From Warehouse to Machine
Analysis of the performance of the internal material delivery
process at SCA/UcM Hoogezand
Author: Jinne Osinga
University of Groningen
Faculty of Business & Economics
MSc. Business Administrations, Operations & Supply Chains
September 2010
Van Heemskerckstraat 3-5
9726 GB Groningen
j.osinga.1@student.rug.nl
Student number: 1662686
Organization: SCA/UcM Hoogezand
Company supervisor: Siebe de Meer
RUG supervisor: Drs. G.D. Soepenberg
I Preface
This thesis is the final part of my time as a student of the Master of Science in Business and administration, following the specialization Operations and Supply Chains.
Finishing this master is the end of a long period of education. From this moment the period of a comfortable life as a student comes to an end, and the life as a working man will start. The period at the University of Groningen has been a period of studying hard and cycling. Although a cycling training always has been a good alternative during the study, I managed to finish my master.
Performing the research at SCA/UcM was a very interesting experience. It was nice and educational to perform a research where a new technique like the AGVs was involved. Before I came here, I had never heard of AGVs and didn’t know the
possibilities. It was very nice to see how the AGVs performed their job and what comes across during their work. Besides that is was very educational that I got the possibility to think along with the project members of Top Gear. Data I had gathered was often used by the project members. Furthermore I really liked the informal culture within the company. All employees where really friendly and were always willing to help me. I would like to thank my supervisors of the University of Groningen, Erik Soepenberg and Gera Welker for their advices and critical and useful feedback. Besides that I would like to thank my supervisor at SCA/UcM, Siebe de Meer, for the good conversations and discussions, information and help during the research. He gave me the opportunity to play an active role in the improvement of the AGV performance.
In addition, great thanks go out to Hans Top, Izebrand Dijkema, and Vanessa
Westerbaan, my ‘roommates’ at SCA. The atmosphere in the office was great; moments of hard working were followed by moments of relaxing, jokes, and laughter. Besides that, the conversations we had led to very useful information for the research. Finally I would like to thank my parents. Although I graduate a bit later than they desired, they always kept supporting me.
Jinne Osinga,
II Management summary
This research aims at analyzing the internal delivery process of SCA/UcM in Hoogezand and to discover the factors that influence the throughput time of the material orders. The main research question this research was based on is as follows:
Which factors cause that the current norm for the internal throughput time of 85 minutes is not achieved, and which recommendations can be given to improve the internal material throughput time?
The research was performed for all three different parts of the internal delivery process: the Warehouse Rearrange-assignments (WR), the Apollo-assignments (AP), and the AGV-assignments (AG). The main goal of the research was to first split the total throughput time into handling, transport, and queuing time and discover which factors influenced these different parts of the throughput time.
The analysis of the performance of the different types of assignments showed that from every part of the process a significant part of the throughput time was queuing time. On average it took 39 minutes and 20 seconds to process a WR-assignment of which 96.4% of the time was queuing time (37 minutes and 51 seconds). The handling time was 26 seconds and the transport time was 1 minute and 3 seconds. For the AP-assignments the average throughput time was 17 minutes and 4 seconds, while the conveyor
normally needs 7 minutes to transport the materials. The queuing time for every order is 10 minutes and 4 seconds. The analysis of the AG-assignments showed an AGV needs 16 minutes and 22 seconds in total to drop a material and return to the conveyor; the total part of queuing time is 19.8% (3 minutes and 14 seconds).
The conclusion of this research is that there are three main causes for the amount of queuing time during the internal delivery process. The capacity of the AGVs is not sufficient at the busy moments during a day. At this moment an AGV can transport 18.3 orders every hour, where an average of 14.2 orders is needed. The variation in the number of orders per hour causes problems, because the throughput times of the AGVs block the flow at the conveyor. This will get worse in the desired state of the AGVs where 800 more orders will be delivered every week. This means an average of 19.1 orders per hour is needed, which is not possible with the current fleet of AGVs. The capacity of an AGV decreases because of breakdowns and the queuing time these breakdowns have as a consequence. Besides that the routing causes queuing time during the transport. Furthermore the lack of capacity of the AGVs increases the throughput times of the AP-assignments. At the moment the conveyor is fully loaded, they can only be transported further when an AGV picks a material from the conveyor. Materials are stuck on a certain position at the conveyor as long as the AGV doesn’t pick any
material. A consequence of this is that conveyor loaders cannot load anything on the conveyor before the materials are moving again. This leads to longer waiting and throughput times for the WR-assignments.
moments, when the conveyor is fully loaded, the time an AGV needs to pick a material from the conveyor decides how long a conveyor loader has to wait before the next material can be loaded. The processing speed of the AGVs cause that the materials on the conveyor cannot flow further and the conveyor loaders have to wait before a new material can be loaded. This means WR-assignments are open for a longer time, which causes longer throughput times.
The length of the breaks of the conveyor loaders also creates an irregular flow for the conveyor and AGVs. All the material orders which arrive during the break of the conveyor have to wait longer before they will be processed. This increases the waiting and throughput time of the WR-assignments. Besides that the pressure on the conveyor and AGVs varies too much. Materials will be longer on the conveyor because the AGV cannot handle the flow quick enough, which causes an increasing throughput time for the AG-assignments.
Finally, several recommendations came up after the research. The recommendations are:
•
The way of ordering materials should change to avoid peaks in the number of materials orders. Not only to improve the current situation, but also because the current way of ordering will lead to more problems when the AGVs will supply all machines. Every material than has a fixed drop location and buffering is not possible any longer. To improve the way of ordering, first the machine coaches should have to become aware of the consequences of peaks in the number of material orders. After that, these coaches will be responsible for instructing the machine operators and the implementation of the desired way of ordering.•
The conveyor loaders should receive a clear instruction about the breaks they take. There should always be one conveyor loader at work in the warehouse to ensure there is a steady and continuous flow of materials on the conveyor. This will result in shorter throughput times for the WR- and AP-assignments. An informative session with the conveyor loaders is needed to make them aware of the importance of a continuous flow of materials on the conveyor.• The capacity should be expanded to deliver all materials on time in the desired state of the system. This can be realized in three ways:
o Using an extra AGV, which requires an investment of € 130, 000,-. o Changing the routing and create waiting positions for the AGV, or separate the current driving route at Panty 6 into a path for loaded AGVs which have to perform the drop and empty AGVs which are returning to the conveyor. Furthermore the advantages of the use of the filter room at the Heracles hall as a return route, or creating waiting positions besides the main route at Astra Middle, Panty 6, and Heracles should be researched.
III Glossary
Term Explanation
Abramskade 4 The address of the central warehouse for the raw materials.
Abramskade 6 The address of the plant on which the manufacturing machines are
located.
AGV Automated Guided Vehicle, a vehicle which receives orders from the warehouse management system. It navigates itself to the machine by constantly checking its position at reflectors on the wall.
Conveyor The transport bridge which connects Abramskade 4 and 6. On this
bridge the materials are transported from warehouse to factory.
RMW Raw Material Warehouse, the warehouse where all raw materials are stored.
WR-assignment Warehouse Rearrange-assignment, the material order from the
machine which is processed by the fork truck drivers at the warehouse; they load the orders on the conveyor.
