The influence of buffering by the use of a circular
conveyor on the output of a sorting system:
A case study in the parcel industry
Master-thesis
June 26, 2017
By S. Hoogendoorn (S2803917) First supervisor: dr. ir. S. Fazi
[email protected] Second supervisor: dr. O.A. Kilic
MSc Technology and Operations Management Faculty of Economics and Business
Master-thesis Technology and Operations Management by S. Hoogendoorn I
Preface
I would like to take the opportunity to write out my appreciation to the people who helping me during this project.
First of all I would like to express my gratitude to my first supervisor dr. ir. S. Fazi who guided me through the process of this thesis and provide me with constructed feedback and suggestions. Also, my thanks go to my second supervisor dr. O.A. Kilic who helped my improving my research.
I want to thank PostNL, which provided me with all the information I needed to build a representative research model. Especially, I would like to thank Mr. J. Jeuring who provided a guided tour through its depot and answered specific questions about the system.
Last but not least, I will thank my family and friends who give me unconditional support during all the ups and down I faced during this process. Without this people I wouldn’t have been able to come to these results.
Master-thesis Technology and Operations Management by S. Hoogendoorn II
Abstract
Master-thesis Technology and Operations Management by S. Hoogendoorn III
Contents
1 Introduction ... 1 2 Literature background ... 3 2.1 Terminology ... 3 2.2 Circular conveyors ... 3 2.3 Sequencing ... 52.4 Conceptual model theory ... 6
3 Problem analysis ... 8
3.1 The case “PostNL” ... 8
3.2 Problem situation ... 8 3.3 Contribution... 11 4 Methodology ... 12 4.1 Research design ... 12 4.2 Data collection ... 12 4.3 Data analysis ... 13 4.3.1 Throughput rate ... 13 4.3.2 Utilization ... 14 4.3.3 Arrival rate ... 14 5 Simulation ... 15 5.1 Model design ... 15 5.2 Assumptions/simplifications. ... 18 5.3 Experimental design ... 19
5.3.1 Scenario 1 (Base case) ... 19
5.3.2 Scenario 2 (Proposed scenario) ... 19
5.4 Experimental settings ... 21
5.5 Output analysis ... 22
5.6 Verification and validation ... 22
6 Numerical results ... 23
6.1 Base case results ... 23
6.2 Scenario 2 ... 24
6.2.1 Results work in progress levels ... 24
6.2.2 Effects of throughput changes ... 25
6.2.3 Effects of conveyor-speed changes ... 27
Master-thesis Technology and Operations Management by S. Hoogendoorn IV
6.2.5 Effects of interarrival time changes... 30
Master-thesis Technology and Operations Management by S. Hoogendoorn 1
1 I
NTRODUCTIONThe e-commerce market has seen a rapid growth in the last few years. For example, the turnover in Europe was 455 billion euros in 2015, and the demand in this market is expected to grow by 31% in 2017 (Ecommerce-News, 2016). Consequently, postal service providers (PSPs) have seen a substantial growth in volumes of parcels to be handled and are under pressure to maintain a reliable delivery. However, not only has the volume of parcels increased, but competition has increased as well. Therefore, it is even more crucial that postal services use their resources as efficiently as possible to outperform competitors in the market (Werners and Wülfing, 2010).
One of the most delicate activities for a PSP is the sorting process, as it has to be performed in a very narrow time span due to the fact that, nowadays, most e-commerce organizations promise next-day delivery to their customers (Mcwilliams, Stanfield and Geiger, 2005). In a sorting center, parcels are directed to destination-specific shipping areas, by means of a conveyor belt. Currently, when the parcels arrive at the shipping areas, they are placed into roll containers, which are the transport units used in this industry, on a first-come, first-served (FCFS) policy. However, internal research at one of the largest PSPs in the Netherlands showed potential cost savings if the utilization of these transportation units can be improved. This study is complementary to a previous study of (Uilhoorn, 2016), who developed a mathematical model, specifically a three-dimensional bin-packing MIP model, to pack roll containers optimally in terms of space. However, the current sorting strategy in this industry is static. In practice, parcels arrive at a distribution center in a random sequence and are only be sorted by their final destination. Therefore, the following situation is proposed: Given a specific packing sequence for a roll container, the system sorts the parcels in the right order and time at the transportation unit. Due to the random arrival order of parcels, some parcels are released directly for packing; others are postponed to re-sequence the product range. Using the circular conveyor in the system as a short time buffer to control the amount postponed parcels is the starting point of this research.
Master-thesis Technology and Operations Management by S. Hoogendoorn 2 in roll containers is maximized, and, in the end, truckload is more utilized for transport between different distribution centers. This research is focused on a case at Post NL, the largest parcel deliverer in the Netherlands, to investigate the problem of efficiency in a dynamic context.
In the current literature, conveyor transport aimed to sort items is applied in several different industries. For example, it is used in baggage handling systems at airports, where customers’ incoming luggage has to be directed to the right airplane on large conveyor networks. A large amount of luggage has to be sorted and checked for security in a small time frame at many different locations where the failure level should be very low (Le, Creighton and Nahavandi, 2007). Another example is the quality selection of items in the agricultural and recycling industry, where decision making concerning sorting hinders the process due to the high speed of the system (Mattone, Campagiorni and Galati, 2000; Li, Wang and Gu, 2002). In the area of distribution centers and automatic warehouses, achieving high efficiency and throughput is the main objective for the sorting activities to deliver a high variety of items on a timely basis (Johnson, 1998). All these problems faced in different areas come together in the dynamic context of PSP (Haneyah et al., 2011). Current research in the area toward parcels sorting has mainly focused on scheduling activities and transport optimization. For example, (Tsui and Chang, 1992) focused on the assignment of outbound dock doors to static destinations or fluctuating destinations due to peak loads. Mcwilliams, Stanfield, and Geiger (2005) performed a simulation-based scheduling algorithm to balance the throughput during peak flows of parcels arriving in a distribution center. However, in the literature, little attention has been paid to how the actual automated sorting process influences the efficiency of PSPs. After having analyzed the aspects influencing the literature toward efficient conveyor sorting, the following research question and sub questions were derived:
How do sorting strategies related to system output affect the system performance of an automated sorting system?
1. What are the most important indicators for system performance? 2. What are the bottlenecks in an automation sorting process?
3. How are the current sorting processes organized?
Master-thesis Technology and Operations Management by S. Hoogendoorn 3
2 L
ITERATURE BACKGROUNDThis section will shed light on the current knowledge about automated sorting processes and the problems faced in this field. In the first part, some important terminology is explained. The second provides details about the current knowledge of circular conveyors. In the third section, insights from sequencing in other industries are discussed. The last part describes a gap in the literature, and the conceptual model is explained.
