out. This means that the machine groups can perform similar processes. The products move in a directed manner through the manufacturing system. Although there is a parallel structure the capacity constrained route (flow > 50%, in the pictures of section 2.1) a conveyor belt follows can be seen as sequential structure. The waiting times on the capacity constrained route are the result of balancing and safety decisions to keep the process economical and running. The balancing decision needs to be made because the uncertain arrival times of assembly parts and differences in subsequent processing times. Buffers will arise to cope with this uncertainty. The safety decision is related to efficient use of capacity to keep prices within limits. To enable this efficient use, buffers are situated in front of most capacity resources.

The points where the most apparent waiting times appear have been visualized. In section 2.1 three types of waiting time have been distinguished; congestion WT, assembly WT and batch WT.

Congestion waiting time is due to materials waiting for capacity con straining operations. This waiting time appears at almost every stage but mainly at the cutting process, the tempering process and riveting process. Especially at the riveting machines the processing times vary significantly, which contributes to congestion.

Assembly waiting time appears in the stages where parts need to be combined into a (semi-)finished product. This is the case for rod covering, hinge mounting and belt connecting and the riveting process.

The batch waiting time is caused by materials waiting to fill a batch or for setting up a machine. The batch waiting time can be found before the (big) presses and the machine that creates a bulge in the rod. The reason for this batch waiting time is the high sequence dependent set up time and the accompanying changeover costs.

(20)

Chapter III: The order and planning process

§3.1 Introduction

The ordering and planning process is coordinated by three departments: sales, procurement and planning. They all have their own responsibility although cooperation is essential to fulfill customer demand. Their roles and activities in the situation prior to the implementation of the WLC concept will be discussed in this chapter, beginning with their roles and activities performed during the order process followed by their roles and activities during the planning process. The average throughput time of the order and planning process activities are described in sections 3.2. In Chapter IV the total average throughput time will be discussed.

§3.2 The order process

How the orders are processed and which tasks are performed by the different parties involved will be discussed next in a stepwise overview of the order process. Currently it can be shown that the interdependencies between sales and planning are sequential in terms of Thompson [1967]. The matching coordination mechanism should be coordination by plan. In reality there is hardly any coordination by plan and sales and planning are coordinating through standardized routines and rules. The level of information exchange in the current system should be adjusted upward to the information need opposed by the sequential interdependencies.

§3.2.1 The order process stepwise

STEP 1 Enquiry: Sales registers each customer enquiry and creates an order number.

Furthermore sales makes a product recipe. The recipe consists of the general customer information, a rough bill of material, the main output characteristics of the ordered product and a required/standard delivery date and a price.

STEP 2 Checking: The recipe made by sales is sent to planning. Planning checks the technical feasibility of the order, completes the bill of material and sends procurement a request to check material availability.

STEP 3 Procurement: Procurement checks material availability. Stock levels are kept within its boundaries by using the order point technique. This results in a enough material available for production at E.A. Broekema and a minimum amount of safety stock that is always present.

STEP 4 Delivery-date setting: If it is technically possible to produce an order and material is available, a due date in terms of weeks is set by planning. This is mainly based on prior experiences and a rough overview of the capacity in the assembly department.

The capacity overview is based on the number of rods that needs to be processed, approximately 40.000 rods per week can be assembled. Complexity of the rods that need to be assembled is not considered. This can result in erroneous capacity estimations.

Actual capacity use, in minutes of work, could be more or less in case more or less complex rods need to be assembled. This could cause delays or overcapacity. If an order

(21)

is new or highly customized a drawing of the product can be included for confirmation by the client. This is the case in maximally 10% of all orders.

STEP 5 Acceptance: Planning provides all the necessary information to sales, which is now able to confirm an order to the client and if necessary adjustments in the order can be made. If the customer wants an earlier date than accounted for it can be regarded as a rush order. This is only accepted for certain customers, for some types of conveyor belts and if it is not disrupting the production process too much.

The order process steps as described above are called the ‘customer enquiry stage’. The enquiry stage takes 1,8 manufacturing day throughput time. This 1,8 manufacturing days makes up 11% of the average total throughput time (16,8 working days). An overview of the average total throughput time and its components will be presented in Chapter IV.

STEP 6 Capacity adjustments: Due to the seasonal pattern E.A. Broekema needs to handle peaks in demand and in capacity use. Therefore the capacity can be adjusted. If for the coming period more work than the 40.000 rods of assembly capacity is accepted the available capacity needs to be adjusted. Planning decides on capacity adjustment.