AP-assignment Apollo-assignment, the material order which is transported on the
conveyor. Apollo is the supplier of the conveyor.
AG-assignment AGV-assignment, the material order which is transported from the
conveyor to the machine by an AGV.
M-assignment Machine-assignment, the material order which is transported from
the conveyor to the machine by a fork truck.
UcM Unicharm Mölnlycke, the company which is a result of the joint-venture with the Japanese company Unicharm.
Heracles A type of machine which is used to manufacture products for the
unit Tena Lady. The materials for the three Heracles machines are supplied by an AGV.
Astra A type of machine which is used to manufacture products for the
unit Tena Lady. The materials for the seven Astra machines are supplied by an AGV.
Shuttle The truck which is used to transport the materials from warehouse
to factory over the road. This shuttle is only used in case of a long breakdown of the conveyor or extremely busy periods.
MapXR The warehouse management systems which is used by the machine
operators to order the materials. Besides that the conveyor loaders process the orders with the help of the system and the AGVs receive their orders from the warehouse management system.
Drop location The location where the AGV drops the material near the machine.
IV Table of contents
I Preface
2
II Management summary
3
III Glossary
5
IV Table of Contents
6
1.
Company
description
7
1.1 The history of SCA/UcM... 7
1.2 Top Gear... 9
2. Problem description
10
2.1 Motive and project information... 10
2.2 Organization problem... 11
2.3 Research design... 11
2.4 Conceptual model... 14
3.
The
Delivery
process
18
4. The current performance
21
4.1 Data analysis...214.2 Influencing factors... 29
5.
Recommendations
and
reflection
40
5.1 Recommendations... 40
5.2 Reflection... 42
References
43
Appendices
1. Organizational chart SCA/UcM 2. Organizational chart unit Logistics 3. Results observation conveyor loaders 4. Throughput time per assignment 5. Drop times per drop location
6. Observation form for throughput time 7. Observation form for one AGV
8. Overview of AGV cycle times
9. Interarrival time of WR-assignments.
1. Company Description
In this chapter the history of SCA/UcM is described. Furthermore information about the
machines and the manufacturing quantity is described. At last a description of the Top Gear and its goals is given.
1.1 The history of SCA/UcM
In 1849 Mölnlycke Väfverie AB, was founded in Råda, near Göteborg in Sweden. The German G.F. Hennig opened a small factory with a weaving machine and a spinning mill. After some time the railway which passed the factory was called Mölnlycke and very soon the whole village built around the factory was called Mölnlycke.
The plant in Hoogezand was part of Mölnlycke until 1975, when SCA merged with the company. SCA is a part of the Swedish SCA Group (Svenska Cellulosa Aktiebolaget). In Europe this is one of the market leaders in manufacturing hygiene products like diapers, tissues, and incontinence products. In 1996 the SCA Group decided to add the name SCA to the name of all the factories for commercial reasons, to show that they all were a part of the SCA Group. For the plant in Hoogezand this meant the name was changed to SCA Mölnlycke.
In 1998 SCA Mölnlycke merged with SCA Hygiene Paper and became SCA Hygiene Products. In 2003 the unit Hygiene Products was dissolved and the Unit Personal Care was established. At this moment SCA/UcM is still part of that business unit. Within SCA, besides the business unit Personal Care, a Global Hygiene, Forest Products, and Packaging unit exists. Worldwide about 50,000 people are working for SCA in 60 different countries (January 2009).
SCA/UcM Hoogezand
In 1963 the factory in Hoogezand was opened, the first factory for Mölnlycke in the Netherlands. From the opening of the factory until the takeover of SCA, the factory was producing female care products (sanitary napkin) and medical products (medical mats). From 1975 the production for medical products was stopped and the factory started with the production of diapers and incontinence products.
In the early nineties the name of SCA was changed to SCA/UcM. The reason for this was a joint venture with the Japanese company Unicharm. UcM stands for Unicharm Mölnlycke. UcM delivers pulp and other raw materials for production. By collaborating with UcM, SCA was ensured of receiving the materials they need for production against a good price.
Machines
At the plant in Hoogezand 23 machines are manufacturing products for the different units. Each machine is attached to one of the units. Table 1.1.1 will make clear which machine (name and number) is manufacturing for which unit.
UcM Baby UcM Inco Tena Lady
Machine (number) Machine (number) Machine (number)
UcM-3 (52) Panty 1 (71) Astra 2 (62)
UcM-4 (53) Panty 2 (72) Astra 3 (64)
UcM-5 (54) Panty 3 (73) Astra 4 (65)
UcM-6/Pegasus 1 (55) Panty 4 (74) Astra 5 (66)
UcM-7/Pegasus 2 (56) Panty 5 (75) Astra 6 (67)
UcM-8/Pegasus 3 (57) Panty 6 (76) Astra 7 (68)
LIPS-1 (81) Astra 8 (69)
Heracles 9 (70)
Heracles 10 (60)
Heracles 11 (61)
Table 1.1.1: Machines per unit
The machines mentioned above all have their own working schedule. This schedule varies between a production week of 72 hours (in two teams) till1 168 hours (in five teams). This means a part of the machines is manufacturing continuously, while others are standing still at fixed (parts of) days. As is visible in table 1.1.2, the machines together produce almost 2.7 billion pieces in a year (2009). The forecast for the total production of 2010 is almost 2.65 billion pieces.
2009 2010
Unit Production Sales Forecasted
production Forecasted Sales
Tena Lady 1,694 1,700 1,715 1,723
UcM Inco 405 439 430 441
UcM Baby 575 575 504 533
Total 2,674 2,714 2,649 2,697
Table 1.1.2: (Forecasted) production and sales for 2009 and 2010 per production unit (in million pieces)
According to Slack, Chambers & Johnson (2007), the production process at SCA/UcM can be described as a mass production process. At SCA/UcM high quantities of
products are fabricated in batches and the variety of products is small. In the basis the products are almost similar, only the composition or size varies. The differences in composition can be explained in the process industries. According to Fransoo and Rutten (1994) companies in this industry add value to materials by mixing, separating, forming or chemical reactions.
At this moment about 650 people are working in Hoogezand, from which the majority is working in production. The products manufactured in Hoogezand are sold and
transported to 90 different countries worldwide. An organizational chart from both the organization and the logistics department is added in the appendix (1 and 2).
1.2 Top Gear
In 2006 the headoffice of SCA Personal Care in Munich came with the announcement that the costs for SCA were rising too much. After that, the plan was made to decrease the costs organization wide. Lower costs lead to a lower cost price, which improves the competitive position of the company and gives SCA/UcM more space to lower the prices if necessary. In Hoogezand the announcement to lower the costs resulted in the project Top Gear, which started in May 2008. The goal of the Top Gear project is as follows:
Design and implement an optimal logistical system for treating the raw materials, from the suppliers’ delivery to the raw material warehouse, to the delivery of the materials at the manufacturing department.
Because of the complexity of the project, the decision was made to split the project into three different attention areas, where the main focus is on cost reduction and clear processes:
• Optimization of the Raw material warehouse (supplier to storage) • Optimization of the customization (strip & turn)
• Optimization internal transport (between RMW and production v.v.)
