Taho Yang*. Yu-Hsiu Hung**. Kuan-Cheng Huang***
*National Cheng Kung University, Tainan, 701 Taiwan (Tel: 886-6-2757575 ext.34225; e-mail: [email protected]).
** National Cheng Kung University, Tainan, 701 Taiwan (Tel: 886-6-2757575 ext.54327; e-mail:[email protected])
*** National Cheng Kung University, Tainan, 701 Taiwan, (e-mail: [email protected]) Abstract: The aim of the study was to demonstrate the effectiveness of Constant Work-in-process (CONWIP) pull system to bike chain production. Two multi-CONWIP production systems (developed by Kaizen, as well as by looking at the bottlenecks) were proposed to prevent WIP accumulations.
Simulation was performed on the average lead time and the total WIP of the two multi-CONWIP systems in comparison with those of a current bike chain production system. Results of the comparisons showed that the Kaizen multi-CONWIP system outperformed all the other systems on the average lead time (with up to 42.8% improvements). The bottleneck multi-CONWIP system were able to reduce the WIP volume by over 50%, outperforming the Kaizen multi-CONWIP system and the current production system. These results show that multi-CONWIP has its potential in reducing production lead time and WIP and can better satisfy customer demands. Copyright © 2019 IFAC
Keywords: Bike chain; CONWIP; Kanban; Pull production; Simulation optimization
1. INTRODUCTION
Bicycle manufacturing is an important industry since bicycle is prevailing world-wide due to both eco-conscious and sports trend. Bicycle chain is a key component of a bicycle.
With respect to the bicycle chain market, its value and compound annual growth rate are presumed to increase continuously to the end of 2022 (Market Research Store, 2017). The prosperity of the bicycle chain market have led to intense competition and falling profitability. Most suppliers are small to medium sized companies that have few resources for innovation. They received little support from bicycle assemblers to enhance their technology capabilities. They faced low-cost challenges, battled to control costs, and attempted to increase service levels. They adopted push manufacturing systems that oftentimes fail to function efficiently (because of the difficulties to accurately forecast the types and numbers of products customers would purchase). Thus, work in progress (WIP) got cumulated easily and grew beyond a predetermined level especially when orders do not come in a consistent manner.
In fact, push production systems had been shown functioning poorly because they schedule work releases on the basis of information outside the system - without a feedback loop communicating the status of WIP (Hopp & Spearman, 2011).
Conversely, pull production systems are robust because they authorize releases based on information from inside the system - controlling the amount of WIP that can be in the system (Pergher & de Almeida, 2017). A pull control strategy has the benefits of reducing manufacturing costs and cycle time variability, as well as improving quality (Hopp & Roof, 1998). It has the potential to resolve the production challenges that the bicycle chain industry are currently facing.
The most straightforward pull strategy to control WIP is to establish a limit on the WIP in the production line, called constant work in progress, i.e., Constant Work-in-process (CONWIP) (Spearman, Woodruff, & Hopp, 1990). The CONWIP protocol means every time a job departs a new job is introduced to the line, which results in a constant WIP level.
The potential benefits of a CONWIP/pull system is well-addressed in the literature (Huang, Chen, Wang, & Shi, 2015).
However, its applicability in solving a practical problem is not well-documented in the literature, particularly, for a complex manufacturing system. To address the gap, this study attempted to solve a pull system design problem of a real world bike chain manufacturer through two CONWIP alternatives. Line segmentation and WIP upper limit decisions are explored to optimize system performance, including lead time and WIP. By the use of a practical application, the conceptual pull control strategy can be realized in a practical sense.
2. LITERATURE REVIEW
CONWIP is introduced by Spearman et al. (1990) as another form of pull production systems. It is one of the most studied pull production control strategies (Onyeocha, Wang, Khoury,
& Geraghty, 2015). One important feature of CONWIP is that it sets an upper limit (WIP cap) on the inventory of a system. When the production line is complex, CONWIP supports monitoring and controlling the performance of WIP.
