- Email: [email protected]
Lean production system design for fishing net manufacturing using lean principles and simulation optimization
Journal of Manufacturing Systems 34 (2015) 66–73
Contents lists available at ScienceDirect
Journal of Manufacturing Systems journal homepage: www.elsevier.com/locate/jmansys
Technical Paper
Lean production system design for fishing net manufacturing using lean principles and simulation optimization Taho Yang a , Yiyo Kuo b,∗ , Chao-Ton Su c , Chia-Lin Hou a a b c
Institute of Manufacturing Information and Systems, National Cheng Kung University, Tainan 701, Taiwan Department of Industrial Engineering and Management, Ming Chi University of Technology, New Taipei City 243, Taiwan Department of Industrial Engineering and Engineering Management, National Tsing Hua University, Hsinchu 300, Taiwan
a r t i c l e
i n f o
Article history: Received 13 September 2012 Received in revised form 1 October 2014 Accepted 24 November 2014 Keywords: Lean manufacturing Simulation Value stream mapping Fishing nets
a b s t r a c t Value stream mapping (VSM) is a useful tool for describing the manufacturing state, especially for distinguishing between those activities that add value and those that do not. It can help in eliminating non-value activities and reducing the work in process (WIP) and thereby increase the service level. This research follows the guidelines for designing future state VSM. These guidelines consist of five factors which can be changed simply, without any investment. These five factors are (1) production unit; (2) pacemaker process; (3) number of batches; (4) production sequence; and (5) supermarket size. The five factors are applied to a fishing net manufacturing system. Using experimental design and a simulation optimizing tool, the five factors are optimized. The results show that the future state maps can increase service level and reduce WIP by at least 29.41% and 33.92% respectively. For the present study, the lean principles are innovatively adopted in solving a fishing net manufacturing system which is not a welladdressed problem in literature. In light of the promising empirical results, the proposed methodologies are also readily applicable to similar industries. © 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
1. Introduction The common challenges that companies face are market competition, increased pressure on inventory, increased service levels, and reduced work in process (WIP). Lean manufacturing is one of the approaches that can help companies respond appropriately to these challenges [11,18]. The focus of the approach is on cost reduction, by eliminating activities that do not add value by linking and balancing work stages, so that products from one stage are consumed directly by the next stage, until the end of the production line is reached [4,5,7,18]. Applications of lean manufacturing have spanned many sectors, including the automotive industry, electronics, white goods and consumer goods [1,21]. However, there is no evidence of work applying lean manufacturing principles to fishing net production. Fishing net manufacturing is a high make-to-order (MTO) environment, because net size and type change according to the ocean environment, fish kinds and ship size.
∗ Corresponding author. Tel.: +886 2 29089899x3118; fax: +886 6 29085900. E-mail addresses: [email protected] (T. Yang), [email protected] (Y. Kuo), [email protected] (C.-T. Su), [email protected] (C.-L. Hou).
MTO manufacturing need not be a pull production system—a lean design principle. Usually, there are many (or some) simultaneous orders in the manufacturing process. These orders may be different in size and specifications. Thus, there is no guarantee that the system will follow a first-in-first-out rule. In fact, it is often quite messy when production is loaded with work-in-process. Thus, it may be effective to adopt lean principles to regulate the flow and to control the work-in-process level. Higher WIP and longer cycle times always result in a lower service level and produce lower customer satisfaction in MTO environments. Moreover the fishing net manufacturing system is distinctive. Through the production procedure the production units become bigger and bigger and the processing time in most steps are quite long (between 3 and 10 days). This provides the motivation for this research, to design a lean manufacturing system for this industry. Value stream mapping (VSM) is a visual tool that facilitates the process of lean production by helping to identify the value-adding steps and eliminating the non-value adding waste [22]. In recent years, VSM has emerged as the superior method to implement lean production in factories, and has been used to identify where waste occurs [3,12,14,15,17,24,25,28]. VSM creates a pictorial representation and common language for the production process, enabling
http://dx.doi.org/10.1016/j.jmsy.2014.11.010 0278-6125/© 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
more purposeful decisions to improve processes. Numerous applications can be found in the literature. For example, Seth and Gupta [22] applied VSM to establish lean production in an Indian twowheeler motor company. Lummus et al. [16] report on a VSM project in a small medical clinic. The new system can increase patient throughput and reduce patient waiting time. Abdulmalek and Rajgopal [1] describe a steel mill case, and apply VSM to identify the opportunities for improvement. Lian and Landeghem [14] propose a generator which automatically yields a simulation model for VSM. The current and future state VSM of a poultry and pig rearing equipment manufacturing system is introduced to demonstrate the proposed generator. Barber and Tietje [2] demonstrate the use of VSM to achieve both greater efficiency and value creation in a sales process. Seth et al. [23] identify and address various sources of waste in the supply chain of an edible cottonseed oil industrial process and use a VSM approach to improve productivity and capacity utilization in an Indian context. Lasa et al. [12,13] redesigned six production systems with the VSM technique, and then analyze the causes for the limited adoption of lean manufacturing concepts. The six production systems are manufacturing systems for kit furniture, water heaters, forging, detonator systems, mechanized and stamped parts, and thermoplastic parts. Rother and Shook [20] proposed seven guidelines based on the concept of lean manufacturing to construct the future VSM which can be implemented in a reasonably short time period without any major investment. According to the seven guidelines, there are several factors that should be considered. Different combinations of the factors would result in different production performance. However, the implementation of the recommendations is likely to be risky. Simulation is a useful tool for evaluating the performance of a new design but it cannot provide the optimal design. Combining simulation with experimental design or intelligent search has been successfully adopted for simulation optimization. Yang et al. [27] use a commercial tool called OptQuest for optimizing an integrated-circuit (IC) packaging system. OptQuest embeds scatter search (SS) in a simulation tool Arena to optimize the simulation models. Yang et al. [26] combine simulation with genetic algorithm for optimizing dispatching rules in a flow shop with multiple processors. Kuo et al. [9] integrate simulation and the Taguchi method to optimize an integrated-circuit (IC) packaging system. The objective of the present study is to model a non-typical production system – a fishing-net manufacturing system – and to propose a lean production system design which is optimized by simulation optimization. The remaining sections of this paper are organized as follows. In Section 2 the lean principles are introduced, and a number of decision factors are highlighted. Then, the fishing-net production system is introduced and the current state VSM is provided in Section 3. Finally, a hybrid experimental design and intelligent search approach is adopted for optimizing the decision factors and then the future state VSM is constructed in Section 4. A summary of results and concluding remarks are presented in Section 5.
67
2.1. Guideline #1: produce to your takt time “Takt time” is used to synchronize the pace of production with the pace of sales. In general, it can be calculated by Eq. (1) Takt time =
available working time per day customer demand rate per day
(1)
However, when the process times of different products are quite different, the takt time calculated by Eq. (1) is not reasonable. For example, suppose there are two types of product, A and B, and the unit process time and demand rate of product A are 120 min and 4 units per day, respectively and the process time and demand rate of product B are 60 min and 8 unit per day. If the available working time is 18 h per day, then the takt time is 90 min (18 × 60/(4 + 8)). This means that one unit should be produced every 90 min, but this is impossible for product A. The present research uses Eqs. (2)–(4) for calculating the takt time. KSi =
The lowest common multiple of processing time for all products Pi
NKi =
Di KSi
Takt time =
(2) (3)
Available working time per day
N
i=1
NKi
(4)
In Eq. (2) KSi indicates the kanban size of product i (i = 1, 2, . . ., N), and Pi indicate the processing time of product i. In Eq. (3) NKi indicates the number of kanbans of product i, and Di indicates the demand for product i per day. Thus, for the same example, the kanban size for product A and B are 1 and 2 units respectively, and the number of kanbans for product A and B are both 4. The takt time is 2.25 h. This means that one unit of product A or two units of product B should be completed every 2.25 h. Each kanban indicates one unit of product A or two units of product B and four kanbans are required per day for both product A and B. It should be noted that the lowest common multiple of processing time for all products in Eq. (2) is an approximate value and should be as small as possible. Managers can increase the process time of some product types to tune it. For example, if there are three product types and their corresponding processing times are 12, 11 and 25 min, then the processing time of the first two product types can be increased to 12.5 min, so that the lowest common multiple of processing time will be 25 min. Corollary 1. The takt time is dependent on the size of a production unit. Corollary 1a. A kanban represent the same required production time regardless of the different product types. 2.2. Guideline #2: develop continuous flow where possible Continuous flow refers to producing one piece at a time to reduce the inventory of WIP and production CT. However continuous flow requires a great deal of creativity to achieve and sometimes it requires plant layout redistribution [12]. In this research, this guideline is not applicable and is not taken into consideration in the case study.