Possibilities for extending capacity on the medium term, e.g. a few weeks, are engaging extra (temporary) staff, working extra shifts, working overtime, outsourcing or relocating staff.

§3.3 The planning process

The orders accepted during the order process need to be dealt with in an efficient and ordered manner. This is the task of planning. Planning releases the work to the production departments and is responsible for achieving promised due dates. The steps in the planning process are shown next.

§3.3.1 The planning process stepwise

STEP 1 Release: After orders have been accepted and confirmed by sales they enter the order pool. For all the orders a production bill has been created with all the details such as due date and number of (external) operations that need to be performed. Planning releases on a daily basis and is considering only orders which are due within four weeks. For releasing the work they will bear in mind the capacity at the assembly department in total amount of rods, equaling 40.000 rods per week. If the assembly department, after releasing the orders of the coming four weeks, is not filled with 40.000 rods per week, orders which are due on a longer term are considered as well. The big orders from the subsidiary company are released separately and are earlier regarded for production than the normal orders. In the off-season E.A. Broekema tries to release the subsidiary-orders earlier to smoothen capacity requirements throughout the year. The rush orders are released right away.

The time after acceptance and before release is called the ‘pool stage’. On average an order for a complete conveyor belt spends 1,7 manufacturing day in the pool. This makes up 6% of total throughput time, see Chapter IV. In short most orders are released right away and are not hold back to reduce work in progress.

STEP 2 Capacity adjustments: If for the coming period more work is released than capacity is available planning can decide for the last short term, e.g. one week, capacity

(22)

adjustments. Options for adjustment on the short term are working overtime and relocating staff.

STEP 3 Priority dispatching: The foremen on the work floor determine urgency by sorting orders on date, number of operations to perform, external operations and by

‘experience’. If all components are available the foremen can decide to start production.

For efficiency purposes the foremen can combine some non-urgent orders in consultation with the planner to reduce set up times. Every Tuesday there is a meeting with the foremen and the planner. They regard the progress of all orders and determine which orders require priority.

(23)

Chapter IV: Performance prior to the implementation of WLC

In this chapter a more detailed analysis of the situation prior to the implementation of the WLC concept at E.A. Broekema will be described. The analysis mainly focuses on the delivery time because this is the main delivery performance aspect to be improved as discussed in section 1.5. First the overall throughput time will be discussed, including the pre-production stage. Second a general analysis of the throughput time using throughput diagrams will be given. Third the throughput diagrams for all the critical machine(group)s are described and analyzed. This is done in detail for one machine, the 100 Ton press. For the other machines, depicted in Appendix A, the analyzes is done more general. Furthermore the expected bottlenecks will be discussed.

§4.2 The overall throughput time

In the previous chapter some throughput time components and their contribution to the average total throughput time are already mentioned. Next a more detailed description of the average total throughput time and its components will follow. As discussed the enquiry stage and the pool stage in total make up 17% of average total throughput time.

The average total throughput time is 16,8 manufacturing days. The division between the several components is depicted in figure 4.1. The total average throughput time consists of three components; ‘enquiry stage’, ‘pool stage’ and ‘production stage’. The critical line operations performed in the production stage are shortening, pressing and assembling. Those operations are depicted as sub-stages. The critical line operations make up 62% of total throughput time.

Figure 4.1: Average total throughput time components and their size at E.A. Broekema.

The remainder of the average total throughput time is built up out of transit and waiting times or operations performed on other machines, such as bulging or operations performed in the special parts department or the covering of rods in the belts department.

Specific data on the waiting times and the other operations are not available.

The measurements involve only the orders for complete conveyor belts produced from December 2004 till April 2005. The processing times are the averages per order, measured in the stated period. The average total throughput time of almost three and a

(24)

half week is not corresponding to the objective formulated in section 1.5. Considering that 67% of orders are for double-belted conveyor belts, with a required maximum throughput time of two weeks, the average should at least be below three weeks. Another remark is that the period was not as busy as projected; this should have resulted in shorter average throughput times. The average total throughput time can be extended by customers requesting orders to be delivered in the future. This is hardly the case at E.A.

Broekema for the orders for complete conveyor belts and the effect it could have on the average total throughput time will be neglected.

The production throughput time is the main object of improvement in reducing the throughput time. However the pre-production throughput time at E.A. Broekema should not increase too much, not to undo the reductions gained in the production throughput time.