2.1 TERMINOLOGY
The function of the current sorting system at postal service centers (PSCs) is quite static, as the system is used to transport items are usually constrained to operate in a come, first-served (FCFS) principle (Johnson, 1998; Wang, Batta and Szczerba, 2005). Currently, the typical technology used in PSC is a cross-belt sorter, which can be described as a series of mini conveyor belts aligned at 90 degrees to the direction of travel and used in applications where gentle treatment of items is required (Johnson and Meller, 2002; Croucher and Baker, 2014). This kind of conveyor system can play a much broader role than just providing orderly transport; it also stores items while they are in transit (Sonderman, 1982). Therefore, the proposed model deals with the sequencing output of the sorting system by delivery items at the load area according to the input of a packing schedule derived from a heuristic model. In this situation, the system has to be smart enough and flexible enough to provide the sequencing and handle all exceptions.
2.2 CIRCULAR CONVEYORS
Master-thesis Technology and Operations Management by S. Hoogendoorn 4 2005), the basic model is extended with multiple loading and unloading stations. This kind of configuration in a closed-loop conveyor system is used in most PSCs (Figure 2.1).
Figure 2.1 - Generic layout of closed-loop parcels sorting process (Haneyah et al., 2011).
Master-thesis Technology and Operations Management by S. Hoogendoorn 5 performance. The results show that the layout of loading and unloading stations around the conveyor system significantly affects system performance.
Waiting time is another aspect that is important in the performance of conveyor sorting systems. When waiting times are known, the queue lengths can be determined and the performance can be evaluated to an acceptance level (Bozer and Hsieh, 2004). Moreover, these waiting times can be used to avoid system stoppage and lengthy queues. Bozer and Hsieh (2004) developed a general model to calculate the waiting times at the loading stations and at every server. An interesting aspect of this article is that this model uses random deterministic arrival times, as is the case for arriving parcels in PSP.
2.3 SEQUENCING
Literature in this area—item sequencing and buffering by the use of a conveyor system—is scarce. Therefore, useful insights from other areas are also discussed. The sorting of items during the production process is often called “re-sequencing,” which is a phenomenon used in assembly lines of the automotive industry to make production more versatile and flexible by using off-line buffers (Battini et al., 2009). Automotive industries consist of mixed model assembly lines with different workstations coupled to buffer stocks. Re-sequencing the production schedule within a buffer adds flexibility in production (due to rework, for example) but adds randomization in the production order. Sequencing rules are used to balance the cycle time and increase the utilization and throughput of a system (Masood, 2006). In a case study in this area, Jayaraman et al. (1997) identified general parameters that affect the sorting processes, such as the configuration of loading areas, the number and capacity of loading areas, and the input and output rules (known as algorithms).
Master-thesis Technology and Operations Management by S. Hoogendoorn 6 transportation units at a loading area using a fixed priority rule or a flexible structure. Meller (1997) used the objective of minimizing the number of item recirculation’s on the circular conveyor as an indicator of the total time for a sorting cycle. Johnson et al. (2002) observed an OAS system used for order-picking approaches in the e-retail industry. This research identifies how different parameters, such as conveyor speed and the arrival induction of an automatic sorting system, influence system capacity. The starting point of this research is that either the arrival process or the sorting conveyor can be the bottleneck in a system and therefore influence the total capacity. Arrival rates were tested at increasing speed levels, confirming that throughput performance is limited to an upper-bound level.
Another area where sequencing occurs is in the classification of yards on railways. In rail classification yards, freight cars are uncoupled, split up into traps to sort them to their final destination, and reassembled into new outbound train blocks. Daganzo et al. (1983) did research on the relative performance of several sorting strategies in the process of railroad classification. In this research, a triangular sorting strategy, which is a three-level sorting process, was shown to not take significantly more processing time for a single item compared to a multi-stage or a single-stage sorting strategy. However, a limitation of this paper is that a small volume throughput level was tested during a sorting cycle.
Although there some work has been done on the performance of sorting systems in PSCs, capacity restrictions due to buffering of items on the conveyor system have barely been investigated. The main reason for this is that only the transport function of the conveyor is currently used (Fedtke and Boysen, 2017). However, they argued for more research on the effects of priority rules on exact problem situations since the rules significantly affect the throughput in their general model. Moreover, Johnson (1998) indicated that a limitation of their model was that the input of the conveyor needs to be controlled by the output of the outbound shipping lanes. Their model faces some idle time when orders need to be completed before new ones are released.
2.4 CONCEPTUAL MODEL THEORY
Master-thesis Technology and Operations Management by S. Hoogendoorn 7 system capacity. The performance is the result of the system output defined in processing time of the parcels and the throughput in the number of parcels per time unit.
Figure 2.2. - Theoretical framework.
Parcel sorting process Prioritize output for
packing sequencing
Systemperformance
(Processing time & throughput)
Master-thesis Technology and Operations Management by S. Hoogendoorn 8
3 P
ROBLEM ANALYSISThis chapter provides an introduction to the case of the company PostNL, followed by a description of the problem situation at PostNL, on which this research is based. Additionally, the practical contribution of this research is explained.
3.1 THE CASE “POSTNL”
The largest Dutch postal service provider has a history of more than 200 years, and there have been many mergers and a lot of cooperation with other partners in the market before it became the current PostNL. With the digitalization of communication policies, such as the Internet and e-mail in the 21-century, there has been a decline in traditional mail (i.e., letters) but an increase in the e-commerce market, resulting in high demand for parcel services (see Figure 3.1). Currently, the parcel market is still growing; in terms of volume, the number of parcels handled has increased from 156 million items in 2015 to 177 million in 2016, which reflects in a growth rate of 13.3% according the company’s annual report (Wortelboer, 2016). Therefore, PostNL has reacted to this market growth by updating their distribution network for parcel services by replacing the 37 old distribution centers for 18 state-of-the-art sorting depots from 2013 to 2015 (Wortelboer, 2016). All of these “new-style” depots are based on the same design with regards to the sorting machinery and processes. This research is part of these improvements, where the focus is on the sorting process within a depot.
Figure 3.1 - Volume development of parcels at PostNL (Wortelboer, 2016).
3.2 PROBLEM SITUATION
Master-thesis Technology and Operations Management by S. Hoogendoorn 9 The first step of a sorting process is to unpack incoming parcels from companies or customers and release them in the automatic sorting system. This is a manual step where a worker picks (randomly) an item from a full roll container or a pallet and places it on an elevating conveyor belt (see set 1 at Figure 3.2). A sorting process contains several of these (unpack) workstations to be able to unload more than one truck at the same time.