2. The lean principles Rother and Shook [20] propose a five phase implementation of VSM. The phases are (1) selection of product family; (2) current state mapping; (3) future state mapping; (4) definition of working plan; and (5) achievement of working plan. The lean techniques to be used are defined in the third phase which contains seven guidelines to define the future state map [12]. The seven guidelines are summarized below:
2.3. Guideline #3: use a supermarket to control production where continuous flow does not extend upstream A “supermarket” is nothing more than a buffer or storage area located at the end of the production process for products that are ready to be shipped [1]. When continuous flow is not possible and batching is necessary, a supermarket can smooth the whole manufacturing process. Supermarkets use a kanban system to fix the
68
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
inventory level. If the number of kanbans in a supermarket is too high, it causes higher inventory cost. On the other hand, the downstream production process will be subject to delays if the number of kanban in the supermarket is too low. Therefore, the number of kanbans in a supermarket is an importation factor when designing the future VSM. Corollary 2. A supermarket is required for the implementation of lean-pull system. 2.4. Guideline #4: try to send customer scheduling to only one production process In Guideline Number 1, for calculating the takt time, the customer demand rate is based on customer scheduling. The process time is set by one of the production processes. That process is called the “pacemaker process”. The pacemaker process synchronizes the pace of the entire manufacturing process and there are no supermarkets downstream of the pacemaker process. Therefore, selecting different workstations as the pacemaker process will have an important influence on the performance of the entire manufacturing system. Corollary 3. A pacemaker process is required for the implementation of lean-pull system. 2.5. Guideline #5: level the production mix Leveling the product mix means dividing the volume of all the product types based on their kanban size, and then producing them evenly over a time period. The more level the product mix is, the greater is the ability to respond to different customer requirements with a short lead time. For example, if there are three product types and the production sequence can be A-A-A-B-B-B-C-C-C or A-BC-A-B-C-A-B-C. The latter is more level and allows the process to proceed smoothly with smaller supermarkets, but it would cause more changeovers. Due to make-to-order (MTO) production environment, an order split to a certain number of production batches should be considered to achieve a leveling production. The use of a larger number of batches with fewer products in each batch results in a more level process. Therefore, the number of batches is also an importation factor in designing future VSM. Moreover, the production sequence also influences the performance of the manufacturing system. That means the managers have to decide between A-B-C-A-B-C-A-B-C and A-C-B-A-C-B-A-C in the example of the production sequence above. Corollary 4. An order splitting strategy is able to improve a leanpull system performance. Corollary 5. The production sequence decision could impact a lean-pull system performance. 2.6. Guideline #6: level the production volume
Fig. 1. The composition of a rope.
decreasing the production unit and so on. It has been addressed in Corollary 4. Lean management is well recognized as a management philosophy rather than a generic tool set. In other word, its principles are generic but its solution methodology is often case specific. For the present study, we adopt the lean principles from the existing literature as an exemplar to identify the five corollaries as the basis of our proposed methodology development which will be discussed in details in the following sections. 3. Fishing-net manufacturing and the current state map The case study was adopted from Hsieh et al. [8]. They introduced the fishing net manufacturing process in detail and then developed a hierarchical rough-cut capacity planning model and demand management system. While they focused on the strategic planning decision, the production control decision is not addressed. The application of lean principle to the fishing-net manufacturing is not found in literature. The solution to this specific application could be applicable to similar industries such as textile industries, variable product-size production, etc. Since the fishing-net manufacturing is rarely seen in literature, we discuss the production process forehead in details as follows. There are various fishing nets, such as gill nets, lift nets, drag nets, surrounding nets (purse seine nets), set nets (trap nets), covering nets, and fish breeding nets. Each net has a unique application which is based on the ocean environment, fish type and ship size. Therefore, the same customer could order nets of different types or different sizes. Each order requires different raw material types, twine sizes, mesh sizes, mesh depths and mesh lengths. The raw materials for fishing nets include nylon, trawl, nylon filament, nylon multi-filament and polyester. Twine size is determined by the composition of a rope as shown in Fig. 1 [8]. Raw material filaments are twisted into yarn, then twined yarns are made into strands. Strands are twisted into twine, and then braided twines into rope. Mesh size is a special unit of a net; different mesh depth and mesh length will result in a different shape and area of fishing net. The relationship between mesh size, mesh depth and mesh length is show in Fig. 2 [8].
Leveling the production volume is related to Guideline number 1. It means that production should be based on a fixed pace, the takt time. It has been addressed in Corollaries 1, 1a, and 4. 2.7. Guideline #7: develop the ability to make “every part every day” (then every shift, then every hour or pallet or pitch) in fabrication processes upstream of the pacemaker process This guideline is related to Guideline number 5. In order to divide the volume of all product types into more batches, some production methods have to be changed, such as reducing the changeover time,
Fig. 2. The mesh size.
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
69
Fig. 3. The current state map.
Net cage is one kind of fishing net which is the main product of the case study company. The main function of the net cage is to be fixed under the sea and cultivate salt water fish in it. According to the size and type of the cultivated fish, the corresponding specifications of net cages are also different. The general diameter varies from 10 m to 30 m, depth varies from 4 m to 10 m, while mesh varies from 2 mm to 30 mm. In the case study company there are six workstations for net cage manufacturing. They are (1) Twisting, (2) Braiding, (3) Net knitting, (4) Dyeing, (5) Heating and (6) Suturing. All manufactured net cages follow the sequence of workstation above, from 1 to 6. First the filament is twisted to form the yarn or strand as the twine in the Twisting workstation. Then the rope is braided from three twines in the Braiding workstation. The rope size is determined by the request of customers. The thicker the rope size, the thicker twine is required, and consequently more twisting jobs are required. In the Net knitting workstation the ropes are knitted to form nets in standard sizes. A knitting machine processes a net continuously. Theoretically, there is no limit to the length of a net, but in practice, it is a multiple of 10 m. There two options for the width of a net: 1 or 2 m. According to the specification of the knitting machines in the case company the standard size can be 10 m × 1 m or 10 m × 2 m. Every knitting machines can knit one single 10 m × 2 m net or two 10 m × 1 m nets simultaneously. In the Dyeing workstation, a number of nets in standard size can be dyed together, but the total weight of the nets cannot exceed the capacity of the dyeing machine. After the Dyeing workstation the net will be heated in the Heating workstation. One single 10 m × 2 m net or two 10 m × 1 m nets can be heated simultaneously. However, the processing time is determined by the rope size. The thicker the rope size, the longer processing time is. Finally, the nets with standard size are sutured to form net cages according to the request of the customers. In other words, a final net is a combination of a certain number of unit nets—either 10 × 2 or 10 × 1 in size. It is the suture process that combines the required number of unit nets into the final product. Because the operators cannot start suturing an order until all the required nets of standard
size are available, the waiting time for nets of standard size in this workstation is very long. The current stat map of the case company is shown in Fig. 3. It should be noted that although the fishing net manufacturing environment is MTO, every workstation has a weekly schedule, because of the long processing time. For example the processing time in the Suturing workstation is 10 days. In the figure, the small boxes represent the processes (workstations) and the numbers inside the boxes present some information about the corresponding processes, such as process time (PT), number of machines, working time (WT) per day, number of shifts and so on. Each process receives its schedule from the department of production control, which is represented by arrows from the department of production control to all the processes. The timeline at the bottom of the map shows value-adding and nonvalue-adding time. The processing time is calculated by adding the time for each process (value-adding time) in the map. The lead-time is obtained by summation of the waiting time (non-value-adding time) for each WIP triangle before each process and process time. The resulting value-adding and non-value-adding times are 25.08 and 19.78 days, respectively. The lead-time is 44.86 days. Thus the value-adding time is 56% of the lead-time. Note that fishing-net manufacturing is a flow shop production line; thus, all products traverse through the same six processing workstations. The process times and lead-times represent their mean values. According to the current state map, reducing the waiting time in front of the each workstation is an opportunity to reduce the lead-time and, consequently, to increase the service level. This is of interest because the focus of the present study is to explore a lean production system design for fishing net manufacturing using lean principles and simulation optimization. However, reducing WIP and its associated non-value-adding time, while maintaining the required system performance, is not straightforward. In fact, it is quite challenging and is the main concern for any scheduling decision and shop floor control system. There may be other noise factors that cause non-value-adding time but they are not the objective of the present study and are not covered here.