CONWIP is suitable for make-to-order production as it does not hold inventory at every stage in the production line (Prakash & Chin, 2015).
In the literature, studies were conducted to discuss CONWIP protocols (Korugan & Gupta, 2014; Pergher & de Almeida,
2017), the advantages (Geraghty & Heavey, 2004), and the potential improvements compared with Kanban and push systems (Gong, Yang, & Wang, 2014). CONWIP had been successfully applied to different manufacturing scenarios, e.g., job-shops, assembly lines, as well as manufacturing layouts, etc. Research showed that CONWIP outperformed other production systems particularly in machine failures (Hopp &
Spearman, 2011) and set-up time/cost reductions (Chang &
Yih, 1994; Herer & Masin, 1997).
Over the years, CONWIP systems had evolved and modified that included new features to fine-tune the overall performance of production systems. The multi-CONWIP system (i.e., segmented CONWIP) is one of the modified systems that drew significant research attention (Kırkavak &
Dinçer, 1999; Yang, Fu, & Yang, 2007). It presents system configurations between conventional CONWIP systems and Kanban systems. In a multi-CONWIP system, a production line is divided into segments - i.e., differing CONWIP systems, each containing several workstations. The movement of WIPs within a segment is governed by each individual CONWIP system. However, the movement of WIPs between segments is governed by a Kanban system.
Simply put, in a multi-CONWIP system, there are more than one but less than the total number of workstation sets of cards circulating in different sections of a line and workload can be controlled and balanced among various workstations.
Little research has been done in investigating how different CONWIP implementation strategies influence system performance (Framinan, Gonzalez, & Ruiz-Usano, 2003), especially that the philosophy for developing reasonable CONWIP alternatives remained unclear (Huang et al., 2015).
Most research for CONWIP/pull strategies employed conceptual models to demonstrate their effectiveness (Yang et al., 2007). They failed to address the integration of CONWIP in real world manufacturing situations (Romagnoli, 2015; Yang, Hsieh, & Cheng, 2011). This suggests the need to bridge the gap between the academia and the reality.
Therefore, further steps become critical in the multi-CONWIP research communities.
3. THE CASE STUDY PROBLEM
To study the efficacy of multi-CONWIP strategies in the real world production system, our study selected a case company in Taiwan that manufactures high-end bicycle chains. A bike chain consists of a series of outside and inside links joined by pins and rollers. The company adopts the make-to-order system to avoid holding WIPs. The company keeps rolls of chains in the warehouse to respond to customer orders. Our study selected one product (occupying 80% of the total production) as the case to study the manufacturing process.
The manufacture of chains in this case company generally involved 13 workstations in the manufacture of chains (Table 1). The manufacturing processes in the case company are presented in Figure 1. The typical batch size is 40 kg.
Outer link
Sx Workstation Transportation
S12
Stock point
Fig. 1. The manufacturing processes/lines for outer link, inner link, pin, and roller
Table 1. Descriptions of the workstations and the batch size involved in the manufacture of chains
Workstation # Batch
The case company adopted a manufacturing Enterprise resource planning (ERP) system to schedule bike chain production. In the production lines, supermarkets/racks were placed between the workstations to store WIP. The major challenges of the chain production were that: (1) the company did not set the WIP caps of the stocks on the racks. The product storage containers were not stacked with a first in, first out manner (causing WIP accumulation and lengthy lead time); (2) the stages of cutting/pressing and stamping (i.e., S1
and S2) involved “rapidly” cutting and pressing steel into the shape of inner and outer links and punching holes. However, the product changeover process took time (typically 30-50 minutes), requiring a big amount of stocks to buffer/maintain continuous production; (3) the stages of quenching #2 and surface treatment (i.e., S8 and S10) were required to simultaneously process chain components from differing stages, which could have caused products waiting in queues;
(4) the processing time of chain components at each workstation was different.