Within these three different areas several projects are executed with the objective to lower the costs. In the raw material warehouse a project is executed to lower the net working capital by lowering the general stock levels. The customization is optimized by building the strip & turn area in the warehouse, to make it possible to repack the
received units in smaller quantities immediately after receiving it from their suppliers. In the old situation this took place at the moment the materials were ordered by the operators at the machine, which caused longer delivery times. The repacking to smaller quantities is necessary to avoid large stocks of a certain material at the machines and returns of large amounts from the machine back to the warehouse. Besides that, repacking is necessary in some cases because the pallets where the products are loaded on are too big to handle for the AGV.
2. The research
In this chapter the organizational problem is described, together with the general information about the research. Furthermore the research question and sub-questions are described and the conceptual model is given. In the last part of this chapter the theoretical framework of this research is given, where the different parts of the conceptual model are described.
2.1 Motive and project information
In May 2008 the Top Gear project was started at SCA/UcM in Hoogezand. The main reason to start this project was a signal from the headquarters of SCA Personal Care in Munich. The costs were too high and reduction was necessary. As mentioned before, a part of this project had to lead to a more efficient way of delivering the raw materials from the warehouse to the production lines, with lower costs.
The plan for the project was to introduce an automatic transportation system for the transport of the raw materials from the different warehouses (one for every production unit) to the machines. This plan had to be changed when it turned out that one of the material warehouses was not fire-resistant anymore. This was the reason to rent a warehouse which was big enough to centralize the stocks for all different production units. First all materials were transported by trucks, which were called shuttles. These trucks were driving between the warehouse (Abramskade 4) and the plant (Abramskade 6). From March 2009 a transport bridge (conveyor) was used to transport the materials from the raw material warehouse to the plant. From the end of the conveyor the
materials are delivered to the machines by fork trucks.
The last step within this process, the delivery to the machines, is automated at SCA/UcM. This was not only a case of changing the transport vehicle, but also the transport routes and the drop locations at the machines had to be redesigned. The decision was made to introduce this at 11 of the total of 23 production lines within the factory. All machines of the unit Tena Lady (Astra and Heracles) are involved in the project and for the Unit UcM Inco only one machine (Panty 6) will be delivered by an AGV. The decision to implement the AGVs at these machines first is based on the difficulty of the implementation. The manufacturing hall where the Astra machines are positioned, is the hall with the smallest ‘roads’ and very little space to manoeuvre for the AGVs. It was assumed that if the implementation would succeed on these machines, it would be possible for all machines at SCA/UcM. Both the Panty 6 and the Heracles machines are situated at the return route for the AGVs to the conveyor, which is the reason they also are involved in the project.
As mentioned before, the fork trucks were replaced by AGVs (Automatic Guided Vehicle). According to the literature, an AGV is a vehicle which is driven by an automatic control system that serves the role of the driver (Ioannou, Jula and Dougherty, 2001). It is a full automatic system, where no human interventions are needed, only when there are disturbances. The AGVs receive their orders by the
The main goal of the Top Gear project is to reduce the costs within the company. The biggest saving is realized by the replacement of the fork truck drivers by five AGVs. The only restriction for the final replacement of the employees is that the performance of the AGVs is sufficient. In this case it means the delivery of 95.0% of the materials should take place within 85 minutes after the order is registered. This time of 85 minutes is based on the norm-times of the different sub-processes:
• From the ordering of the material to loading the conveyor: 45 minutes; • The time the material is on the conveyor: 20 minutes;
• From the conveyor to the delivery by AGV: 20 minutes. 2.2 Organizational problem
As mentioned before, the intention is to introduce the automated transport and delivery of the raw materials also at the other production lines (phase two). But before they will make a start with the implementation of the second phase, the management wants a stable and reliable process for the delivery of the machines in phase one. After a few months, this still isn’t the case. There is a big difference between the performance of the process and the desired performance according to the stated norms.
The performance of the delivery process in a certain number of weeks needs a closer look. Hereby the focus will be on the on-time delivery performance of the total internal delivery process of the unit Tena Lady. This is the unit where the AGVs are responsible for the final part of the process: the delivery at the machine. This research should give more information about the performance of the AGVs and the exact reason why the performances are below the stated norms. Researching this will deliver information which can be used to finetune the current processes and it will lead to recommendations to improve the results for the different parts of the delivery process.
The final product of this research will be a report about the performance of the delivery process where the AGVs are involved. Taking a closer look at the process and the different parts of the process will deliver knowledge about the actual performance of the delivery process. By researching and analyzing the data of the three sub processes, it is possible to clarify the performance and recommend improvements within these
processes.
2.3 Research design
Within this research I will focus on the causes for the current delivery speed at the different machines within the unit Tena Lady. From the moment the AGVs are introduced, the performance has never been on the desired level of 95.0% of the deliveries within 85 minutes. The number of 85 minutes is the sum of the norm-times for the three different assignment types:
• WR-assignment: within 45 minutes; • AP-assignment: within 20 minutes; • AG-assignment: within 20 minutes.
AP- and AG-assignments are based on simulations which were executed before the implementation of the AGVs. During this simulation it turned out that the time from the start of the conveyor to one of the two pick-up locations for the AGV is seven minutes. For the AG-assignments another simulation was executed; it turned out that on average the AGV needs eight minutes from unloading the conveyor at the pick-up position to the drop-zone at the machine. By taking into account queuing time and/or disturbances for both AP- and AG-assignments, the norm of 20 minutes is determined. Besides the WR-, AP-, and AG-assignments, M-assignments are visible in the system. For all machines which aren’t supplied by AGVs, the fork truck is responsible for the delivery of materials. For all orders which are supplied by a fork truck, an
M-assignment is created in MAPxr. The objective of this research is:
Analyze the causes for the current internal material delivery
performance and recommend improvements within the process.
To realize the objective of this research, I will answer the following research questions:
Which factors cause that the current norm for the internal
throughput time of 85 minutes is not achieved, and which
recommendations can be given to improve the internal material
throughput time?
To conduct this research and answer the main research questions, the following sub questions are stated:
• What is the delivery performance of the different parts of the material delivery process?
• Which factors influence the delivery performance of the different parts of the process?
• How can SCA/UcM improve the delivery performance of the process of material delivery?
Part/Research question Gathering method Gathering location
What is the delivery performance of the different parts of the material delivery process?
Data analysis SCA-systems and
employees Which factors influence the
delivery performance of the different parts of the process?
Literature research, Data
analysis, observations Books, articles from library (RUG). How can SCA/UcM improve the
delivery performance of the process of material delivery?
Literature, Data analysis,
observations Books, articles from library (RUG).
Table 2.5.1: How and where to gather data?
The data which is needed to answer the first question is available in MAPxr, the system which is used within SCA/UcM for the material management. For every time-period (month, week, day, shift) the data about internal material orders can be requested for every machine and every unit. The logistics department designed a special Excel sheet in which it is possible to divide the number of orders into the three earlier mentioned types of assignments:
• WR-assignments; • AP-assignments; • AG-assignments.