70
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
Fig. 4. Simulation interface.
Two performance indicators, average WIP level and average service level, are taken into consideration in this case study. WIP level is defined as the total volume of semi-finished goods in the production system. Service level is defined as the percentage of orders that are completed before their corresponding due date. As the company did not collect data on service level and WIP, this research developed a simulation model to evaluate it. Commercial software, Arena [29], is adopted as the simulator for the present study. The resulting example of the model interface, in which all the operation logics are verified by the managers of the case study company, is shown in Fig. 4. To evaluate the performance of each scenario, a run length of 750 days with 3–12 replications was adopted for the simulation model. The number of replications depends on 10% of the mean for a 95% confidence level. The estimated service level and WIP are 68% and 63,971 kg respectively. The results of the simulation run are validated by the managers of the case study company.
4. Future-state map optimization Although the case factory has already streamlined its production line prior to the present study, its distinctive manufacturing process causes a prolonged lead-time and high work-in-process level as discussed in the previous section. Because of the effectiveness of lean system design in lead-time reduction, it is the objective of the present study to use the lean principles in an innovative way to
solve the case problem. In light of the five corollaries identified in Section 2, we identify five key controllable factors as: (1) production unit size, (2) the pacemaker process, (3) the number of batches for an order split, (4) the production sequence, and (5) supermarket size, to optimize the proposed fishing-net manufacturing system. The production units in net cage manufacturing process have changed and are becoming bigger and bigger. In the beginning, the filaments change to twines and the twine size depends on the customer’s requirements. Then the twines change to ropes. In the Net knitting workstation the production units become nets with standard size. Finally the net cages are sutured or stitched. A manufacturing system with different production units requires a complicated production strategy and is difficult to control. According to the guidelines for designing the future state map, the production unit should be as small as possible to level the product mix. However, in the net cage manufacturing process, reducing the production unit will not only cause a greater number of changeovers but will also increase the processing time in the Suturing workstation. In this research, the production unit of each workstation depends on the standard net size. For different customer orders, the specifications such as the net weight and rope size, are also different. Thus the processing times of the standard net size could be different in every workstation. As regard the pacemaker process, this research selects Knitting workstation, Dyeing workstation and Suturing workstation as the candidates. Knitting workstation usually has the highest machine utilization. Dyeing workstation is batch processing. Its capacity
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
71
Table 1 The design factors for lean manufacturing. Factors
Description (unit)
A B C D
Production unit (m) Pacemaker process Production sequence Number of batches
Factor levels Level 1
Level 2
Level 3
10 m × 1 m Suture workstation EDD 1
10 m × 2 m Knitting workstation FIFO 2
Dyeing workstation SPT 3
is batch-size dependent. Suturing workstation is the most labor intensive workstation. Its throughput is limited by its staffing level. For the sake of cost saving, it is often the company policy to staff as few as possible, and thus, often makes it as the bottleneck of the line. In summary, the potential bottleneck could shift among the three workstations which are thus considered as the candidate pacemaker locations. The number of batches to be allowed will decide the transportation volume between workstations. More batches allow for a more level product mix, but will cause more changeovers. The changeover times between different products are different. An appropriate production sequence can reduce the total changeover time. This research selects earliest due date (EDD), first in fist out (FIFO) and shortest process time (SPT) as alternative criteria for designing the future state map. As regard the supermarket size, this research uses the Arena embedded optimizer, OptQuest, as the optimization tool which embeds scatter search in a simulation model to optimize the supermarket size for all lean manufacturing strategies. OptQuest is a proven tool that can avoid entrapment in a local optimal solution and is not sensitive to its initial solution. Glover et al. [6], Laguna and Martí [11] give details of scatter search and the OptQuest algorithms. Thus, four factors (production unit, pacemaker process, number of batches and production sequence) are taken into consideration in the experimental approach to designing the future state map. For each scenario, OptQuest is used to find the best supermarket size and the corresponding performances are viewed as the performance of each scenario. The design levels of each factor are shown in Table 1. The Taguchi method aims to find an optimal combination of parameters that have the smallest variance in performance. The signal-to-noise ratio (S/N ratio, ) is an effective way to find significant parameters by evaluating minimum variance. A higher S/N
ratio means better performance for combinatorial parameters [9]. The S/N ratio can then be defined as
j = −10 × log
1 1 r v2 k=1
or j = −10 × log
r
1 2 vjk r r
(5)
jk
Eq. (5) is used for the ‘larger-the-better’ responses and Eq. (6) is used for the ‘smaller-the-better’ responses. An L18 (21 × 37 ) orthogonal array has the least number of treatments to account for one two-level and more than three three-level control factors. Therefore, an L18 (21 × 37 ) orthogonal array was used to collect the experimental data. The experimental scenarios are shown in columns 2–5 of Table 2. The two levels and three levels of all factors were denoted as 1–2 or 1–3 (the lower, the middle and the upper levels). For each scenario, OptQuest was used to search for the best supermarket size to maximize the service level. The service level and WIP for each scenario with the best supermarket size are shown in columns 6 and 7 of Table 2. Service level is to be maximized, and the WIP is to be minimized. Let j be the S/N ratio of scenario j and let vjk be the simulation result for scenario j, in the kth replication. r is the total number of replications.
Table 2 The experimental scenarios and results. Scenario
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
L18
Mean value
S/N ratio
A
B
C
D
Service level
WIP (kg)
Service level
WIP
1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2
1 1 1 2 2 2 3 3 3 1 1 1 2 2 2 3 3 3
1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
1 2 3 1 2 3 2 3 1 3 1 2 2 3 1 3 1 2
0.83 0.85 0.80 0.82 0.82 0.77 0.81 0.78 0.84 0.85 0.83 0.89 0.82 0.84 0.82 0.84 0.83 0.83
34,887.11 36,725.24 35,266.78 34,578.18 35,939.26 36,267.69 35,765.61 39,411.85 37,659.27 41,036.15 38,989.49 36,979.23 35,661.37 37,201.86 39,286.82 40,285.11 39,868.90 38,042.93
19.99 20.23 19.71 19.91 19.90 19.39 19.74 19.41 20.06 20.16 20.01 20.61 19.92 20.10 19.92 20.05 19.97 20.00
−89.94 −88.64 −91.08 −90.59 −89.84 −91.21 −91.11 −91.95 −90.03 −92.29 −90.14 −88.73 −89.78 −90.19 −91.91 −92.12 −89.56 −91.65
(6)
k=1
Fig. 5. The effect of S/N ratio of service level.
Fig. 6. The effect of S/N ratio of WIP level.