4. THE PROPOSED METHODOLOGY AND EMPIRICAL ILLUSTRATION
To deal with the challenges and to ensure the shortest lead time, this study utilized Rother and Shook (2003)’s lean production guidelines as the basis to modify the current production processes. On top of the guidelines, this study employed two multi-CONWIP strategies to investigate line segmentations that presumably were able to prevent WIP accumulation in the company’s long production lines. In our study, four steps were proposed: (1) Determine the takt time;
(2) Select the pacemaker process; (3) Use multi-CONWIP and supermarkets to prevent over-production; (4) Leveling
Production: reduce unevenness to ensure production efficiency. The following explained the two employed multi-CONWIP strategies:
4.1 The Kaizen strategy
Currently the case company was running Kaizen events on a monthly basis. According to the production requirements, S1
and S2 had to perform batch production because of the long setup time and the short processing time for cutting, pressing, and stamping the steel to form the required product features.
Stock points were thus set up next to S1 and S2 to store the batch products (WIP). In addition, a stock point was also set up at the last workstation of the production line for every chain element to control and regulate the production rhythm.
The production segmentation plan is illustrated in Figure 2.
W1 to W8 represent the information flows that describe the amount of WIP in each CONWIP loop.
S1 S2
Sx Transportation Information flow
W1 W2 W3
Workstation Stock point
Fig. 2. The production segmentations based on the status quo Kaizen approach
4.2 The bottleneck strategy
According to the Theory of Constraint (Goldratt, 1990), the throughput of a system is determined by the bottleneck. To increases the throughputs and the efficiency of the process flows, in this approach, stock points were set up in front of the bottleneck workstations of the manufacturing lines of the four bike chain components (for multi-CONWIP
segmentation). To find out the bottlenecks in the manufacturing lines, one of the ways is to identify the workstations where WIP highly cumulates. However, in the real world environments, bottlenecks could have shifted by the various processing times of different types of products (Chakravorty & Atwater, 2006).
In this study, we performed discrete event simulation (with Rockwell Arena ® 14.0, simulation software) to determine the bottlenecks in the case company’s push production system. Note that other simulation models may also viable as a potential future investigation (Terkaj, Toliov, & Urgo, 2015). Results of the simulation showed that the bottleneck for the outer link production line was S5; the bottlenecks for the inner link, the pin, and the roller production lines was S8. In Figure 3, bottleneck workstations are marked with slashes;
stock points are inserted in the front as buffer. Figure 3 also shows that there is no stock point in front of S8 in the roller
production lines because we assumed unlimited supply of raw materials. W1 to W7 represent the information flows that describe the amount of WIP in each CONWIP loop.
S1 S2
Sx bottleneck
W1 W2
W4 W3
W5 W6
W7
Sx Transportation Information flow
Outer link
Inner link
Pin
Roller
Workstation Stock point
Fig. 3. The production segmentations based on the bottleneck approach
For the purpose of the study, the WIP (i.e., the WIP cap－the number of Kanban, each representing 40 kg) in each
CONWIP loop was estimated using Little’s law (Hopp &
Spearman, 2011) (WIP = cycle time × throughput). The WIP caps however for the Kaizen approach were determined by the existing Kanbans in the shop floor. For WIP calculation, the cycle times of the workstations were obtained from the case company. The throughput was determined by the throughput of S8 as it was the bottleneck workstation for manufacturing both pins and rollers.