For every order it is visible how long every part of the process took and which
percentage of the total number of orders is performed within the stated norms. Based on these data it is possible to find out which part(s) of the process has problems with
achieving the goals.
2.4 Conceptual model
Figure 2.4.1: Conceptual model
Throughput time
The focus in this research is on the throughput time of the internal delivery of the materials from warehouse to the machines by an AGV. According to Hopp and Spearman (2000) the throughput time can be described as:
“The time from the moment of ordering of a job at the beginning of the routing, until it reaches an inventory point at the end of the routing”
important factor. In these articles the focus is on manufacturing strategies. Translated to the internal transport at SCA/UcM, the processing time can be seen as the handling time of an order. Within the delivery process there is no set-up time, which is the reason that set-up time has no influence on the delivery performance.
To measure the throughput time, data from Map XR is available. Furthermore a special Excel sheet is designed to specify the delivery time of the different types of assignments within the delivery process. For every ordered material and for every delivery, data is available about the time every different assignment took. The number of orders delivered late can be calculated for both the total delivery process and for the different types of assignments within the process.
The total throughput time is influenced by three different types of time. I will describe these different types.
Queuing time
According to Riezebos, van Donk, and Bokhorst (2008) queuing time can be described as
“The time between the moment of arrival and the moment of starting the service; the time spent in the queue”.
Riezebos et al. (2008) describe four different types of queues, but only one of these types occurs during the delivery process; this is congestion. Congestion means the throughput time of the process is temporarily higher then the intensity in which the flow units are arriving at the server.
The total queuing time can be influenced by several factors. Riezebos et al. (2008) describe five different factors which influence the total queuing time of an
order/customer, including:
• Interarrival time: the interarrival time can be described as the time that elapses between the arrivals of two different flow units (orders/customers). A situation where the interarrival time is constant shows another order arrival than an arrival where the interarrival times strongly differs.
• Processing time: the time that is needed to process an order. At SCA/UcM this can be the processing of new material orders (WR-assignments), materials which are processed by the conveyor (AP-assignments) or materials which are processed by the AGVs (AG-assignments). Based on the processing time the capacity per hour can be calculated: 1 hour/processing time (= number of units per hour).
• Number of parallel servers: the number of servers helping processing the flow unit at the same time. Within SCA the WR-assignments are processed by two conveyor loaders on fork trucks, there is one conveyor, and there are five AGVs to handle orders at the same time.
has to wait until a few small orders are finished. There are four different methods to handle the incoming orders:
o FIFO: The order of arrival is the order of processing;
o Arbitrary: An order is picked accidentally from the open orders; o LIFO: The last arrived orders are processed first;
o SPT: The customer/order with the shortest processing time is processed first.
• Waiting room: Here two things play an important role:
o The capacity of the waiting room: how many flow units can be in the queue? Is this a finite or infinite amount (Adan & Resing, 2002)? o The question whether there is one waiting room for all servers, or one
waiting room for every server.
Besides the factors mentioned above, Slack et al (2007) mention the utilization of the servers. This is the percentage of the time the server is actually in use. This can be calculated by processed orders / (parallel servers*processing rate).
Houtzeel (1982) describes queuing time as the time which is usually the largest of the four components (set-up time, queuing time, move time, processing time). In some systems, the queuing time is 90% of the total throughput time of a process. Bozer & Hsieh (2004) mention the total queuing time is a good indicator in deciding whether the process’ performance is acceptable.
At SCA, queuing time can occur in each of the different parts of the process. First an order has to wait before it is treated by one of the fork truck drivers. After that the material possibly has to wait, while standing on the conveyor or when one of the AGVs has to pick the material from it. While the material is on the AGV, queuing time can occur when it has to wait for one of the other AGVs to proceed with the execution of the order.
Transport time
Johnson (2003) and Plossl (1988) describe the transport time as below:
“The total transport time within the process is the sum of times spent moving a material between different workstations in the routing”.
Within the three different types of assignments of the internal delivery process, transport time occurs. During the handling of the WR-assignment the transport time exists of the time needed to drive with an unloaded fork truck to the ordered material and back to the start of the conveyor with the material loaded at the forks. From the moment the material is loaded on the conveyor the transport time for the
The total transport time depends on two different factors. The distance is the first factor; the further the pick or drop of a material is away, the longer it takes to transport to the correct location. The second factor is the speed; the faster the transportation vehicle is, the faster the pick or drop of a material can be performed.
Handling time
In the literature the handling time is described by Johnson (2003) as:
“The sum of the times spent processing a part at each workstation required in the routing”. Within the total internal delivery process at SCA/UcM, there is only one type of assignment where handling time occurs during the processing of the order: the WR-assignments. The handling time here exist of opening the order in MapXR, scanning the materials and loading the material on the forks of the fork truck.
3. The delivery process
In this chapter the internal delivery process is described in detail. For every step during the process is described which activities should be performed, and who or what should perform these activities.
With the introduction of the AGVs and the conveyor, the process of internal material delivery was changing. The conveyor and the new, central warehouse both had a big influence on the design of the new process. Figure 3.1.1 shortly describes the current delivery process.
Figure 3.1.1: The internal material delivery process 3. Transport on conveyor
2.2 Picking material and loading the conveyor 2.1 Open orders in MAPxr 4.1 AGV delivery 4.2 Fork truck delivery 1. Material order at machine 1. Material ordering
The machine operators order the materials they need in the material management system (MAPxr). A machine operator orders a material at the moment the drop location is empty or almost empty. In the last situation the operator has to be sure the drop location is empty at the moment the AGV arrives with the new material. Only a small car or stand has to be present to drop the materials and the route should be free of obstacles.
An order exists of three types of information: a material number, the ordering machine (= delivery machine) and the time the order is entered. The machine operators have the opportunity to order several different materials at once. The ordered quantity for the majority of the materials for Tena Lady is one unit; only for the frequently used
is the limited space at the drop location, where in almost all cases only space exists for one material unit, except for the frequently used materials. At some machines for the material pulp there is a double drop location. The machine operators at the units UcM Baby and UcM Inco have the possibility to order an unlimited amount of a material. During the registration of an order in MAPxr, the operators can add a remark. This can be the type of material or that the order should be processed as fast as possible.
2. Order picking and loading
All materials ordered by the different machines appear on the screen in the fork trucks as WR-assignments in MAPxr (2.1). This means not only the orders for the AGVs, but also the orders for the fork trucks will appear. The fork truck drivers only see of which machine the operator(s) ordered, how many units, which time, and if the machine
operator added a remark (which is also visible). For the material pulp the addition ‘pulp!’ is visible, this is automatically added by the system because it is the main material from the production process.
The system displays the orders in a chronological sequence at the fork trucks, with the oldest orders at the top and the newest orders below. The orders are treated according to the FIFO principle; the oldest orders have to be processed first. By opening an order, the materialnumber of the ordered material becomes visible. The fork truck driver drives to the location where the material is stored. The materials’ SSCC number is scanned and the system identifies the material-unit and checks whether it is the correct material. If it is the correct material, it is loaded on the forks of the truck and
transported to the conveyor (2.2). The fork truck driver scans the materials from the warehouse to ‘zone 60’ (= the conveyor). The internal norm for the processing of WR-assignments is 45 minutes after the order is registered in MAPxr.