§4.3 Introduction throughput diagrams

Throughput diagrams are built up out of lines which represent cumulative input or output (in processing time) for each stage, see figure 4.2. Figure 4.2 distinguishes three stages: acceptation, release and completion. The date is put on the x-axis and the processing time is put on the y-axis. In this case the processing time is represented in riveting minutes, e.g. the amount of minutes contributed to the total in the assembly department. The curves (acceptation, release and completion) are cumulative and start on the 17^th of January. Every day a consecutive amount of processing time is accepted, released or completed it will be added to the cumulative total on that day. The stepwise pattern is the result of changing totals per day. Totals are changed at the beginning of each day and the day will end at the same cumulative level as it started at, representing a horizontal line for that day. The changes of the next day are again projected at the beginning of the day.

The positions where the cumulative curves are horizontally show periods where no work is accepted, released or completed. This could be a period of only one day, a weekend, a holiday period or a period where no work is done. Reasons for having no work done are machine breakdowns, machine changeovers or insufficient work available.

The horizontal distance between acceptation and completion gives and indication about the total throughput time at the regarded moment. As depicted in table 1.2 the average total throughput time is 20,4 days. The vertical distance between acceptation and completion gives an indication about the work in progress (WIP) at the regarded moment.

On average the WIP was 29.767 riveting minutes of processing time.

The linear black lines depict approximately the average processing speed on input and output in the period 17-1-2005 till 15-4-2005. This equals a three month period including the production weeks 3 till 15. The top line depicts the average processing speed on input; the bottom line depicts the average processing speed on output.

(25)

Figure 4.2: Basic throughput diagram for E.A. Broekema.

By studying the curves in more detail reveal that in most cases the accepted work is released almost immediately. Orders are on average released within 1,7 day after acceptation, see section 4.2. As an example the orders accepted on Thursday 3 February are released on Friday 4 February and the orders accepted on Friday 4 February are released on Monday 7 February. Another remark is that the horizontal difference between release and completion, e.g. the throughput time, is considerable long and varies widely.

In January the actual processing speed at acceptation and completion was rather slow compared to the average processing speed. In February the acceptation started to increase but the completion stayed on the same level or even slowed down a little. This has a lengthening effect on the throughput times at that moment. Only in the end completion started to pick up and acceptation slowed down slightly. The convergence of those lines in the end reduce the throughput time again.

Due to the irregularity in those curves the actual throughput time and actual work in progress varies significantly from the averages. The average throughput time is too high and there is too much variation in throughput times, considering the required delivery time. WLC should result in a lower average throughput times and show less variation in throughput times. This is accomplished because WLC signals an increase in input which requires production to respond by increasing output. If production is already producing maximally, WLC will prevent acceptation of an order that exceeds the maximum for that period.

(26)

Figure 4.3 will give a more detailed picture of the throughput diagrams at E.A.

Broekema. The critical machines are included and the enquiry stage is added.

Figure 4.3: Complex throughput diagram for E.A. Broekema.

The top black line is a manual estimation of the processing speed on the input side of the process. As can be seen input is slowing down in this period. The bottom black line is an estimation of the processing speed on the output side of the process. The output curve is sloping up. The black lines prove that the manufacturing system is only slowly reacting to changes on the input side of the process. Either there is no signal from enquiry that more work is arriving or production is not picking up the signal and increasing output directly. The backlog grows almost the whole period and it is hard to reduce this backlog.

A quicker reaction in production output could prevent the build up of backlog and keep the average throughput times shorter and more constant. This proves the planner’s statement that if it is now getting busy at shortening it will get busy two to three weeks later at riveting. The other way around is also true, if shortening workload reduces it takes two to three weeks for riveting to respond.

Furthermore it can be seen that a considerate amount of the throughput time is consumed after completion on the presses and before arrival on the riveting machines.

This could be explained because in most cases the rods do not leave directly from the rods department to the assembly department. Sometimes the rods are processed in other departments before entering the assembly department. Another explanation could be that the rods are waiting in the assembly department for other assembly parts to arrive. Only if the package is complete the order is typified as arrived for riveting.

(27)

§4.4 Detailed throughput diagrams

First the throughput diagram of the 100 Ton machine in the rods department is depicted. Second the throughput diagrams of the other critical machine(group)s are described. Schönsleben [2000, p. 497] distinguishes three operational states of a workstation qualifying the amount of work in progress in the system: underload, the transitional range and overload. Schönsleben is used to describe the machine(group)’s (main)state in the depicted period.