72
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
Table 3 The analysis of variance in service level. Factors
Sun of squares
A B C D
0.320 0.259 0.002 0.215
Degrees of freedom
Mean square
F
1 2 2 2
0.320 0.130 0.001 0.107
6.065 2.455 0.018 2.037
0.053
Error
0.527
10
Total
1.324
17
Table 4 The analysis of variance in WIP. Factors A B C D
Sun of squares
Degrees of freedom
Mean square
F 0.257 1.544 1.648 4.361
0.218 2.614 20,792 70,387
1 2 2 2
0.218 1.307 1.396 3.694
Error
8.469
10
0.847
Total
21.481
17
Eqs. (5) and (6) respectively are used for calculating the signal to noise ratio (S/N) of service level and WIP. The S/N ratios of all scenarios are shown in columns 8 and 9 of Table 2. According to the S/N ratios in Table 2, the effect of each factor level on service level and WIP are calculated and shown in Figs. 5 and 6. The analysis of variance in service level and WIP are shown in Tables 3 and 4. Fig. 5 shows that A2 B1 C1 D2 is the best design for the future state map to maximize the service level. Fig. 6 shows that A1 B1 C2 D2
is the best design for the future state map to minimize the WIP. However, in Table 3 it can be seen that the contribution of factor C (production sequence) to the variance of service level is very small, and in Table 4 it can be seen that the contribution of factor A (production unit) to the variance of WIP is very small. This research, therefore, chose the level of factor C based on the effect of S/N ratio on WIP level (Fig. 6) and chose the level of factor A based on the effect of S/N ratio on service level (Fig. 5). Thus the A2 B1 C2 D2 is the design of future state map. The best supermarket size for A2 B1 C2 D2 is found using the simulation optimization tool, OptQuest. The corresponding future state maps are illustrated in Fig. 7. The optimizing results produced by OptQuest show that the design of A2 B1 C2 D2 increases service level from 68% to 90% and reduces WIP from 63,971 kg to 42,269.31 kg. The improvements from the current state map are 32.35% and 33.92%, respectively. Moreover the non-value adding time reduces from 44% (19.78 days) to 35.46% (13.78 days). Note that the optimal design of production unit (factor A) is level 2 (10 × 2 m) which is bigger than level 1 (10 × 1 m). Although the guideline number 7 encourages the production unit to be smaller the better. The size of a final net cage range from 120 to 900 m2 . The smaller production unit would cause more processing time for suturing process and would deteriorate the suturing workstation performance. There are many challenges for the present study, as discussed in this paper. Although the lean principles are generic, their application is usually innovative, as is the case in the present study. Since the results of the case study are promising and are readily applicable to similar industries, the contribution of the present study is justified.
Fig. 7. The future state map of A2 B1 C2 D2 .
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
5. Conclusions Lean manufacturing has been applied successfully in many manufacturing industries. It focuses on cost reduction by eliminating non-value adding activity. In general, waiting is the most common non-value adding activity. An appropriate production strategy can effectively reduce the total waiting time. Moreover, VSM is a tool that helps managers to visualize value adding and non-value adding activity. In this research, a net cage manufacturing system is proposed as the case study. Although net cage manufacturing operates in an environment of MTO, because of the long manufacturing lead time, every workstation has its weekly schedule and the material flows between workstations are calculated using the push strategy. These characteristics are quite different from traditional MTO manufacturing systems. Moreover, the production units are changed through the manufacturing system, and it becomes challenging to determine the size of moving batches. Based on the guidelines for implementing lean manufacturing, some important production factors are selected for designing the future state VSM. Using experimental design and a simulation optimization tool, these important factors are optimized. According to analysis based on the simulation, the future state map not only increases service level but also reduces the WIP. In the case study, the selected factors can be changed to any level without extra investment. That means the case company can implement the future state map to achieve lean manufacturing without any financial pressure. This could be the first step to achieving lean manufacturing. Although the lean principles are often generic, the proposed case study is rarely seen in literature and is an innovative application. In light of the promising results, the proposed methodologies are also readily applicable to similar industries. Acknowledgments The authors thank the anonymous company for providing the case study. This work was supported, in part, by the National Science Council of Taiwan, Republic of China, under grant NSC-1012221-E-131-043 and NSC-101-2221-E-006-137-MY3. References [1] Abdulmalek FA, Rajgopal J. Analyzing the benefits of lean production and value stream mapping via simulation: a process sector case study. Int J Prod Econ 2007;107:223–36. [2] Barber CS, Tietje BC. A research agenda for value stream mapping the sales process. J Pers Sell Sales Manag 2008;28:155–65. [3] Braglia M, Carmignani G, Zammori F. A new value stream mapping approach for complex production systems. Int J Prod Res 2006;44:3929–52.
73
[4] Doolen TL, Hacker ME. A review of lean assessment in organizations: an exploratory study of lean practices by electronics manufacturers. J Manuf Syst 2005;24:55–67. [5] Elmoselhy SAM. Hybrid lean-agile manufacturing system technical facet, in automotive sector. J Manuf Syst 2013;32:598–619. [6] Glover F, Laguna M, Martí R. Scatter search. In: Ghosh A, Tsutsui S, editors. Theory and applications of evolutionary computations: recent trends. New York: Springer-Verlag; 2000. [7] Hobbs DP. Lean manufacturing implementation: a complete execution manual for any size manufacturer. Florida: J. Ross Publishing; 2004. [8] Hsieh HC, Yang T, Su CT, Lin CT. The development of a hierarchical roughcut capacity planning model and demand management system for fishing-net manufacturing. Eur J Ind Eng 2012;6(4):422–40. [9] Kuo Y, Yang T, Huang GW. The use of a grey-based Taguchi method for optimizing multi-response simulation problems. Eng Optim 2008;40:517–28. [11] Laguna M, Martí R. Neural network prediction in a system for optimizing simulations. IIE Trans 2001;34:273–82. [12] Lasa IS, Castro RD, Laburu CO. Extent of the use of lean concepts proposed for a value stream mapping application. Prod Plan Control 2009;20:82–98. [13] Lasa IS, Laburu CO, Castro RD. Evaluation of value stream mapping in manufacturing system redesign. Int J Prod Res 2008;46:4409–30. [14] Lian YH, Landeghem HV. Analyzing the effects of lean production using a value stream mapping-based simulation generator. Int J Prod Res 2007;45:3037–58. [15] Lu J-C, Yang T, Wang C-Y. A lean pull system design analyzed by value stream mapping and multiple criteria decision-making method under demand uncertainty. Int J Comput Integr Manuf 2011;24:211–28. [16] Lummus RR, Vokurka RJ, Rodeghiero B. Improving quality through value stream mapping a case study of a physician’s clinic. Total Qual Manag 2006;71:1063–75. [17] McDonald T, Van Aken EM, Rentes AF. Utilizing simulation to enhance value stream mapping: a manufacturing case application. Int J Logist 2002;5:213–32. [18] Meade DJ, Kumar S, Houshyar A. Financial analysis of a theoretical lean manufacturing implementation using hybrid simulation modeling. J Manuf Syst 2006;25:137–52. [20] Rother M, Shook J. Learning to see: value stream mapping to add value and eliminate MUDA. Brookline: The Lean Enterprise Institute; 1998. [21] Saurin TA, Ribeiro JLD, Vidor G. A framework for assessing poka-yoke devices. J Manuf Syst 2012;31:358–66. [22] Seth D, Gupta V. Application of value stream mapping for lean operations and cycle time reduction: an Indian case study. Prod Plan Control 2005;16: 44–59. [23] Seth D, Seth N, Goel D. Application of value stream mapping (VSM) for minimization of wastes in the processing side of supply chain of cottonseed oil industrial in Indian context. J Manuf Technol Manag 2008;19:529–50. [24] Sullivan WG, McDonald TN, Van Aken EM. Equipment replacement decisions and lean production. Robot Comput Integr Manuf 2002;18:255–65. [25] Yang T, Hsieh C-H, Cheng B-Y. Lean-pull strategy in a reentrant manufacturing environment: a pilot study for TFT-LCD array manufacturing. Int J Prod Res 2011;49:1511–29. [26] Yang T, Kuo Y, Cho C. A genetic algorithms simulation approach for the multi-attribute combinatorial dispatching decision problem. Eur J Oper Res 2007;176:1859–73. [27] Yang T, Kuo Y, Chou P. Solving a multiresponse simulation problem using a dual-response system and scatter search method. Simul Model Pract Theory 2005;13:356–69. [28] Yang T, Lu J-C. The use of multiple attribute decision-making method and value stream mapping in solving pacemaker location problem. Int J Prod Res 2011;49:2793–817. [29] Kelton WD, Sadowski RP, Sturrock DT. Simulation with Arena. 4th ed. New York: McGraw-Hill; 2007.