3. Transport at the conveyor
The conveyor transports the materials from the RMW (Abramskade 4) to the factory at Abramskade 6. The conveyor is split up in different zones and in the software it is possible to follow where on the conveyor the materials are. At the end of the conveyor there are three exits, one for the materials which will be delivered by fork trucks, one for pallets transported by AGVs, and one for reels transported by AGVs. For all three exits, the materials follow their own route.
At the end of the conveyor, the materials for an AGV which are loaded on a pallet follow another route than the reels. For the AGV it is impossible to pick a reel when it is standing vertically. The reels first have to be turned over, before the AGV has the possibility to pick them. A special unit is installed to turn over the reels, so they are situated horizontally and the AGV can pick them from the conveyor.
Based on the material number and machine number which are present in the order in MAPxr, the conveyor can decide if an order is for the fork truck or AGV and whether it is a pallet or a reel.
4. Delivery at the machine
order should be picked and the final position of the drop. The pick position ‘1101’ is for the reels, ‘1102’ for the materials on pallets. The AGV which receives the order can be one that is parked and available for a new order, or one which is returning from a drop and which is closest to the conveyor.
After the AGV picks the material from the conveyor, it will transport the material to the machine where it’s ordered. For every different type of material there is a different drop location. Dependent on the type of material and whether it is a reel or pallet, the drop is performed on the floor (glue, boxes), on a small car (wadding), or on a stand (pulp, bags). The drop can only be carried out if no materials are present at the drop location and if the car or stand is present.
4. The current performance
In this chapter the current performance of the WR-, AP-, and AG-assignments is described. The throughput times are split up, for the handling time, transport time, and queuing time is described how these are calculated. Furthermore the factors which influence the throughput times of the different types of assignment are described.
4.1 Data analysis
From the beginning of the year 2010, the area within the plant which is supplied by AGVs slowly increased. The transport of materials on AGVs started for a few drop locations of the Astra-machines. Other drop locations for the supply by AGV were added in the system over time. In the weeks where all machines were supplied by AGVs (week 9 and 10), it turned out that the reliability of this way of delivering was not acceptable. There were too many late deliveries and the number of minutes machines had to stop as a consequence of a lack of material was extremely high.
From that moment the decision was made to reduce the number of drop locations for AGVs and only supply the seven Astra machines by AGV. From the moment this situation showed reliable throughput times, the number of drops was slowly expanded. In order to prevent things from going wrong, and as a support for the loaders of the conveyor, a shuttle was driving between the warehouse and the factory. The aim of this shuttle was to reduce the pressure on the conveyor and conveyor loaders as long as the performance wasn’t sufficient.
The shuttle mainly transported materials for UcM Inco and UcM Baby, the materials which are normally transported to the machine by fork trucks. During the transport of the shuttle, the AGVs kept driving between conveyor and the machines of Tena Lady and Panty 6. Slowly the number of materials which were transported according to the normal AGV-process increased. In week 19 the majority of the materials was
transported the normal way; everything was transported over the conveyor. The fork trucks and AGVs delivered the materials at the machine.
Week nr. WR-ass. AP-ass. AG-ass. M-ass
21 5,752 5,496 2,074 3,680 22 6,489 5,971 2,529 3,922 23 6,002 5,163 2,330 3,690 24 6,741 6,271 2,542 4,205 25 6,509 6,339 2,469 4,025 Total 31,493 29,240 11,944 19,522 Percentage 100% 92.8% 38.9% 61.1% Average 6,299 5,848 2,389 3,904
Table 4.1.1: Number of assignments per week
small difference, which means a part of the materials is transported with the shuttle. Furthermore figure 4.1.1 shows the number of AG- and M-assignments (orders transported by fork trucks) is about the same as the number of WR-assignments. The area which is supplied by AGVs is responsible for 38, 9% of the total number of deliveries within the factory.
The data used for this research is from week 21 until week 25. In these weeks the Astra and Heracles machines and Panty 6 are all supplied by AGVs. Table 4.1.2 shows the total number of material orders in this period fluctuated between 5,752 and 6,741 orders. The delivery performance (delivery within 85 minutes) fluctuated between 59.3% and 79.9%.
Week Order Quantity On-time Avg. del. time
21 5,752 orders 79.9% 63 min.
22 6,489 orders 72.8% 69 min.
23 6,002 orders 59.3% 80 min.
24 6,741 orders 62.2% 80 min.
25 6,509 orders 78.1% 65 min.
Total 31,493 orders 70.3% 72 min.
Table 4.1.2: WR-assignments, on-time performance and average delivery time
Table 4.1.3 shows the dispersal of the material orders over the different production units. In the desired state of phase one of the implementation of the AGVs, the number of AG-orders will be higher than in the current state. At this moment a certain number of materials are still delivered by a fork truck because there were problems with the delivery by AGV. Besides extra materials the return of waste and left-over materials will be performed by the AGVs. This means they receive an order from a machine to pick up the materials/waste and transport it to the waste containers or conveyor.
Week nr. Tena Lady % of Total UcM Baby % of Total UcM Inco % of Total
21 2,389 41.53% 1,045 18.17% 2,318 40.30% 22 2,671 41,16% 1,285 19,80% 2,533 39,04% 23 2,613 43.54% 1,059 17.64% 2,330 38.82% 24 2,748 40.77% 1,188 17.62% 2,805 41.61% 25 2,689 41.31% 1,217 18.70% 2,603 39.99% Total 13,110 41.63% 5,794 18.40% 12,589 39.97%
Week AGV (Astra) Trucks (Astra) AGV (Panty 6) Trucks (Panty 6) Total
21 1,948 orders 565 orders 126 orders 328 orders 2,967 orders 22 2,415 orders 375 orders 114 orders 401 orders 3,305 orders 23 2,220 orders 512 orders 110 orders 385 orders 3,227 orders 24 2,225 orders 463 orders 142 orders 442 orders 3,272 orders 25 2,350 orders 476 orders 119 orders 304 orders 3,249 orders
Total 11,158 0rders 2,391 orders 611 orders 1,860 orders 16,020 orders
Average 2,232 orders 478 orders 122 orders 372 orders 3,204 orders
Table 4.1.4: Orders which will be delivered by AGVs in the future.
Table 4.1.4 shows what the amount of AG-assignments will be in the future. The average number of orders for the AGV machines during these five weeks is 3,204 orders, while the average number of AG-assignments during the measured period was 2,389 orders (Table 4.1.1). Compared to the current state, this means the number of AG-assignments will increase with 34.1% in the desired state.
Below I will describe which part of the throughput time of the WR-, AP-, and AG-assignments exists of handling time, transport time, and queuing time.
WR-assignments
The total time needed to process a WR-assignment exists of the three types of time mentioned in the conceptual model: queuing time, transport time and handling time. Here the queuing time is the time an order has to wait before it is processed by one of the two fork truck drivers. The time starts at the moment the material is ordered by one of the machine operators. The handling time is the time needed to scan and load the order on the fork truck. The transport time is the time needed to drive from the conveyor to the warehouse location of the ordered material and back to the conveyor.
Week WR-ass. On-time % < 45 min.