Figure 3.2 - Schematic layout PSC adapted from: Fedtke and Boysen (2017).
Master-thesis Technology and Operations Management by S. Hoogendoorn 10 Other parcels continue their way on the circulating conveyor until passing the exit station, a chute related to its final destination (step 3 at Figure 3.2). At this stage, a decision is made to actually “chute” the item or postpone that action for another round on the conveyor. When the item is chuted, the parcel leaves the automatic sorting system, ending at a packing station where five resources (roll containers) are available to place items on. The placement of items on the resource is a manual job; a worker at this station takes the item from the chute and places it on one of these roll containers. Here, the worker checks the destination code, and then one of the five available roll containers is selected according the insight of the workers for the most optimal position. Once the parcel is packed, it will leave the depot by a truck or van. The conveyor-loop in this process contains 44 of these packing stations.
Currently, the parcels are loaded randomly in roll containers. Every parcel is unique in dimensions and weight, resulting in a random product mix. When these items are randomly processed, the result would be an output of containers where packing rate is not constant and optimized (see Figure 3.3).
Figure 3.3A - Unstructured container packing (Current situation ).
Figure 3.2B - Layer packing (Proposed situation).
Master-thesis Technology and Operations Management by S. Hoogendoorn 11
3.3 CONTRIBUTION
As explained, the current process at the PSC is not able to organize the container packing process in an efficient way. This research considers the ideal situation at a PSC where a set of parcels will be sorted through the automatic sorting system since it has been proved that decision support system outperforms the manual task (Uilhoorn, 2016). The system considers the input given by a bin packing algorithm for the ideal position of parcels in a roll container. Together with a software model and a packing configuration, the automatic sorting system will be controlled to deliver the parcels in the appropriate sequence at the roll container. The physical process is the same as it is currently in terms of arriving items, sorting equipment, and manual container packing, but a “smart” software element will be added to the automatic sorting system (Figure 3.4). The desired outcome of the sorting process would be a roll container that has a minimal waste of unused packing capacity Figure 3.2B.
Model and algoritm Packing configurations Information of incomming parcels
Conveyor system Roll container packing Parcels
Software
Master-thesis Technology and Operations Management by S. Hoogendoorn 12
4 M
ETHODOLOGYThe methodology section gives an overview of the methodology used to test the effects of system performance. First, it explains what type of analytical approach was used. Second, this section shows how the data was gathered from the case situation and transformed into useful data for the simulation study.
4.1 RESEARCH DESIGN
“Many real life problems in optimization are too complex to be solved by mathematical programming” (Allaoui and Artiba, 2004). In particular, small package sorting is a complex dynamic process in the PSC environment (Werners and Wülfing, 2010). System complexity is there for many reasons. First, due to the large amount of items carried in a day, where distribution of the dimension characteristics of parcels is random, the complexity of the system increases. Second, many operational decisions need to be made on destination assignment problems based on priority of items. A third reason for this has to do with the dynamics of the system, where many inbound and outbound shipping lanes influence the throughput rate of the system. For example, a typically distribution center of a PSP can consist of 40–100 packing gates and several loading areas (Werners and Wülfing, 2010). A simulation study of the sorting process for a parcel sorting center is the most likely approach to solve the problem of measuring system performance using priority rules.
The simulation model has been built with the Tecnomatix Plant Simulation software, which is based on a discrete event simulation approach using the SimTalk programming language. “Plant Simulation provides all necessary functionality to model, analyze, and maintain large and complex systems in an efficient way” (Bangsow, 2010). As argued in the previous paragraph, a postal sorting process contains all these complexity characteristics. Discrete event simulation can be defined as a system where state-change events happen at discrete moments in time, but events take zero time to happen (Varga, 2001). Priority rules in the sortation process are typically examples of state-dependent events.
4.2 DATA COLLECTION
Master-thesis Technology and Operations Management by S. Hoogendoorn 13 category B data is not yet available but collectable, and category C is not available and not collectable.
Contextual data in this research is mainly gathered by non-recorded and semi-structured interviews with experts who have experience in the field of PSCs (e.g., depot managers and project supervisors). Besides the interviews, a visit to a sorting center of the case company brought many insights to the process. The information from this collection method was mainly to get insights into the main characteristics of the sorting process. Category B data consists of data to set system parameters for the simulation. Most data about the arrival times, even as the dimensions of the parcels, are available in a TrackandTrace analysis of a sorting cycle over the span of a week at one of the PostNL distribution centers. System characteristics, such as conveyor length and speed as well as the capacities of the buffers, were not directly available but were collected via an interview with the system designers. For category C data, assumptions have been made, which are described in the assumptions section of the chapter five.
4.3 DATA ANALYSIS
Process data of the TrackandTrace system at a sorting depot from PostNL were analyzed to create the most realistic base scenario for the simulation study. This dataset consists of information from parcels handled at a specific depot of three typical processing days, where the total amount of processed entities was around 119,000. The data file includes the time of each arriving parcel entering the sorting process, the three standard dimensions (length, width, and height), the lifting conveyor belt on which parcels are loaded into the system, the destination chutes where parcels are released on base of the final destination, and the source of the parcel (received from which customer). This data is assumed to be generalizable for an average PSC.
4.3.1 Throughput rate
Master-thesis Technology and Operations Management by S. Hoogendoorn 14
Nominal cycle length (hours) Number of items (N) Process rate (Items/min) Utilization (U) Cycle 1 5.77 32, 704 88 0.66 Cycle 2 8.40 42,330 98 0.74 Cycle 3 4.28 25,054 94 0.72 Cycle 4 3.61 18,957 84 0.64 Average 91 0.69
Table 4.1- Analysis cycle length and system utilization of real data.
4.3.2 Utilization
From the data, it could be determined that the throughput capacity is around 8,000 units per hour, which is a maximum throughput rate of 133 items/hour. Having the data of maximum throughput and average processing rate, the utilization of the system can be determined using the following formula:
𝑈𝑡𝑖𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 =𝑝𝑟𝑜𝑐𝑒𝑠𝑠𝑖𝑛𝑔 𝑟𝑎𝑡𝑒
𝑐𝑎𝑝𝑎𝑐𝑖𝑡𝑦 (Hopp and Spearman, 2011) (1)
The results of the utilization for every process cycle are presented in Table 4.1. The average utilization is 0.69 by an average processing rate of 91 items per hour.
4.3.3 Arrival rate
The loading process of items into the sorting system shows a batch pattern, where a constant interarrival time of 30 seconds between each batch has been observed. The size of these batches seems to be non-constant over time. Therefore, a frequency-histogram was created to estimate a distribution for the batch sizes in Figure 4.1. A normal distribution was fitted to this histogram with a mean of 13.7 and a standard deviation of 5.6.