5. SIMULATION AND EMPIRICAL RESULTS A current state simulation model was built in Rockwell Arena ® 14.0. The output analyzer was used to monitor the volume of WIP. The output analyses were performed under a steady-state condition. The simulation was performed in a discrete-event simulator, replications of experiments were required to ensure the variation of the simulated throughputs did not go beyond an appropriate confidence interval. The verification of the simulation model was performed and confirmed using the animation feature in Arena ® 14.0 that checked whether a product was processed in the same path as the one in the real manufacturing process. The validation of the simulation model was performed and confirmed by comparing the output of the simulation with the throughput of the real manufacturing system. The
simulation-optimization model we used is shown in the following to minimize the number of Kanbans (i.e., the volume of WIP):
Min � � 𝑊𝑊𝑗𝑗 𝑁𝑁 𝑗𝑗 =1 4
𝑖𝑖=1
Decision variables: i: i=1 means outer link, i=2 means inner link, i=3 means pin, i=4 means roller; j: j denotes the jth CONWIP for manufacturing the ith chain component, j=1, 2, 3,…Ni; k: k denotes the kth product of the ith chain component, k=1, 2, 3, 4, 5, 6, 7, 8, 9; Ni: the number of workstations for manufacturing the ith chain component; Wij: the number of Kanban for the jth CONWIP under the ith chain component, ∀ j=1,2,3,…Ni; Fk: the volume of the kth product undelivered to
the customer; Ok: the volume of customer order for the kth product.
Subject to
𝑂𝑂𝑘𝑘− 𝐹𝐹𝑘𝑘
𝑂𝑂𝑘𝑘 ≥ 0.95, ∀𝑘𝑘 = 1,2, 3, 4, 5, 6, 7, 8, 9 Ni, Wij, Fk, Ok ≥0, ∀ 𝑖𝑖, 𝑗𝑗, 𝑘𝑘 = integer
Table 2 shows the total WIP and the average lead times by implementing differing multi-CONWIP strategies. In Table 2, the total WIP was the total number of Kanbans times 40 (kg).
From Table 2, two multi-CONWIP strategies perform better than the current push manufacturing system in the total WIP volume and the average lead time. The improvements on WIP using multi-CONWIP approaches are generally over 40%. The reason that the impact of using the Kaizen
approach is 43.3% is that, for the longest production line (i.e., manufacturing pins), the Kaizen approach (compared with the bottleneck strategy) only inserts one stock point to the end of the production line, making its impact on WIP reduction smaller. Table 2 also shows that the Kaizen strategy outperforms the bottleneck strategy in the lead time reduction (up to 42.8%). The reason is that the bottleneck strategy deals with more line segmentation than the Kaizen approach. Thus, the impact of the Kaizen strategy on extending the lead time is relatively smaller.
Table 2. The WIPs for the two multi-CONWIP strategies - Before and after the simulation (initial vs. minimal), and the
improvements Current push
mfg. The Kaizen
strategy The bottleneck strategy
6. CONCLUSIONS
The purpose of this study was to introduce CONWIP pull system to the bike chain production line of a case company manufacturing high-end bike chains. The onsite observation revealed that the company did not set the upper stock limit among the workstations which caused the waste of
overproduction and excessive amounts of WIP in production.
Two multi-CONWIP strategies are proposed to conduct line segmentation that prevent WIP accumulation in the
production lines. Simulation was performed on the
performance of the two multi-CONWIP strategies regarding the average lead time and the total WIP of all production lines. The outcomes of the simulation was compared with the
performance of the case company’s real production system.
Results of the comparisons showed that using the Kaizen strategy and the bottleneck strategy are more effective than the current production system in reducing the average lead time and the WIP volume. These results demonstrated that a multi-CONWIP pull system can better satisfy customer demand. Despite that the outcomes of the study were based on the presumptions that there was no equipment failure, no defects in production with unlimited raw material supply, and no material handling, the results are valuable and suggest the applicability of adopting multi-CONWIP in enhancing the performance of the production systems in the real world.
ACKNOWLEDGEMENT
The authors thank the anonymous company for providing the case study. This work was supported, in part, by the Ministry of Science and Technology, Taiwan, under grant MOST-106-2221-E-006-162-MY3.
REFERENCES
Chang, T.-M., & Yih, Y. (1994). Generic kanban systems for dynamic environments. The International Journal of Production Research, 32(4), 889-902.