21 5,752 4,491 78.1% 22 6,489 4,407 67.9% 23 6,002 3,866 64.4% 24 6,741 4,400 65.3% 25 6,509 4,791 73.6% Overall 31,493 21,955 69.7%
Table 4.1.5: WR-assignments and on-time performance
In the measured period the average number of material orders was 6,300 orders per week. This means every day 900 material orders have to be handled by the conveyor loaders, which are 37.5 orders per hour. This means on average every 96 seconds a material is loaded on the conveyor. As mentioned before, the WR-assignments have to be handled within 45 minutes after the order is placed. The performance in the
measured period is visible in table 4.1.5. With an average on-time performance of 69.7%, in the measured weeks it is far below the norm of 95.0% within 45 minutes.
throughput time of a WR-assignment is 39 minutes and 20 seconds. It can be concluded that the average throughput time is very high, because it is very close to the norm time of 45 minutes. Besides the high average, figure 4.1.6 shows there is a wide dispersion of the throughput times. Furthermore, the frequency of the categories with a throughput time above 45 minutes is too big. This also counts for the categories with a throughput time bigger than 30 minutes. These categories have too much influence on the average throughput time. The minimal throughput time of a WR-assignment is 33 seconds and the maximum throughput of a WR-assignment is 15 hours and 55 minutes.
Dispersal WR-assignment time
0 500 1000 1500 2000 2500 3000 3500 0: 05 0: 15 0: 25 0: 35 0: 45 0: 55 1: 05 1: 15 1: 25 1: 35 1: 45 1: 55 Mo re Throughput tim e Fr e q ue n c y 0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% Frequency Cumulative %
Figure 4.1.6: Dispersal of throughput time WR-assignments
To determine which parts of the total throughput time exist of transport time and handling time, I observed the conveyor loaders during their work. Therefore I used the stopwatch method (Riezebos et al., 2009). I measured the time they needed to load an order on the conveyor. The measured time is the sum of the transport and handling time. In appendix 3 the results of the observation are visible for all of the three fork trucks. Here the interarrival time shows the time elapsed after the previously loaded order. The observation showed there are differences per material; the materials which are ordered the most are stored close to the conveyor, which means the transport time is shorter for these materials. For materials which are ordered rarely, it is the other way around. The average time needed to transport and handle an order is 1 minute and 29 seconds.
The part handling time within the mentioned time is 26 seconds per order, which can be divided into the following parts:
• Pick up: 12 seconds; • Scan: 8 seconds;
• Load on conveyor: 6 seconds.
This means the average total transport time per order is 1 minute and 3 seconds (1:29 – 0:26). Furthermore it means the average queuing time per order is 37 minutes and 51 seconds (39:20 – 1:29), which is 96.3% of the total throughput time.
The sum of the handling time and transport time of 1 minute and 29 seconds leads to a capacity of 40 material units per conveyor loader (3,600 seconds/89 seconds), which means the conveyor loaders together can handle 80 WR-assignments per hour.
AP-assignments
Just like the WR-assignments, the time used to process an AP-assignment can be split up in different types of times. For the conveyor these only are transport and queuing time. Transport time is the time the material is actually moving on the conveyor, queuing time is the time the material has to wait on the conveyor for further transport.
Week AP-ass. On-time % on-time
21 5,496 4,299 78.2% 22 5,971 4,312 72.2% 23 5,163 2,977 57.7% 24 6,271 4,441 70.8% 25 6,339 4,603 72.6% Overall 29,240 20,632 70.6%
Table 4.1.7: AP-assignments and on-time performance
As mentioned before not all WR assignments are transported on the conveyor in the measured period, 92.8% of the orders are transported as AP-assignment. Table 4.1.7 shows the number of AP-assignments in the measured period and the performance of this part of the process. With an average on-time performance of 70.6%, the
performance is far below the stated norm of 95.0% within 20 minutes.
The conveyor had to transport 29,240 orders in total, which is 5,848 AP-assignments per week or 34.8 AP-assignments per hour. In theory the conveyor has the capacity to load a material order every 50 seconds. In this case the elevator of the conveyor is the bottleneck: it takes 50 seconds in total to load the elevator, to lift the unit, to transport the material to the next section of the conveyor and go down again. With a theoretical interarrival time of 50 seconds the total capacity of the conveyor is 72 AP-assignments per hour.
In the data (Excel) sheet where the overall delivery performance is measured and split up to the different parts of the process, the throughput time of every assignment is visible for every week, day, or shift. An example of this Excel sheet is visible in
are dispersed. 70.6% of the AP-assignments are processed within the stated norm-time of 20 minutes. This sheet also shows the time which is needed to transport the materials to the last position of the conveyor for the AGVs. Without interruptions that is 7 minutes.
On average it takes 17 minutes and 4 seconds for the conveyor to process an
AP-assignment. This average throughput time is very high; it is very close to the norm time of 20 minutes for an AP-assignment. Besides that the average is over 100% more than the time normally needed to transport the materials (7 minutes). Furthermore, the dispersal of the throughput times is big, a lot of categories in figure 4.1.8 have a frequency higher than 1,000.
Dispersal AP-assignments 0 1000 2000 3000 4000 5000 6000 0: 06 0: 08 0: 10 0: 12 0: 14 0: 16 0: 18 0: 20 0: 22 0: 24 0: 26 0: 28 0: 30 0: 32 0: 34 0: 36 0: 38 0: 40 Mor e Throughput time Fr e q ue n c y 0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% Frequency Cumulative %
Figure 4.1.8: Dispersal of throughput time AP-assignments
The speed at the conveyor is the same for all materials, but the distances of fork-truck orders, AGV pallet orders, and AGV reel orders do vary. The first part of the conveyor is the same for every material. At the moment the orders arrive at the factory side of the plant, the difference is made between orders for fork trucks and AGVs, which both have different pick positions. The materials for the fork trucks go straight ahead and follow the shortest route. The orders for AGVs are transported to the left, where the difference is made between reels and materials on pallets. The distance for the reels is the longest of all materials. This means the transport time of the reels for the AGVs is the longest, and the transport time for the materials for the fork truck is the shortest.
conveyor is 10 minutes and 4 seconds (17 minutes and 4 seconds – 7 minutes). This is 59.0% of the total throughput time.
AG-assignments
The throughput time of an AG-assignment exists of both transport and queuing time. The transport time is the time which is needed for the actual transport of a material. Queuing time is the time the AGV is standing still while it is loaded with a material.
Week AG-ass. On-time % on-time
21 2,074 1,822 87.9% 22 2,529 2,209 87.4% 23 2,330 2,014 86.4% 24 2,542 2,184 85.9% 25 2,469 2,162 87.6% Overall 11,944 10,391 87,0%
Table 4.1.9: AP-assignments and on-time performance
Table 4.1.9 shows the number of AG-assignments of every week. The total amount of AG-assignments is 11,944, which means 40.8% (11,944/29,240) of the orders
transported over the conveyor is delivered at the machine by an AGV. From the total number of material orders 37.9% (11,944/31,493) is supplied by an AGV. 87.0% of the AG-assignments from week 21 till week 25 are delivered within the norm time of 20 minutes. Just like the WR- and AP-assignments, this performance is far below the norm for on-time delivery of 95.0%.