Figure 4.1 - Approximation of arrival distribution of real data. 0 0,02 0,04 0,06 0,08 0,1 0 5 10 15 20 25 30 P ro ba bi lit y (P ) Batch size (N)
Frequency distribution batches
Master-thesis Technology and Operations Management by S. Hoogendoorn 15
5
S
IMULATIONThis section explains how the simulation model was built and how the different solution scenarios were composed. Next, different validation approaches are discussed, including the experimental design and output measures.
5.1 MODEL DESIGN
The simulation model is focused on the practical situation, as described in the previous section. Since some aspects of the system are beyond the scope of this research and have no influence of the sorting performance, the model has been simplified for practical situations. Therefore, a conceptual model was created where the scope and detail levels were justified, as shown in Appendix II.
The simulation model can be split into two parts: the conveyor loop in combination with the loading station and the chutes combined with the packing station. In the first part (see Figure 5.1), batches of parcels were created and loaded in the system, where each arriving parcel is assigned a set of attributes related to the packing schedule, including the number of circulations, chute assignment, container location (three levels), alternative size, and packing policy, with items facing a first decision moment before the circular conveyor. From here, the parcels can be placed on the circular conveyor when there is capacity available; if not, they are placed in a buffer. Data about each parcel are recorded for every single item on the circular conveyor.
Parcels loading
Parcels loading Circular conveyorCircular conveyor
Capacity available on sorting conveyor Capacity available on sorting conveyor buffer buffer No Yes Container packing schedule Container packing schedule Data colletion: - Item circulations - WIP Data colletion: - Item circulations - WIP
Figure 5.1 – Flowchart of decision process at the loading site of the automatic sorting system.
Master-thesis Technology and Operations Management by S. Hoogendoorn 16 available on the circular conveyor; if not, an item is directly assigned to a chute on to preserve the system from blocking. Secondly, there should be capacity available in the chute to buffer. Thirdly, the location of an item should be available at the packing station, which is threefold, for example, “position 1.2.6”. How an item position number is constructed is shown in Appendix IV. Therefore, the information about the attributes for the location in the conveyor has to be read by the system. If an item meets all location conditions, it is assigned to a resource and leaves the system. Data are recorded about the container filling. A full representation of the flowchart is shown in Appendix III.
Chute .. (buffer) Chute .. (buffer) Capacity available in buffer Capacity available in buffer Parcel destination of this chute Parcel destination of this chute Destined container available at station Destined container available at station
Layer and item available for
destination Layer and item
available for destination Circulair conveyor Pack on container Pack on container Yes Yes Yes Yes
Container leaves the system Container leaves the
system WIP on conveyor loop below capacity constraint WIP on conveyor loop below capacity constraint Yes Parcel destination of this chute Parcel destination of this chute Capacity available in buffer Capacity available in buffer No Yes Yes Data colection: -Number of parcels -Placed accourding schedure or not. -Cycletime of parcel. Data colection: -Number of parcels -Placed accourding schedure or not. -Cycletime of parcel.
Master-thesis Technology and Operations Management by S. Hoogendoorn 17 When the system starts loading parcels on the circular conveyor, the first item intended for the first chute is released and the parameter is set for the first container loaded at chute 1. For example, when a parcel destined for chute 1 contains location number 8.2.3, the first container to be loaded is container 8. The same goes for the second container at chute 1 unless the container number is different from the first container since two containers are located at station 2.
The probability of selecting a parcel from the random distribution which meet the conditions in the base scenario is larger compared to the proposed situation. Therefore, it is expected that the circular conveyor will reach the capacity at some moment in time and therefore resulting in a higher utilization. As explained in the theory section, the WIP of the system needs to be controlled to stop the system from blocking (Bozer and Hsieh, 2005). To be able to control the WIP, an extra condition is added to the decision statement, which only allows sequential filling when capacity utilization of the circular conveyor is below a threshold value of 80%. Since it is known from the practical situation that the one-level priority rule will result in a stable system, terminating the scheduled packing will return the system in a stable position again, called “overload status”. As a consequence of the terminated schedule, the next parcel which meet and intended for the chute is packed on the container. The packing of this overload item results in a decreased capacity of the container of a random amount of location positions between 1 and 3. For example, one roll container can be packed with 12 sequential packed items and 1 non-sequential packed item with a size of three location positions. The container will therefore be packed with a maximum amount of 13 items (Figure 5.3 A and B) instead of a fully utilized container of 15 items. Until the system is stabilized, the FCFS policy is applied. 1.3.3 1.3.4 1.3.5 1.2.5 2.1.4 1.2.5 1.3.1 1.3.2 1.2.2 1.2.3 1.2.4 1.2.2 1.2.3 1.2.4 8.2.3 1.1.4 1.1.5 1.2.1 1.1.4 1.1.5 1.2.1 1.1.1 1.1.2 1.1.3 1.1.1 1.1.2 1.1.3
Master-thesis Technology and Operations Management by S. Hoogendoorn 18
5.2 ASSUMPTIONS/SIMPLIFICATIONS.
Simulating the model with the parameters of the real system would result in too complex a model with a too long a run length. Therefore, a scaled version of the real situation was created. To preserve the dynamics of the process, every parameter was factorized as shown in Table 5.1.
Real system Simulation model Capacity (N items) 380 100
Conveyor speed (m/s) 2.4 2.4
Average throughput rate (items/min) 90 18
Average container capacity (items) 30 15
Container packed at one station simultaneously
5 2
Chutes 48 10
Load stations 7 1
Buffer capacity 30 15
Mean batch sizes 13.7 6
Table 5.1 - Scale parameters for simulation model.
Some additional assumptions and simplifications for the model have been made with respect to the practical problem situations due to uncertainties or restrictions of the available data. The following list contains the assumptions and simplifications made for the model:
Dimensions of the parcels are not taken into account, but an parcel is seen as an item.
One item takes one location, and a container has the capacity of 15 locations.
All the characteristics of the parcels are known upfront with respect to the final
destination as well as to the destination on the roll container. This will make it possible to create a container packing scheme in advance before sorting.
All parcels entering the system fit within the system boundaries. The current sorting
systems have a chute where items that does not fit within the system boundaries or destinations are unknown will be directed to the chute. These are not taken into account since this does not influence the container packing in this simulation design.
All containers and parcels have a single destination, which means that one parcel is
only assigned to one container.
There are always containers available for packing at the chute station; no change over
Master-thesis Technology and Operations Management by S. Hoogendoorn 19
The order of placing items on the layer does not have an impact on the sorting process,
so it makes no difference if parcels are packed in sequence 1234 or 4213, and so on.