§ 4.4.1 Throughput diagram 100 Ton

I-O Diagram 100 Ton

0 5000 10000 15000 20000 25000 30000 35000 40000

17-01-05 27-01-05 06-02-05 16-02-05 26-02-05 08-03-05 18-03-05 28-03-05 07-04-05 Date

Cum. processing time (100 Ton min.)

Enquiry Acceptation Release

Completion (shortening) Arrival (100 Ton) Completion (100 Ton) Order completion

Figure 4.4: Throughput diagram 100 Ton.

The 100 Ton processes approximately 25.000 100 Ton minutes of output in the depicted period, see figure 4.4. In the beginning the 100 Ton can keep up with the arriving load. Later the mean work in process is in the transitional range with a small period of overload from half March. When the enquiry rate increases the production rate increase at the 100 Ton appears much later. This is extending the throughput time. If production is increased slightly right away the amount of extra backlog that will appear is limited. For comparison the internal data on utilization levels that E.A. Broekema records are used, see figure 4.5. The utilization level is based on an average amount of rods the machine normally can produce compared to the actual production. The utilization level is measured each week. The diagram shows the utilization level depicted on every Wednesday of each week. The non-production line gives an actual image of machine downtime due to change-over time, setup time and breakdowns. The utilization level shows if there is an increase in enquiry speed there is no slack to cope with that increase.

(28)

The 100 Ton is already working most of the time above maximum capacity. This means that it is possible to work faster and produce more than the average without employing extra capacity. For the planner this dilemma is always apparent.

100 Ton

0,0%

20,0%

40,0%

60,0%

80,0%

100,0%

120,0%

19-1-2005 2-2-2005 16-2-2005 2-3-2005 16-3-2005 30-3-2005 13-4-2005

Utilisation Non-production

Figure 4.5: Utilization curve 100 Ton.

§ 4.4.1 Throughput diagrams other machine(group)s

The other machine(group)s show similar patterns as described with the 100 Ton.

Especially the slow reaction pattern to increase production when input increases is similar. The diagrams of the machine(group)s are depicted in Appendix A. A summary of the characteristics is given in table 4.1. The most apparent characteristics are listed in table 4.1 because changes during the depicted period in operational state and utilization level appear.

Department Machine(group) Processing Operational state Utilization Rods Shortening 40.000 min. Trans./overload Moderate

Flattening-2 7.000 min. Underload Low

160 Ton 10.000 min. Overload High

Bulging 15.000 min. Trans./underload Moderate Assembly Riveting-2 70.000 min. Trans./overload Moderate

Riveting-3 20.000 min. Overload/trans. Low

Riveting-4 6.500 min. Underload Moderate

Table 4.1: Machine characteristics

The throughput diagrams of the machine(group)s in the rods department show that in most cases by using better signaling devices and quick responses the throughput time can be reduced. Problems realizing delivery dates could be expected especially at the 100 Ton and 160 Ton because of their state of overload and high utilization levels. In the future the 400 Ton will help to reduce the utilization levels on the 100 Ton and 160 Ton.

The throughput diagrams of the machine(group)s in the assembly department show also possibilities for reducing throughput times. The appropriate reaction is taken later than desired. Problems realizing delivery dates could be expected at the riveting-2 and riveting-3 machine(group)s because their state of overload and they appear to have difficulties coping with increasing enquiry speed. This is rather strange because utilization levels at those machine(group)s are not particular high. Reason for the current utilization levels at those machine(group)s is that the utilization level is an average of

(29)

several machines. If one machine is not in use this will lower the utilization level significantly.

It must be remarked that E.A. Broekema deals with a seasonal pattern in demand. This may cause varying bottleneck capacities throughout the season because of changing order types and customer demands.

§4.5 Conclusion

Throughput diagrams are a useful device to determine bottlenecks in production.

Operational states of underload, transitional range and overload are detected by analyzing the throughput diagrams. At E.A. Broekema a detailed analysis of the machine(group)s is made. The 100 Ton, 160 Ton, riveting-2 and riveting-3 are primarily detected as constraining machine(group)s in the measured period. The utilization levels of the 100 Ton and 160 Ton are most of the time close to the maximum. In the future the 400 Ton can be used to reduce the utilization levels of the 100 Ton and 160 Ton. Bottleneck variations may appear due to the seasonal pattern of demand.