Dispersal AG-assignments 0 500 1000 1500 2000 2500 0: 02 0: 04 0: 06 0: 08 0: 10 0: 12 0: 14 0: 16 0: 18 0: 20 0: 22 0: 24 0: 26 0: 28 0: 30 0: 32 0: 34 0: 36 0: 38 0: 40 Mo re Throughput time Fr e q ue nc y 0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% Frequency Cumulative %
In the measured period the AGVs transported 11,944 orders in total, which is 2,389 AG-assignments per week, or 14.2 per hour. The data from MAPxr shows the
performance of all different AG-assignments which are processed. Figure 4.1.10 shows how the throughput times of the different orders are dispersed. The biggest category in this graph is the category of orders which are delivered at the machine within a time between 4 and 6 minutes (16.3%) after they arrive at the last position of the conveyor. Thus, the mentioned time is not only the drop itself, but also the time it is present at the latest position of the conveyor. Figure 4.1.10 shows almost 50% of the materials is delivered within 10 minutes, after that the frequency of every category is decreasing. Furthermore it shows the dispersion of the throughput times of the AG-assignments is big.
Within the plant the distance between conveyor and machines or drop locations differs a lot. This is the reason I decided to split the different machines into six areas. Below the different areas are mentioned with the number of drop locations within the area:
- Heracles East: Partly Heracles 9, 10, 11 – 7 drop locations; - Heracles West: Partly Heracles 9, 10, 11 – 31 drop locations; - Panty 6 – 17 drop locations;
- Astra East: Partly Astra 2, 3, 4 – 12 drop locations;
- Astra Central: Partly Astra 2,3,4,5,6,7,8 – 36 drop locations; - Astra West: Partly Astra 5, 6, 7, 8 – 19 drop locations.
Within these different areas, the average of the shortest times per drop location for an AG-assignment is seen as the minimal time needed to drive to one of the areas
(transport time). I assume the extra time needed for the drop can be seen as queuing time. Scarabee, the supplier of the AGVs, delivered the data with the minimal, maximal and average times of the drops per machine or drop location. The number of drop locations per area differs. In Appendix 5 an overview is added of the drop locations and the number of drops per location. Furthermore the minimum drop time is visible, and the average per drop and area is calculated.
Area Average minimum Average drop-time Difference Drops
Astra Central 3:57 5:05 1:08 1,242 Astra East 4:25 6:46 2:21 435 Astra West 2:41 4:33 1:52 763 Heracles East 8:54 14:32 5:38 322 Heracles West 6:59 12:02 5:03 1,261 Panty 6 6:13 8:56 2:43 214 Total 4:54 8:08 3:14 4,237
Table 4.1.11: Drop times per area
Table 4.1.11 shows information about the length of the AG-assignments in the 27th and
28th week of 2010. The average drop-time is the average of all performed drops within
while the average minimum is 4 minutes and 54 seconds. This leads to an average queuing time during the transport of 3 minutes and 14 seconds per drop.
After the drop the AGV has to return to the conveyor to pick another order. To find out how long this takes, I observed how many time elapses between the moment an AGV picks a material and returns to the conveyor for the pick of the following order (cycle time). The forms I used are added as appendix 6 and 7. Appendix 6 shows the order picks of all AGVs during the observed time. The time registered is the time an AGV picks an order. Based on all picks the average cycle time of an AGV is calculated. Appendix 7 shows the data of an observation of a specific AGV. For every AGV the picks at the conveyor and drops at the machine are registered. Besides that the time the AGV parks, charges the battery, replaces the battery, or is in a breakdown are
measured. Based on the measured data the average cycle time can be calculated.
It turned out that on average an AGV needs 16 minutes and 22 seconds to return at the conveyor after it picked an order. An overview of all measured cycle times per AGV and calculation of the average is added as Appendix 8. With the earlier mentioned average drop-time of 8 minutes and 8 seconds, it means after every drop the AGV needs 8 minutes and 14 seconds to return to the conveyor. Thus, the total transport time within the cycle time is 13 minutes and 8 seconds and the queuing time is 3 minutes and 14 seconds. For the AG-assignments, the queuing time is 19.8% of the total throughput time. The lap time of 16 minutes and 22 seconds means the 5 AGVs can process 18.3 material orders per hour.
Transport Time Handling Time Queuing time
Type of
Assignment Time Percentage Time Percentage Time Percentage Through-put time
WR-ass 1:03 2.6% 0:26 1.1% 37:51 96.3% 39:20
AP-ass 7:00 41.0% --- --- 10:04 59.0% 17:04
AG-ass 13:08 80.2% --- --- 3:14 19.8% 16:22
Table 4.1.12: The assignments split up to the different types of time
Table 4.1.12 gives a summary of the different times the throughput time exist of. It shows that for every type of assignment the queuing time is a significant part of the total throughput time; for the WR- and AP-assignments it is even the biggest part and above 50% of the total. To find out which factors influence the queuing time and thus the total throughput time of a material order, further research is needed. The results of that will be described in paragraph 4.2.
4.2 Influencing factors
As turned out in paragraph 4.1, the part of queuing time for the different types of assignments is large. According to the literature described in paragraph 2.4 there are six different factors which can have influence on the total queuing time. For the internal delivery process of SCA/UcM two of these factors play a role, which I will describe in this paragraph: the utilization of the servers of the different processes and the
Utilization of servers
The utilization is the actual output from a process or facility to its designed capacity (Slack, 2007). Table 4.1.2 shows the utilization of the different servers within the internal delivery process. The numbers in the table are mentioned in chapter 4.1 of this report.
Transport unit Capacity Needed quantity Utilization
Conveyor loaders 80 orders 37.5 orders 46.9%
Conveyor 72 orders 34.8 orders 48.6%
AGV(s) 18.3 orders 14.2 orders 77.6%
Table 4.2.1: Utilization of the servers
The table shows the capacity of the conveyor loaders and the conveyor is sufficient to process all orders, the capacity they have to process will not cause any problems. For the AGVs the utilization is high, it cannot be afforded to work with four instead of five AGVs. This will not only lead to more pressure on the AGVs, but also to longer queuing times for the AP- and WR-assignments.
In paragraph 4.1 I described what the average drop-time and cycle time of an AG-assignment is. The average cycle time of 16 minutes and 22 seconds can be separated in 13 minutes and 8 seconds transport time and queuing time of 3 minutes and 14 seconds. As mentioned before, this means the average queuing time of an AGV during a drop is 3 minutes and 14 seconds.
To gather more information about the transport and queuing time of the AGVs I observed the five AGVs several times. The total observed time was 50 hours; the actual time per AGV is mentioned in Table 4.2.2. During these observations it turned out the queuing time for the AGVs has two causes: breakdowns and the routing. The
breakdowns do not only cause problems for the broken AGV, but also for the following AGVs which have to wait because they cannot pass. The queuing time caused by the routing occurs in the areas where only one AGV is allowed at the same time. Here, AGVs have to wait for another AGV to leave an area before they can enter the specific area.