The sequence of packing the layers in containers is randomly based on arriving parcels
at the circular conveyor. For example, item 1 for layer 2 can be packed first, and then parcel 5 for layer 1 can be packed.
Containers at a chute can be packed randomly, so container 5 can be packed first
where after container 1 can be packed.
A constant conveyor speed is assumed during the sorting cycle.
5.3 EXPERIMENTAL DESIGN
The simulation is outlined into two different scenarios. Scenario 1 is based on the current sorting process as used at the case company, called the base case. Scenario 2 is the scenario where a second level of priority rule is added to the decision procedure, named the proposed scenario.
5.3.1 Scenario 1 (Base case)
In the current system, the sorting process is based on one selection rule for the output, related to the chute assignment. This base situation was used as a reference to validate the model and parameters to the practical settings (Table 4.1). Parcels arrived with a constant arrival rate but a normal distributed batch size with a mean of six items. As can be seen from Table 5.2, the conveyor capacity was set on 100 items, where a throughput level of 2,000 items was tested, as this volume is close to the practical throughput of a sorting cycle. A second experiment tested the capacity of the system to be able to determine the utilization, creating an overload situation of arriving parcels that gave the maximum throughput rate.
Throughput level (demand)
Interarrival rate (sec.)
Arrival rate Batch size Conveyor speed (Items/sec)
Experiment 1 2,000 18 Constant 6 2.4
Experiment 2 2,000 3 Constant 6 2.4
Table 5.2 - Experimental settings of the base case.
5.3.2 Scenario 2 (Proposed scenario)
Master-thesis Technology and Operations Management by S. Hoogendoorn 20 proposed scenario. In order to determine the effect of the proposed scenario of sequential container packing on system performance, a sensitivity analysis was conducted. The function for a sensitivity analysis is to test the robustness of the proposed results (Robinson, 2014, p. 67).
Throughput level (demand)
Interarrival rate (sec.)
Variance Conveyor speed (m/s) Experiment 1 500 18 2.775 2.4 Experiment 2 1000 18 2.775 2.4 Experiment 3 2000 18 2.775 2.4 Experiment 4 500 18 2.775 3 Experiment 5 500 18 2.775 1.5 Experiment 6 500 18 none 2.4 Experiment 7 500 10 2.775 2.4
Table 5.3 - Experimental settings of proposed scenario.
The sensitivity analysis was conducted in the following manner. Resulting from the literature review, four factors deemed important for the performance of the system were tested, including throughput level (demand), interarrival rate, variance of batch sizes, and conveyor speed. A full experiment was completed, where each factor was tested for multiple parameter levels in different experimental settings.
The capacity of a system is equal to the maximum throughput of a system (Hopp and Spearman, 2011). The maximum throughput in a conveyor system with rs induction stations,
or the number of arriving stations, is represented by the following formula provided by Johnson et al. (2002):
(2𝑟𝑠
𝑟𝑠+1) 𝑠 (2)
Master-thesis Technology and Operations Management by S. Hoogendoorn 21 according to the work of Johnson and Meller (2002), who did experiments with different normal distributions in a system with similar kinds of settings.
Conveyor speed is also a way to influence the throughput of a system, but the acceleration of the conveyor is limited due to practical issues. Fedtke and Boysen (2017) did some practical research toward conveyor speed levels and came to a range varying from 2–3 m/s. Therefore, the maximum speed level was set to 3 m/s in the experiments. Moreover, the case situation used a speed level of 2.4 m/s and was therefore set for the second speed level. A third speed level, 1.5 m/s, was set below the practical situation to see what the effects of a low conveyor speed would be.
The throughput levels were empirically determined until differences in output appeared. The starting point was the practical throughput level during a sorting cycle, which included 2,000 items, relative to the simulation model. Several levels down to just above the system capacity were chosen—2,000, 1,000, and 500 items, respectively.
5.4 EXPERIMENTAL SETTINGS
The simulation model designed can be classified as a terminating simulation, meaning that there is a natural end point that determines the length of a run (Robinson, 2014, p. 168). The end point of the simulation model can be seen as the completion of the input of the packing schedule. A general problem in the analyses of output from terminating simulation models is the need to determine the number of replications needed to construct a confidence interval from the simulation (Itami et al., 2005). The number of replications was determined for every experimental factor, and it was found that the WIP level over time was leading by determining the replication numbers. A replication number of 7 should be enough for a stable representation. Appendix V shows the outcome of the measurement.
Master-thesis Technology and Operations Management by S. Hoogendoorn 22
5.5 OUTPUT ANALYSIS
To validate the base case, the output measures used were throughput and the cycle time of the experiment. Using the output of a capacitated run and of a run of the practical situation produced data for determining the utilization of the simulation model. The utilization of the total system was determined from the base case and compared to the results from analysis of the real system.
The output of the simulation of the proposed scenario was fivefold: capacity utilization of a container, distribution of the percentage of parcels filled according the two-level sorting strategy, average work in progress (WIP) over time, total cycle time (CT), and distribution of item circulations over time.
5.6 VERIFICATION AND VALIDATION
This research was approached with a qualitative research design which has been selected on the base of the objective for the research as explained in section 4.1. For a valid and thorough analysis of the problem situation, the results of the simulation models were compared to the actual data received from the case company PostNL. The comparison of simulation data with real world data is called black-box validation, which ensures that the model represents the actual data with enough accuracy (Robinson, 2014).
Master-thesis Technology and Operations Management by S. Hoogendoorn 23
6 N
UMERICAL RESULTSThe result section is divided in two main sections. In the first section numerical results of the base case experiments are compared and analyzed, where the second section is dedicated to the numerical results of the proposed scenario.
6.1 BASE CASE RESULTS
To validate the simulation model, results of this model are compared to the real situation. A first experiment has been executed with only output assignment on the conveyor based on the final destination of an item. Parameters are applied as described in section 5.3 for the base case. Table 6.1 shows the outcomes of the first two experiments. In the second column the cycle time is presented, this value is used to determine the utilization of the simulation model. The utilization of the simulation model is shown in the last column, 0.76 respectively. From the data analysis, an utilization ranging between 0.69 and 0.74 has been determined which is slightly lower than the simulation model. This difference can be explained by the assumptions and simplifications which have been made concerning the simulation model, for example, the worker speed is assumed to be constant.
In addition to this results, two other validation parameters are measured to ensure the model does not take into account the sequential filling decisions. The first validation variable, container filling utilization in the third column, explains to what extend the container is filled compared to its capacity. A filled container in this situation means that a random item is assigned to a location (not taking into account dimensions) and only considering the final destination. So every random parcel intended for the particular chute will fit in the roll container, therefore, a hundred percent filling could be achieved. The second validation variable, percentage of sequential filling, measures if a parcels is assigned according a one level priority or a two level priority decision. Results validate the full use of a single level decision process.