(30)

Chapter V: Assessment of the proposed WLC system

In the past a choice has been made to implement WLC for improving the delivery performances, mainly focusing on delivery time. The WLC ideas are best symbolized by the statement below:

“The key concept of WLC is to control queues at an acceptable level so that lead times are controlled and delivery dates can be met, whilst good use is made of the available capacity” [Kingsman and Hendry, 2002].

The delivery performances (delivery time, delivery reliability and the delivery flexibility) are influenced by using the WLC concept. Before rushing into implementation first the applicability of the WLC must be tested.

§5.2 Applicability

Henrich et al. [2004] have developed an evaluation framework for the applicability of WLC in small to medium-sized make-to-order (MTO) companies. The framework is a combination of characteristics described in literature on WLC. The grey areas mark the scores which apply the best with WLC according to literature. In the paper of Henrich et al. [2004] E.A. Broekema has been analyzed by using the framework. The results are depicted in table 5.1. Only the extreme values are marked. When the company is scoring on average on a certain indicator the indicator is not marked. It is expected that an average score will not influence the applicability negatively.

Characteristics Indicators Low High

Arrivals Arrival intensity X Inter-arrival time variability Due dates Due date tightness

Variability of due date allowance X Operations Processing time lumpiness

Processing time variability

Set up/processing time ratio X Routings Routing sequence variability

Routing length Routing length variability

Routing flexibility X Level of convergence X

‘Best fit’ for applying WLC

X Extreme values for E.A. Broekema

Table 5.1: Evaluation framework for E.A. Broekema [Henrich et al., 2004, p. 21].

(31)

Most of the extreme values apply with the desired indicators for implementing WLC.

Only the set up/processing time ratio, due to some of the high situation specific setup times, and the level of convergence, due to the assembly structure, might cause problems.

E.A. Broekema already started a program to reduce set up times and needs to continue the program to reduce set up times even more. The level of convergence is reduced by regarding the rods department and the assembly department as constraining departments.

The benefit of this system restriction is that those departments are sequential, see section 2.1, and minimize the level of convergence. WLC ideas could be implemented and contribute to a solution for E.A. Broekema in reducing throughput times. Next a discussion of the WLC concept and its characteristics is described.

§5.3 The WLC concept

The WLC concept is based on the principles of input/output control as defined by Plossl and Wight [1973]. In a job shop in front of each station, an arriving job finds a queue of jobs waiting to be processed. WLC tries to control the length of these queues by making input/output decisions. The input and output decisions can be made on three hierarchical levels but the main decision to control the shop queues is the release function. The first level is the job entry level, the second level is the job release level and the third level is the priority dispatching level, see figure 5.1. The means for controlling input and output are depicted in figure 5.1. Input control decisions regulate the allowance of jobs to the next level. Output control decisions adjust capacity in order to change the production volume to process the allowed jobs. Both the input and output control are captured in an information system.

“An important and obvious quality of the WLC concept is that it buffers the shop floor against the dynamics of the order flow” [Breithaupt et al., 2002, p. 630].

Figure 5.1: The hierarchical WLC concept [Land & Gaalman, 1996].

The input/output control is related to workload norms, e.g. the maximum amount of workload admitted for the shop floor. For E.A. Broekema these norms are determined for every critical capacity group. On the short term one of these critical capacity groups will

(32)

be the delivery date determining. The capacity of this group will be filled most of the time. Work can only be accepted on basis of the limited available capacity on that machine. The WLC concept contributes in identifying this capacity group. Next the levels are discussed in more detail and consequences for E.A. Broekema are indicated.

§5.3.1 Job entry level

Figure 5.1 shows the three decisions for control at the entry level: acceptance, medium term capacity adjustment, due date assignment. Although the three decisions for control are discussed separately interdependencies are evident.

§5.3.1a Acceptance

The acceptance is related to the method presented in Bechte [1994]. Sales and planning have a task in accepting work. Sales will be dealing with simple orders in the future, while planning can focus on the current and more complex orders. This division leaves more time for planning to identify accurate were constraints appear. First the roles of sales and planning are discussed followed by the procedure for calculation delivery dates.

The simple orders are taken care of by sales. Sales is completely responsible for order acquisition. Sales can only accept maximally 60% of the riveting capacity without planning to interfere. Orders that exceed the 60% limit or the more complex orders always need to pass planning. Until 60% of riveting capacity is filled no constrains in the rods department will appear. In the future the 60% limit needs to be reviewed for up or down scaling.