Queuing Time
AGV Observed time Drops Breakdown Wait for Breakdown Routing Total
1 9 h 00 min 18 0:56 0:09 0:10 1:15 2 9 h 00 min 25 0:15 0:20 0:16 0:51 3 11h 30 min 27 0:12 0:32 0:21 1:05 4 9 h 00 min 20 0:26 0:21 0:18 1:05 5 11h 30 min 32 0:34 0:44 0:22 1:40 Total 50 h 00 min 122 2:23 2:06 1:27 5:56
Table 4.2.2: Queuing time per category and AGV
the time an AGV is waiting for an AGV that is broken down, and the column routing shows the time an AGV has to wait for another AGV that is processing its order or on its way back to the conveyor.
The queuing time as a consequence of the breakdowns occurs for different reasons. In the simplest situation a machine operator walks in front of an AGV and the AGV will restart itself. Besides that there were breakdowns caused by waste on the floor, or stands or cars which were not present at the drop location. Furthermore there were some inexplicable breakdowns. For these types of breakdowns a mechanic had to come to the AGV to solve the problem.
Normally the machine operators have to report the breakdown at the moment they notice it. During the observation of the AGVs it turned out this rarely happened, which led to increasing queuing times and in some cases other AGVs waiting behind it. The best example is a situation where a breakdown was not reported and it took 14 minutes before it was solved. As a consequence of this long breakdown, the other AGVs were waiting for 45 minutes in total before they could proceed with their order(s).
The total queuing time caused by breakdowns during the observations was 2 minutes and 12 seconds per drop (4:29/122), which is 68% of the total queuing time of the AGV. The fact inexplicable breakdowns occur makes it almost impossible to reduce the total queuing time caused by breakdowns. If it is only possible to reduce this queuing time with 50%, this will lead to an increasing capacity of 1.4 orders per hour. This means the utilization will decrease to 72.3%, which means there is less pressure on the system and the AGVs have to work less hard to process all orders.
As visible in table 4.2.2 the routing is the other cause for the queuing times of the AGVs. Within the factory there are three areas where the major part of this queuing time occurs: Astra Middle, Heracles, and Panty 6. The problem with Astra Middle and Heracles is that there is only one AGV allowed at the same time and the second AGV has to wait before it can enter the area. For Astra Middle this means the waiting AGV is blocking the main route of the AGVs, which causes a queue at that point. For Heracles the same counts, with the difference that not the first, but the second waiting AGV blocks the main route. At Panty 6 the AGVs block the main route at the moment a drop is performed, because the drops for these machines are located next to the main route. Besides these specific delivery areas, at the end of the tunnel between the Panty- and Astra hall queuing time was measured. This is the crossing for the AGVs returning to the conveyor and the AGVs which just left the conveyor.
The average queuing time of a drop has a big influence on the number of drops which can be performed by the AGVs. As mentioned before, with a lap-time of 16 minutes and 22 seconds, 18.3 drops can be performed every hour. This means every 3 minutes and 17 seconds a new material is picked from the conveyor. The observation showed that when the buffer of the conveyor is full of materials, the flow through of materials on the conveyor stops. When this occurs, the queuing times for the AP-assignments increase, because the conveyor only moves at the moment a material is picked by an AGV and only than the materials on the conveyor are transported further. Furthermore the queuing times for the WR-assignments increase, because the conveyor loaders can only load a new material on the conveyor after the AGV has picked one from the conveyor. An example of this situation is the 28th of May from 9 a.m. till 9 p.m. During this period
the conveyor did not have the opportunity to create a continuous flow of materials, because it had to wait for the AGVs to pick materials from the last position. Table 4.2.3 gives a short summary of the data of that day. During the mentioned 12 hours, the average number of AG-assignments was 16.3 per hour, which means the utilization of the AGVs was 89.1%. Table 4.2.4 shows these high utilization causes longer throughput times for the AP-assignments.
Hour AG-assignments Average AP-time
9:00 – 10:00 13 assignments 0:16:22 10:00 – 11:00 12 assignments 0:15:33 11:00 – 12:00 21 assignments 0:35:14 12:00 – 13:00 16 assignments 0:28:24 13:00 – 14:00 16 assignments 0:42:51 14:00 – 15:00 19 assignments 0:38:34 15:00 – 16:00 16 assignments 0:33:27 16:00 – 17:00 17 assignments 0:25:56 17:00 – 18:00 19 assignments 0:35:12 18:00 – 19:00 19 assignments 0:36:14 19:00 – 20:00 11 assignments 0:19:11 20:00 – 21:00 17 assignments 0:21:14
Table 4.2.3: Average AP-time at the 28th of May
At this moment, on average the utilization of the AGVs is sufficient to deliver all orders at the machine. However, as described in paragraph 4.1 the AGVs have to deliver more orders in the desired state (3,204 orders per week). Besides that the waste- and return orders should be processed by the AGVs and they need time to charge or change their batteries.
The numbers mentioned above make clear that the capacity of the AGVs in the desired situation is not sufficient to deliver the material demand of the machine operators. This means the capacity of the AGV must increase to ensure the different parts of the
delivery process will be processed within the stated norm times.
Interarrival time of WR-assignments
As described by Riezebos et al. (2008) the interarrival time is the time that elapses between the arrivals of two different flow units (orders/customers). To calculate the average interarrival time for the WR-assignments I combined the data of week 21 till week 25. Appendix 9 shows a short overview of a small part of the data sheet. Here the arrival times of the assignments and the difference in time between two consecutive orders are visible. Figure 4.2.4 shows how these interarrival times of the WR-assignments are dispersed.
Dispersal of interarrival WR-assignments
0 2000 4000 6000 8000 10000 12000 14000 16000 0:00: 30 0:01 :00 0:01: 30 0:02: 00 0:02: 30 0:03: 00 0:03: 30 0:04: 00 0:04: 30 0:05: 00 0:05: 30 0:06: 00 0:06: 30 0:07: 00 0:07 :30 0:08: 00 0:08: 30 0:09 :00 0:09: 30 0:10: 00 Mor e Interarrival time Fre q ue nc y 0,00% 20,00% 40,00% 60,00% 80,00% 100,00% 120,00% Frequency Cumulative %
Figure 4.2.4: Dispersal of interarrival time WR-assignments
The graph shows that 43.8% (= 13,781 orders) of the total number of WR-assignments arrives within 30 seconds after the previous order. Remarkable for the data within this category (< 30 seconds) is that there are 8,264 orders within this category (26.2% of the total number of WR-assignments) where the interarrival time is 0 seconds. This is a consequence of material orders which contain order lines for several materials or contain a quantity bigger than one.
in this situation the arrival of orders is not constant, but it strongly differs. Figure 4.2.4 shows there is a big dispersion of interarrival times, which means the number of
arriving orders also differs every hour.
WR-assignments per hour (week 25)
0 50 100 150 200 250 300 350 400 450 500 1 0 2 1 3 2 4 3 5 4 6 5 7 6 8 7 9 8 10 9 11 10 12 11 13 12 14 13 15 14 16 15 17 16 18 17 19 18 20 19 21 20 22 21 23 22 24 23 Hour Nu m b er o f W R 's
Figure 4.2.5: Material orders per hour in week 25