Experimental settings Cycle time (hours) Container filling utilization
Percentage of sequential filling
Utilization
Experiment 1 (Maximum throughput level) 3,69 100% 0%
Experiment 2 (Practical parameters) 2,84 100% 0% 0,76
Data analysis (Real situation) 0,69-0,74
Master-thesis Technology and Operations Management by S. Hoogendoorn 24 Not only the cycle time of the sorting process is analyzed for the base case, the number of circulations for every items on the conveyor belt is also measured as indicator for the cycle time per item (Figure 6.1). Results shows that every parcel will be assigned to a container within the first round, implying that the buffering function of the conveyor is not used. Taking into account this results it could be assumed that the input rate is lower than the output rate of the system.
Figure 6.1 - Average number of item circulations during base case.
Figure 6.2 - Average WIP level of cirular conveyor base case.
The final experimental factor in this scenario, WIP, is shown as a representation over time to illustrate the system behavior in Figure 6.2. The graphs clearly shows a constant level of WIP during the process cycle, which also confirms that no items are buffered in the system during processing.
6.2 SCENARIO 2
6.2.1 Results work in progress levels
The analysis of work in progress over time shows that these levels behave the same over time for all the experimental settings in the proposed scenario, therefore only one experimental comparison of the WIP level over time from the base scenario and the proposed scenario (experiment 3) is analyzed further in detail in this section. Figure 6.3 shows clearly a major difference in average WIP levels for both scenarios. In the base scenario there is a very low level of WIP during the whole process cycle, 9 items on average. This low level is an indication that the circular conveyor is utilized on a very low level, around 10 percent. Since, the utilization of the complete system was calculated on 0.76 in section 6.1, the circular conveyor can be seen as the none bottleneck in whole sorting system. However, the results of
0 1 0 1 2 N u m b er o f circu lat ion s Time (hours)
Average number of circulations over time 0 5 10 15 20 0 1 2 N u m b er o f it em s (N) Time (hours)
Work in progress (WIP)
Master-thesis Technology and Operations Management by S. Hoogendoorn 25 the proposed scenario shows that the level of WIP increase immediately to a level around 80 items, which was set as capacity constraint for scheduled packing. Since the circular conveyor meets is capacity, the circular conveyor can be seen as the bottleneck in the system in this scenario which is the most interesting result from this analysis. Another important characteristic to analysis the stage around 1.8 hours where the WIP decrease from capacity level to zero at 3 hours. At this stage the scheduled packing policy is terminated and packing according the base scenario is further applied to empty the system.
Figure 6.3 - Comparison of WIP levels from base case and proposed scenario.
6.2.2 Effects of throughput changes
The experimental factor, throughput, was tested for three levels ranging from 500 up to 2000 items during a process cycle in a comprising analysis of experiment 3 ,4, and 5. The results of output measures; capacity utilization of a container, sequential filling, total processing time are presented in (Figure 6.4). The graph shows an increase in throughput results in an increase in processing time, where Figure 6.5 shows that this growth of processing time is almost linear. As the consistency in the ability of the sorting system to fill containers according the predefined schedule is the most important outcome of the model, these outcomes are outlined against the total processing time and total container utilization. Figure 6.4 also shows that by a low throughput level of 500 items the system is able to fill container according its predefined schedule by 38 percent. When this throughput level has been increased, this sequential filling level decreased to 15 percent, which is an decrease of 22 percent in performance compared to a low throughput level. The bars of the average utilization of the containers, indicate to which extend the model was able to fill containers to the capacity. When the throughput is 500 items a container is filled for more than 70 percent of its capacity but when the throughput is
0 20 40 60 80 100 0 0,5 1 1,5 2 2,5 3 3,5 4 N u m b er o f it em s (N) Time (min.)
Work in progress over time (WIP)
Master-thesis Technology and Operations Management by S. Hoogendoorn 26 increased the utilization decreases down to 58 percent. A representation over time for the container utilization is shown in Appendix VI.
Figure 6.4 - Results simulation for experimental factor Throughput.
Figure 6.5 - Processing time by increasing throughput.
The second outcome measure in this scenario, the number of item circulations on the conveyor, used as an indicator for the processing time of every individual item. A squatter plot of the item circulations are shown in Figure 6.6 for the three different throughput levels. The graph shows that the majority of the items stays within 0 and 15 circulations but when the throughput is increased the upper bound increases up to 30 circulations per item. To be able to compare these results a trend line is drawn for the different throughput levels, resulting in an average value which is the same for all the three levels. The level of throughput will have no effect on the average circulations but does effect the upper and lower bound.
throughput 500 thoughput 1000 throughput 2000
Capacity utilization container 0,74 0,58 0,55
According predefined schedule 0,35 0,15 0,15
Total processing time 0,85 1,64 3,02
0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 Ti m e (hour s) U ti liz at ion
Container utilization and processing time
Master-thesis Technology and Operations Management by S. Hoogendoorn 27
Figure 6.6 - Item circulations on circular conveyor experimental factor throughput.
6.2.3 Effects of conveyor-speed changes
The second experimental factor, conveyor speed, has been tested for three levels varying from 1.5 up to 3 m/s. The results of experiments 3, 5, and 6. A throughput of 500 items was used in this experiments. Results of the total processing time, filling according schedule and container utilization are presented in Figure 6.7. The results shows barely no effect in the utilization of the container packing, however, performances of the ability of the system to pack items according the predefined schedule declined by 9 percent. Even the processing times remain more or less constant with a varying conveyor-speed.
Figure 6.7 - Results simulation output for experimental factor Speed (by throughput 500). 0 5 10 15 20 25 30 35 0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 N u m b er o f circu lat ion s Time (hours)
Scatter plot of items circulations over time
Throughput 500 Throughput 1000 Throughput 2000
Poly. (Throughput 500) Poly. (Throughput 1000) Poly. (Throughput 2000)
speed 1,5 m/s
speed 2,4
m/s speed 3 m/s
Capacity utilization container 0,66 0,67 0,68
According predefined schedule 0,40 0,31 0,31
Total processing time 0,82 0,85 0,84
0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00 Ti m e (hour s) U ti liz at ion
Master-thesis Technology and Operations Management by S. Hoogendoorn 28 The item circulations for the different conveyor speed levels are presented in Figure 6.8. Interestingly is that the average circulations over time will not change when the speed of the conveyor is increased but when the speed is lowered a decrease of item circulations is shown. This result implies that the conveyor is the bottleneck in the system by a high speed.