Contents lists available at ScienceDirect
Journal of Manufacturing Systems journal homepage: www.elsevier.com/locate/jmansys
Technical Paper
Lean production system design for fishing net manufacturing using lean principles and simulation optimization Taho Yang a , Yiyo Kuo b,∗ , Chao-Ton Su c , Chia-Lin Hou a a b c
Institute of Manufacturing Information and Systems, National Cheng Kung University, Tainan 701, Taiwan Department of Industrial Engineering and Management, Ming Chi University of Technology, New Taipei City 243, Taiwan Department of Industrial Engineering and Engineering Management, National Tsing Hua University, Hsinchu 300, Taiwan
a r t i c l e
i n f o
Article history: Received 13 September 2012 Received in revised form 1 October 2014 Accepted 24 November 2014 Keywords: Lean manufacturing Simulation Value stream mapping Fishing nets
a b s t r a c t Value stream mapping (VSM) is a useful tool for describing the manufacturing state, especially for distinguishing between those activities that add value and those that do not. It can help in eliminating non-value activities and reducing the work in process (WIP) and thereby increase the service level. This research follows the guidelines for designing future state VSM. These guidelines consist of five factors which can be changed simply, without any investment. These five factors are (1) production unit; (2) pacemaker process; (3) number of batches; (4) production sequence; and (5) supermarket size. The five factors are applied to a fishing net manufacturing system. Using experimental design and a simulation optimizing tool, the five factors are optimized. The results show that the future state maps can increase service level and reduce WIP by at least 29.41% and 33.92% respectively. For the present study, the lean principles are innovatively adopted in solving a fishing net manufacturing system which is not a welladdressed problem in literature. In light of the promising empirical results, the proposed methodologies are also readily applicable to similar industries. © 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
1. Introduction The common challenges that companies face are market competition, increased pressure on inventory, increased service levels, and reduced work in process (WIP). Lean manufacturing is one of the approaches that can help companies respond appropriately to these challenges [11,18]. The focus of the approach is on cost reduction, by eliminating activities that do not add value by linking and balancing work stages, so that products from one stage are consumed directly by the next stage, until the end of the production line is reached [4,5,7,18]. Applications of lean manufacturing have spanned many sectors, including the automotive industry, electronics, white goods and consumer goods [1,21]. However, there is no evidence of work applying lean manufacturing principles to fishing net production. Fishing net manufacturing is a high make-to-order (MTO) environment, because net size and type change according to the ocean environment, fish kinds and ship size.
∗ Corresponding author. Tel.: +886 2 29089899x3118; fax: +886 6 29085900. E-mail addresses: [email protected] (T. Yang), [email protected] (Y. Kuo), [email protected] (C.-T. Su), [email protected] (C.-L. Hou).
MTO manufacturing need not be a pull production system—a lean design principle. Usually, there are many (or some) simultaneous orders in the manufacturing process. These orders may be different in size and specifications. Thus, there is no guarantee that the system will follow a first-in-first-out rule. In fact, it is often quite messy when production is loaded with work-in-process. Thus, it may be effective to adopt lean principles to regulate the flow and to control the work-in-process level. Higher WIP and longer cycle times always result in a lower service level and produce lower customer satisfaction in MTO environments. Moreover the fishing net manufacturing system is distinctive. Through the production procedure the production units become bigger and bigger and the processing time in most steps are quite long (between 3 and 10 days). This provides the motivation for this research, to design a lean manufacturing system for this industry. Value stream mapping (VSM) is a visual tool that facilitates the process of lean production by helping to identify the value-adding steps and eliminating the non-value adding waste [22]. In recent years, VSM has emerged as the superior method to implement lean production in factories, and has been used to identify where waste occurs [3,12,14,15,17,24,25,28]. VSM creates a pictorial representation and common language for the production process, enabling
http://dx.doi.org/10.1016/j.jmsy.2014.11.010 0278-6125/© 2014 The Society of Manufacturing Engineers. Published by Elsevier Ltd. All rights reserved.
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
more purposeful decisions to improve processes. Numerous applications can be found in the literature. For example, Seth and Gupta [22] applied VSM to establish lean production in an Indian twowheeler motor company. Lummus et al. [16] report on a VSM project in a small medical clinic. The new system can increase patient throughput and reduce patient waiting time. Abdulmalek and Rajgopal [1] describe a steel mill case, and apply VSM to identify the opportunities for improvement. Lian and Landeghem [14] propose a generator which automatically yields a simulation model for VSM. The current and future state VSM of a poultry and pig rearing equipment manufacturing system is introduced to demonstrate the proposed generator. Barber and Tietje [2] demonstrate the use of VSM to achieve both greater efficiency and value creation in a sales process. Seth et al. [23] identify and address various sources of waste in the supply chain of an edible cottonseed oil industrial process and use a VSM approach to improve productivity and capacity utilization in an Indian context. Lasa et al. [12,13] redesigned six production systems with the VSM technique, and then analyze the causes for the limited adoption of lean manufacturing concepts. The six production systems are manufacturing systems for kit furniture, water heaters, forging, detonator systems, mechanized and stamped parts, and thermoplastic parts. Rother and Shook [20] proposed seven guidelines based on the concept of lean manufacturing to construct the future VSM which can be implemented in a reasonably short time period without any major investment. According to the seven guidelines, there are several factors that should be considered. Different combinations of the factors would result in different production performance. However, the implementation of the recommendations is likely to be risky. Simulation is a useful tool for evaluating the performance of a new design but it cannot provide the optimal design. Combining simulation with experimental design or intelligent search has been successfully adopted for simulation optimization. Yang et al. [27] use a commercial tool called OptQuest for optimizing an integrated-circuit (IC) packaging system. OptQuest embeds scatter search (SS) in a simulation tool Arena to optimize the simulation models. Yang et al. [26] combine simulation with genetic algorithm for optimizing dispatching rules in a flow shop with multiple processors. Kuo et al. [9] integrate simulation and the Taguchi method to optimize an integrated-circuit (IC) packaging system. The objective of the present study is to model a non-typical production system – a fishing-net manufacturing system – and to propose a lean production system design which is optimized by simulation optimization. The remaining sections of this paper are organized as follows. In Section 2 the lean principles are introduced, and a number of decision factors are highlighted. Then, the fishing-net production system is introduced and the current state VSM is provided in Section 3. Finally, a hybrid experimental design and intelligent search approach is adopted for optimizing the decision factors and then the future state VSM is constructed in Section 4. A summary of results and concluding remarks are presented in Section 5.
67
2.1. Guideline #1: produce to your takt time “Takt time” is used to synchronize the pace of production with the pace of sales. In general, it can be calculated by Eq. (1) Takt time =
available working time per day customer demand rate per day
(1)
However, when the process times of different products are quite different, the takt time calculated by Eq. (1) is not reasonable. For example, suppose there are two types of product, A and B, and the unit process time and demand rate of product A are 120 min and 4 units per day, respectively and the process time and demand rate of product B are 60 min and 8 unit per day. If the available working time is 18 h per day, then the takt time is 90 min (18 × 60/(4 + 8)). This means that one unit should be produced every 90 min, but this is impossible for product A. The present research uses Eqs. (2)–(4) for calculating the takt time. KSi =
The lowest common multiple of processing time for all products Pi
NKi =
Di KSi
Takt time =
(2) (3)
Available working time per day
N
i=1
NKi
(4)
In Eq. (2) KSi indicates the kanban size of product i (i = 1, 2, . . ., N), and Pi indicate the processing time of product i. In Eq. (3) NKi indicates the number of kanbans of product i, and Di indicates the demand for product i per day. Thus, for the same example, the kanban size for product A and B are 1 and 2 units respectively, and the number of kanbans for product A and B are both 4. The takt time is 2.25 h. This means that one unit of product A or two units of product B should be completed every 2.25 h. Each kanban indicates one unit of product A or two units of product B and four kanbans are required per day for both product A and B. It should be noted that the lowest common multiple of processing time for all products in Eq. (2) is an approximate value and should be as small as possible. Managers can increase the process time of some product types to tune it. For example, if there are three product types and their corresponding processing times are 12, 11 and 25 min, then the processing time of the first two product types can be increased to 12.5 min, so that the lowest common multiple of processing time will be 25 min. Corollary 1. The takt time is dependent on the size of a production unit. Corollary 1a. A kanban represent the same required production time regardless of the different product types. 2.2. Guideline #2: develop continuous flow where possible Continuous flow refers to producing one piece at a time to reduce the inventory of WIP and production CT. However continuous flow requires a great deal of creativity to achieve and sometimes it requires plant layout redistribution [12]. In this research, this guideline is not applicable and is not taken into consideration in the case study.