The planning department checks the technical feasibility of the accepted orders and can fill capacity up to 90% leaving 10% of capacity free for rush orders. Rush orders are sent directly to planning. Planning determines whether the rush order will be accepted. Both sales and planning department need to work with the same capacity overview, see figure 5.2 for a general example. This overview is generated for each critical capacity group.

The percentages show the maximum amount of processing time sales and planning can accept and their responsibilities areas. The curve depicts the amount of work already accepted either by sales or by planning.

Figure 5.2: Capacity overview for one capacity group and the acceptance of a new order.

(33)

Next the procedure for calculating delivery dates is described. For a new order a ‘free date’ is determined. This is the first date the critical capacity groups have sufficient capacity to produce the order. As can be seen the determined free date by sales is different from the free date of planning (providing no other capacity group becomes constraining), see figure 5.2. If a customer wants a quicker delivery than sales can promise the order should pass planning. Not only the capacity availability of the critical capacity groups is determining the free date. The material availability, the duration of the vulcanization trajectory, the duration of the rod covering operation and the fitting of special parts can determine the free date. This last group will be referred to as ‘limitation- group’.

The above mentioned elements will result in a latest free date on one of these elements.

Adding the remaining norm throughput time of the following operations to this free date results in a realistic estimated delivery date. The latest free date of all groups can be depicted as in figure 5.3.

Figure 5.3: Delivery date using free date, routing and throughput time [Bechte, 1994].

Workcenter C has the delivery date determining free date. To this free date the norm throughput time on the workcenters D en E must be added to result in a realistic delivery date. Orders can be accepted if the requested delivery date exceeds the estimated delivery date. For determining the free date cumulative capacity overviews must be generated, see figure 5.2, and the processing time contribution of an order needs to be calculated. The calculation technique for the processing time contribution of an order is discussed in Chapter VI.

One difficulty in the acceptance procedure is the fact that until the order is accepted and confirmed by the customer the workload is not added to the system. When too many orders are in this customer enquiry stage there is no accurate overview of total assigned capacity, because no capacity is assigned to the order until it is accepted. Some projects use strike rates [Kingsman et al., 1989] to determine the amount of processing time in the

(34)

enquiry stage that will eventually enter the pool. At E.A. Broekema this processing time will be neglected because on average it only spends 1,8 day in the enquiry stage.

Customers need to know that a quick confirmation is expected.

The interdependencies between sales and planning shift from sequential to reciprocal [Thompson, 1967]. Both sales and planning are now dependent on each others work. In order to ensure quality and effectiveness they should coordinate through mutual adjustment. The information system will enable the increased information exchange due to the increased information need.

§5.3.1b Medium term capacity adjustment

Capacity will be entered on a week level for the coming 40 days. When the cumulative processing time (minutes) of accepted work will reach capacity limits repeatedly, medium to long term capacity adjustments need to be regarded. Options for capacity extensions on the long term, e.g. several months to years, are acquiring new machinery or engaging extra personnel. Possibilities for extending capacity on the medium term, e.g. a few weeks, are engaging extra (temporary) staff, working extra shifts, working overtime, outsourcing or relocating staff.

§5.3.1c Due date setting

When all the constraints, e.g. free date of capacities and limitation-group, are known the information system calculates an earliest delivery date. The information system uses the routing information and norm throughput time on each operation the order is visiting for calculating the earliest delivery date. The norm throughput time is set to realize the required throughput time for each type of belt, see section 1.5. The routing will be manually imported by the planner, because no data on the exact routing information is available at E.A. Broekema. This will be handled in section 6.3.

The norm throughput time is set for every operation and shown in table 5.2, notice that not every operation is mentioned separately. Also the availability of the limitation-groups can be set.

Operation Department Norm throughput time (days) Material availability Procurement Set by procurer (limitation-group)

Shortening Rods 1

Bulging Rods 1

Pressing Rods 1

Tempering Rods 2

Vulcanization Externally Set by planner (limitation-group) Various Special parts Set by planner (limitation-group) Belt cutting/punching Belts 1 day together

Rivet inserting Belts 1

Rod covering Belts Set by planner (limitation-group)

Various Assembly 1

Assembling Assembly 1

Table 5.2: Norm throughput times in the information software and limitation groups.

medium sized Make-To-Order company