However when the speed is lowered the average circulations of items has decreased.
Figure 6.8 - Item circulations on circular conveyor experimental factor Speed.
Since there was an effect in scheduled packing of varying speed by a throughput of 500 items, Another experiment has been executed by a throughput of 2000 items to identify if these patterns are present by the practical throughput, shown in Figure 6.9. An analyzation of the graph shows no that the packing performance decrease to a level around 10 percent and remain constant. The pattern of processing time remain constant which is equal to the patterns of a throughput of 500 items.
Figure 6.9 - Results simulation output for experimental factor Speed (by throughput 2000). 0 5 10 15 20 25 30 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 N u m b er o f circu lat ion s Time (hours)
Scatter plot of items circulations over time
Speed 2.4 m/s Speed 3 m/s Speed 1,5 m/s
Poly. (Speed 2.4 m/s) Poly. (Speed 3 m/s) Poly. (Speed 1,5 m/s)
speed 1,5 m/s speed 2,4 m/s speed 3 m/s
Capacity utilization container 0,55 0,55 0,55
According predefined schedule 0,09 0,10 0,10
Total processing time 2,99 3,02 3,06
0,00 0,50 1,00 1,50 2,00 2,50 3,00 3,50 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00 Ti m e (ho ur s) Ut ili zat io n
Master-thesis Technology and Operations Management by S. Hoogendoorn 29
6.2.4 Effects of variance changes
Another experimental factor which was tested during the sensitivity analysis is the variance of the batch sizes. In this analyzes, experiments 3 and 8 has been compared. In the first experiment there was a high variation in batch sizes (2.775) where in the second experiment no variation in batches was applied. Figure 6.10 shows the results of the outcomes of the first three output measures. Interesting is that the variance shows no effect in one of the three output measures.
Figure 6.10 - Results simulation output for experimental factor batch variance (by throughput 500).
The number of circulations for both variance analyses are presented in Figure 6.11, showing that the average levels for both experiments show no effects similar to the outcomes in the other output measures of variance.
Figure 6.11 - Item circulations on circular conveyor experimental factor batch variance.
Variance 2,775 Constant arrival single items
According predefined schedule 0,68 0,68
Capacity utilization container 0,31 0,31
Total processing time 0,82 0,81
0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00 Ti m e (hour s) U ti liz at ion
Container utilization and processing time
0 5 10 15 20 25 30 35 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 N u m b er o f circu lat ion s Time (hours)
Scatter plot of items circulations over time
Master-thesis Technology and Operations Management by S. Hoogendoorn 30
6.2.5 Effects of interarrival time changes
The fourth experimental factor tested, was the interarrival time between each arriving batch in the conveyor system. In this analysis experiment 3 and 8 has been tested where the outcomes are presented in Figure 6.12. A interarrival time of 10 seconds and 18 seconds of batches has been compared. The results show that, logically, when the interarrival time has been increased the total processing time increased as well. However, remarkable is that performance
scheduled container packing declined.
Figure 6.12 - Results simulation output for experimental factor batch interarrival time (by throughput 500).
Figure 6.13 shows the number of circulations for both experiments compared to each other. An obvious result from this analysis is that the average number of circulations decrease when the interatrial time is lowered. This effect can be explained by the fact that when a bottleneck element in the system is fed by a lower input rate but the output rate remained the same, items must stay longer in the process.
Figure 6.13 - Results simulation output for experimental factor batch interarrival time.
Interarrival time 10 sec Interarrival time 18 sec
Capacity utilization container 0,68 0,67
According predefined schedule 0,30 0,23
Total processing time 0,52 0,85
0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 0,00 0,10 0,20 0,30 0,40 0,50 0,60 0,70 0,80 0,90 1,00 Ti m e (hour s) U ti liz at ion
Container utiliztion and processing time
0 10 20 30 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 N u m b er o f circu lat ion s Time (hours)
Scatter plot of items circulations over time
Interarival time 10 sec. Interarrivaltime 18 sec.
Master-thesis Technology and Operations Management by S. Hoogendoorn 31
7 D
ISCUSSIONIn this thesis, it has been evaluated how the performance of an automatic sorting system changed when a buffering function was added to the circular conveyor in an automatic sorting system. In a simulation model, an first attempt was made to catch the changes of the behavior in fundamental parts of system performances. Outcomes in the study are sometimes mixed and does not always result in the desired improvements but some useful insights could be derived from the proposed system for further research.
7.1 THEORETICAL/MANAGERIAL IMPLICATION
Whereas the literature indicates that controlling the capacity of the automatic sorting system is fundamental to the performance of the system, this report shows to which degree adding constraints to output selection of a circular effect the output performance of the entire system. By systematically examine the aspects of the sorting system a number of lessons could be learned by analyzing the simulation in this research.
In a first analysis of WIP level of the circular conveyor over time shows that machine as non-bottleneck machine in the base scenario has been shifted to the non-bottleneck machine in the scenario where the buffering aspect is applied to the system. This results is in conformity with the results of (Xue and Proth, 1987) who did a study towards short term buffering function during in a manufacturing setting and found the same result of increased WIP.
In a sensitivity analysis, it was aimed to getting more insights into changed performance of fundamental parts in system behavior when the bottleneck has shifted from the output area to the circular conveyor. First, by testing increasing volume levels during a process cycle and keeping arrival rates constant. Results has shown a linear increase in processing time, however, the performance of packing area declined in terms of the ability to pack containers according the predefined schedule. Evaluating this result brought us the insight that, this decreased performance is the effect of the sequential packing constraint. When the WIP level of the circular conveyor exceeds the system capacity the scheduled packing was eliminated. However, the longer the system exceed its capacity the worse the performance of the container packing.
Master-thesis Technology and Operations Management by S. Hoogendoorn 32 in the least circulations of items on the conveyor. However, when the conveyor speed has been increased, used as a trigger to increase the throughput tare, the average cycle time per parcel has increased. Increasing the speed even more brings the system to a state where an upper bound level have been reached where packing performance plunged to its lowest performance. This effect explains itself by the fact that the balance between capacity and input of the circular conveyor is lowered, resulting in less available WIP which the buffering function of the circular conveyor hinders. This results are in agreement with the study of (Johnson, 1998) who proved that the throughput rate of a system is hindered by conveyor speed. However, remarkable is that the performance of scheduled packing results in opposite results. Where a lower conveyor speed will result in better performances of scheduled container packing compared to a higher speed. Moreover, when the speed is increased, performance of scheduled container packing will decline and stabilize to a constant performance level. We would explain this effect by being the result of an overload on the circular by which the buffering function is terminated and only transportation of items to packing stations based on final destination. More close examination of this effect needs to confirm this thought.