2. The lean principles Rother and Shook [20] propose a five phase implementation of VSM. The phases are (1) selection of product family; (2) current state mapping; (3) future state mapping; (4) definition of working plan; and (5) achievement of working plan. The lean techniques to be used are defined in the third phase which contains seven guidelines to define the future state map [12]. The seven guidelines are summarized below:
2.3. Guideline #3: use a supermarket to control production where continuous flow does not extend upstream A “supermarket” is nothing more than a buffer or storage area located at the end of the production process for products that are ready to be shipped [1]. When continuous flow is not possible and batching is necessary, a supermarket can smooth the whole manufacturing process. Supermarkets use a kanban system to fix the
68
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
inventory level. If the number of kanbans in a supermarket is too high, it causes higher inventory cost. On the other hand, the downstream production process will be subject to delays if the number of kanban in the supermarket is too low. Therefore, the number of kanbans in a supermarket is an importation factor when designing the future VSM. Corollary 2. A supermarket is required for the implementation of lean-pull system. 2.4. Guideline #4: try to send customer scheduling to only one production process In Guideline Number 1, for calculating the takt time, the customer demand rate is based on customer scheduling. The process time is set by one of the production processes. That process is called the “pacemaker process”. The pacemaker process synchronizes the pace of the entire manufacturing process and there are no supermarkets downstream of the pacemaker process. Therefore, selecting different workstations as the pacemaker process will have an important influence on the performance of the entire manufacturing system. Corollary 3. A pacemaker process is required for the implementation of lean-pull system. 2.5. Guideline #5: level the production mix Leveling the product mix means dividing the volume of all the product types based on their kanban size, and then producing them evenly over a time period. The more level the product mix is, the greater is the ability to respond to different customer requirements with a short lead time. For example, if there are three product types and the production sequence can be A-A-A-B-B-B-C-C-C or A-BC-A-B-C-A-B-C. The latter is more level and allows the process to proceed smoothly with smaller supermarkets, but it would cause more changeovers. Due to make-to-order (MTO) production environment, an order split to a certain number of production batches should be considered to achieve a leveling production. The use of a larger number of batches with fewer products in each batch results in a more level process. Therefore, the number of batches is also an importation factor in designing future VSM. Moreover, the production sequence also influences the performance of the manufacturing system. That means the managers have to decide between A-B-C-A-B-C-A-B-C and A-C-B-A-C-B-A-C in the example of the production sequence above. Corollary 4. An order splitting strategy is able to improve a leanpull system performance. Corollary 5. The production sequence decision could impact a lean-pull system performance. 2.6. Guideline #6: level the production volume
Fig. 1. The composition of a rope.
decreasing the production unit and so on. It has been addressed in Corollary 4. Lean management is well recognized as a management philosophy rather than a generic tool set. In other word, its principles are generic but its solution methodology is often case specific. For the present study, we adopt the lean principles from the existing literature as an exemplar to identify the five corollaries as the basis of our proposed methodology development which will be discussed in details in the following sections. 3. Fishing-net manufacturing and the current state map The case study was adopted from Hsieh et al. [8]. They introduced the fishing net manufacturing process in detail and then developed a hierarchical rough-cut capacity planning model and demand management system. While they focused on the strategic planning decision, the production control decision is not addressed. The application of lean principle to the fishing-net manufacturing is not found in literature. The solution to this specific application could be applicable to similar industries such as textile industries, variable product-size production, etc. Since the fishing-net manufacturing is rarely seen in literature, we discuss the production process forehead in details as follows. There are various fishing nets, such as gill nets, lift nets, drag nets, surrounding nets (purse seine nets), set nets (trap nets), covering nets, and fish breeding nets. Each net has a unique application which is based on the ocean environment, fish type and ship size. Therefore, the same customer could order nets of different types or different sizes. Each order requires different raw material types, twine sizes, mesh sizes, mesh depths and mesh lengths. The raw materials for fishing nets include nylon, trawl, nylon filament, nylon multi-filament and polyester. Twine size is determined by the composition of a rope as shown in Fig. 1 [8]. Raw material filaments are twisted into yarn, then twined yarns are made into strands. Strands are twisted into twine, and then braided twines into rope. Mesh size is a special unit of a net; different mesh depth and mesh length will result in a different shape and area of fishing net. The relationship between mesh size, mesh depth and mesh length is show in Fig. 2 [8].
Leveling the production volume is related to Guideline number 1. It means that production should be based on a fixed pace, the takt time. It has been addressed in Corollaries 1, 1a, and 4. 2.7. Guideline #7: develop the ability to make “every part every day” (then every shift, then every hour or pallet or pitch) in fabrication processes upstream of the pacemaker process This guideline is related to Guideline number 5. In order to divide the volume of all product types into more batches, some production methods have to be changed, such as reducing the changeover time,
Fig. 2. The mesh size.
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
69
Fig. 3. The current state map.
Net cage is one kind of fishing net which is the main product of the case study company. The main function of the net cage is to be fixed under the sea and cultivate salt water fish in it. According to the size and type of the cultivated fish, the corresponding specifications of net cages are also different. The general diameter varies from 10 m to 30 m, depth varies from 4 m to 10 m, while mesh varies from 2 mm to 30 mm. In the case study company there are six workstations for net cage manufacturing. They are (1) Twisting, (2) Braiding, (3) Net knitting, (4) Dyeing, (5) Heating and (6) Suturing. All manufactured net cages follow the sequence of workstation above, from 1 to 6. First the filament is twisted to form the yarn or strand as the twine in the Twisting workstation. Then the rope is braided from three twines in the Braiding workstation. The rope size is determined by the request of customers. The thicker the rope size, the thicker twine is required, and consequently more twisting jobs are required. In the Net knitting workstation the ropes are knitted to form nets in standard sizes. A knitting machine processes a net continuously. Theoretically, there is no limit to the length of a net, but in practice, it is a multiple of 10 m. There two options for the width of a net: 1 or 2 m. According to the specification of the knitting machines in the case company the standard size can be 10 m × 1 m or 10 m × 2 m. Every knitting machines can knit one single 10 m × 2 m net or two 10 m × 1 m nets simultaneously. In the Dyeing workstation, a number of nets in standard size can be dyed together, but the total weight of the nets cannot exceed the capacity of the dyeing machine. After the Dyeing workstation the net will be heated in the Heating workstation. One single 10 m × 2 m net or two 10 m × 1 m nets can be heated simultaneously. However, the processing time is determined by the rope size. The thicker the rope size, the longer processing time is. Finally, the nets with standard size are sutured to form net cages according to the request of the customers. In other words, a final net is a combination of a certain number of unit nets—either 10 × 2 or 10 × 1 in size. It is the suture process that combines the required number of unit nets into the final product. Because the operators cannot start suturing an order until all the required nets of standard
size are available, the waiting time for nets of standard size in this workstation is very long. The current stat map of the case company is shown in Fig. 3. It should be noted that although the fishing net manufacturing environment is MTO, every workstation has a weekly schedule, because of the long processing time. For example the processing time in the Suturing workstation is 10 days. In the figure, the small boxes represent the processes (workstations) and the numbers inside the boxes present some information about the corresponding processes, such as process time (PT), number of machines, working time (WT) per day, number of shifts and so on. Each process receives its schedule from the department of production control, which is represented by arrows from the department of production control to all the processes. The timeline at the bottom of the map shows value-adding and nonvalue-adding time. The processing time is calculated by adding the time for each process (value-adding time) in the map. The lead-time is obtained by summation of the waiting time (non-value-adding time) for each WIP triangle before each process and process time. The resulting value-adding and non-value-adding times are 25.08 and 19.78 days, respectively. The lead-time is 44.86 days. Thus the value-adding time is 56% of the lead-time. Note that fishing-net manufacturing is a flow shop production line; thus, all products traverse through the same six processing workstations. The process times and lead-times represent their mean values. According to the current state map, reducing the waiting time in front of the each workstation is an opportunity to reduce the lead-time and, consequently, to increase the service level. This is of interest because the focus of the present study is to explore a lean production system design for fishing net manufacturing using lean principles and simulation optimization. However, reducing WIP and its associated non-value-adding time, while maintaining the required system performance, is not straightforward. In fact, it is quite challenging and is the main concern for any scheduling decision and shop floor control system. There may be other noise factors that cause non-value-adding time but they are not the objective of the present study and are not covered here.