Chakravorty, S. S., & Atwater, J. B. (2006) Bottleneck management: theory and practice, Production Planning
& Control, 17(5), 441-447.
Framinan, J. M., Gonzalez, P. L., & Ruiz-Usano, R. (2003).
The CONWIP production control system: review and research issues. Production Planning & Control, 14(3), 255-265.
Geraghty, J., & Heavey, C. (2004). A comparison of hybrid push/pull and CONWIP/pull production inventory control policies. International Journal of Production Economics, 91(1), 75-90.
Goldratt, E. M. (1990). Theory of constraints: North River Croton-on-Hudson.
Gong, Q., Yang, Y., & Wang, S. (2014). Information and decision-making delays in MRP, KANBAN, and CONWIP. International Journal of Production Economics, 156, 208-213.
Herer, Y., & Masin, M. (1997). Mathematical programming formulation of CONWIP based production lines; and relationships to MRP. International Journal of Production Research, 35(4), 1067-1076.
Hopp, W. J., & Roof, M. L. (1998). Setting WIP levels with statistical throughput control (STC) in CONWIP production lines. International Journal of Production Research, 36(4), 867-882.
Hopp, W. J., & Spearman, M. L. (2011). Factory physics:
Waveland Press.
Huang, G., Chen, J., Wang, X., & Shi, Y. (2015). A
simulation study of CONWIP assembly with multi-loop in mass production, multi-products and low volume and OKP environments. International Journal of Production Research, 53(14), 4160-4175.
Kırkavak, N., & Dinçer, C. (1999). The general behavior of pull production systems: The allocation problems.
European Journal of Operational Research, 119(2), 479-494.
Korugan, A., & Gupta, S. M. (2014). An adaptive CONWIP mechanism for hybrid production systems. The International Journal of Advanced Manufacturing Technology, 74(5-8), 715-727.
Market Research Store. (2017). Global Bike Chains Market Research Report 2017 Retrieved from Deerfield Beach, FL: http://www.marketresearchstore.com/report/global-bike-chains-market-research-report-2017-2022-163169 Onyeocha, C. E., Wang, J., Khoury, J., & Geraghty, J. (2015).
A comparison of HK-CONWIP and BK-CONWIP control strategies in a multi-product manufacturing system. Operations Research Perspectives, 2(Supplement C), 137-149.
Pergher, I., & de Almeida, A. T. (2017). A multi-attribute decision model for setting production planning
parameters. Journal of Manufacturing Systems, 42, 224-232.
Prakash, J., & Chin, J. F. (2015). Modified CONWIP systems:
a review and classification. Production Planning &
Control, 26(4), 296-307.
Romagnoli, G. (2015). Design and simulation of CONWIP in the complex flexible job shop of a Make-To-Order manufacturing firm. International Journal of Industrial Engineering Computations, 6(1), 117-134.
Rother, M., & Shook, J. (2003). Learning to see: value stream mapping to add value and eliminate muda: Lean Enterprise Institute.
Spearman, M. L., Woodruff, D. L., & Hopp, W. J. (1990).
CONWIP: a pull alternative to kanban. The
International Journal of Production Research, 28(5), 879-894.
Terkaj, W., Tolio, T., Urgo, M. (2015) A virtual factory approach for in situ simulation to support production and maintenance planning. CIRP Annals Manufacturing Technology, 64(1), 451-454.
Yang, T., Fu, H.-P., & Yang, K.-Y. (2007). An evolutionary-simulation approach for the optimization of multi-constant work-in-process strategy—A case study.
International Journal of Production Economics, 107(1), 104-114.
Yang, T., Hsieh, C.-H., & Cheng, B.-Y. (2011). Lean-pull strategy in a re-entrant manufacturing environment: a pilot study for TFT-LCD array manufacturing.
International Journal of Production Research, 49(6), 1511-1529.