Master-thesis Technology and Operations Management by S. Hoogendoorn 33 who appoint that an increase in system capacity does not lead to an increased system performance.
7.2 PRACTICAL IMPLICATION
As discussed in the previous section, performance gives the best results for lower volume throughput during a processes, up to five times of the system capacity by maintaining current arrival rates. This makes the solution applicable for system such as automatic warehouses of distribution centers where more homogeneity in items is present. However, in PSCs’ it is common to maximize the utilization of the system during a sorting cycle in order to minimize the uses of resources such as labor and energy. This result throughput rates which are maximized. The current model is not able to deal with this high rates and therefore implementation in a PSC environment needs more optimization.
Throughout all experiments, performance of scheduled container packing are shown to be very sensitive to changes of system parameters. Mainly, due to the fact that applying the buffering function in the circular conveyor would increase the WIP level to a status where it is continuously balancing just below or above system capacity. Since the a simulation model is a simplified version, a practical situation would be more dynamic and therefore more sensitive for disruptions. Since PSCs’ maintains high service levels, disruptions in the sorting process would affect these service levels significantly. The potential cost savings due to an improvement in transportation efficiency would probably not outweigh the costs caused by process disruptions.
7.3
LIMITATIONSThere are several limitations to this study. First, due to the fact that the area of this research is not well covered in literature since the industry is booming in the last few years. Moreover, insights from other areas are hard to implement because systems differ significant. Second, the information provided by the case company was very limited since their system contains lots of aspect which are patented and therefore kept secret. For generalizability we choose to not make the research confidential, as a result some fundamental aspect of the system were approximated. Also due to the lack of information no comparison could be made of packing performance in the current setting which affect the validation of the study.
Master-thesis Technology and Operations Management by S. Hoogendoorn 34 packing. This result in a lot of chaos in the packing process, for example, placing an item in the transportation unit where it was original not scheduled for, destroyed the schedule for another transportation unit too. When the system is able to adjust this packing schedule while processing, performance could have been increased. Additionally, the packing schedule assumes one best solution for sequencing items on the conveyor, where in a practical solution multiple combinations will lead to more valid solutions. Even the 3d-bin packing model gives a heuristic solution in where multiple solutions are possible as feasible solutions. Since many items will have the same or almost the same size, they can be placed in more than one transportation provided that their final destination is the same. Therefore when one item could have multiple options for placing in the container performance of scheduled packing could improve more.
Master-thesis Technology and Operations Management by S. Hoogendoorn 35
8 C
ONCLUSIONIn this research, a first attempt is made towards efficiency improvement of an automatic sorting system in a PCS setting in order to manage the utilization of transportation units used. As a starting point, we formulated following the research question:
How sortation strategies related to system output affect the system performance of an automated sorting system?
In order to answer the research question, three sub questions where created: to get insight in the current system; to identify important performance indicators to analyze, to identify the bottleneck in the system. An literature study identified important parameters, such as conveyor speed, system throughput and arrival rate. A simulation model has been created test these dynamic elements of speed, throughput and capacity at different stages in a typical automatic sorting process. Results show a potential opportunity to apply a packing schedule as input for an automatic sorting process by low throughput rates, however, increasing the throughput show that the dynamics in the system has been increased dramatically and limit performance due to capacity of the system.
Master-thesis Technology and Operations Management by S. Hoogendoorn 36
9 R
EFERENCESAllaoui, H. and Artiba, A. (2004) ‘Integrating simulation and optimization to schedule a hybrid flow shop with maintenance constraints *’, Computer and Idustrial Engineering, 47(4), pp. 431–450. Bangsow, S. (2010) Manufacturing simulation with plant simulation and simtalk: Usage and
programming with examples and solutions. Springer Science & Business Media.
Bastani, A. S. (1988) ‘Analytical solution of closed-loop conveyor system with discrete and deterministic material flow’, European Journal of Operational Research, 35(2), pp. 187–192.
Battini, D., Faccio, M., Persona, A. and Sgarbossa, F. (2009) ‘Balancing – sequencing procedure for a mixed model assembly system in case of finite buffer capacity’, International Journal of Advanced
Manufacturing Technology, 44(3), pp. 345–359.
Bozer, Y. A. and Hsieh, Y. (2004) ‘Expected waiting times at loading stations in discrete-space closed-loop conveyors’, European Journal of Operational Research, 155(2), pp. 516–532.
Bozer, Y. A. and Hsieh, Y. (2005) ‘Throughput performance analysis and machine layout for discrete-space closed-loop conveyors’, IIE Transactions, 37(1), pp. 77–89.
Croucher, P. R. A. and Baker, P. (2014) The handbook of logistics and distribution managment:
Understanding the supply chain. Edited by K. P. Publischers. Kogan Page Publishers.
Daganzo, C. F., Dowling, R. G. and Hall, R. W. (1983) ‘Railroad clasification througput: The case of multistage traingular sorting.’, Transportation Research, 17(2), pp. 95–106.
Ecommerce-News (2016) Ecommerce in Europe to reach €509.9 billion in 2016, 31 May, 2016. Fedtke, S. and Boysen, N. (2017) ‘Layout Planning of Sortation Conveyors in Parcel Distribution Centers Layout Planning of Sortation Conveyors in Parcel Distribution Centers’, Transportation
Science, 51(1), pp. 3–18.
Han, Y.-H., Zhou, C., Bras, B., McGinnis, L., Carmichael, C. and Newcomb, P. J. (2003) ‘Paint line color change reduction in automobile assembly through simulation’, Proceedings of the 2003 Winter
Simulation Conference, 2003., 2(August 2014), pp. 1204–1209.
Haneyah, S., Zijm, H., Schutten, M. and Schuur, P. (2011) ‘Generic Planning and Control of
Automated Material Handling Systems : Practical Requirements Versus Existing Theory’, Computers
in Industry, 64(3), pp. 177–190.
Hopp, W. J. and Spearman, M. L. (2011) Factory Physics Foundations of Manufacturing Managment. Third Edit. Waveland Press, Inc., United States of Amerika.
Itami, R. M., Zell, D., Grigel, F. and Gimblett, R. (2005) ‘Generating confidence intervals for spatial simulations – determining the number of replications for spatial terminating simulations’, In
International Congress on Modelling and Simulation: Advances and Applications for Management and Decision Making, (MODSIM05).
Jayaraman, A., Naryanaswamy, R. and Gunal, A., K. (1997) ‘A sortation system model. In Prodeedings of the 29th conference on Winter simulation’, IEEE Computer Society, pp. 866–871. Johnson, M. E. (1998) ‘The impact of sorting strategies on automated sortation system performance’,