70
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
Fig. 4. Simulation interface.
Two performance indicators, average WIP level and average service level, are taken into consideration in this case study. WIP level is defined as the total volume of semi-finished goods in the production system. Service level is defined as the percentage of orders that are completed before their corresponding due date. As the company did not collect data on service level and WIP, this research developed a simulation model to evaluate it. Commercial software, Arena [29], is adopted as the simulator for the present study. The resulting example of the model interface, in which all the operation logics are verified by the managers of the case study company, is shown in Fig. 4. To evaluate the performance of each scenario, a run length of 750 days with 3–12 replications was adopted for the simulation model. The number of replications depends on 10% of the mean for a 95% confidence level. The estimated service level and WIP are 68% and 63,971 kg respectively. The results of the simulation run are validated by the managers of the case study company.
4. Future-state map optimization Although the case factory has already streamlined its production line prior to the present study, its distinctive manufacturing process causes a prolonged lead-time and high work-in-process level as discussed in the previous section. Because of the effectiveness of lean system design in lead-time reduction, it is the objective of the present study to use the lean principles in an innovative way to
solve the case problem. In light of the five corollaries identified in Section 2, we identify five key controllable factors as: (1) production unit size, (2) the pacemaker process, (3) the number of batches for an order split, (4) the production sequence, and (5) supermarket size, to optimize the proposed fishing-net manufacturing system. The production units in net cage manufacturing process have changed and are becoming bigger and bigger. In the beginning, the filaments change to twines and the twine size depends on the customer’s requirements. Then the twines change to ropes. In the Net knitting workstation the production units become nets with standard size. Finally the net cages are sutured or stitched. A manufacturing system with different production units requires a complicated production strategy and is difficult to control. According to the guidelines for designing the future state map, the production unit should be as small as possible to level the product mix. However, in the net cage manufacturing process, reducing the production unit will not only cause a greater number of changeovers but will also increase the processing time in the Suturing workstation. In this research, the production unit of each workstation depends on the standard net size. For different customer orders, the specifications such as the net weight and rope size, are also different. Thus the processing times of the standard net size could be different in every workstation. As regard the pacemaker process, this research selects Knitting workstation, Dyeing workstation and Suturing workstation as the candidates. Knitting workstation usually has the highest machine utilization. Dyeing workstation is batch processing. Its capacity
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
71
Table 1 The design factors for lean manufacturing. Factors
Description (unit)
A B C D
Production unit (m) Pacemaker process Production sequence Number of batches
Factor levels Level 1
Level 2
Level 3
10 m × 1 m Suture workstation EDD 1
10 m × 2 m Knitting workstation FIFO 2
Dyeing workstation SPT 3
is batch-size dependent. Suturing workstation is the most labor intensive workstation. Its throughput is limited by its staffing level. For the sake of cost saving, it is often the company policy to staff as few as possible, and thus, often makes it as the bottleneck of the line. In summary, the potential bottleneck could shift among the three workstations which are thus considered as the candidate pacemaker locations. The number of batches to be allowed will decide the transportation volume between workstations. More batches allow for a more level product mix, but will cause more changeovers. The changeover times between different products are different. An appropriate production sequence can reduce the total changeover time. This research selects earliest due date (EDD), first in fist out (FIFO) and shortest process time (SPT) as alternative criteria for designing the future state map. As regard the supermarket size, this research uses the Arena embedded optimizer, OptQuest, as the optimization tool which embeds scatter search in a simulation model to optimize the supermarket size for all lean manufacturing strategies. OptQuest is a proven tool that can avoid entrapment in a local optimal solution and is not sensitive to its initial solution. Glover et al. [6], Laguna and Martí [11] give details of scatter search and the OptQuest algorithms. Thus, four factors (production unit, pacemaker process, number of batches and production sequence) are taken into consideration in the experimental approach to designing the future state map. For each scenario, OptQuest is used to find the best supermarket size and the corresponding performances are viewed as the performance of each scenario. The design levels of each factor are shown in Table 1. The Taguchi method aims to find an optimal combination of parameters that have the smallest variance in performance. The signal-to-noise ratio (S/N ratio, ) is an effective way to find significant parameters by evaluating minimum variance. A higher S/N
ratio means better performance for combinatorial parameters [9]. The S/N ratio can then be defined as
j = −10 × log
1 1 r v2 k=1
or j = −10 × log
r
1 2 vjk r r
(5)
jk
Eq. (5) is used for the ‘larger-the-better’ responses and Eq. (6) is used for the ‘smaller-the-better’ responses. An L18 (21 × 37 ) orthogonal array has the least number of treatments to account for one two-level and more than three three-level control factors. Therefore, an L18 (21 × 37 ) orthogonal array was used to collect the experimental data. The experimental scenarios are shown in columns 2–5 of Table 2. The two levels and three levels of all factors were denoted as 1–2 or 1–3 (the lower, the middle and the upper levels). For each scenario, OptQuest was used to search for the best supermarket size to maximize the service level. The service level and WIP for each scenario with the best supermarket size are shown in columns 6 and 7 of Table 2. Service level is to be maximized, and the WIP is to be minimized. Let j be the S/N ratio of scenario j and let vjk be the simulation result for scenario j, in the kth replication. r is the total number of replications.
Table 2 The experimental scenarios and results. Scenario
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
L18
Mean value
S/N ratio
A
B
C
D
Service level
WIP (kg)
Service level
WIP
1 1 1 1 1 1 1 1 1 2 2 2 2 2 2 2 2 2
1 1 1 2 2 2 3 3 3 1 1 1 2 2 2 3 3 3
1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3
1 2 3 1 2 3 2 3 1 3 1 2 2 3 1 3 1 2
0.83 0.85 0.80 0.82 0.82 0.77 0.81 0.78 0.84 0.85 0.83 0.89 0.82 0.84 0.82 0.84 0.83 0.83
34,887.11 36,725.24 35,266.78 34,578.18 35,939.26 36,267.69 35,765.61 39,411.85 37,659.27 41,036.15 38,989.49 36,979.23 35,661.37 37,201.86 39,286.82 40,285.11 39,868.90 38,042.93
19.99 20.23 19.71 19.91 19.90 19.39 19.74 19.41 20.06 20.16 20.01 20.61 19.92 20.10 19.92 20.05 19.97 20.00
−89.94 −88.64 −91.08 −90.59 −89.84 −91.21 −91.11 −91.95 −90.03 −92.29 −90.14 −88.73 −89.78 −90.19 −91.91 −92.12 −89.56 −91.65
(6)
k=1
Fig. 5. The effect of S/N ratio of service level.
Fig. 6. The effect of S/N ratio of WIP level.
72
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
Table 3 The analysis of variance in service level. Factors
Sun of squares
A B C D
0.320 0.259 0.002 0.215
Degrees of freedom
Mean square
F
1 2 2 2
0.320 0.130 0.001 0.107
6.065 2.455 0.018 2.037
0.053
Error
0.527
10
Total
1.324
17
Table 4 The analysis of variance in WIP. Factors A B C D
Sun of squares
Degrees of freedom
Mean square
F 0.257 1.544 1.648 4.361
0.218 2.614 20,792 70,387
1 2 2 2
0.218 1.307 1.396 3.694
Error
8.469
10
0.847
Total
21.481
17
Eqs. (5) and (6) respectively are used for calculating the signal to noise ratio (S/N) of service level and WIP. The S/N ratios of all scenarios are shown in columns 8 and 9 of Table 2. According to the S/N ratios in Table 2, the effect of each factor level on service level and WIP are calculated and shown in Figs. 5 and 6. The analysis of variance in service level and WIP are shown in Tables 3 and 4. Fig. 5 shows that A2 B1 C1 D2 is the best design for the future state map to maximize the service level. Fig. 6 shows that A1 B1 C2 D2
is the best design for the future state map to minimize the WIP. However, in Table 3 it can be seen that the contribution of factor C (production sequence) to the variance of service level is very small, and in Table 4 it can be seen that the contribution of factor A (production unit) to the variance of WIP is very small. This research, therefore, chose the level of factor C based on the effect of S/N ratio on WIP level (Fig. 6) and chose the level of factor A based on the effect of S/N ratio on service level (Fig. 5). Thus the A2 B1 C2 D2 is the design of future state map. The best supermarket size for A2 B1 C2 D2 is found using the simulation optimization tool, OptQuest. The corresponding future state maps are illustrated in Fig. 7. The optimizing results produced by OptQuest show that the design of A2 B1 C2 D2 increases service level from 68% to 90% and reduces WIP from 63,971 kg to 42,269.31 kg. The improvements from the current state map are 32.35% and 33.92%, respectively. Moreover the non-value adding time reduces from 44% (19.78 days) to 35.46% (13.78 days). Note that the optimal design of production unit (factor A) is level 2 (10 × 2 m) which is bigger than level 1 (10 × 1 m). Although the guideline number 7 encourages the production unit to be smaller the better. The size of a final net cage range from 120 to 900 m2 . The smaller production unit would cause more processing time for suturing process and would deteriorate the suturing workstation performance. There are many challenges for the present study, as discussed in this paper. Although the lean principles are generic, their application is usually innovative, as is the case in the present study. Since the results of the case study are promising and are readily applicable to similar industries, the contribution of the present study is justified.
Fig. 7. The future state map of A2 B1 C2 D2 .
T. Yang et al. / Journal of Manufacturing Systems 34 (2015) 66–73
5. Conclusions Lean manufacturing has been applied successfully in many manufacturing industries. It focuses on cost reduction by eliminating non-value adding activity. In general, waiting is the most common non-value adding activity. An appropriate production strategy can effectively reduce the total waiting time. Moreover, VSM is a tool that helps managers to visualize value adding and non-value adding activity. In this research, a net cage manufacturing system is proposed as the case study. Although net cage manufacturing operates in an environment of MTO, because of the long manufacturing lead time, every workstation has its weekly schedule and the material flows between workstations are calculated using the push strategy. These characteristics are quite different from traditional MTO manufacturing systems. Moreover, the production units are changed through the manufacturing system, and it becomes challenging to determine the size of moving batches. Based on the guidelines for implementing lean manufacturing, some important production factors are selected for designing the future state VSM. Using experimental design and a simulation optimization tool, these important factors are optimized. According to analysis based on the simulation, the future state map not only increases service level but also reduces the WIP. In the case study, the selected factors can be changed to any level without extra investment. That means the case company can implement the future state map to achieve lean manufacturing without any financial pressure. This could be the first step to achieving lean manufacturing. Although the lean principles are often generic, the proposed case study is rarely seen in literature and is an innovative application. In light of the promising results, the proposed methodologies are also readily applicable to similar industries. Acknowledgments The authors thank the anonymous company for providing the case study. This work was supported, in part, by the National Science Council of Taiwan, Republic of China, under grant NSC-1012221-E-131-043 and NSC-101-2221-E-006-137-MY3. References [1] Abdulmalek FA, Rajgopal J. Analyzing the benefits of lean production and value stream mapping via simulation: a process sector case study. Int J Prod Econ 2007;107:223–36. [2] Barber CS, Tietje BC. A research agenda for value stream mapping the sales process. J Pers Sell Sales Manag 2008;28:155–65. [3] Braglia M, Carmignani G, Zammori F. A new value stream mapping approach for complex production systems. Int J Prod Res 2006;44:3929–52.
73
[4] Doolen TL, Hacker ME. A review of lean assessment in organizations: an exploratory study of lean practices by electronics manufacturers. J Manuf Syst 2005;24:55–67. [5] Elmoselhy SAM. Hybrid lean-agile manufacturing system technical facet, in automotive sector. J Manuf Syst 2013;32:598–619. [6] Glover F, Laguna M, Martí R. Scatter search. In: Ghosh A, Tsutsui S, editors. Theory and applications of evolutionary computations: recent trends. New York: Springer-Verlag; 2000. [7] Hobbs DP. Lean manufacturing implementation: a complete execution manual for any size manufacturer. Florida: J. Ross Publishing; 2004. [8] Hsieh HC, Yang T, Su CT, Lin CT. The development of a hierarchical roughcut capacity planning model and demand management system for fishing-net manufacturing. Eur J Ind Eng 2012;6(4):422–40. [9] Kuo Y, Yang T, Huang GW. The use of a grey-based Taguchi method for optimizing multi-response simulation problems. Eng Optim 2008;40:517–28. [11] Laguna M, Martí R. Neural network prediction in a system for optimizing simulations. IIE Trans 2001;34:273–82. [12] Lasa IS, Castro RD, Laburu CO. Extent of the use of lean concepts proposed for a value stream mapping application. Prod Plan Control 2009;20:82–98. [13] Lasa IS, Laburu CO, Castro RD. Evaluation of value stream mapping in manufacturing system redesign. Int J Prod Res 2008;46:4409–30. [14] Lian YH, Landeghem HV. Analyzing the effects of lean production using a value stream mapping-based simulation generator. Int J Prod Res 2007;45:3037–58. [15] Lu J-C, Yang T, Wang C-Y. A lean pull system design analyzed by value stream mapping and multiple criteria decision-making method under demand uncertainty. Int J Comput Integr Manuf 2011;24:211–28. [16] Lummus RR, Vokurka RJ, Rodeghiero B. Improving quality through value stream mapping a case study of a physician’s clinic. Total Qual Manag 2006;71:1063–75. [17] McDonald T, Van Aken EM, Rentes AF. Utilizing simulation to enhance value stream mapping: a manufacturing case application. Int J Logist 2002;5:213–32. [18] Meade DJ, Kumar S, Houshyar A. Financial analysis of a theoretical lean manufacturing implementation using hybrid simulation modeling. J Manuf Syst 2006;25:137–52. [20] Rother M, Shook J. Learning to see: value stream mapping to add value and eliminate MUDA. Brookline: The Lean Enterprise Institute; 1998. [21] Saurin TA, Ribeiro JLD, Vidor G. A framework for assessing poka-yoke devices. J Manuf Syst 2012;31:358–66. [22] Seth D, Gupta V. Application of value stream mapping for lean operations and cycle time reduction: an Indian case study. Prod Plan Control 2005;16: 44–59. [23] Seth D, Seth N, Goel D. Application of value stream mapping (VSM) for minimization of wastes in the processing side of supply chain of cottonseed oil industrial in Indian context. J Manuf Technol Manag 2008;19:529–50. [24] Sullivan WG, McDonald TN, Van Aken EM. Equipment replacement decisions and lean production. Robot Comput Integr Manuf 2002;18:255–65. [25] Yang T, Hsieh C-H, Cheng B-Y. Lean-pull strategy in a reentrant manufacturing environment: a pilot study for TFT-LCD array manufacturing. Int J Prod Res 2011;49:1511–29. [26] Yang T, Kuo Y, Cho C. A genetic algorithms simulation approach for the multi-attribute combinatorial dispatching decision problem. Eur J Oper Res 2007;176:1859–73. [27] Yang T, Kuo Y, Chou P. Solving a multiresponse simulation problem using a dual-response system and scatter search method. Simul Model Pract Theory 2005;13:356–69. [28] Yang T, Lu J-C. The use of multiple attribute decision-making method and value stream mapping in solving pacemaker location problem. Int J Prod Res 2011;49:2793–817. [29] Kelton WD, Sadowski RP, Sturrock DT. Simulation with Arena. 4th ed. New York: McGraw-Hill; 2007.